CEBIT 2004

[Gast]

Hi, da ich mich ein bisschen gelangweilt habe, machte ich mich einfach mal aus Spass an die Übersetzung des Cebit Artikels. Beginnend mit Seite 1. Seite 2 ist auch schon fertig. Seite 3 hab ich jetzt das erste Drittel. Vielleicht kann ja einer schon mal drüber schauen.

CeBIT 2004 Report
23th March 2004 / by Leonidas / Side 1 of 3

Last Friday and Saturday we (Leonidas, aths, Xmas and CapM[Webmaster]) were together on this year's Cebit. By courtesy of a good preparation we could overcome the two days halfway stress-free because we haven't arranged too much appointments and we've had enough time for the additionally on the CeBIT arranged appointments. At this CeBit a lot of interesting things we could hear in the discussion with the manufacturers. This report shall refer about it.

About nVidia and specially about the NV40, we don't want to talk so much about it or not at all speculate in. We will be present at the market launch of NV40 with deep informations about the technology of the NV40-GPU and maybe with a benchmark part. This means that my hands are tied a bit till the specific date of the market launch (insider will know how it is mean). But up to the middle of April the time isn't so long and we can promise that the wait on our article will be worthwhile.

On the MSI exhibition stand you can snatch in a certain manner "official" Informations about the NV40 because MSI introduce their corresponding Videocards to a selected specialized public (probably distributors and OEMs). With these information can be confirmed that the NV40-Chip has an internal AGPx8 interface and NV40 Videocards are planned with a 128 and 256 MB memory assembly at present - however obviously still no 512 MB.

Otherwise we could see some pictures of GeForce PCX Videocards. These Videocards with which nVidia makes the best known chipsets NV18, NV34, NV36 and NV38 "ready" for the new connection by using a PCI express Bridge. So far nVidia released only one picture of a GeForce PCX 4300 on NV18-Basis. Following we see now a GeForce PCX 5950 on NV38-Basis as well as a GeForce PCX 5750 on NV36-Basis:

On both Videocards you see the passively cooled PCI Express Bridge component near under the GPU. These Videocards with PCI Express will come into the market probably shortly after the launch of the first PCI express Mainboard-Chipsets anyhow some could see at several board manufacturers.

Unfortunately we were not so successful with informations concerning ATi and/or Testsamples to their NextGen-Projekt. Therefore we can only speculate together with the Web about the R420. We have however no information from first hand to offer. Very probably the R420 was running in the Cebit Backroom. And perhaps whether there something is moveable up to the R420-Launch :-).

Now, at least we could hold once again a high-interesting conversation with Sapphire, one of the most important board manufacturers of ATi Videocards. Thus we could see also a R420-Cooler of the manufacturer Arctic Cooling for the Sapphire version of the R420-Chips (sorry, unfortunately no picture). This reminded of the bare sight not insignificant to the NV30-"leaves blower". ;-) Us it was assured that the design of the entire cooling construction was so enormously only be able to set the fan speed very low so the R420-Cooler is contrary to its appearance rather "silent". The standard cooling solution from ATi is another more ordinary version. Sapphire will go a special way with their cooling solution for the R420.

For the R420 it remains thereby that it provide "only" Pixel Shader 2.x as here already announced. ATi will simply try to omit the Pixel Shader 3.0 and want to be primarily the performance leader with the R420 (internal AGPx8 interface) and/or R423 (internal PCI Express x16 interface). In this connection in the Web finally 12 Pixel Shader units were announced for R420. NV40 should have 16 of it (we do not want to commentate the last number).

Many read a coming victory of the nVidia hardware already from this comparison. Apparently it does not seem in such a way ATi counted on a solid/huge Design of nVidia. Nevertheless we think that this duel will become very scarce probably once again, with which the final clock rates of R420 and NV40 will definitively be the determining factor which are apparent not yet fixed however at present.

Much will depend surely also on the selected memory clocks, because Samsung GDDR3-Speicher is available up to physical 800 MHz finally. But to take advantage of this possibility would make a Videocard also very expensive. Samsung could surely initially this very highly clocked memory balanced out with gold. If R420 and/or NV40 should rise equal with 800 MHz memory clock into the ring, besides then the clock area for the refresh chips - following sometime in the autumn - would become clearly smaller. Here it applies to be waiting simply, for which possibility the two videocard developers will decide.

Concerning the R420 we could snap open another interesting rumor, according to which the R420 could be an original design with 16 Pixel Shader units. Due to - alleged - yield problems in production however ATi would deliver this chip initially only with 12 active pixels Shader units. The other 4 units would be deactivated. If this should be correct in such a way, ATi could possibly, after solution of these initial yield problems, in a few months attack again also with a R420 with 16 active Pixel Shader units. Thats an interesting rumor, but at present naturally no more than a rumor. :-)

And in the long run we could make a snapshot at the Sapphire exhibition stand of their Dual-R300-Videocard sample which is in former times already once in our news mentioned even if it remained nevertheless then with this design concept. Also the coming GPUs of ATi and nVidia will be again technically prepared for this multi-chip possibility but because of missing economy no board manufacturer will dare itself to such constructions.

[Gast]

XGI on the other hand presented a new LowCost DirectX9-Design with the Volari V3XT - as already in the news reported. Likewise as already in the news assumed, it can be said now surely that it concerns here the XG42-GPU and not the XP5-GPU descending from Trident, which is used by the Volari V3. However the Volari V3XT is probably interesting really only for the absolute LowCost market because it is equipped with only one pixel Shader 2.0 unit. XGI wants to attack the GeForceFX 5200 with this chip probably exclusively - but from down :-). On the other hand of the chips with high order of XGI there is unfortunately nothing new to report. A little more gave it from S3, whereby we inferred much information here from the discussion with Club3D. So we could photograph the GammaChrome-Videochip with internal PCI Express Interface, introduced recently by S3:

That Club3D could show here a first (although single) GammaChrome-Board gives hope in the direction that S3 can bring this chip in time into the market - at present the third quarter is planned as market entrance for GammaChrome. And obviously it concerns with this - as already assumes - the regular Refresh-Chip DeltaChrome. Beside the PCI express interface the chip will bring along also internal improvements (however probably few) and possibly higher clock rates, particularly the latter would be very interesting because already the delta chrome scaled outstanding with rising clock.

For the DeltaChrome there is now also hope concerning the deliverability in our degrees of latitude. Club3D received recently a few Samples from S3 - finished Videocards are to come now in approx. three weeks on our market. If S3 can improve here still the drivers, which quite already are now on a good way, could become the DeltaChrome - in its market segment upto 150 Euro - quite a passable alternative. And in the long run still a picture of a ATi RV380-Samples was to be seen at Club3D (Club3D manufactures Videocards with ATi -, nVidia, S3 and XGI Chipsets), whereby appropriate RV380-Samples was to be seen just as also with some other manufacturers:

The RV380-Chip, like here already described, will be the first GPU at the market with native PCI Express interface to be implemented, because it is to start in the second quarter. Otherwise it concerns however "only" one otherwise unchanged RV360 alias Radeon 9600XT.

With the Mainboards our interest primarily applied for these boards, whatever have a PCI express x16 and an AGPx8 port on the PCB and thus the choice leave to the buyer, which Videocard he wants to use.
VIA here the outrider seems to be, because of them beginning PT890-Chipsatz possesses shortly this possibility in the Chipset (although the ports are not at the same time usable) and also the reference Design of the PT890 is equipped with both card locations. One of the first manufacturers of which this solution offered - as already reports - was Chaintech:

A similar board had to have been seen with Soltek, have seen in such a way is also with Asus be, and in the long run MSI plans to equip according to unofficial instructions its PT890-Mainboard likewise in such a way.Possibly then also the other Mainboard manufacturers pull tight accordingly and build their PT890-Boards correspond the Via reference PCB with a PCI express x16 and a AGPx8 card location. Us besides it was said that VIA with the AMD counterpart of the PT890, the K8T890-Chipset, a same configuration is planned thus again with both card locations (picture at Beyond3D). Unfortunately the K8T890 is however only expected for the third quarter, while PT890-Platinen are to already come in the second quarter into the market.

Otherwise one could see many Grantsdale (i915) and Alderwood Mainboards (i925), because Intel will shortly bring these new chipsets with PCI Express card location, Socket 775 and DDR2 support into the market. With most memory manufacturers therefore also DDR2-Speicher in the forms DDR2/400 and DDR2/533 was laid out which is cheaper than predicted and costs only 50% more than DDR1. DDR2/667 stands on the plan for the first quarter of the next year although with the very high latencies of 4-4-4.

As we however communicated by MemorySolution got, faster clocked DDR2-Speicher than DDR2/533 may come probably still in this year into the market (perhaps DDR2/667, perhaps also intermediate variants like DDR2/600) perhaps still air for clock increases is with the production of DDR2. The performance of the current DDR2/533 will exceed naturally particularly because of the high latencies hardly those from good DDR1 memory due to the marketing-effective name nevertheless most memory manufacturers set significant on DDR2.

Not however so OCZ, which keep out at present completely from DDR2, they want to set only on DDR1. OCZ shows DDR1/600 on their exhibition stand which is to come shortly into the market. Besides we experienced in the discussion that OCZ could supply DDR1/650 if they wanted and besides up to the end of the year, DDR1/700 (!) will be able to reach. On DDR2 OCZ was not at all good to speak and accused to this memory constructional lack which prevented in the long run high real achievements with this memory.

If we are honest, the information with the DDR1/700 throws naturally some over the heap which we thought so far concerning DDR2-Speicher. Naturally it was also clear that DDR2 needs a good piece more clock due to its worse latencies, in order to be able to be same fast as DDR1. But at the latest with DDR2/667 the effect should adjust itself nevertheless that DDR2 is better than DDR1 - so the theory, based always on fastest DDR1 memory modules thus DDR1/550. The initial frequencies of DDR2 would be relatively senseless, but at least with the high-clocked versions one would have an advantage.

But if DDR1/700 is possible for OCZ and will come also into the market, DDR2 is redundant in our opinion. The fastest DDR2-Art should be DDR2/800, which 2005/2006 are to come sometime into the market. But this DDR2/800 might not be faster due to the high latencies of the memory in the long run than DDR1/700 - and thus then also the last advantage of DDR2 disappears. The memory type might replace however probably on a long-term basis DDR1, simply because they are better to merchandise themselves out by name and higher clock frequencies. In addition, a complete failure of DDR2 in the market is possible, if the problem of the difference in prices is not solved.

We received high-interesting from OCZ also about GDR, GDDR2, DDR3 and GDDR3. So far we assumed for this that GDDR2 and GDDR3 are only pure videochip modifications of the standards DDR2 and DDR3. With GDDR2 this may be correct still halfway, but with GDDR3 this does not apply. Because DDR3 is not specified yet at all, the memory manufacturers won GDDR3 thus exclusively from GDDR2 and there turned off the mistakes, which were made with DDR2 and/or GDDR2, as far as possible.

Therefore one has much more clock possibilities with GDDR3. For example Samsung stopped the production of GDDR2 and delivers GDDR3 at present up to 800 MHz. Thus the coming DDR3-Standard can be quite differently than the GDDR3-Standard which is finally no correct standard, but only a self-creation of the memory manufacturers with actually misleading name.

[Gast]

First of all, Intel confirmed us that the upcoming D0-Stepping indeed was to serve a lower energy dissipation for the Prescott processor. Furthermore the upcoming rating system was definitely confirmed on the part of Intel that all desktop and mobile processors starting from the second quarter will have this additional designation (more exact, although still unofficial information at present see here).

Intel does not want to indicate the achievement with this processor number, separate it is alone a clear number for each processor to be found, since the clock frequency is not sufficient as difference any longer due to that presently many clock-same processors by Intel. And naturally thereby also the position of the Pentium-M processor is to be improved which in the final customer market due to its low clock some sales problems to have is :-). Thus the "Processor Number" will follow also no determined pattern, the clock, FSB, Cache size or other one cannot be interpolated from these number, because Intel make a rough guess to select the numbers

Concerning the 64-Bit-Enhancements of the Prescott processor Intel gave itself rather clam which is somehow logical, the feature comes nevertheless from the competitor AMD. Intel waits at present for the appearance of 64 bits operating systems and applications then they will activate this feature for the Prescott which carries already inactively this feature at present in itself.

Besides we spoke with Intel also about future Intel-processors, whereby Intel however hardly moved out some new informations. Interesting is if necessary that Intel considers to use the Prescott core and/or its Xeon-Variation in the future also for the Pentium of 4 Extreme Edition - naturally then again with an appropriate Level3 Cache in contrast to the normal Pentium 4 Prescott. In the discussion about future Intel cores arose besides, that the Intel representative pronounced the Nehalem processor (which is coming up after the Tejas and under what a completely new architecture is assumed) in a different way like us and thus the existence of this code name indirectly admitted :-))).


***Anmerken möchte ich noch, daß ich kein Experte des englischen bin. Der eine oder andere technische Ausdruck ist sicher falsch. Auch bei Gramatik und den Zeiten hapert es bei mir doch manchmal kräftig. Hab mir aber wirklich mühe gegeben das halbwegs sinnvoll zu übersetzen. Übrigens ist es mir schon klar, das andere Artikel hätten bevorzugt werden sollen, aber eigentlich hab ich den Artikel nicht übersetzt damit er nachher "Online geht", sondern nur so für mich zum Spass und zur Übung.

Mit freundlichen Grüßen

ein netter Gast, der sich freut das die Regpflicht noch mal abgewendet wurde.

[huha]

Hmm, sieht mir irgendwie sehr stark nach maschineller Übersetzung aus?! Kann das jemand bestätigen?

-huha

[Aqualon]

Original geschrieben von huha
Hmm, sieht mir irgendwie sehr stark nach maschineller Übersetzung aus?! Kann das jemand bestätigen?
Ich bezweifle, dass es eine maschinelle Übersetzung gibt, die aus Leonidas Schreibstil so einen Text erzeugen kann.

1. Absatz von Gast

Last Friday and Saturday we (Leonidas, aths, Xmas and CapM[Webmaster]) were together on this year's Cebit. By courtesy of a good preparation we could overcome the two days halfway stress-free because we haven't arranged too much appointments and we've had enough time for the additionally on the CeBIT arranged appointments. At this CeBit a lot of interesting things we could hear in the discussion with the manufacturers. This report shall refer about it.

1. Absatz von Google-Übersetzung

On last Friday and Saturday (19 and 20 March) we were, let us speak Leonidas, aths, Xmas and CapM (Web master), together on the CeBIT of this year. By means of a good preparation in that form, than that we had planned not all too much at dates, we could master the two days halfway stress-free and also additionally the dates still developing in the course of such a fair notice. In the case of these CeBIT really once very much interesting in the discussion with the manufacturers resulted, about which this report is to report.

Auch wenn es stellenweise noch holprig ist, denke ich schon, dass der Gast es von Hand übersetzt hat.

Aqua