Zhaoxin x86-Prozessoren

[Tralalak]

27:58 Why Intel Wanted Centaur

omyOeMP7sjQ

[Tralalak]

Performance predictions of the Zhaoxin server Kaisheng KH-40000 Series SoC. and desktop Kaixian KX-7000 Series SoC.


http://tiebapic.baidu.com/forum/w%3D580%3B/sign=1797a8348b1b9d168ac79a69c3e5b58f/a9d3fd1f4134970a511f2ac4d0cad1c8a6865de8.jpg

https://tb-video.bdstatic.com/tieba-smallvideo-transcode-cae/5658734_c376cd6efc17266eba212ceacbcd4cb9_0_cae.mp4?vt=0&pt=3&ver=&cr=3&cd=3&sid=&ft=2&tbau=2021-11-15_fa81a523734bb0adc170d80b3496bf693311e89d68b88004e8713718eea912d9&ptid=

[robbitop]

To me this looks like intel hiring more engineers without Centaur's patents and stuff. This way they get more manpower and circumvent any monopoly rules, becuase Centaur still exists. Doesn't matter if it is an empty husk.


Makes me wonder how dependent ZhaoXhin is on Centaur nowadays? Or can they stand on their own making complete CPUs?
Hiring people is a different thing than buying something. Slavery isn't a thing anymore. ;)
You can just poach people offering better conditions, salary etc. Nearly everybody has a price tag to leave a company. That shouldn't cost anything near that amount for the amount of people Centaur has in the RnD department.
Intel seemed to have acquired some kind of a business group out of VIA. Question is what was part of that deal. For sure it is not just staff.

[Kinoco]

Performance predictions of the Zhaoxin server Kaisheng KH-40000 Series SoC. and desktop Kaixian KX-7000 Series SoC.


Very interested to see KX-7000 predictions, but it returns "403 Forbidden" and "401 Authorization Required".

[rasti666]

https://www.tomshardware.com/news/chinese-xindong-fenghua-gpu-announced

Another manufacturer after VIA eH1 plans to enter a graphics card on PCI-E?

[y33H@]

It's PowerVR.

[Tralalak]

Centaur Trots Off Into the Sunset
After a 26-year run, Centaur Technology has reached the end of the trail. Glenn Henry’s processor-design team, known for its x86 expertise, has become part of Intel.
source: https://www.linleygroup.com/mpr/index.php?j=MPR

[Benutzername]

27:58 Why Intel Wanted Centaur

https://youtu.be/omyOeMP7sjQ


Thank you. They are mostly guessing as well, but the angle for the very efficient x86 cores that Centaur was working on for machine learning and stuff to get tighter stacking in server racks makes some sense in combination with the Xe stuff for data centres.


I wonder why the FTC has not said anything yet? There are only two x86 makers left in the USA now. Whereas it looks very hard at the nVidida ARM deal.

[Loeschzwerg]

Centaur Trots Off Into the Sunset
After a 26-year run, Centaur Technology has reached the end of the trail. Glenn Henry’s processor-design team, known for its x86 expertise, has become part of Intel.
source: https://www.linleygroup.com/mpr/index.php?j=MPR

==>

Centaur Trots Off Into the Sunset

December 7, 2021
Author: Linley Gwennap

After a 26-year run, Centaur Technology has reached the end of the trail. Glenn Henry’s processor-design team, known for its x86 expertise, has become part of Intel. That company agreed to pay Via Technologies, which has owned Centaur since 1999, $125 million to recruit its engineers. Via retains Centaur’s intellectual property (IP), including its recent CHA design, although we don’t expect the company to market that server processor without an engineering team to support it.

...

Anyone has a subscribtion for Linley Group News to get the full information?

[Tralalak]

Rumours: Zhaoxin KH-40000 Series and KX-6000G comming soon!

[Loeschzwerg]

Yes, rumors say official launch of KX-6000G and KH-40000 is this month. I guess we will see products mid/late 2022.
Sidenote: It is also rumored that Glenfly is releasing it's GPU based product this month.

Well, we will see :)

[Tralalak]

Centaur Technology Global Online Auction
January 25–27, 2022
Austin, TX
source: http://www.siliconhillsauctions.com/centaur-sale-details.php

C'est La Vie... :comfort:

It would be interesting to have it in your collection. If is of course, CentaurHauls.

http://www.siliconhillsauctions.com/images/centaur-auction/17.jpg


http://www.siliconhillsauctions.com/images/centaur-auction/16.jpg

[Loeschzwerg]

Centaur Technology Global Online Auction
January 25–27, 2022
Austin, TX
source: http://www.siliconhillsauctions.com/centaur-sale-details.php

C'est La Vie... :comfort:

It would be interesting to have it in your collection. If is of course, CentaurHauls.

http://www.siliconhillsauctions.com/images/centaur-auction/17.jpg


http://www.siliconhillsauctions.com/images/centaur-auction/16.jpg

Sad to see it end and seeing those images reminds me of the closure of Fujitsu factory in Augsburg Germany.
Looking at the dual CPU board it says ZX-200 and the mounting is 2011ish style, so I'm sure it's CHA.

[davidzo]

Thats really Sad

So they are closing their own facilities and moving over to Intels Austin Texas offices? That means that their own design capability basically ceases to exist. They might transfer over some of the IP to some sales and lawyer office to generate some additional sales, but basically their engineering department is now becoming an implementation+testing workshop for Intel.
That makes sense for Intel, because they have so many concurrent projects that they could easily give them something less critical, like chipsets, telco, etc.
Currently Intel is execution limited and has to catch up to its own schedules. Icelake-SP, Rocketlake, Sapphire rapids, etc. have all been late because of insufficient ressources in implementation and testing. And this concurrent project count is only rising further down the roadmap since Gelsinger is in charge. Giving one project to the experienced engineers of centaur to free up the ressources for the others is a good idea. It is easier to by a team that knows how to complement each other than to create a new one from scratch.

I mean its also nice for the centaur engineers, because they will finally have access to the best equipment in the world and almost unlimited ressources, something which never happened during the last two decades at centaur. But its also sad for them because they give up their independence and freedom of design decisions.

[Skysnake]

If Intel need some more Chip designers, just send me a job offer ;)

[Kinoco]

Does anybody know the memory speed KX-7000 is planned to support?
I read it was initially planned for DDR5, but is now planned for DDR4 as KX-6000.
But KX-6000 is limited to DDR4 2666. Maybe KX-7000 will support DDR4 3200? (no XMP obviosuly)

Although from the article below, maybe other improvements (such as cache) may be required in order to get benefit from increased memory speed.
https://chipsandcheese.com/2021/11/16/zhaoxin-part-3-a-sort-of-anti-climax/

For anyone that missed it, that is 3rd and final part of the article linked on page 36 (part 1 and part 2)

[Tralalak]

New x86 processors from ZHAOXIN @ 2022

https://0.rc.xiniu.com/g4/M00/31/F6/CgAG0mHUCViAM3L8AADOm06re3M859.jpg
source: https://www.zhaoxin.com/news_view.aspx?nid=2&typeid=267&id=1397

[Kinoco]

New x86 processors from ZHAOXIN @ 2022
source: https://www.zhaoxin.com/news_view.aspx?nid=2&typeid=267&id=1397

Thanks for the link.

"In 2022, server processors and desktop processors based on a new independent architecture design are ready to come out, which will further enrich our product line and provide strong guarantees for the smooth migration of users' applications and information security."

They talk about new, independent and migration, which could suggest not x86. But you state "x86", so is it only the awaited KX-7000, or is it a change of path?

Coming in 2022 means this was in the pipeline already, and not a reply to the recent Centaur/Intel deal.

[Tralalak]

Kinoco: The Kaisheng KH-40000 Series server processor will reportedly be based on the Centaur CNS microarchitecture.

The Kaixian KX-6000G desktop and mobile processor will reportedly evolve the KX-6000 Series with a new integrated graphics processor.

[Loeschzwerg]

I'm very excited about both and I hope we will also see a KX-7000 with CNS design.

Processor shown on the left is KH-40000 and on the right I'd guess KX-6000G.

[y33H@]

Oh wow, that's a nice addition to all the other CES stuff - thanks!

[Kinoco]

Kinoco: The Kaisheng KH-40000 Series server processor will reportedly be based on the Centaur CNS microarchitecture.

The Kaixian KX-6000G desktop and mobile processor will reportedly evolve the KX-6000 Series with a new integrated graphics processor.

Thanks Tralalak. I hope official announcement soon raher than late 2022, so that we may be able to buy it still 2022.
If KX-7000 is still 1 year till official annoucement, I guess we may not be able to buy it before 2024.

Anyway I am curious to find the offical specs of KH-40000 and KX-6000G.

[Loeschzwerg]

A few boards from the Centtech auction, showen by @brutuscat2 (https://twitter.com/brutuscat2) on Twitter.

VIA VT5971A "Dual Nano":
https://pbs.twimg.com/media/FLwQklTWUAogo7j?format=jpg&name=large

Centaur/Zhaoxin HA001EA1 "CHA/CNS bring-up board" (maybe?):
https://pbs.twimg.com/media/FLwM8wBWYAAkkU5?format=jpg&name=large

Centaur CHA001-C "CHA/CNS mainboard":
https://pbs.twimg.com/media/FLvx0oQXoAQBLUe?format=jpg&name=large

[Loeschzwerg]

Two more boards from @brutuscat2...

Centaur/Zhaoxin CHX001 with KX-5640
https://pbs.twimg.com/media/FLzpWXjVgAQYRjE?format=jpg&name=large

Centaur/Zhaoxin HA006-B "Dual CHA/CNS"
https://pbs.twimg.com/media/FL0_ghwWUAgyD57?format=jpg&name=large

[Loeschzwerg]

Centtech CHA SoC:
https://pbs.twimg.com/media/FL6yQ-gWUAEOaNe?format=jpg&name=large

[=Floi=]

Sieht heftig aus. gibt es dazu daten?

edit
https://www.computerbase.de/forum/threads/via-centaur-cha-ncore-ai-cpu-ein-gesockeltes-lga-paket-ein-neuer-mitbewerber-auf-dem-cpu-markt.1926038/

[Loeschzwerg]

https://browser.geekbench.com/v5/cpu/12878360

Hier gibt es ein frisches Geekbench Ergebnis.

Edit:
CB R23 (8C @ 2.2GHz) -> SC 477 / MC 3802
https://pbs.twimg.com/media/FL-xEOeXwA4kxLI?format=png&name=small

Quelle wie gehabt von @brutuscat2

[y33H@]

EDIT: hatte Loeschzwerg schon ^^

[Loeschzwerg]

Für 2.2GHz ist das R23 Ergebnis recht OK. SC ist durchaus zwischen Ivy und Haswell, taktbereinigt.

[Tralalak]

The new x86 VIA / Centaur CHA processor has been tested. Performance spoils the beats, the release is probably quite uncertain
source: https://www.cnews.cz/novy-x86-procesor-via-centaur-cha-byl-otestovan-vykon-kazi-takty-vydani-je-asi-dost-nejiste/

[konkretor]

https://twitter.com/brutuscat2/status/1506049246016987143?s=20&t=IJr4yWfkifJ-a6nVktrMNg

[Loeschzwerg]

Very interesting articel from Chips&Cheese about the Centaur CNS:
https://chipsandcheese.com/2022/03/23/via-part-4-a-deep-dive-into-centaurs-last-cpu-core-cns/

[konkretor]

https://pc.watch.impress.co.jp/docs/column/hothot/1399623.html



Verified whether the x86 "KX-U6780A" manufactured by Zhaoqin, China, has "Core i5-class performance

[Tralalak]

Sneak peak: ZHAOXIN Kaisheng KH-40000 Series

https://0.rc.xiniu.com/g4/M00/47/15/CgAG0mJXwJKANRaOAAG8BesnTC0690.jpg

source: https://www.zhaoxin.com/news_view.aspx?nid=2&typeid=267&id=1419
source: https://tieba.baidu.com/p/7796218400

[Tralalak]

7-ZIP single-thread benchmark @ KH-40000 (16nm, 12C, unknown but my estimate "2.0GHz or 2.2GHz") vs KX-6580 (16nm, 8C, 2.5GHz)

4 277 MIPS vs 2 848 MIPS
+50,18% vs. 66,59%

source: https://tieba.baidu.com/p/7790881954
source: https://i.imgur.com/yP9wVUv.png


The results show the difference of a big core based on the Centaur CNS microprocessor architecture @ KH-40000 Series (ZX-F)

[Tralalak]

CPU-Z 2.01 and Glenfly Arise GT10C0 GPU. April 13th, 2022
CPU-Z version 2.01 has been released. This version adds the support of the Glenfly Arise-GT10C0 GPU. Hatched from Zhaoxin, Glenfly is an independant company that designed the Arise GPU and proposes it on its own PCIe video cards. The GT10C0 GPU is manufactured with a 28 nm process, and works with 2 GBytes of DDR4 memory.

source: https://www.cpuid.com/news/74-cpu-z-2-01-and-glenfly-arise-gt10c0-gpu.html
source: https://www.cpuid.com/softwares/cpu-z.html

HWiNFO upcomming changes
Added Glenfly Tech GF ARISE GPU
source: https://www.hwinfo.com/version-history/

[Tralalak]

Examining Centaur CHA’s Die and Implementation Goals


https://i0.wp.com/chipsandcheese.com/wp-content/uploads/2022/04/cns_on_hsw-e.jpg?resize=768%2C822&ssl=1
mspaint-assisted visualization of a high core count CNS implementation. This image would occupy 372 mm2

source: https://chipsandcheese.com/2022/04/30/examining-centaur-chas-die-and-implementation-goals/?fbclid=IwAR1ItPFfuJvj4um-3dLq0o_K6WjG-nMbJaCenThF9-_4fub0l5YZp9PUqp0

P.S. Centaur CNS cores @ (Zhaoxin Kaisheng KH-40000 Series adopded this x86 microarchitecture up to 32 cores on two die up to 64 cores)

[Loeschzwerg]

An interesting move, "Russia to use Zhaoxin":
https://www.techpowerup.com/295016/russia-to-use-chinese-zhaoxin-x86-processors-amidst-restrictions-to-replace-intel-and-amd-designs

Good for Zhaoxin I guess :)

[konkretor]

https://www.golem.de/news/prozessor-russisches-motherboard-mit-x86-cpus-von-zhaoxin-angekuendigt-2205-165500.html

[Sardaukar.nsn]

DDR4, M2 Slot, 8 Kerne klingt ja erstmal okay für einen "normalen" PC. Kann man die Leistung ungefähr mit einem älteren AMD bzw. Intel vergleichen?

[Unicous]

Das Ding ist auf Niveau 2010/2011, das ist eine Verzweifelungstat. Ob man die restlichen Komponenten zusammenbekommt ohne gegen die Sanktionen zu verstoßen halte ich auch für Wunschdenken.

[][immy]

Das Ding ist auf Niveau 2010/2011, das ist eine Verzweifelungstat. Ob man die restlichen Komponenten zusammenbekommt ohne gegen die Sanktionen zu verstoßen halte ich auch für Wunschdenken.
Nicht nur das. Wenn sie die Sanktionen unterwandern, landen sie ziemlich schnell auf der Sanktionsliste der USA, was sie dann auch die Lizenzen kosten dürfte. Frag mal Huawei wie das so ist dort drauf zu stehen.
Zumal dann auch Zulieferer Gefahr laufen dort drauf zu kommen.

Und ich muss ehrlich sagen, ich habe in den letzten 20 Jahren abseits von AMD & Intel keinen stabilen x86er-Kompatiblen Prozessor gesehen, der auch stabil mit bestehender Software läuft.

[konkretor]

https://www.phoronix.com/scan.php?page=news_item&px=Zhaoxin-LuJiaZui-GCC-May


Zhaoxin Tries Again To Upstream Their "LuJiaZui" CPU Support Within GCC

[Tralalak]

De ja vu for me

-KHxXS7wHuU

[Tralalak]

Glenfly Arise1020 GPU with KaiXian processor
source: https://browser.geekbench.com/v5/compute/4887398

_______


SIV - System Information Viewer Version 5.64 released on 14-May-2022

Added Glenfly Arise-GT-10C0 (http://rh-software.com/siv_glf.png) + Arise-GT-1040 + Arise1020 + Arise1010 Graphics Controller support.
Added Nuvoton NCT6799D support. Added Zhaoxin KX-6000 (Lujiazui) CPU + ZX-E chipset support.
Added ASRock X399 Taichi + Shanghai Zhaoxin HX002EA motherboard support.

source: http://rh-software.com/

[Loeschzwerg]

bkHk9w8mE1E

Further benchmark results an images can be found here:
https://twitter.com/ultrawide219

https://pbs.twimg.com/media/FUhuDU9UAAAeYXh?format=jpg&name=large
https://pbs.twimg.com/media/FUhuEC4UcAA5ffA?format=jpg&name=large
https://pbs.twimg.com/media/FUblP_fUAAAIkVJ?format=jpg&name=large

[Tralalak]

First image Glenfly

https://tieba.baidu.com/photo/p?kw=%E5%85%86%E8%8A%AF&ie=utf-8&flux=1&tid=7874098618&pic_id=a6b9bcfc5266d016d584bff1d22bd40734fa3529&pn=1&fp=2&see_lz=1

[Loeschzwerg]

Original source:
https://www.bilibili.com/read/cv17048665

Device manager is said to report "Zhaoxin VGA bios". This specific Arise-GT10C0 card is equipped with 2GB (@ 64bit) and sadly no drivers are currently available. GPU was produced week 42 year 2021.

[Tralalak]

ZHAOXIN_ZHANGJIANG = ZX-C / ZX-C+
ZHAOXIN_WUDAOKOU = KX-5000 / KH-20000
ZHAOXIN_LUJIAZUI = KX-6000 / KH-30000
ZHAOXIN_YONGFENG = KH-40000 / KX-7000

[Loeschzwerg]

China’s Domestic Zhaoxin KH-4000 16 Core CPU Offers Single-Core Performance Close To AMD’s 1st Gen Zen Chips
https://wccftech.com/china-domestic-zhaoxin-kh-4000-16-core-cpu-offers-single-core-performance-close-to-amd-1st-gen-zen-chips/

https://www.forum-3dcenter.org/vbulletin/attachment.php?attachmentid=79690&stc=1&d=1655891926

Original Source:
https://news.mydrivers.com/1/840/840971.htm

[Tralalak]

Centaur CHA performance results

Cinebench R11.5 / R20
WPrime
SuperPi

source: https://www.techpowerup.com/forums/threads/new-high-performance-x86-compatible-microprocessors-from-centaur-via.259161/page-6#post-4780211

[Loeschzwerg]

Zhaoxin chip history:
https://www.forum-3dcenter.org/vbulletin/attachment.php?attachmentid=79703&stc=1&d=1656082097

It's a screenshot from an animated gif of the following news (at the bottom):
https://news.mydrivers.com/1/841/841373.htm

---

[Tralalak]

In the middle is the KH-40000

[Loeschzwerg]

Probably both if you ask me.

[Tralalak]

In 2022 only KH-40000

[Loeschzwerg]

That's unfortunately true :( And still no further signs of KX-6000G.

[Tralalak]

KX-6000G is still LUJIAZUI and only new iGPU ZX C1080.
Not much interesting for me... in 2022/2023...

Here is results ZHAOXIN KaiXian KX-6000G @ 3.39 GHz with ZX C1080 iGPU
source: https://browser.geekbench.com/v5/cpu/compare/7767362?baseline=15619785

[Loeschzwerg]

I'm mostly interested in the iGPU =)

[Tralalak]

I know :up:

[Loeschzwerg]

New KH40000 results:
https://browser.geekbench.com/v5/cpu/15706425

[Tralalak]

New KH40000 results:
https://browser.geekbench.com/v5/cpu/15706425
CentaurHauls Family 6 Model 71 Stepping 2 vs CentaurHauls Family 7 Model 11 Stepping 3

CNS vs KH-40000

L1 Instruction Cache is 64KB per core and this is differend from Centaur CNS

L3 Cache is 8MB per eigt core cluster and this is differend from Centaur CNS with 16MB L3 cache per eight core

source: https://browser.geekbench.com/v5/cpu/compare/12878360?baseline=15706425

[Loeschzwerg]

Official specs of the Glenfly Arise GT 10C0 graphics card have been released.

https://www.glenfly.com/index.php?catid=22

- 28nm
- 500MHz GPU
- 1200MHz Mem
- 2GB/4GB DDR4 (64/128bit)
- 1.5TFLOPs FP32
- 48GPixels/sec
- 96GTexels/sec
- PCIe 3.0 x8
- DX11, OGL4.5, OCL 1.2

https://www.forum-3dcenter.org/vbulletin/attachment.php?attachmentid=79949&stc=1&d=1658064112

[DrFreaK666]

Ganz schön altbacken

[konkretor]

Bin gespannt bis es die zum kaufen gibt

[Loeschzwerg]

Der Markierung auf dem Bild nach ist die GPU schon fast ein Jahr alt (KW 33 2021).

---

Edit:
Further images/info about the Glenfly Arise GT-10C0:
https://bxc.com.cn/productinfo25.html

According to this site the card does 3241 points in Glmark2. Resolution used is unclear, but a Radeon 520 on the same site scores 1055.

Power consumption of the GT-10C0 is 45W.

79959 79960 79957 79958

=> https://twitter.com/Loeschzwerg_3DC/status/1548906355469258753?cxt=HHwWgsC4wfaH6f4qAAAA

[konkretor]

https://www.computerbase.de/2022-07/glenfly-arise-gt10c0-specification/

https://videocardz.com/newz/chinese-glenfly-arise-gt-10c0-gets-official-specifications-1-5-tflops-of-28nm-gpu-power

[Benutzername]

Ganz schön altbacken


Naja 28 nm ist 2011 auf den markt gekommen. diese grakas gab es damals:

https://www.3dcenter.org/artikel/grafikkarten-marktueberblick-dezember-2011


die 500 mhz sind ein bisschen lahm von der glenfly.

[Tralalak]

But there are also 28nm graphics cards

GeForce 900 series

• • Mid-range
• GeForce GTX 950
• GeForce GTX 960

• • High-end
• GeForce GTX 970
• GeForce GTX 980

• • Enthusiast
• GeForce GTX 980 Ti
• GeForce GTX Titan X

source: https://en.wikipedia.org/wiki/GeForce_900_series


Radeon HD 7000 series

• • Entry-level
73xx - 76xx

• • Mid-range
• 7750
• 7770
• 7790
• 7850
• • High-end
• 7870
• 7870 XT
• 7950
• 7970

• • Enthusiast
• 7990

source: https://en.wikipedia.org/wiki/Radeon_HD_7000_series

But yes, it would need at least 16nm

[Tralalak]

KX-6000G
Zhaoxin ZX-1080C iGPU benchmarked
source: https://firstxw.com/view/300728.html

[ling dian]

Next month, there will be a conclusion meeting of Zhaoxin to summarize the past two years. It may release 6000g/40000, and there will be surprises next year. I don't know if I can understand this paragraph

[ling dian]

Gt10c0 graphics card Glmark2 running score 3241 resolution probability is 1280×1024, see the running score is GTX660 / HD7850, but the actual performance under the DX internal message is GTX750, may be the driver or D4 video memory is not good, anyway, the minimum guarantee also has 1030 D4 performance, This card has been shipped to the government, private to rely on second-hand platform outflow

[konkretor]

https://videocardz.com/newz/chinese-zhaoxin-based-gpu-is-reportedly-as-fast-as-10-year-old-geforce-gt-630

[Tralalak]

ZHAOXIN KaiXian KX-6000G (4C 3.2GHz, 4MB L2) with ZX C1080 GPU (SM6.0, 512MB, PCI) (D3D 11)
source: https://ranker.sisoftware.co.uk/show_device.php?q=c9a598c28acb84dc95dbfbb0d1b8e089e886a6edb598ae9eae9ed9ff98a588 b99fedd0e0c6af92a385edd0e5c3bb86b791f491ac9cbac9f4cc&l=en
source: https://ranker.sisoftware.co.uk/show_device.php?q=c9a598c29abaf9c8f8c0f0d097c792b4d3eec3f2d4a69bab8de4d9ff97aa8c f4c9ef8aefd2e2c4b78ab2&l=en


Video Shader Compute - SisoftSandra ranker
Zhaoxin ZX C1080 GPU (SM6.0, 512MB, PCI) (D3D 11) 172.99Mpix/s (+1526,83%)
Zhaoxin ZX C960 GPU (SM5.0, 512MB, PCI) (D3D 11) 11.33Mpix/s (6,55%)

[Tralalak]

OpenBLAS 0.3.21 released, preliminary support for Zhaoxin/Centaur KH40000 processors.
source: https://www.oschina.net/news/205955/openblas-0-3-21-released

[Tralalak]

Rumours
Dual Die KH-40000 Series processor with 2x 16C/16T = 32C/32T

http://tiebapic.baidu.com/forum/w%3D580/sign=2c9100fd262762d0803ea4b790ed0849/e6530054b319ebc43238c947c726cffc1f17160b.jpg?tbpicau=2022-09-14-05_91628e24829197d034ecac2424592b76

P.S. dual socket solution 2x 32C/32T = 64C/64T

source: https://tieba.baidu.com/p/8016496564

[Loeschzwerg]

Review of the Centaur CHA from Golem.de:
https://www.golem.de/news/centaur-cha-im-test-der-letzte-x86-prozessor-seiner-art-2209-167757.html

[y33H@]

Danke, Loeschzwerg! übrigens mein letzter Artikel für Golem, ich wechsle auf die Dunkle Seite.

[Skysnake]

?????

[Tangletingle]

Ab in die Industrie. PR team eines größeren semis würde ich annehmen.

[Skysnake]

Ja das ist mir auch klar, das es in die Industrie geht. Die Frage war eher wohin und was ;)

Aber definitiv ein Verlust

[Loeschzwerg]

https://forum.golem.de/kommentare/sonstiges/centaur-cha-im-test-der-letzte-x86-prozessor-seiner-art/bye-bye-marc-alles-gute/156338,6351567,6351567,read.html
ich gehe zu Intel als EMEA Comms Specialist for Enterprise und bin somit im selben Unternehmen tätig wie die ehemaligen Centaur Leute ... sofern deren künftiges Tun in einer Server CPU landet (Sierra Forest?), arbeite ich vll sogar mehr oder weniger direkt mit einigen von ihnen zusammen.

Das wäre natürlich cool =) Ich hoffe ja immer noch dass man noch ein paar weitere Details und Hintergrundinfos zum "Deal" bekommt. Aber auch ganz allgemeine Sachen zur Arbeit bei Centaur wären interessant.

[Skysnake]

Wäre ja lustig, wenn daraus was werden würde wie mit den Core i aus Israel

[Badesalz]

Comms for Enterprise??
ich wechsle auf die Dunkle SeiteDas könnte hinhauen. An sich wird es wohl um PR gegen AMD gehen (?) :tongue:

@Skysnake
Jein. Intel will damit die E-cores pimpen um das gegen Arm aufzufahren. Eine CPU für Massiv-MT ohne allzugroße loads pro Kern. Halbwegs spannend, aber eher nicht mit der Story von Penryn vergleichbar. Schätze ich jedenfalls.

[y33H@]

Comms for Enterprise?
Xeon usw ...

[Badesalz]

DAS ist mir klar. Das ist ja da wo AMD grad auch alles kleinzuhauen droht. Da kann man jedwede Verstärkung gut gebrauchen ;)

[Ex3cut3r]

Danke, Loeschzwerg! übrigens mein letzter Artikel für Golem, ich wechsle auf die Dunkle Seite.

Wohin gehts? :redface:

Dunkle Seite? Nvidia?

[y33H@]

siehe #831

[=Floi=]

Es wäre mal interessant irgendwo mehr über die fortschritte der letzten (10?) jahre im serverumfeld zu erfahren.
Auch beispiele aus der preixis über heavy load wäre mal super interessant. zB SAP server eines großen konzerns.

Interviews mit den leuten aus israel zur core geschichte, standort und warum man dort hinging wären sicher auch was für die öffentlichkeit. Das ist der unterschätzteste standort überhaupt imho.

[y33H@]

Das war lustigerweise in Arbeit, hab's aber zeitlich nimmer geschafft.

[Badesalz]

Es wäre mal interessant irgendwo mehr über die fortschritte der letzten (10?) jahre im serverumfeld zu erfahren.Auf x86 bezogen tat sich da nicht viel mehr als für den Consumer. Bis Epycs kamen.
Sonst ist das Thema wie immer das interne Netzwerk an sich ("Interconnects") wichtiger als irgendwelche CPUs, da es weiter einer der Flaschenhälse ist.

Auch beispiele aus der preixis über heavy load wäre mal super interessant. zB SAP server eines großen konzerns.So grob aus der Ecke kommend frag ich mich was es da besonderes zu erfahren gäbe? Wobei, wahrscheinlich massiv mehr abenteuerliches als wenns um IBM Z gehen würde :usweet:

Interviews mit den leuten aus israel zur core geschichte, standort und warum man dort hinging wären sicher auch was für die öffentlichkeit.Wer ging dahin? Da arbeiteten nicht überwiegend Inder und Kanadier. Da ging keiner hin. Man war da ;) Intel hat da so einige hochbegabte gefunden von denen nur sehr wenige auswandern wollten. Also machte man eben da auf um die zu kriegen. Warum gibt es IBM Böblingen denn?

Die Israelis bleiben die ewigen Helden der x86 Welt. Sie haben uns davor bewahrt an Itanium zu verenden. Ich halte den Beitrag von AMD64 für nicht weniger wichtig, aber für kleiner als den der Israelis. Hätte Intel keine solche Alternative, gäbe es nicht die interne Auseinandersetzung und sie hätten Itanic bis aufs Blut verteidigt.

Das ist der unterschätzteste standort überhaupt imho.
Nach Pentium-M ist das ein unterschätzter Standort? Ja, für Leute die sich dafür nicht interessieren. Denen ist das aber eben auch schnuppe.

[Ex3cut3r]

siehe #831

Ups, sorry, da war ich wohl blind. Viel Glück und Erfolg dort. :smile:

Dem 3DCenter bleibst du ja erhalten.

[Tralalak]

siehe #831
Viel Glück

[Badesalz]

Ja. Das kann man besser auch nicht ausdrücken.

Viel Glück :up:

[Loeschzwerg]

KX-6000G prototype with C1080 IGP:
https://www.forum-3dcenter.org/vbulletin/attachment.php?attachmentid=80884&stc=1&d=1665672344

Source:
https://tiebac.baidu.com/p/7951506552

[Loeschzwerg]

KX-6000G running with 3.2GHz:
https://www.forum-3dcenter.org/vbulletin/attachment.php?attachmentid=80938&stc=1&d=1665933382
https://www.forum-3dcenter.org/vbulletin/attachment.php?attachmentid=80939&stc=1&d=1665933382

[Tralalak]

Hell froze over: VIA AMOS-3007 ultra-compact fanless system s 1.5GHz Intel® Elkhart Lake Atom x6413E processor (https://www.intel.com/content/www/us/en/products/sku/207908/intel-atom-x6413e-processor-1-5m-cache-up-to-3-00-ghz/specifications.html) :up:

[Loeschzwerg]

Geekbench 5 Score of the upcoming KX-6000G:
https://www.forum-3dcenter.org/vbulletin/attachment.php?attachmentid=81171

[Loeschzwerg]

81217 81218

Zhaoxin KX-6000G and KH-40000 have been officially released:
https://www.zhaoxin.com/news_view.aspx?nid=2&typeid=268&id=1491

KX-6000G product page:
https://www.zhaoxin.com/prod_view.aspx?nid=3&typeid=581&id=2410

KH-40000 product page:
https://www.zhaoxin.com/prod_view.aspx?nid=3&typeid=582&id=2411

[Tralalak]

Too bad the KX-6000G is only a quad core. But it was known.
So even the Zhaoxin KH-40000 Series presented today did not transfer the AVX-512 support from the Centaur CNS microarchitecture or. Centaur CHA processor.


The 7nm KX-7000 could be big.LITTLE ("Yongfeng + Lujiazui") :D

[robbitop]

Who is acutally designing the DX12 IGP? Is that still S3 Graphics (are they still located in Tremont?). Or is it a GPU licence?

[Loeschzwerg]

Designed by Glenfly, a subsidiary company of Zhaoxin. It's their own IP and not a licensed design.

[Tralalak]

1PSmbjKsXuE

[Tralalak]

https://0.rc.xiniu.com/g4/M00/6D/30/CgAG0mNsopSATA7HAAkNRQWXrt4145.png

Universal Server Donghai XRS302-4162212 2U
Based on 2x Zhaoxin Kaisheng® KH-40000/16 server processor up to 2.7GHz and 32MB L3 cache.

source: https://www.zhaoxin.com/case_view.aspx?nid=4&typeid=230&id=1487

[Tralalak]

Andreas Schilling ����
@aschilling

Unreleased CHA SoC with CNS cores. Was expected to launch for production in late 2020. It was designed for TSMC 16 nm. Die size is 14.59 x 13.68 = 199.59 mm².

- eight CNS cores
- 64/256/2,048 kB L1/L2/L3$
- one NCORE NPU
- 16 MB SRAM
- Ring interconnect
- PCIe 3.0
- 8C DDR4-3200

https://pbs.twimg.com/media/FiPvxmeXoAE5T2T?format=jpg&name=small


source: https://twitter.com/aschilling/status/1595382343996350464
source: https://www.flickr.com/photos/130561288@N04/52498302214/in/photostream/

[Tralalak]

Zhaoxin KaiXian KX-7000 Series ?

https://0.rc.xiniu.com/g4/M00/74/5B/CgAG0mO86UuAeVPpAADMeVut9eo987.jpg
source: https://www.zhaoxin.com/news_view.aspx?nid=2&typeid=267&id=1518

[rasti666]

I am waiting for KX-10000 processors and a Zhaoxin eH-10 graphics card with DirectX 13.

The current KINO-KX-U6580-R10 is the only one available on the market and probably the last one to buy in Europe.

For years I used VIA/S3 processors and graphics cards. Today, it's just a search for Chinese-government devices.

Western sanctions and the Russian invasion of Ukraine, and the support of Putin by China, are taking away the desire to use these pseudo-projects.

Taiwan's VIA - is for me the last thing I valued. Greetings to all fans of VIA and S3. Maybe someday - the world will return to normal - maybe we will live to see normal times...

Thank you all. Farewell.

[Tralalak]

rasti666: I see it the same way.
It was a great pity when VIA was looking for an investor for a joint venture in 2011 and 2012, so Europe fell asleep again.
We could have had European x86/x86-64 processors...

We in Slovakia work in a joint venture Slovak-US company on Tachyum processors.

So I believe that not only the Slovak super computer will be built on these processors.

[Loeschzwerg]

Haier launches the Boyue G43 laptop equipped with KX-6000G:
https://wccftech.com/haier-reveals-boyue-g43-laptops-using-zhaoxin-kx-6000g-cpu-with-integrated-graphics/
https://www.zhaoxin.com/case_view.aspx?nid=4&typeid=227&id=1486

[Tralalak]

Zhaoxin „Kaixian“ PC/embedded processors and „Kaisheng“ server processors

https://0.rc.xiniu.com/g4/M00/84/BB/CgAG0mRIjsWAXNVpAAH8poHuL-E907.jpg
source: https://www.zhaoxin.com/news_view.aspx?nid=2&typeid=272&id=1554

[konkretor]

should we need an extra thread?


https://wccftech.com/chinese-chipmaker-launches-powerstar-x86-cpus-performance-several-times-higher-than-other-domestic-chips/


http://duchuang.sznews.com/content/mb/2023-05/06/content_30215325.html



Powerstar x86

[y33H@]

Rebranded Intel chip? :freak:

[Corny]

Die Mainboards dazu wären spannend - bestimmt nebenbei Intel-kompatibel

[Loeschzwerg]

@konkretor: Ja, vielleicht ein eigenes Sammelthema, ähnlich "China-GPUs" nur eben "China-CPUs".

@y33h@: Davon gehe ich stark aus und wäre letztendlich nicht das erste mal dass da Restbestände irgendwelcher Art (Samples) neu verwurstet werden. Siehe auch die "neuen" Mainboards für alte Sockel.

@Corny: Sind vermutlich gängige Mainboards mit allerhöchstens einem angepassten Bios. Ähnliche Geschichten hatten wir z.B. mit den Mobile CPUs auf Trägern für reguläre Sockel wie z.B. hier: https://de.aliexpress.com/item/1005004587885067.html

[Corny]

Für eine Resteverwertung wäre das aber eine sehr großspurige Präsentation.



Das mit den Mobile-CPUs auf Trägern für Sockel war mir neu.

[Loeschzwerg]

Auch nicht großspuriger als Zhaoxin mit Produkten/Stückzahlen in homöopathischen Dosen. Vielleicht musste Intel irgendein Lager leeren bzw. irgendein Zwischenhändler und dann kam es zu dieser "Idee".
Aber warten wir es mal ab, irgendwann wird sicher jemand genaue Einblicke liefern.

[Fritter]

Kann man Zhaoxin auch wo kaufen ?

[konkretor]

Für eine Resteverwertung wäre das aber eine sehr großspurige Präsentation.



Das mit den Mobile-CPUs auf Trägern für Sockel war mir neu.


scam , wtf

https://wccftech.com/china-powerstar-x86-cpus-confirmed-rebadged-intel-10th-gen-comet-lake-chips/

[Loeschzwerg]

Leider wie befürchtet. Interessant wäre jetzt nur noch zu wissen wie genau es zu diesem Produkt gekommen ist.

[Tralalak]

Glenfly Arise 1020 review
source: https://www.tieba.com/p/8404074287

[Loeschzwerg]

KX-7000 seems to be released December this year:
https://tiebac.baidu.com/p/8688015060

SingleCore 2.6GHz, AllCore 3.0GHz. IPC is expected to be around ZEN1.

[=Floi=]

Ist SC wirklich geringer als MC?
8 kerne?
cache?
Gibt es schätzungen zur TDP?

[Loeschzwerg]

Ist SC wirklich geringer als MC?

Oh nee, das ist ein Tippfehler von mir. Sind 3.6GHz im SC. TDP und Cache sind mir zumindest noch unbekannt.
Generell scheint KX7000 nicht der Überflieger zu werden.

[hikimiru]

KX-7000 is dead. It was CNS on 16nm.
There are only sign of KX-8000. https://lore.kernel.org/lkml/20230731055932.4336-1-TonyWWang-oc@zhaoxin.com/
Not sure, but it is either Yongfeng on 7nm, or new family 8, which sign was already seen https://github.com/OpenMathLib/OpenBLAS/pull/3565/files.

[Tralalak]

Zhaoxin/VIA x86 processors CPU-Z performance from 2005 to 2022 Version 17.01.64

ZX-G / KH-40000 Series / YongFeng 16-core server CPU released by Zhaoxin in 2022 has a single-core 2.5G and a full-core 2.2G.
SC: 252
MC: 3 746,2


vs


ZX-F/CHA / Centaur CNS not released in 2020, 8-core CPU, 2.0G without overclock, 2.5G for single core and 2.5G for all cores after overclock.
SC: 270,1
MC: 2 206,9


vs


ZX-E / KX-6000 Series / LuJiaZui released in 2019, 8-core CPU, single-core 2.7G, full-core 2.7G.
SC: 180,8
MC: 1 421


vs


ZX~D / KX-5000 Series / WuDaoKou released in 2017, 4-core CPU, single-core 2.0G, full-core 2.0G.
SC: 135,7
MC: 524,7


vs


ZX-C+ / Centaur CN-R / VIA Nano X4 released in 2016 has a 4-core CPU, single-core 2.0G and full-core 2.0G. This configuration exceeds the FSB.
SC: 132
MC: 489,4


vs


VIA Nano X2/ZX-A/ZX-B / Centaur CNC released in 2012-2014, 2-core CPU, single-core 1.73G, full-core 1.73G.
SC: 101,3
MC: 202,3


vs


VIA Nano / Centaur CNA released in 2008, single-core CPU, single-core 1.6G.
SC: 86,1
MC: 89


vs


VIA C7-M / Esther released in 2005, single-core CPU, single-core 1.6G.
SC: 5,8
MC: 1

source: https://tieba.baidu.com/p/8647996529

[Tralalak]

ZHAOXIN KaiSheng KH-40000/16 2.2GHz (16C/16T, 2.5 GHz Turbo, 4x8MB L3 cache) benchmarked
source: https://www.zhihu.com/question/538943703/answer/3292939952

[Tralalak]

Zhaoxin Kaixian® KX-7000 Series Processor


https://0.rc.xiniu.com/g5/M00/06/A3/CgAGbGV31GOASTNvAAS2g6SRSgk001.jpg


Zhaoxin launches Kaixian KX-7000 series high-performance desktop processors with x86 Century Avenue architecture


• 8 cores / 8 threads
• maximum working frequency of 3.7 GHz
• 4 MB L2 cache and 32 MB L3 cache
• supports dual-channel DDR5 memory with a maximum capacity of 128 GB
• integrated high-performance C-1190 iGPU graphics card, supports the latest APIs such as DX12, OpenCL 1.2 and OpenGL 4.6, support H.265 HEVC/H.264 AVC. DP/HDMI/VGA output, supports dual-channel 4K
• support for 24x PCIe 4.0, 2x USB 4 + 4x USB 3.2 Gen2 + 2x USB 2.0, 3x SATA Gen3 interface
• compatible with x86 32/64-bit instructions, SSE4.2/AVX/AVX2 extended instructions, SM2, SM3 and SM4 accelerated instructions
• supports virtualization technologies such as VT-x and VT-d 2.5
• LGA 1700 (45 mm x 37.5 mm)


https://0.rc.xiniu.com/g5/M00/06/A1/CgAGbGV3zT2AGRftAAHNBqd8smc615.jpg


source: https://www.zhaoxin.com/prod_view.aspx?nid=3&typeid=593&id=2757

[Loeschzwerg]

Depending on clock KX-7000 SC performance is roughly between 350-400 (CPU-Z).

[scx]

I've been meaning to write a longer analysis for some time, but every time I get closer to finishing the work, I make more and more digressions, so the topic expands infinitely. Currently I'm working on:
- Analysis of the Vulkan driver included in the latest Zhaoxin driver.
- Comparison of Vulkan implementation in Chinese and Western GPU drivers.
- Overview of Zink, including its weaknesses despite full support for OpenGL 4.6, and analysis of which Vulkan drivers are likely to support it in the near future.
- Comparison of Chinese GPUs, primarily in terms of OpenGL support.
- Comparison of American and Chinese GPU/GPGPU solutions, including an overview of CUDA-like ecosystems (NVIDIA [GeForce]: CUDA; AMD [Radeon]: CUDA-like ROCm; Intel [Arc]: oneAPI vs Moore Threads [MTT]: CUDA-like MUSA; Muxi [MXC]: CUDA-like MXMACA; Cambrian [Siyuan]: CUDA-like CNToolKit; Haiguang [DCU]: CUDA-like ROCm; Biren [BR]: CUDA-like BIRENSUPA; Tianshu Zhixin [Zhikai]: CUDA-like corex), importance of NVPTX and SPIR(-V), analysis of what is wrong with Intel oneAPI/SYCL and AMD ROCm/HIP and what is missing in pure OpenCL and Vulkan compute functionality.
- Comparison of SPIR(-V) support in Chinese GPU/GPGPU drivers and analysis of libcx4spirv2nir.so in the Zhaoxin CX4-series driver.
- Evolution of the S3G drivers, including the evolution of the OpenCL driver and related tools since the Destination microarchitecture, and clarification of the relationship between Zhaoxin (Shanghai Zhaoxin Semiconductor), GlenFly (Glenfly Technology Company) and JARI (Jiangsu Automation Research Institude) of the 716 Institute of CSIC (China Shipbuilding Industry Corporation).
- Evolution of S3G-derivatives GPUs from Carrera to Elite 4000AI and CHX005, with a small digression about Number Nine.
- Evolution of Centaur-derivatives CPUs from WinChip to KX-8000.
- Comparison of retro graphics APIs: ATI ATI3DCIF (ATI 3D C InterFace) [ATI3DCIF.DLL], Matrox MSI (Matrox Simple Interface) [MSI95.DLL], NVIDIA NVlib (nVIDIA multimedia library) [NVVIDMOD.DLL]+[NVAPI.DLL], PowerVR SGL (Super Graphics Library) [SGL.DLL][SGL2.DLL], Rendition RRedline [VERITE.DLL]+[REDLINE.DLL], S3 S3d (S3 3D graphics engine) [S3DTKW.DLL] & MeTaL [METAL.DLL], as well as early versions of Direct3D [DDRAW.DLL]+[D3DIM.DLL][D3DIM700.DLL], Glide [GLIDE.DLL][GLIDE2X.DLL][GLIDE3X.DLL] and OpenGL [OPENGL32.DLL].
- Comparison of wrappers and software implementations of 3D APIs, including ATI3DCIF (Euan Kirkhope's ATI3DCIF Wrapper, GLCIF, GLRage), Direct3D (3Dmigoto, d3d8-gl, d3d8to9, D3D9On12, D3D11On12, D8VK, d912pxy, DDrawCompat, dgVoodoo2, DX9Res, DXGL, DXOverride, DXUP, DXVK, DxWnd, ENBConvertor, DxWrapper, Mesa D3D1X, Mesa Gallium Nine, Pixomatic, SoftWire, Special K, TransGaming SwiftShader, VKD3D, VKD3D-Proton, VorpX, WARP, WineD3D), Glide (BigRRed, Boost Wrapper, Clide, Creative Labs Unified Driver, dgVoodoo, dgVoodoo2, DirectGlide, eVoodoo, GL2Ideal, GL-Glide, GliD3D, GlideGL64, GlideXP, Glidos, Hacktarux's Glide Wrapper, Kshaikh's Glide Wrapper, Glitch64/Glide64, MGlide, MUGlide, nGlide, OpenGlide, psVoodoo, Sven's Glide Wrapper, Trey Harrison's Glide Wrapper, uglide/micro-glide, XGl200/Scott's Glide Wrapper, XGlide, Zeckensack's Glide Wrapper), Mantle (grvk), Vulkan (Metal), OpenGL (3dfxVGL, AltOGL, ANGLE, DawnWrapper, dgles, GateGL, GL4ES, GLDirect, GLon12, GLOverride, Mesa D3D12, Mesa LLVMpipe, Mesa SoftPipe, Mesa SWR, Mesa VirGL, Mesa Zink, Mesa3D for Windows + MesaInjector, MesaFx, OpenGLOn12, QindieGL, s3mesa, SGI Cosmo, ThunderGL, TitaniumGL, VKGL, WickedGL), RRedline (RReady), SGL (PowerSGL to SGL Direct2 Backward Compatibility Layer) and Vulkan (Google SwiftShader, Mesa DZN, Mesa LVP, Mesa VN), including a side-by-side comparison of WARP and Mesa LLVMpipe & SoftPipe & SWR performance in PC gaming.
- Analysis of the new compiler from Zhaoxin: ZX-GCC, based on GCC 11.3.
- Analysis of the new version of zx-gmilib - ZHAOXIN GMI library (GuoMi Instruction), with SM2/SM3/SM4 support.

[scx]

Vulkan

Some time ago I informed about the availability of a new Zhaoxin graphics driver with support for Vulkan API, but I guess the news did not reach here.
Anyway, the zhaoxin-cx4-linux-graphics-driver-dri_26.00.00-0kylin1_amd64.deb driver for the Zhaoxin C-1080 GPU was build on October 19 (2023-10-19) from the "source_gfx_Rel_CHX004_CHX005-linux-release" sources and was released on November 2nd (2023-11-02).
It supports Vulkan 1.1. Below I will try to present a slightly broader perspective.

Chinese GPUs

CSMicro (LingJoWing/CSMicro Corporation): http://www.ljmicro.cn

- csmicro-deb 1.494.8-0kylin5 loongarch64 (2022-06-27) [GP101/GP102] - Vulkan driver not found
- csmicro-deb 1.494.10-0kylin5 arm64/mips64el (2022-06-27) [GP101/GP102] - Vulkan driver not found
- ljmgp 2.2.4.6-0kylin1 amd64/arm64/loongarch64 (2023-10-25) [GP200] - Vulkan driver not found

ICube (ICube Corporation): http://www.icubecorp.com

- icubegpu 3.2.1-0kylin1 arm64 (2023-08-08) - Vulkan driver not found

Innosilicon (Innosilicon Technology) Fantasy II - based on modified PowerVR: https://www.innosilicon.com

- innogpu-g0 2.1.4-1kylin1 amd64/arm64 (2022-08-25) - "/etc/vulkan/icd.d/icdconf.json" ("libVK_INNO.so": "/usr/lib/{x86_64,aarch64}-linux-gnu/libVK_INNO.so"): Vulkan 1.2.70
- innogpu-g0 2.2.33-2 amd64/arm64/loongarch64 (2023-09-21) - "/etc/vulkan/icd.d/innoconf.json" ("libVK_INNO.so": "/data/usr/lib/{x86_64,aarch64,loongarch64}-linux-gnu/libVK_INNO.so"): Vulkan 1.2.70
- innogpu-fh2m 2.4-0kylin1_amd64/arm64/loongarch64 (2023-06-06) - "/etc/vulkan/icd.d/innoconf.json" ("libVK_INNO.so": "/usr/lib/{x86_64,aarch64,loongarch64}-linux-gnu/innogpu-fh2m/libVK_INNO.so"): Vulkan 1.2.70

Jingjia Micro (Changsha Jingjia Microelectronics): https://www.jingjiamicro.com

- mwv206-dev 2.2.9-0kylin1 amd64/arm64/loongarch64/mips64el (2023-01-01) - Vulkan driver not found
- mwv207-dev 1.4.4-0kylin1 amd64/arm64/loongarch64 (2023-06-26) - Vulkan driver not found

Loongson (Loongson Technology): https://www.loongson.cn

- loonggl 0.1.0-0kylin2 loongarch64 (2023-08-17) - Vulkan driver not found

Moore Threads (Moore Threads Technology) MTT MUSA - based on heavily modified PowerVR: https://www.mthreads.com

- musa 1.0.5-0kylin3 arm64 (2022-09-16) - "/etc/vulkan/icd.d/musaicdconf.json" ("libVK_MT.so": "/usr/lib/aarch64-linux-gnu/musa/libVK_MT.so"): Vulkan 1.0.0
- musa 2.1.1-0kylin1 amd64/arm64/loongarch64 (2023-06-27) - "/etc/vulkan/icd.d/musaicdconf.json" ("libVK_MT.so": "/usr/lib/{x86_64,aarch64,loongarch64}-linux-gnu/musa/libVK_MT.so"): Vulkan 1.0.0

Phytium (Phytium Technology) X100 - based on vanilla PowerVR: https://www.phytium.com.cn

- x100-gpu-ddk 1.16-3kylin2k22.19 arm64 (2022-09-08) - "/etc/vulkan/icd.d/icdconf.json" ("libVK_IMG.so": "/usr/lib/aarch64-linux-gnu/pvr/libVK_IMG.so"): Vulkan 1.0.0

RuiXinView (Nanjing Ruixin View Technology) BF200 - based on PowerVR: http://www.ruixinview.com

- rxview-bf200 0.99.2-0kylin1 arm64 (2022-04-13) - "/etc/vulkan/icd.d/rxv-icdconf.json" ("libVK_IMG.so": "/usr/lib/aarch64-linux-gnu/rxview/libVK_IMG.so"): Vulkan 1.0.0
- rxview-bf200 0.99.4-0kylin1 arm64 (2022-11-08) - "/etc/vulkan/icd.d/rxv-icdconf.json" ("libVK_IMG.so": "/usr/lib/aarch64-linux-gnu/rxview/libVK_IMG.so"): Vulkan 1.0.0

Sietium (Xintong Semiconductor Technology) GenBu: https://www.sietium.com

- sietium-umod-drv 0.1.21kylin0 amd64/arm64 (2023-09-15) - "/etc/vulkan/icd.d/genbuconf.json" ("/usr/lib/{x86_64,aarch64}-linux-gnu/genbu02/libvulkan_genbu.so"): Vulkan 1.2.182
- sietium-umod-drv 0.1.21-1kylin0 amd64/arm64 (2023-11-07) - "/etc/vulkan/icd.d/genbuconf.json" ("/usr/lib/{x86_64,aarch64}-linux-gnu/genbu02/libvulkan_genbu.so"): Vulkan 1.2.182

XDX (XDX Computing Technology) - based on PowerVR: http://www.xdxct.com

- xdxgpu 1.10.012-0kylin1 arm64 (2023-02-01) - "/etc/vulkan/icd.d/xdxgpu_icd.json" ("libvlk_xdxgpu.so": "/usr/lib/aarch64-linux-gnu/xdxgpu/libvlk_xdxgpu.so"): Vulkan 1.0.0
- xdxgpu 1.10.139-0kylin1 amd64/loongarch64 (2023-09-08) - "/etc/vulkan/icd.d/xdxgpu_icd.json" ("libvlk_xdxgpu.so": "/usr/lib/{x86_64,aarch64,loongarch64}-linux-gnu/xdxgpu/libvlk_xdxgpu.so"): Vulkan 1.0.0

Zhaoxin (Shanghai Zhaoxin Semiconductor) - based on heavily modified S3G: https://www.zhaoxin.com & http://www.jariec.com & https://www.glenfly.com

- jari-linux-graphics-driver-dri 23.00.12-0kylin1 arm64 (2020-11-10) [JARI G12] - Vulkan driver not found
- zxe3k-linux-graphics-driver-dri 23.00.20-0kylin1 amd64/arm64/mips64el (2021-08-04) [JARI G12] - Vulkan driver not found
- jari-linux-graphics-driver-dri 23.00.20-0kylin3 arm64 (2022-06-16) [JARI G12] - Vulkan driver not found
- zhaoxin-cx4-linux-graphics-driver-dri 24.00.11-0kylin1 amd64 (2022-06-22) [ZX C-1080] - Vulkan driver not found
- zhaoxin-cx4-all-linux-graphics-driver-dri 24.00.25 amd64 (2023-09-06) [ZX C-1080] - Vulkan driver not found
- zhaoxin-cx4-linux-graphics-driver-dri 24.00.25-0kylin1 amd64 (2023-09-22) [ZX C-1080] - Vulkan driver not found
- gf-arise-linux-graphics-driver-dri 25.00.00-00-0kylin2 amd64/arm64 (2021-12-03) [Glenfly Arise] - Vulkan driver not found
- gf-arise-linux-graphics-driver-dri 25.00.02-00-0kylin1 amd64/arm64/loongarch64/mips64el (2022-03-07) [Glenfly Arise1020/Arise] - Vulkan driver not found
- arise-linux-graphics-driver-dri 25.00.10 amd64 (2022-12-09) [Glenfly Arise1020/Arise-GT1040/Arise1010/Arise-GT10C0] - Vulkan driver not found
- gf-arise-linux-graphics-driver-dri 25.00.22-0kylin1 amd64/arm64/loongarch64/mips64el (2023-10-25) [Glenfly Arise1020/Arise-GT1040/Arise1010/Arise-GT10C0t/Arise-GT10C0] - Vulkan driver not found
- zhaoxin-cx4-linux-graphics-driver-dri 26.00.00-0kylin1 amd64 (2023-10-19) [ZX C-1080] - "/usr/share/vulkan/icd.d/cx4_icd.x86_64.json" ("/usr/lib/x86_64-linux-gnu/libvulkan_cx4.so"): Vulkan 1.1.107



Chinese licensed GPUs

HUAWEI (Huawei Technologies) - based on ARM (ARM Holdings) Mali Valhall: https://www.arm.com/products/silicon-ip-multimedia & https://www.huawei.com

- hw-mali-valhall-r23 1.3.4 arm64 (2022-01-10) - "/etc/vulkan/icd.d/mali_vulkan.json" ("/usr/lib/aarch64-linux-gnu/libmali.so"): Vulkan 1.2.131
- hw-mali-valhall-r23 1.3.6 arm64 (2023-08-17) - "/etc/vulkan/icd.d/mali_vulkan.json" ("/usr/lib/aarch64-linux-gnu/libmali.so"): Vulkan 1.2.131



Western GPUs

AMD (Advanced Micro Devices) Radeon: https://www.amd.com

- vulkan-amdgpu 19.50-967956 amd64/i386 (2019-12-18) - "/opt/amdgpu/etc/vulkan/icd.d/amd_icd{64,32}.json" ("/opt/amdgpu/lib/{x86_64,i386}-linux-gnu/amdvlk{64,32}.so"): Vulkan 1.1.127
- vulkan-amdgpu-pro_19.50-967956 amd64/i386 (2019-12-18) - "/opt/amdgpu-pro/etc/vulkan/icd.d/amd_icd{64,32}.json" ("/opt/amdgpu-pro/lib/{x86_64,i386}-linux-gnu/amdvlk{64,32}.so"): Vulkan 1.1.129
- mesa-vulkan-drivers 23.0.4~kylin1~f amd64 (2023-06-04) - "/usr/share/vulkan/icd.d/radeon_icd.x86_64.json" ("/usr/lib/x86_64-linux-gnu/libvulkan_radeon.so"): Vulkan 1.3.238
- amdvlk 2023.Q4.2 amd64/i386 (2023-11-24) - "/etc/vulkan/icd.d/amd_icd{64,32}.json" ("/usr/lib/{x86_64,i386}-linux-gnu/amdvlk{64,32}.so"): Vulkan 1.3.269

Intel (Intel Corporation) Graphics/Arc: https://www.intel.com

- mesa-vulkan-drivers 23.0.4~kylin1~f amd64 (2023-06-04) - "/usr/share/vulkan/icd.d/intel_icd.x86_64.json" ("/usr/lib/x86_64-linux-gnu/libvulkan_intel.so"): Vulkan 1.3.238

NVIDIA (NVIDIA Corporation) GeForce/Quadro: https://www.nvidia.com

- libnvidia-gl-535 535.129.03-0kylin0.20.04.1 amd64/arm64/i386 (2023-11-04) - "/usr/share/vulkan/implicit_layer.d/nvidia_layers.json" ("libGLX_nvidia.so.0": "/usr/lib/{x86_64,aarch64,i386}-linux-gnu/libGLX_nvidia.so.0"): Vulkan 1.3.242



Vulkan specification release date:

- Vulkan 1.0: 2016-02-16
- Vulkan 1.1: 2018-03-07
- Vulkan 1.2: 2020-01-15
- Vulkan 1.3: 2022-01-25


Mesa Vulkan driver reference implementation release date:

- Vulkan 1.0: 2016-07-08 (Mesa 12.0)
- Vulkan 1.1: 2018-05-18 (Mesa 18.1)
- Vulkan 1.2: 2020-02-19 (Mesa 20.0)
- Vulkan 1.3: 2022-11-30 (Mesa 22.3)


Mesa Vulkan driver by hardware (https://mesamatrix.net) (2023-12-29):

- Vulkan incomplete implementation: hasvk (Intel Gen7-8), pvr (PowerVR Rogue), panvk (ARM Mali Midgard/Bifrost)
- Vulkan 1.0: v3dv (Broadcom VideoCore VI+)
- Vulkan 1.1: nvk (NVIDIA Turing+)
- Vulkan 1.3: radv (AMD GCN/RDNA), anv (Intel Gen9+), tu (Qualcomm Adreno)


Vulkan ICDs (Installable Client Driver):
https://tinyurl.com/vulkan2023

Possible Vulkan extensions in Zhaoxin 26-series driver:

$ grep -HERi "VK_(KHR|EXT)" zhaoxin-cx4-linux-graphics-driver-dri_26.00.00-0kylin1_amd64 2>/dev/null
Binary file zhaoxin-cx4-linux-graphics-driver-dri_26.00.00-0kylin1_amd64/data/usr/lib/x86_64-linux-gnu/libvulkan_cx4.so matches
$ strings zhaoxin-cx4-linux-graphics-driver-dri_26.00.00-0kylin1_amd64/data/usr/lib/x86_64-linux-gnu/libvulkan_cx4.so | grep -E '^VK(_[A-Za-z0-9]+)+$'
VK_KHR_surface
VK_KHR_xcb_surface
VK_KHR_xlib_surface
VK_KHR_device_group_creation
VK_KHR_display
VK_EXT_debug_report
VK_EXT_direct_mode_display
VK_EXT_swapchain_colorspace
VK_KHR_get_physical_device_properties2
VK_KHR_get_surface_capabilities2
VK_KHR_external_memory_capabilities
VK_KHR_external_semaphore_capabilities
VK_KHR_external_fence_capabilities
VK_NV_external_memory_capabilities
VK_EXT_display_surface_counter
VK_KHR_get_physical_device_properties2
VK_KHR_surface
VK_KHR_xcb_surface
VK_KHR_xlib_surface
VK_KHR_display
VK_KHR_get_surface_capabilities2
VK_EXT_swapchain_colorspace
VK_KHR_sampler_mirror_clamp_to_edge
VK_KHR_descriptor_update_template
VK_AMD_shader_explicit_vertex_parameter
VK_AMD_negative_viewport_height
VK_AMD_texture_gather_bias_lod
VK_AMD_mixed_attachment_samples
VK_KHR_storage_buffer_storage_class
VK_EXT_shader_viewport_index_layer
VK_KHR_get_memory_requirements2
VK_KHR_shared_presentable_image
VK_KHR_sampler_ycbcr_conversion
VK_NV_device_generated_commands
VK_EXT_vertex_attribute_divisor
VK_KHR_separate_depth_stencil_layouts
VK_ZX_KHR_descriptor_update_template_and_KHR_push_descriptor
VK_ZX_KHR_device_group_and_KHR_swapchain
VK_KHR_shader_draw_parameters
VK_KHR_swapchain
VK_KHR_maintenance1
VK_KHR_maintenance2
VK_KHR_maintenance3
VK_KHR_relaxed_block_layout
VK_KHR_dedicated_allocation
VK_KHR_external_memory
VK_KHR_external_memory_fd
VK_KHR_device_group
VK_KHR_bind_memory2
VK_KHR_external_semaphore
VK_KHR_external_semaphore_fd
VK_AMD_rasterization_order
VK_AMD_shader_ballot
VK_AMD_shader_trinary_minmax
VK_AMD_gcn_shader
VK_AMD_draw_indirect_count
VK_KHR_draw_indirect_count
VK_AMD_gpu_shader_half_float
VK_AMD_shader_info
VK_AMD_shader_fragment_mask
VK_EXT_sample_locations
VK_EXT_debug_marker
VK_AMD_gpu_shader_int16
VK_EXT_shader_subgroup_ballot
VK_KHR_16bit_storage
VK_EXT_shader_stencil_export
VK_KHR_image_format_list
VK_KHR_external_fence
VK_KHR_external_fence_fd
VK_EXT_global_priority
VK_KHR_display_swapchain
VK_KHR_push_descriptor
VK_NV_clip_space_w_scaling
VK_EXT_display_control
VK_GOOGLE_display_timing
VK_EXT_discard_rectangles
VK_EXT_hdr_metadata
VK_EXT_validation_cache
VK_EXT_external_memory_host
VK_EXT_sampler_filter_minmax
VK_EXT_transform_feedback
VK_EXT_conditional_rendering
VK_KHR_create_renderpass2
VK_KHR_depth_stencil_resolve
VK_EXT_separate_stencil_usage
VK_KHR_multiview
VK_KHR_variable_pointers
VK_EXT_host_query_reset
VK_KHR_imageless_framebuffer
VK_KHR_timeline_semaphore
VK_EXT_4444_formats
VK_EXT_line_rasterization
VK_EXT_custom_border_color
VK_EXT_calibrated_timestamps
VK_KHR_sampler_mirror_clamp_to_edge
VK_KHR_shader_draw_parameters
VK_KHR_swapchain
VK_KHR_maintenance1
VK_KHR_maintenance2
VK_KHR_maintenance3
VK_KHR_relaxed_block_layout
VK_KHR_dedicated_allocation
VK_KHR_descriptor_update_template
VK_KHR_external_memory
VK_KHR_external_memory_fd
VK_KHR_device_group
VK_KHR_bind_memory2
VK_KHR_external_semaphore
VK_KHR_external_semaphore_fd
VK_AMD_rasterization_order
VK_AMD_shader_ballot
VK_AMD_shader_trinary_minmax
VK_AMD_shader_explicit_vertex_parameter
VK_AMD_gcn_shader
VK_AMD_draw_indirect_count
VK_KHR_draw_indirect_count
VK_AMD_negative_viewport_height
VK_AMD_gpu_shader_half_float
VK_AMD_shader_info
VK_AMD_shader_fragment_mask
VK_AMD_texture_gather_bias_lod
VK_AMD_mixed_attachment_samples
VK_EXT_sample_locations
VK_EXT_debug_marker
VK_AMD_gpu_shader_int16
VK_EXT_shader_subgroup_ballot
VK_KHR_16bit_storage
VK_KHR_storage_buffer_storage_class
VK_EXT_shader_subgroup_ballot
VK_EXT_shader_stencil_export
VK_EXT_shader_viewport_index_layer
VK_KHR_get_memory_requirements2
VK_KHR_image_format_list
VK_KHR_external_fence
VK_EXT_global_priority
VK_KHR_display_swapchain
VK_KHR_push_descriptor
VK_KHR_shared_presentable_image
VK_KHR_sampler_ycbcr_conversion
VK_NV_device_generated_commands
VK_NV_clip_space_w_scaling
VK_EXT_display_control
VK_GOOGLE_display_timing
VK_EXT_discard_rectangles
VK_EXT_hdr_metadata
VK_EXT_validation_cache
VK_EXT_external_memory_host
VK_EXT_sampler_filter_minmax
VK_EXT_vertex_attribute_divisor
VK_EXT_transform_feedback
VK_EXT_conditional_rendering
VK_KHR_create_renderpass2
VK_KHR_depth_stencil_resolve
VK_EXT_separate_stencil_usage
VK_KHR_multiview
VK_KHR_variable_pointers
VK_EXT_host_query_reset
VK_KHR_imageless_framebuffer
VK_KHR_timeline_semaphore
VK_KHR_separate_depth_stencil_layouts
VK_EXT_4444_formats
VK_EXT_line_rasterization
VK_EXT_custom_border_color
VK_ZX_KHR_descriptor_update_template_and_KHR_push_descriptor
VK_ZX_KHR_device_group_and_KHR_swapchain
$


Possible Vulkan extensions in Zhaoxin 26-series driver (sorted):

$ strings zhaoxin-cx4-linux-graphics-driver-dri_26.00.00-0kylin1_amd64/data/usr/lib/x86_64-linux-gnu/libvulkan_cx4.so | grep -E '^VK(_[A-Za-z0-9]+)+$' | sort -ufVd
VK_AMD_draw_indirect_count
VK_AMD_gcn_shader
VK_AMD_gpu_shader_half_float
VK_AMD_gpu_shader_int16
VK_AMD_mixed_attachment_samples
VK_AMD_negative_viewport_height
VK_AMD_rasterization_order
VK_AMD_shader_ballot
VK_AMD_shader_explicit_vertex_parameter
VK_AMD_shader_fragment_mask
VK_AMD_shader_info
VK_AMD_shader_trinary_minmax
VK_AMD_texture_gather_bias_lod
VK_EXT_4444_formats
VK_EXT_calibrated_timestamps
VK_EXT_conditional_rendering
VK_EXT_custom_border_color
VK_EXT_debug_marker
VK_EXT_debug_report
VK_EXT_direct_mode_display
VK_EXT_discard_rectangles
VK_EXT_display_control
VK_EXT_display_surface_counter
VK_EXT_external_memory_host
VK_EXT_global_priority
VK_EXT_hdr_metadata
VK_EXT_host_query_reset
VK_EXT_line_rasterization
VK_EXT_sample_locations
VK_EXT_sampler_filter_minmax
VK_EXT_separate_stencil_usage
VK_EXT_shader_stencil_export
VK_EXT_shader_subgroup_ballot
VK_EXT_shader_viewport_index_layer
VK_EXT_swapchain_colorspace
VK_EXT_transform_feedback
VK_EXT_validation_cache
VK_EXT_vertex_attribute_divisor
VK_GOOGLE_display_timing
VK_KHR_16bit_storage
VK_KHR_bind_memory2
VK_KHR_create_renderpass2
VK_KHR_dedicated_allocation
VK_KHR_depth_stencil_resolve
VK_KHR_descriptor_update_template
VK_KHR_device_group
VK_KHR_device_group_creation
VK_KHR_display
VK_KHR_display_swapchain
VK_KHR_draw_indirect_count
VK_KHR_external_fence
VK_KHR_external_fence_capabilities
VK_KHR_external_fence_fd
VK_KHR_external_memory
VK_KHR_external_memory_capabilities
VK_KHR_external_memory_fd
VK_KHR_external_semaphore
VK_KHR_external_semaphore_capabilities
VK_KHR_external_semaphore_fd
VK_KHR_get_memory_requirements2
VK_KHR_get_physical_device_properties2
VK_KHR_get_surface_capabilities2
VK_KHR_image_format_list
VK_KHR_imageless_framebuffer
VK_KHR_maintenance1
VK_KHR_maintenance2
VK_KHR_maintenance3
VK_KHR_multiview
VK_KHR_push_descriptor
VK_KHR_relaxed_block_layout
VK_KHR_sampler_mirror_clamp_to_edge
VK_KHR_sampler_ycbcr_conversion
VK_KHR_separate_depth_stencil_layouts
VK_KHR_shader_draw_parameters
VK_KHR_shared_presentable_image
VK_KHR_storage_buffer_storage_class
VK_KHR_surface
VK_KHR_swapchain
VK_KHR_timeline_semaphore
VK_KHR_variable_pointers
VK_KHR_xcb_surface
VK_KHR_xlib_surface
VK_NV_clip_space_w_scaling
VK_NV_device_generated_commands
VK_NV_external_memory_capabilities
VK_ZX_KHR_descriptor_update_template_and_KHR_push_descriptor
VK_ZX_KHR_device_group_and_KHR_swapchain
$


Let's sort this out a bit more and compare with Mesa:

Vulkan 1.1:

- VK_KHR_16bit_storage - present, also available in Mesa (anv/gen8+, dzn, hasvk, lvp, nvk, radv, tu/a650, v3dv, vn)
- VK_KHR_bind_memory2 - present, also available in Mesa (anv, dzn, hasvk, lvp, nvk, pvr, radv, tu, v3dv, vn)
- VK_KHR_dedicated_allocation - present, also available in Mesa (anv, dzn, hasvk, lvp, nvk, radv, tu, v3dv, vn)
- VK_KHR_descriptor_update_template - present, also available in Mesa (anv, dzn, hasvk, lvp, nvk, panvk, radv, tu, v3dv, vn)
- VK_KHR_device_group - present, also available in Mesa (anv, dzn, hasvk, lvp, nvk, tu, v3dv, vn)
- VK_KHR_device_group_creation - present, also available in Mesa (anv, hasvk, nvk, dzn, lvp, tu, v3dv, vn)
- VK_KHR_external_fence - present, also available in Mesa (anv, hasvk, lvp, nvk, pvr, radv, tu, v3dv, vn)
- VK_KHR_external_fence_capabilities - present, also available in Mesa (anv, hasvk, lvp, nvk, pvr, radv, tu, v3dv, vn)
- VK_KHR_external_memory - present, also available in Mesa (anv, dzn, hasvk, lvp, nvk, pvr, radv, tu, v3dv, vn)
- VK_KHR_external_memory_capabilities - present, also available in Mesa (anv, dzn, hasvk, lvp, nvk, pvr, radv, tu, v3dv, vn)
- VK_KHR_external_semaphore - present, also available in Mesa (anv, dzn, hasvk, lvp, nvk, pvr, radv, tu, v3dv, vn)
- VK_KHR_external_semaphore_capabilities - present, also available in Mesa (anv, dzn, hasvk, lvp, nvk, pvr, radv, tu, v3dv, vn)
- VK_KHR_get_memory_requirements2 - present, also available in Mesa (anv, dzn, hasvk, lvp, nvk, pvr, radv, tu, v3dv, vn)
- VK_KHR_get_physical_device_properties2 - present, also available in Mesa (anv, dzn, hasvk, lvp, nvk, panvk, pvr, radv, tu, v3dv, vn)
- VK_KHR_maintenance1 - present, also available in Mesa (anv, dzn, hasvk, lvp, nvk, radv, tu, v3dv, vn)
- VK_KHR_maintenance2 - present, also available in Mesa (anv, dzn, hasvk, lvp, nvk, radv, tu, v3dv, vn)
- VK_KHR_maintenance3 - present, also available in Mesa (anv, dzn, hasvk, lvp, nvk, radv, tu, v3dv, vn)
- VK_KHR_multiview - present, also available in Mesa (anv, dzn, hasvk, lvp, nvk, radv, tu, v3dv, vn)
- VK_KHR_relaxed_block_layout - present, also available in Mesa (anv, dzn, hasvk, lvp, nvk, radv, tu, v3dv, vn)
- VK_KHR_sampler_ycbcr_conversion - present, also available in Mesa (anv, hasvk, lvp, nvk, radv, tu, v3dv, vn)
- VK_KHR_shader_draw_parameters - present, also available in Mesa (anv, dzn, hasvk, lvp, nvk, radv, tu, vn)
- VK_KHR_storage_buffer_storage_class - present, also available in Mesa (anv, dzn, hasvk, lvp, nvk, panvk, radv, tu, v3dv, vn)
- VK_KHR_variable_pointers - present, also available in Mesa (anv, hasvk, lvp, nvk, panvk, radv, tu, v3dv, vn)

Vulkan 1.2:

- VK_KHR_8bit_storage - not found, but available in Mesa (anv/gen8+, dzn, hasvk, lvp, nvk, radv, v3dv, vn)
- VK_KHR_buffer_device_address - not found, but available in Mesa (anv/gen8+, hasvk, lvp, nvk, radv, tu, v3dv, vn)
- VK_KHR_create_renderpass2 - present, also available in Mesa (anv, dzn, hasvk, lvp, nvk, radv, tu, v3dv, vn)
- VK_KHR_depth_stencil_resolve - present, also available in Mesa (anv, dzn, hasvk, lvp, nvk, radv, tu, v3dv, vn)
- VK_KHR_draw_indirect_count - present, also available in Mesa (anv, dzn, hasvk, lvp, nvk, radv, tu, vn)
- VK_KHR_driver_properties - not found, but available in Mesa (anv, dzn, hasvk, lvp, nvk, radv, tu, v3dv, vn)
- VK_KHR_image_format_list - present, also available in Mesa (anv, dzn, hasvk, lvp, nvk, pvr, radv, tu, v3dv, vn)
- VK_KHR_imageless_framebuffer - present, also available in Mesa (anv, dzn, hasvk, lvp, nvk, radv, tu, v3dv, vn)
- VK_KHR_sampler_mirror_clamp_to_edge - present, also available in Mesa (anv, dzn, hasvk, lvp, nvk, radv, tu, v3dv, vn)
- VK_KHR_separate_depth_stencil_layouts - present, also available in Mesa (anv, dzn, hasvk, lvp, nvk, radv, tu, v3dv, vn)
- VK_KHR_shader_atomic_int64 - not found, but available in Mesa (anv/gen9+, lvp, nvk, radv, vn)
- VK_KHR_shader_float16_int8 - not found, but available in Mesa (anv/gen8+, dzn, hasvk, lvp, radv, tu, vn)
- VK_KHR_shader_float_controls - not found, but available in Mesa (anv/gen8+, dzn, hasvk, lvp, nvk, radv, tu, v3dv, vn)
- VK_KHR_shader_subgroup_extended_types - not found, but available in Mesa (anv/gen8+, hasvk, lvp, radv, tu, vn)
- VK_KHR_spirv_1_4 - not found, but available in Mesa (anv, dzn, hasvk, lvp, nvk, radv, tu, v3dv, vn)
- VK_KHR_timeline_semaphore - present, also available in Mesa (anv, dzn, hasvk, lvp, nvk, pvr, radv, tu, v3dv, vn)
- VK_KHR_uniform_buffer_standard_layout - not found, but available in Mesa (anv, dzn, hasvk, lvp, nvk, pvr, radv, tu, v3dv, vn)
- VK_KHR_vulkan_memory_model - not found, but available in Mesa (anv, hasvk, lvp, nvk, radv, tu, v3dv, vn)
- VK_EXT_descriptor_indexing - not found, but available in Mesa (anv/gen9+, dzn, lvp, nvk, radv, tu, vn)
- VK_EXT_host_query_reset - present, also available in Mesa (anv, hasvk, lvp, nvk, pvr, radv, tu, v3dv, vn)
- VK_EXT_sampler_filter_minmax - present, also available in Mesa (anv/gen9+, lvp, nvk, radv, tu, vn)
- VK_EXT_scalar_block_layout - not found, but available in Mesa (anv, dzn, hasvk, lvp, nvk, pvr, radv/gfx7+, tu, vn)
- VK_EXT_separate_stencil_usage - present, also available in Mesa (anv, dzn, hasvk, lvp, nvk, radv, tu, v3dv, vn)
- VK_EXT_shader_viewport_index_layer - present, also available in Mesa (anv, hasvk, lvp, nvk, radv, tu, vn)

Vulkan 1.3:

- VK_KHR_copy_commands2 - not found, but available in Mesa (anv, hasvk, lvp, nvk, panvk, pvr, radv, tu, v3dv, vn)
- VK_KHR_dynamic_rendering - not found, but available in Mesa (anv, dzn, hasvk, lvp, nvk, radv, tu, vn)
- VK_KHR_format_feature_flags2 - not found, but available in Mesa (anv, hasvk, lvp, nvk, pvr, radv, tu, v3dv, vn)
- VK_KHR_maintenance4 - not found, but available in Mesa (anv, hasvk, lvp, nvk, radv, tu, v3dv, vn)
- VK_KHR_shader_integer_dot_product - not found, but available in Mesa (anv, dzn, hasvk, lvp, nvk, radv, tu, v3dv, vn)
- VK_KHR_shader_non_semantic_info - not found, but available in Mesa (anv, hasvk, nvk, radv, tu, v3dv, vn)
- VK_KHR_shader_terminate_invocation - not found, but available in Mesa (anv, hasvk, lvp, nvk, radv, tu, v3dv, vn)
- VK_KHR_synchronization2 - not found, but available in Mesa (anv, dzn, hasvk, lvp, nvk, panvk, radv, tu, v3dv, vn)
- VK_KHR_zero_initialize_workgroup_memory - not found, but available in Mesa (anv, hasvk, lvp, radv, tu, v3dv, vn)
- VK_EXT_4444_formats - present, also available in Mesa (anv, hasvk, lvp, nvk, radv, tu, v3dv, vn)
- VK_EXT_extended_dynamic_state - not found, but available in Mesa (anv, hasvk, lvp, nvk, radv, tu, vn)
- VK_EXT_extended_dynamic_state2 - not found, but available in Mesa (anv, hasvk, lvp, nvk, radv, tu, vn)
- VK_EXT_image_robustness - not found, but available in Mesa (anv, hasvk, lvp, nvk, radv, tu, v3dv, vn)
- VK_EXT_inline_uniform_block - not found, but available in Mesa (anv, hasvk, lvp, nvk, radv, tu, v3dv, vn)
- VK_EXT_pipeline_creation_cache_control - not found, but available in Mesa (anv, hasvk, lvp, radv, tu, v3dv, vn)
- VK_EXT_pipeline_creation_feedback - not found, but available in Mesa (anv, hasvk, lvp, radv, tu, v3dv, vn)
- VK_EXT_private_data - not found, but available in Mesa (anv, hasvk, lvp, nvk, pvr, radv, tu, v3dv, vn)
- VK_EXT_shader_demote_to_helper_invocation - not found, but available in Mesa (anv, hasvk, lvp, nvk, radv, tu, vn)
- VK_EXT_subgroup_size_control - not found, but available in Mesa (anv, hasvk, lvp, nvk, radv, tu, vn)
- VK_EXT_texel_buffer_alignment - not found, but available in Mesa (anv, hasvk, lvp, nvk, pvr, radv, tu, v3dv, vn)
- VK_EXT_texture_compression_astc_hdr - not found, but available in Mesa (vn)
- VK_EXT_tooling_info - not found, but available in Mesa (anv, hasvk, nvk, pvr, radv, tu, v3dv, vn)
- VK_EXT_ycbcr_2plane_444_formats - not found, but available in Mesa (lvp, nvk, vn)

Khronos extensions that are not part of any Vulkan version:

- VK_KHR_acceleration_structure - not found, but available in Mesa (anv/gfx12.5+, radv/gfx10.3+)
- VK_KHR_android_surface - not found, not available in Mesa
- VK_KHR_calibrated_timestamps - not found, but available in Mesa (radv)
- VK_KHR_cooperative_matrix - not found, but available in Mesa (radv/gfx11+)
- VK_KHR_deferred_host_operations - not found, but available in Mesa (anv, hasvk, radv)
- VK_KHR_display - present, also available in Mesa (anv, pvr, radv, tu, v3dv)
- VK_KHR_display_swapchain - present, but not available in Mesa
- VK_KHR_external_fence_fd - present, also available in Mesa (anv, hasvk, nvk, pvr, radv, tu, v3dv, vn)
- VK_KHR_external_fence_win32 - not found, not available in Mesa
- VK_KHR_external_memory_fd - present, also available in Mesa (anv, dzn, hasvk, lvp, nvk, pvr, radv, tu, v3dv, vn)
- VK_KHR_external_memory_win32 - not found, but available in Mesa (dzn)
- VK_KHR_external_semaphore_fd - present, also available in Mesa (anv, dzn, hasvk, nvk, pvr, radv, tu, v3dv, vn)
- VK_KHR_external_semaphore_win32 - not found, but available in Mesa (dzn)
- VK_KHR_fragment_shader_barycentric - not found, but available in Mesa (nvk/Turing+, radv/gfx10.3+)
- VK_KHR_fragment_shading_rate - not found, but available in Mesa (anv/gen11+, radv/gfx10.3+)
- VK_KHR_get_display_properties2 - not found, but available in Mesa (anv, pvr, radv, tu, v3dv)
- VK_KHR_get_surface_capabilities2 - present, also available in Mesa (anv, lvp, nvk, pvr, radv, tu, v3dv, vn)
- VK_KHR_global_priority - not found, but available in Mesa (radv, tu)
- VK_KHR_incremental_present - not found, but available in Mesa (anv, hasvk, lvp, radv, tu, v3dv, vn)
- VK_KHR_maintenance5 - not found, but available in Mesa (anv, lvp, radv, tu)
- VK_KHR_map_memory2 - not found, but available in Mesa (anv, nvk, radv)
- VK_KHR_performance_query - not found, but available in Mesa (anv/gen8+, radv/gfx10.3+, tu, v3dv)
- VK_KHR_pipeline_executable_properties - not found, but available in Mesa (anv, nvk, hasvk, radv, tu, v3dv)
- VK_KHR_pipeline_library - not found, but available in Mesa (anv, lvp, radv, tu, vn)
- VK_KHR_present_wait - not found, but available in Mesa (anv, radv, tu, x11/display)
- VK_KHR_push_descriptor - present, also available in Mesa (anv, hasvk, lvp, nvk, radv, tu, vn)
- VK_KHR_ray_query - not found, but available in Mesa (anv/gfx12.5+, radv/gfx10.3+)
- VK_KHR_ray_tracing_maintenance1 - not found, but available in Mesa (anv/gfx12.5+, radv/gfx10.3+)
- VK_KHR_ray_tracing_pipeline - not found, but available in Mesa (anv/gfx12.5+, radv/gfx10.3+)
- VK_KHR_shader_clock - not found, but available in Mesa (anv, hasvk, lvp, nvk, radv, vn)
- VK_KHR_shader_subgroup_uniform_control_flow - not found, but available in Mesa (anv, hasvk, radv)
- VK_KHR_shared_presentable_image - present, but not available in Mesa
- VK_KHR_surface - present, also available in Mesa (anv, dzn, hasvk, lvp, nvk, panvk, pvr, radv, tu, v3dv, vn)
- VK_KHR_surface_protected_capabilities - not found, but available in Mesa (anv, lvp, nvk, radv, v3dv, vn)
- VK_KHR_swapchain - present, also available in Mesa (anv, dzn, hasvk, lvp, nvk, panvk, pvr, radv, tu, v3dv, vn)
- VK_KHR_swapchain_mutable_format - not found, but available in Mesa (anv, hasvk, lvp, nvk, radv, tu, v3dv, vn)
- VK_KHR_vertex_attribute_divisor - not found, but available in Mesa (radv)
- VK_KHR_wayland_surface - not found, but available in Mesa (anv, dzn, lvp, nvk, panvk, radv, tu, v3dv, vn)
- VK_KHR_win32_keyed_mutex - not found, not available in Mesa
- VK_KHR_win32_surface - not found, but available in Mesa (dzn, lvp)
- VK_KHR_workgroup_memory_explicit_layout - not found, but available in Mesa (anv, nvk, hasvk, radv, v3dv)
- VK_KHR_xcb_surface - present, also available in Mesa (anv, dzn, hasvk, lvp, nvk, radv, tu, v3dv, vn)
- VK_KHR_xlib_surface - present, also available in Mesa (anv, dzn, hasvk, lvp, nvk, radv, tu, v3dv, vn)
- VK_EXT_acquire_drm_display - not found, but available in Mesa (anv, radv, tu, v3dv)
- VK_EXT_acquire_xlib_display - not found, but available in Mesa (anv, lvp, nvk, radv, tu, v3dv)
- VK_EXT_attachment_feedback_loop_dynamic_state - not found, but available in Mesa (lvp, radv)
- VK_EXT_attachment_feedback_loop_layout - not found, but available in Mesa (lvp, nvk, radv, tu, v3dv)
- VK_EXT_border_color_swizzle - not found, but available in Mesa (anv, hasvk, lvp, nvk, radv/gfx10+, tu, v3dv, vn)
- VK_EXT_buffer_device_address - not found, but available in Mesa (anv/gen8+, hasvk, nvk, radv)
- VK_EXT_calibrated_timestamps - present, also available in Mesa (anv, hasvk, lvp, radv, vn)
- VK_EXT_color_write_enable - not found, but available in Mesa (anv, hasvk, lvp, nvk, radv, tu, v3dv, vn)
- VK_EXT_conditional_rendering - present, also available in Mesa (anv, hasvk, lvp, nvk, radv, tu, vn)
- VK_EXT_conservative_rasterization - not found, but available in Mesa (anv/gen9+, radv, vn)
- VK_EXT_custom_border_color - present, also available in Mesa (anv, hasvk, lvp, nvk, panvk, radv, tu, v3dv, vn)
- VK_EXT_debug_marker - present, also available in Mesa (radv)
- VK_EXT_debug_report - present, also available in Mesa (anv, dzn, lvp, nvk, pvr, radv, tu, v3dv)
- VK_EXT_depth_bias_control - not found, but available in Mesa (anv, nvk, radv)
- VK_EXT_depth_clamp_zero_one - not found, but available in Mesa (anv, radv)
- VK_EXT_depth_clip_control - not found, but available in Mesa (anv, hasvk, lvp, nvk, radv, tu, v3dv, vn)
- VK_EXT_depth_clip_enable - not found, but available in Mesa (anv, hasvk, lvp, nvk, radv, tu, vn)
- VK_EXT_depth_range_unrestricted - not found, but available in Mesa (anv/gen20+, radv, lvp)
- VK_EXT_descriptor_buffer - not found, but available in Mesa (lvp, radv, tu)
- VK_EXT_device_fault - not found, but available in Mesa (radv)
- VK_EXT_device_memory_report - not found, but available in Mesa (vn)
- VK_EXT_direct_mode_display - present, also available in Mesa (anv, lvp, radv, tu, v3dv)
- VK_EXT_discard_rectangles - present, also available in Mesa (radv)
- VK_EXT_display_control - present, also available in Mesa (anv, hasvk, radv, tu)
- VK_EXT_display_surface_counter - present, also available in Mesa (anv, lvp, radv, tu)
- VK_EXT_dynamic_rendering_unused_attachments - not found, but available in Mesa (anv, nvk, radv, vn)
- VK_EXT_extended_dynamic_state3 - not found, but available in Mesa (anv, lvp, nvk, radv, tu, vn)
- VK_EXT_external_memory_acquire_unmodified - not found, but available in Mesa (radv)
- VK_EXT_external_memory_dma_buf - not found, but available in Mesa (anv, hasvk, nvk, pvr, radv, tu, v3dv, vn)
- VK_EXT_external_memory_host - present, also available in Mesa (anv, hasvk, lvp, radv)
- VK_EXT_filter_cubic - not found, but available in Mesa (tu/a650)
- VK_EXT_fragment_shader_interlock - not found, but available in Mesa (anv/gen9+, radv/gfx9+, vn)
- VK_EXT_global_priority - present, also available in Mesa (anv, hasvk, radv, tu)
- VK_EXT_global_priority_query - not found, but available in Mesa (anv, hasvk, radv, tu)
- VK_EXT_graphics_pipeline_library - not found, but available in Mesa (anv, lvp, radv, tu, vn)
- VK_EXT_hdr_metadata - present, but not available in Mesa
- VK_EXT_image_2d_view_of_3d - not found, but available in Mesa (anv, hasvk, lvp, nvk, radv, tu, vn)
- VK_EXT_image_compression_control - not found, but available in Mesa (radv)
- VK_EXT_image_drm_format_modifier - not found, but available in Mesa (anv, hasvk, radv/gfx9+, tu, v3dv, vn)
- VK_EXT_image_sliced_view_of_3d - not found, but available in Mesa (anv, nvk, radv/gfx10+)
- VK_EXT_image_view_min_lod - not found, but available in Mesa (anv, hasvk, nvk, radv, tu, vn)
- VK_EXT_index_type_uint8 - not found, but available in Mesa (anv, hasvk, nvk, lvp, panvk, radv/gfx8+, tu, v3dv, vn)
- VK_EXT_line_rasterization - present, also available in Mesa (anv, hasvk, nvk, lvp, radv, tu, v3dv, vn)
- VK_EXT_load_store_op_none - not found, but available in Mesa (anv, nvk, radv, tu, v3dv, vn)
- VK_EXT_memory_budget - not found, but available in Mesa (anv, hasvk, lvp, radv, tu, v3dv, vn)
- VK_EXT_memory_priority - not found, but available in Mesa (lvp, radv)
- VK_EXT_mesh_shader - not found, but available in Mesa (anv/gfx12.5+, lvp, radv)
- VK_EXT_multi_draw - not found, but available in Mesa (anv, hasvk, lvp, nvk, radv, tu, vn, v3dv)
- VK_EXT_multisampled_render_to_single_sampled - not found, but available in Mesa (lvp)
- VK_EXT_mutable_descriptor_type - not found, but available in Mesa (anv, lvp, nvk, radv, tu, vn)
- VK_EXT_nested_command_buffer - not found, but available in Mesa (lvp)
- VK_EXT_non_seamless_cube_map - not found, but available in Mesa (anv, hasvk, lvp, nvk, radv, tu, vn)
- VK_EXT_pageable_device_local_memory - not found, but available in Mesa (lvp)
- VK_EXT_pci_bus_info - not found, but available in Mesa (anv, hasvk, nvk, radv, vn)
- VK_EXT_physical_device_drm - not found, but available in Mesa (anv, hasvk, nvk, radv, tu, v3dv, vn)
- VK_EXT_pipeline_library_group_handles - not found, but available in Mesa (anv, radv)
- VK_EXT_pipeline_robustness - not found, but available in Mesa (anv, radv, v3dv)
- VK_EXT_post_depth_coverage - not found, but available in Mesa (anv/gfx10+, lvp, radv/gfx10+)
- VK_EXT_primitives_generated_query - not found, but available in Mesa (anv, hasvk, lvp, nvk, radv, tu, vn)
- VK_EXT_primitive_topology_list_restart - not found, but available in Mesa (anv, hasvk, lvp, nvk, radv, tu, v3dv, vn)
- VK_EXT_provoking_vertex - not found, but available in Mesa (anv, hasvk, lvp, nvk, radv, tu, v3dv, vn)
- VK_EXT_queue_family_foreign - not found, but available in Mesa (anv, hasvk, radv, tu, vn)
- VK_EXT_rasterization_order_attachment_access - not found, but available in Mesa (lvp, tu, vn)
- VK_EXT_robustness2 - not found, but available in Mesa (anv, hasvk, lvp, nvk, radv, tu, vn)
- VK_EXT_sample_locations - present, also available in Mesa (anv, hasvk, nvk, radv/gfx9-, tu/a650)
- VK_EXT_shader_atomic_float - not found, but available in Mesa (anv, hasvk, lvp, radv)
- VK_EXT_shader_atomic_float2 - not found, but available in Mesa (anv/gen9+, lvp, radv)
- VK_EXT_shader_image_atomic_int64 - not found, but available in Mesa (nvk, radv)
- VK_EXT_shader_module_identifier - not found, but available in Mesa (anv, hasvk, radv, tu, v3dv)
- VK_EXT_shader_object - not found, but available in Mesa (lvp)
- VK_EXT_shader_stencil_export - present, also available in Mesa (anv/gen9+, lvp, radv, tu, vn)
- VK_EXT_shader_subgroup_ballot - present, also available in Mesa (anv, dzn, hasvk, lvp, nvk, radv, vn)
- VK_EXT_shader_subgroup_vote - not found, but available in Mesa (anv, dzn, hasvk, lvp, nvk, radv)
- VK_EXT_swapchain_colorspace - present, also available in Mesa (radv, anv, tu)
- VK_EXT_transform_feedback - present, also available in Mesa (anv, hasvk, lvp, nvk, radv, tu, vn)
- VK_EXT_validation_cache - present, but not available in Mesa
- VK_EXT_vertex_attribute_divisor - present, also available in Mesa (anv, dzn, hasvk, lvp, nvk, panvk, radv, tu, v3dv, vn)
- VK_EXT_vertex_input_dynamic_state - not found, but available in Mesa (anv, lvp, nvk, radv, tu, vn)
- VK_EXT_ycbcr_image_arrays - not found, but available in Mesa (anv, hasvk, lvp, nvk, radv)
- VK_AMD_buffer_marker - not found, but available in Mesa (radv, tu)
- VK_AMD_device_coherent_memory - not found, but available in Mesa (radv)
- VK_AMD_draw_indirect_count - present, also available in Mesa (radv)
- VK_AMD_gcn_shader - present, also available in Mesa (radv)
- VK_AMD_gpu_shader_half_float - present, also available in Mesa (radv)
- VK_AMD_gpu_shader_int16 - present, also available in Mesa (radv)
- VK_AMD_memory_overallocation_behavior - not found, but available in Mesa (radv)
- VK_AMD_mixed_attachment_samples - present, also available in Mesa (radv)
- VK_AMD_rasterization_order - present, also available in Mesa (radv)
- VK_AMD_shader_ballot - present, also available in Mesa (radv)
- VK_AMD_shader_core_properties - not found, but available in Mesa (radv)
- VK_AMD_shader_core_properties2 - not found, but available in Mesa (radv)
- VK_AMD_shader_early_and_late_fragment_tests - not found, but available in Mesa (radv)
- VK_AMD_shader_explicit_vertex_parameter - present, also available in Mesa (radv)
- VK_AMD_shader_fragment_mask - present, also available in Mesa (radv/gfx10.3-)
- VK_AMD_shader_image_load_store_lod - not found, but available in Mesa (radv)
- VK_AMD_shader_info - present, but not available in Mesa
- VK_AMD_shader_trinary_minmax - present, also available in Mesa (radv)
- VK_AMD_texture_gather_bias_lod - present, also available in Mesa (radv)
- VK_ANDROID_external_memory_android_hardware_buffer - not found, but available in Mesa (anv, radv, vn)
- VK_ANDROID_native_buffer - not found, but available in Mesa (anv, radv, tu, v3dv, vn)
- VK_ARM_rasterization_order_attachment_access - not found, but available in Mesa (lvp, tu)
- VK_GOOGLE_decorate_string - not found, but available in Mesa (anv, hasvk, lvp, radv)
- VK_GOOGLE_display_timing - present, but not available in Mesa
- VK_GOOGLE_hlsl_functionality1 - not found, but available in Mesa (anv, hasvk, lvp, radv)
- VK_GOOGLE_user_type - not found, but available in Mesa (anv, hasvk, radv)
- VK_IMG_filter_cubic - not found, but available in Mesa (tu/a650)
- VK_INTEL_shader_integer_functions2 - not found, but available in Mesa (anv, hasvk, radv)
- VK_NV_clip_space_w_scaling - present, but not available in Mesa
- VK_NV_compute_shader_derivatives - not found, but available in Mesa (anv, hasvk, radv)
- VK_NV_device_generated_commands - present, but not available in Mesa
- VK_VALVE_mutable_descriptor_type - not found, but available in Mesa (anv, hasvk, radv, tu, vn)

Deprecated Vulkan extensions:

- VK_NV_external_memory_capabilities - present, but not available in Mesa // deprecated by VK_KHR_external_memory_capabilities

Obsoleted Vulkan extensions:

- VK_AMD_negative_viewport_height - present, but not available in Mesa // obsoleted by VK_KHR_maintenance1

Probably not Vulkan extensions:

- VK_ZX_KHR_descriptor_update_template_and_KHR_push_descriptor - present, but not available in Mesa // probably an environmental variable
- VK_ZX_KHR_device_group_and_KHR_swapchain - present, but not available in Mesa // probably an environmental variable

[scx]

Zink

Zink driver is a Gallium driver which can be seen as implementation of OpenGL on top of Vulkan (OpenGL 2-4 to Vulkan). So, it is somehow similar to other wrappers (https://emulation.gametechwiki.com/index.php/Wrappers), like dgVoodoo2 (Glide 2-3 & Direct3D 2-9 to Direct3D 11-12), DXVK (Direct3D 9-11 to Vulkan) or MoltenVK (Vulkan to Metal).

Zink (https://docs.mesa3d.org/drivers/zink.html) driver requirements:

OpenGL 2.1:

- VK_KHR_maintenance1: present
- VK_KHR_create_renderpass2: present
- VK_KHR_imageless_framebuffer: present
- VK_KHR_timeline_semaphore: present
- VK_EXT_custom_border_color: present
- VK_EXT_provoking_vertex: not found
- VK_EXT_line_rasterization: present
- VK_KHR_swapchain_mutable_format: not found
- VK_EXT_border_color_swizzle: not found
- VK_KHR_descriptor_update_template: present
- VK_KHR_external_memory: present
- VK_EXT_scalar_block_layout: not found

OpenGL 3.0:

- VK_EXT_transform_feedback: present
- VK_EXT_conditional_rendering: present

OpenGL 3.2:

- VK_EXT_depth_clip_enable: not found

OpenGL 3.3:

- VK_EXT_vertex_attribute_divisor: present

OpenGL 4.0:

- VK_KHR_maintenance2: present

OpenGL 4.2:

- VK_EXT_image_2d_view_of_3d: not found
- VK_KHR_shader_draw_parameters: present

OpenGL 4.4:

- VK_KHR_sampler_mirror_clamp_to_edge: present

OpenGL 4.6:

- VK_KHR_draw_indirect_count: present



Although the above analysis doesn't take into account all Zink requirements, and the listed Vulkan extensions supported by the Zhaoxin driver are only a prediction, it should give a rough picture of the situation. Anyway, this means that we can expect even OpenGL 4.6 support for Zhaoxin GPUs via the Zink driver pretty soon. This is especially useful on Linux. On Windows, this implementation is a bit problematic. You can definitely do (https://github.com/pal1000/mesa-dist-win) it, but some games and apps just doesn't like it. I'll try to explain it later.

[Tralalak]

CentaurHauls Family 7 Model 11 Stepping 1
Zhaoxin Kaixian KX-7000’s GeekBench scores revealed, performance doubled compared to previous generation
source: https://article.pchome.net/article/content-2173598.html

[Loeschzwerg]

DIE-Shots, pictures and benchmarks of the KX-7000:
87928

87929

87930

87931

Source:
@9550pro (Twitter) via a now deleted bilibili video.

[Loeschzwerg]

Further snapshots of that bilibili video I got sent via DM on Twitter:
87932

87933

87934

87935

87936

Vendor of that KX-7000 seems to be ASUS.

[Kinoco]

Kann man Zhaoxin auch wo kaufen ?
Only place I know is taobao.com, although registration is not so easy.
Does anybody know other places to buy Zhaoxin Boards+CPU?

[Kinoco]

CentaurHauls Family 7 Model 11 Stepping 1
Zhaoxin Kaixian KX-7000’s GeekBench scores revealed, performance doubled compared to previous generation
source: https://article.pchome.net/article/content-2173598.html

For me even more important than CPU double performance, is the iGPU 4x performance and the H264 and H265 encode and decode.
Because on the practical usages tests of KX-6000 I have seen, the main cause of bottleneck was the iGPU.

[Kinoco]

Only place I know is taobao.com, although registration is not so easy.
Does anybody know other places to buy Zhaoxin Boards+CPU?

Also found this board from Taiwanese IEI for the international market, selling in Germany on compmall.de:
https://www.heise.de/select/ct/2021/20/2117513415912404118
https://www.compmall.de/KINO-KX-U6580-R10-507373

The IEI KINO -KX board uses AMI BIOS. I am not sure what BIOS is used in the chinese boards.

But I guess that now that KX-7000 is coming out, the KX-6000 has little interest.

[Kinoco]

Also found this board from Taiwanese IEI for the international market, selling in Germany on compmall.de:
https://www.heise.de/select/ct/2021/20/2117513415912404118
https://www.compmall.de/KINO-KX-U6580-R10-507373

The IEI KINO -KX board uses AMI BIOS. I am not sure what BIOS is used in the chinese boards.

But I guess that now that KX-7000 is coming out, the KX-6000 has little interest.

Notice just now, this is the board Tralalak uses :biggrin:

[Loeschzwerg]

A comparison of DIE shots:
87970

87971

87972

Source: https://tiebac.baidu.com/p/8996713724

[Kinoco]

A comparison of DIE shots:

Source: https://tiebac.baidu.com/p/8996713724

Following your source for the pictures, I found this quote:
"The KX-7000 has a new front end that introduces out-of-order execution and better caching."
So it seems that the doubling in performance is coming mostly from the out-of-order-execution, and a smaller part from faster DDR speeds.

From the statement above, I am not totally sure if KX-6000 did not have any out-of-order-execution and KX-7000 is introducing it for the first ime or if this is just a major revision on that. Do you know the answer to this?

One of the things I don't like about Intel and AMD architecture, is their agressive implementation of branch prediction and out-of-order execution, that causes so many critical issues. I woiuld hope that VIA/Zhaoxin architecture could keep pure in that regard, as Intel was on the day. But maybe KX-6000 was not already?

[Loeschzwerg]

KX-6000 is also an out-of-oder design. VIA Isaiah core (Nano) was the first implementation of an out-of-order architecture. KX-7000 is probably just improving the execution and branch prediction.

Can't wait for an in-depth review from Chips&Cheese like they always do:
https://chipsandcheese.com/2021/09/22/the-weird-and-wacky-world-of-via-part-2-zhaoxins-not-quite-electric-boogaloo/

[y33H@]

VEQAgOe9i6Y

[Tralalak]

https://youtu.be/VEQAgOe9i6Y
After a decade, Linus imitated Joel Hruska by changing the CPUID Centaur to Intel?
source: https://arstechnica.com/gadgets/2008/07/atom-nano-review/6/

[Tralalak]

Rumours
Zhaoxin KaiSheng KH-50000 Series (ZX-I) server x86 processors should have 30% higher IPC than desktop KX-7000 Series (ZX-H)

• 5 nm
• higher frequency
• up to 96C
• dual socket up to 192C
• can support big.LITTLE
• 2H24

source: https://zhuanlan-zhihu-com.translate.goog/p/532115524?_x_tr_sl=zh-CN&_x_tr_tl=sk&_x_tr_hl=sk&_x_tr_pto=sc
source: https://www.zhihu.com/question/637817841/answer/3473035158

[Kinoco]

KX-6000 is also an out-of-oder design. VIA Isaiah core (Nano) was the first implementation of an out-of-order architecture. KX-7000 is probably just improving the execution and branch prediction.

Can't wait for an in-depth review from Chips&Cheese like they always do:
https://chipsandcheese.com/2021/09/22/the-weird-and-wacky-world-of-via-part-2-zhaoxins-not-quite-electric-boogaloo/

Than you so much for the clarification and the link to chipsandcheese. I just read all 4 articles, including Isaiah’s Branch Predictor on part 1 and Lujiazui Branch Predictor on part 2. All very interesting to me.
Overall they point several issues/limitations and therefore points to be improved.
So I now understant your interest on their next article about KX-7000 and their take on changes from Lujiazui. I too am now very curious to read it.

[Kinoco]

Anyone has any news on KX-7000 production/availability?
What impact will have the (forced) refusal of ASML to provide maintenance on SMIC machines?

[Tralalak]

Zhaoxin KX-7000 review
source: https://pc.watch.impress.co.jp/docs/column/hothot/1626253.html

[Kinoco]

Thank you so much Tralalak.
So it confirms it is being produced.
I will get one. It is fine for me. I only do Browsing, office work, and 4K video playback. I never use 3D.
Windows drivers is also not an issue, I use Linux Debian testing, which I am hoping will have all required drivers already included.
The motherboard is also fine for me, being a regular desktop motherboard.
I searched on tabao and can't find any KX-7000, only the the usual KX-6780. So it seems I will have to wait some longer till those appear for sale there, as I don't know another place to buy KaiXian from Europe.
Thanks again for the update.

[Tralalak]

A live report on the performance of the Chinese-developed x86-compatible CPU "Zhaoxin KX-7000"! Its performance is... [Delivered at 9:00 PM on Tuesday, November 12th]​
source: https://pc.watch.impress.co.jp/docs/news/1638514.html


oeC0Mj3k56Q

[Tralalak]

Finally the KX-7000 in notebooks:

The first KX-7000 notebook! Lenovo Kaitian N8 Pro is launched with all the chips

https://0.rc.xiniu.com/g5/M00/2E/52/CgAGbGdJNeCAF7AeAABpxvV-dvs674.jpg
source: https://www.zhaoxin.com/news_view.aspx?nid=2&typeid=277&id=1740

[Kinoco]

A live report on the performance of the Chinese-developed x86-compatible CPU "Zhaoxin KX-7000"! Its performance is... [Delivered at 9:00 PM on Tuesday, November 12th]​
source: https://pc.watch.impress.co.jp/docs/news/1638514.html


https://youtu.be/oeC0Mj3k56Q

Thank you Tralalak for the link with the video review . Only now I had the proper time to see it 3 times in order to get all details. Youtube translation to English worked quite well.
I am decided to buy the KX-7000 now, but this "ASUS board" XC-KX700M D4 does not fit me as it has only 2 SATA ports and I need at least 3 SATA ports. Also it does not seem to have Front panel USB3 header. I would love the 2 digit display for code report though. In any case this board can not be found for sale on Taobao.
There is another board for sale there, but I am not sure how trusted that board can be in terms of quality. But it has the 3 SATA ports and Front panel USB3 header that I want. Can anyone advise for or against this board?
Also it seems another development board, with serial port connector. Do you people think there will be proper commercial boards for KX-7000 sometime soon or is it likely that this type of developments boards will be the only ones available for some time?
My main concern is the lack of BIOS updates from these devopment boards compared to proper commercial boards. Or do you think BIOS updates is not likely from any kind of board that may come for KX7000?
Thanks in advance for any advise.

[Tralalak]

"New" Kaixian® KX-U6980S and KX-6940S processors

The Kaixian® KX-U6980S and KX-6940S processors, which use domestic manufacturing processes, have been upgraded to support dual-channel DDR4-3200 memory, and have added SM2, SM3 and SM4 national secret algorithm acceleration instructions. At the same time, energy efficiency control has been further optimized, which not only brings a better user experience to various computing platforms, but also promotes the coordinated development of independent chip research and development, manufacturing, packaging and testing, and independent supply chain, improves the level of independent industrial chain supply chain, and promotes the sustainable innovation and development of the industry.


https://0.rc.xiniu.com/g5/M00/30/C2/CgAGbGd2DYyAU_JaAAECyAFHf0c727.jpg
source: https://www.zhaoxin.com/news_view.aspx?nid=2&typeid=278&id=1764
source: https://www.zhaoxin.com/prod_view.aspx?nid=3&typeid=129&id=327



P.S. Next-generation processors, please stay tuned KH-50000 Series Server Processors

Zhaoxin's next-generation server processor, KH-50000, adopts a chiplet interconnection architecture. A single processor can have a larger number of high-performance CPU cores. At the same time, the IO is upgraded to PCIe5 and DDR5, and it brings a new ZPI 5.0 interconnection, providing greater data bandwidth, lower latency, and lower power consumption to further meet the application needs of cloud computing, data centers, artificial intelligence, etc., and provide a high-performance and highly reliable computing power foundation for the digital transformation and upgrading of the industry.


https://0.rc.xiniu.com/g5/M00/30/C2/CgAGbGd2DrCAF_DWAAChliYCzOs815.jpg
source: https://www.zhaoxin.com/news_view.aspx?nid=2&typeid=278&id=1764

[Kinoco]

So here it is the KX-7000:

90826

I like this board: BIOS with selectable language (Chinese or English), and some features the ASUS board does not: VRM cooler, 3 SATA ports, 3 fan headers (although all share same PWM, not controllable separately). Although it lacks from the ASUS the Code Display and the USB4.
I still need to change the cooler that came with the board for a good dual tower cooler. Temperature may be limiting CPU and/or integrated graphics.

[Kinoco]

This BIOS (KunlunTech) reports 3.2 GHz CPU while on the youtube review, the "ASUS" BIOS reported 3.6GHz.

However under Windows 11, I get the exact same clocks as the youtube review: CPUZ reports only 3GHhz, then Task Manager reports 3.6GHz as "Base clock" with a maximum real case clock when overloaded of exactly 3.45GHz.
So it seems the CPU is the same.

According to Zhaoxin website specs, there are only 3.6 and 3.7 models for the desktop (plus 3.5 for SODIMM).
And then on taobao there is the chart below that specifies the standard (base?) and peak values for each model, stating the 3.6 is peak clock with base of 3.0 and the 3.7 is also peak but with base of 3.2.

So, it seems to me the "ASUS" board chooses to report the peak clock and KunlunTech board chooses to report the standard clock. On software side, it seems CPUZ chooses to report the standard clock and Task Manager the Peak clock.

[Tralalak]

Zhaoxin’s KX-7000
source: https://chipsandcheese.com/p/zhaoxins-kx-7000

[Tralalak]

Lenovo Kaitian N8 Pro laptop review: The mobile version of Zhaoxin KX-7000 is even more efficient than Intel/AMD
source: https://znyj.ofweek.com/news/2025-05/ART-12300-8330-30662920.html

[ChaosTM]

Ich hab keine Ahnung, wie gut der Chip sein wird, aber wir sollten uns Sorgen machen..

[Sardaukar.nsn]

Warum Sorgen? Ein ernsthafter neuer Wettbewerber kann den Markt nur bereichern.