Interesting comment on Columbia and NV30 from Anand

[KimB]

No, NVIDIA has had a record number of tapeouts. A tape-out is the approval of a physical design layout that is sent for masking, to produce the mask set for manufacturing. If a design needs revisions that's one thing, but "tape-out" refers to the first time it is sent out to the mask house.

Guess I'm not entirely up to date on that terminology, then.

But, it shouldn't change much of anything, as I'm pretty certain that the "record number of tapeouts" was stated significantly after the release of the GeForce4 cards...

 

[3dcgi]

But...
If it is a massively parallel programmeable design, with a bunch of identical execution units, perhaps it's not as huge an undertaking as it seems. I mean for the chip after this one with 200million transistors,
how are they going to model/debug/verify the chip? I'm assuming that besides being nice for the 3d developers, programmability makes the chip designers job easier as well...

These chips are only getting harder to design and I believe programmability makes the chip even more difficult. Routing becomes a lot more difficult because the output of say an arithmetic unit might not go to just one place like in a fixed function pipeline. Not to mention there's more overhead in control logic. Also, I would be shocked in nvidia got it right with the first tape out. That would be extraordinary for a chip of that complexity.

 

[KimB]

I don't think there's any such thing as getting it right with the first tapeout with any remotely-complex processor...

 

[ActionNews]

Hmmm....but VIA is interested in PowerVR-tech:

( http://prohardver.hu/ph_cikk.php?id=98&p=1 )

" There were many gossips about VIA and ST Microelectronics cooperating on the graphics market in one way, or another, through PowerVR's Kyro series graphics chips. Could you briefly sum up what's happening and what we should expect from VIA in this area?"

"The Kyro technology is certainly very interesting, and we are continuing to evaluate it. At the same time, however, S3 Graphics is executing very well on its roadmap and will be introducing its new low power Zoetrope core very shortly. This will be the first step in achieving our long term objective of returning to the mainstream graphics market. "

This interview is from May!

And this is from www.digitimes.com :

VIA eyeing ST’s graphics chip division?

Charles Chou and Emma Wang, Taipei; Christy Lee, DigiTimes.com [Wednesday 6 March 2002]

In response to a recent rumor that it is planning to acquire the graphics chip division of STMicroelectronics (ST), chipset designer VIA Technologies on March 5 only confirmed that it has indeed been in contact with some international graphics chip companies, but refused to comment further on the progress of the plan, and more importantly, the identity of the candidates.

VIA said that its graphics chip acquisition, S3, may have leading advantages in the notebook sector, but to strengthen its desktop graphics chip product lines it has started reaching out to some graphics chip designers for possible cooperation. The company further noted that acquisition is not necessarily the only way to obtain related technology.

Industry sources said that they would be conservative about the actual value of the purchase if ST’s division is truly a candidate VIA is considering. They indicated that the market response to ST’s Kyro II line, launched in 2001, was rather weak.

ST transferred its graphics chip technology from Nvidia, and had manufactured graphics chips for Nvidia, four to five years ago. The production cooperation only lasted for just over a year. As for the driver program for its graphics chips, ST has adopted the technology of UK-based VideoLogic Systems. Given the moderate performance of VideoLogic’s driver program, compared to those of Nvidia and ATI Technologies, it is believed that VIA may not purchase the driver program segment if an acquisition were to be proposed.

VIA obtained S3’s Savage graphics core technology by acquiring the company in 2000. With S3’s expertise in notebook graphics chips and graphics-incorporated integrated chipsets, VIA successfully entered the notebook sector. At present, the Savage core technology has been updated from the Savage 4 to Savage 8. VIA’s chipsets, such as the P4M266, PN266T and KN266, are all based on the Savage 8 graphics core.

Currently, VIA and S3 are designing a new-generation graphics core titled Zoetrope, set for launch in mid-2002.

So....am i right, that VIA says that they need a partner for desktop graficscards?

CU ActionNews

 

[mboeller]

"The Kyro technology is certainly very interesting, and we are continuing to evaluate it. At the same time, however, S3 Graphics is executing very well on its roadmap and will be introducing its new low power Zoetrope core very shortly. This will be the first step in achieving our long term objective of returning to the mainstream graphics market. "

You can "translate" this also like
"PowerVR is nice and good but thanks a lot we are not interested, cause S3 is by far good enough for the foreseeable future. They have developed the Zoetrope for the notebook market already and are on the way to develop (DX9)chips good enough for the desktop market also, so we don't need PVR at all"


By the way, anandtech has changed the text about NV30 and Columbia already so the quote above is the last sign that something is coming from VIA.

 

[Nexus]

I don't think there's any such thing as getting it right with the first tapeout with any remotely-complex processor...

You think too much and know too little..

http://www.anandtech.com/cpu/showdoc.html?i=1591&p=4

 

[Dave Baumann]

I don't think there's any such thing as getting it right with the first tapeout with any remotely-complex processor...

You think too much and know too little..

http://www.anandtech.com/cpu/showdoc.html?i=1591&p=4

And given what I've just seen I definitly know that not to be the case...

 

[nAo]

And given what I've just seen I definitly know that not to be the case...

C'mon, don't tease us

 

[PiNkY]

"...record number of tapeouts..."

Wouldn`t most of you agree that nvidias most important high-end part currently is nv2a + mcp in xbox (prop. 5m parts / year) and their respective engineering team is under immense pressure to cut cost on that chip(set) (remember the Nvidia vs. Microsoft pricing dispute). Also with Sony about to shrink GS + EE to 0.13 and integrate into one single package (i am not so much into consoles, but thats what you keep reading in the by3d forums) and the recent price cuts just prior to E3 this must be a major issue to them and could potentialy hinder their pc performance more than people imagine (remember that Carmack hint about them loosing ~ half year of their tech. advantage because of X-Box). Also there are at least 2 pc chipsets in developement (AMD k8/K7) as well as their shrinked versions of nv17/nv25. imho the importance of the latter two chips for nvidia is often overlooked, because margins in the then mid-range to value dx8 product range are bound to be much lower with all the new players entering the market and that range has been nvidias cash cow for the last 3 years (think tnt2 + gf2mx). Also the tapeout comment was made on their shareholders conference as explanation for higher than anticipated engineering costs. This rather suggests they are really starting to feel the pressure now in comparision to their last 2 years of unchallenged rule.

 

[Ailuros]

"The Kyro technology is certainly very interesting, and we are continuing to evaluate it. At the same time, however, S3 Graphics is executing very well on its roadmap and will be introducing its new low power Zoetrope core very shortly. This will be the first step in achieving our long term objective of returning to the mainstream graphics market. "

You can "translate" this also like
"PowerVR is nice and good but thanks a lot we are not interested, cause S3 is by far good enough for the foreseeable future. They have developed the Zoetrope for the notebook market already and are on the way to develop (DX9)chips good enough for the desktop market also, so we don't need PVR at all"


By the way, anandtech has changed the text about NV30 and Columbia already so the quote above is the last sign that something is coming from VIA.

mboeller,

It didn't cross your mind that Anand may have been asked to remove that part hm?

According to the few things I know - yet can't get into specifics - if VIA would ever licence anything from ImgTech then it wouldn't be a low end part for sure. What S3 has in the works despite the possible compliance isn't anything above budget AFAIK, but could be wrong either. If it would have been "good enough" then VIA wouldn't have been "snooping around" for alternatives all along. Don't expect anyone to verify that negotiations actually took place after all too

If there is a possibility of a future cooperation of the two it is far too early to announce anything before the licencee will be able to evaluate any real product. In terms of performance/compliance Series4 was quite ahead of Zoetrope; why it didn't end up to an agreement? Ask ST for that.

 

[ActionNews]

??? The text about Columbia is still in the article! I found it some minitues ago!

But something other at Anandtech:

I read over at RIVA Station of some rumors concerning S3's supposed upcoming 3D chip dubbed Columbia which will feature a programmable Tile engine called Chameleon. Here's a clip:

Apparently, S3's Paul Ayscough has said "You can't have a low end part anymore without a high end part." The programmable tile render is rumoured to be named Chameleon.

Columbia graphics core:

• 
  All new 3D Engine supporting:
• Quad pipe, dual texture
• Full DX8/ 9
• support from View frustrum clipping down (Pixel shader, anisotropic filtering, cubic environment
  mapping, 3D textures, point sprites, Hierarchial Z and deferred rendering, Pixel rate of 1.2 GP/ s w/ 2UV, trilinear filter)
• 300MHz Engine Clock
• 128-bit DDR @ 300MHz DDR
• Process: 0.13 micron
• Engine samples - Q2'2001

When you look at the combination of the Tile renderer and Hierarhical Z with 600 MHz memory, we can clearly state that this could be a very advanced card to be shipped with ATI's R300 and Nvidia NV30 in spring 2002.

But it's from May 2001!
If it's true that Clombia is a deferred render, i think i'll take it if PowerVR doesn't get VIA or someone else as a new partner!
But i also ask me why then VIA looks for a partner in grafics-sector?

CU ActionNews

 

[Ailuros]

If those specs were for real Action News, then it's not a Tiler, rather an IMR combining deferred rendering via software and a hierarchical Z-buffer.

If that's supposed to be even a half way dx9 product then it's extremely underwhelming. Not enough fillrate/bandwidth and not enough textures per pass to even call it competitive with what is about to launch. That looks rather low end to mid end to me.

.13um and samples in Q2' 2001? ROFL

 

[3dcgi]

I don't think there's any such thing as getting it right with the first tapeout with any remotely-complex processor...

You think too much and know too little..

http://www.anandtech.com/cpu/showdoc.html?i=1591&p=4

Those Hammer chips were not "getting it right." The clock speeds are lower than the final production chip will be. Getting it right means original tape out to production with one spin. On a complex design there are always some problems. It's rumored that the GeForce4 only took one rev, but it also wasn't a completely new design.

 

[Pete]

"Those Hammer chips were not "getting it right." The clock speeds are lower than the final production chip will be. "

But the rumors are that AMD is purposefully limiting all samples to 800MHz, to not reveal their full power (they have 1667MHz samples working as well)?