you think ATI R400 will double the transistor count again?

[OpenGL guy]

OpenGL guy... don't take everything so personally man... relax... I was trying to show that everything is depending scale.

and besides, I don't know the transistor count G550, so that tripled was just a plain guess really. But I know that Parhelia has something like about 79 million Transistors and I am pretty sure that G550 had less than 30 million...

So the R300 has about 4x what the G550 has You're right, it's all a matter of scale

so IMO, limits are made to be broken in this market.

This is true... mainly because you can't let your competitors get a competitive advantage.

 

[megadrive0088]

Don't people already planning what 1 billion transistor CPU going to be like. I read over Toms Hardware, that Intel is going multi core as well for their 1 billion transistor CPU.

Do you think 1 billion transistor GPU will go multi core ? with huge cache ?
My hunch, GPU will hit 1 billion transistor count first before CPU.

I believe the first GPU with 1 billion or more transistors, that makes it into a consumer product, will be the Graphics Synthesizer 3 in Playstation3. Wether it is multi-core or not, I have no idea. but its massive transistor count will be so high most likely because of its huge embedded memory cache.

Sony and Simplex already produce a GS "1.5" with 32MB eDRAM. forgot the transistor count on that sucker. Sony was ment to have EE2 and GS2 out this year for workstation use. Then around the 2005 timeframe, EE3 and GS3 would be completed to form the basis of PS3 in 2005-2006. that according to Sony.

 

[megadrive0088]

If the R400 doubles the transistor count again, it's not coming our next summer/fall.

that was my thought as well. there is no way that R400 will be around before summer 2003. probably late summer/early fall like R300/R9700.
Between now and R400, we will probably see the R350 with 2 TMUs per pipe and DDR-II as ATI's answer to NV30.

 

[Vince]

EDIT: and of course, we have to remember that in 1999 (or was it 2000? I need to check this one.) Rendition / Micron already demoed with V4400 chip having more than 128 Million Transistors.

PixelFusion was similar, in that it had >100MTransistors devoted to it's SIMD/eDRAM arrays on a 0.25um process IIRC back in '99 or 2000. Pretty cool actually, but the per die costs were astronomical.

Sony and Simplex already produce a GS "1.5" with 32MB eDRAM. forgot the transistor count on that sucker.

The Graphic Synthesizer I-32 has 280-odd Million transistors, it's 0.18um and it's package is >20cm*20cm.

I don't think it has much, if any, relevance to what ATi is producing for the consumer marketplace though, probobly best to leave it outta this conversation.

 

[megadrive0088]

thanks for the info on that 32MB GS, I had forgotten that it was called
GS I-32. man, that's a HUGE transistor count. almost all of it is from the that 32MB of eDRAM. I would imagine if GS3 is similar in concept, that it could easily exceed a billion transistors.

getting back to ATI, with what has been said about process technology, I cannot see them constantly doubling their transistor count year after year, like they have done twice already.

 

[RussSchultz]

Except, of course, that the R200 happened at the beginning of last year, and the R100 about 18 months before that.

Hhhmmm, sounds a bit like moore's law, no?

 

[MfA]

Ok I found the reference to the EE3, of course by the same account the EE2 should be ready this year and appear in a workstation 100 times more powerfull than the original playstation ... and a workstation should have come out in 2000 based on the EE which was 10 times more powerfull than the original playstation. Given that string of failures with his predictions (Im assuming neither the EE2 nor the workstations will ever see the light of day) and since it was before the IBM partnership I think we can safely say that the 500 million transistor number is utterly and completely meaningless

 

[PC-Engine]

 

[Kristof]

The Graphic Synthesizer I-32 has 280-odd Million transistors, it's 0.18um and it's package is >20cm*20cm.

I think that's a typo... cm -> mm ?!

K-

 

[Simon F]

Sony is going for 500M transistors on .10 micron process with production to begin in 2004 at the latest (in time for PS3 in 2005) that is the main CPU chip I believe. The PS3 graphics chip could very well be even larger than 500M transistors, as PS2's GS was about 4x larger than the EE (43M vs 10.5M)

Whether this is correct or not, it's not really relevant. Sony can afford to lose a very large amount of money on each system because they get a cut from all the software packages sold.

PC IHVs have to make a profit.

 

[Dio]

Actually, the idea that a console manufacturer can burn money indefinitely on console hardware because of the software subsidy is somewhat of a fallacy.

The reasoning is that the average console owner buys something like 10 games over the life of the console (I can't remember the exact figure, but I was surprised, it was significantly lower than I expected). The software subsidy isn't that great - only a few £ per game - so you can't afford to subsidise the hardware assuming a £200 kickback from the software side.

Even if the figure was up in the 20-game region that I would have thought more normal for life sales, that's still probably under £100 to subsidise the hardware with.

The general rule is that the first few million units (about the first year of the product) can afford to be sold as a loss leader, but as console volume ramps you have to shrink your chips, cheapen components, etc. and get the hardware to break even. Then in the later life of the product - after a second or third shrink - the hardware should start making a modest profit (or at least should have room to do so before competitor price-cutting leads you down that road too).

 

[Randell]

Except, of course, that the R200 happened at the beginning of last year, and the R100 about 18 months before that.

Hhhmmm, sounds a bit like moore's law, no?

product launches were actually

R300 was launched July 2002
R200 was previewed August 2001/reviewed October 2001
R100 was launched July 2000

did the tape outs etc occur in a different relative time frame (~12 months bewteen each product?)

according to you timescale R100 occurred summer 1999.

 

[Simon F]

Dio,
I was only assuming a few dollars/Euros/whatever per chip. Something that could be achieved with the sale of a couple of games. (BTW the market that the Console companies want to get into is online gaming, where they can get more regular income (BTW I'm not implying that is bad in any way))

The thing is that the HW vendors in the PC market don't have that option at all.

 

[Ailuros]

As a layman I wouldn't compare PC graphics chips with console chips, simply because of the lifespan each can have. I might keep a console for a 5 year period which wouldn't be much of a problem, while on the contrary keeping even a high end graphics board for 5 years would make me a hero *ahem*.

 

[Johnathan256]

megadrive0088 said:

I like that idea alot since ATI would have 256-bit bus AND DDRII, while NV30 only has 128-bit bus with DDRII while being equal in the number of pipes and TMUs.

Where is the link to the site where NVIDIA said that NV30 will have a 128-bit memory bus? I am really curious as to where you get your facts.

 

[Megadrive1988]

Where is the link to the site where NVIDIA said that NV30 will have a 128-bit memory bus? I am really curious as to where you get your facts.

There are no links to Nvidia saying that NV30 will for certain have a 128-bit bus, only hints of that. Wasn't it David Kirk who said that 256-bit
was overkill? That's pretty strong indication, IMHO, that Nv30 won't have a 256-bit memory bus. Now I'd absolutely LOVE for NV30 to have a 256-bits bus since you can never have too much bandwidth. So I guess we'll just wait and see right?

 

[DadUM]

Where is the link to the site where NVIDIA said that NV30 will have a 128-bit memory bus? I am really curious as to where you get your facts.

There are no links to Nvidia saying that NV30 will for certain have a 128-bit bus, only hints of that. Wasn't it David Kirk who said that 256-bit
was overkill? That's pretty strong indication, IMHO, that Nv30 won't have a 256-bit memory bus. Now I'd absolutely LOVE for NV30 to have a 256-bits bus since you can never have too much bandwidth. So I guess we'll just wait and see right?

David Kirk said 6 months ago 256 bit was overkill "for the moment." But you must also realize that he also made this statement right after the paper launch of Parhelia which had a 256 bit memory interface. Which product does better, the GF4 Ti4600 or the Parhelia?

I am sick of that statement being offered as "proof" that NV30 must be 128 bit. Fact: I don't know what memory interface NV30 will have and I would not be surprised with either.

 

[Grall]

JonathanFanBoi:

There are no facts NV30 will have a 256-bit bus either, so don't get your panties all in a bind over nothing. Like Dadum pointed out, Nvidia representatives (Kirk, or somebody else) called 256-bit "overkill", which I certainly does not see them say if their next product features it.

Doesn't mean of course it won't have a 256-bit bus, but in most people's opinion it is a fairly strong indication it might not have it. Well, there's always Alexok's assurances it will have it of course, if you believe him.........

Anyway... Why does it matter to you or anyone else with half a brain if NV30 has a 4-bit or 2048-bit memory interface as long as it delivers kickass performance and IQ at an at least fairly reasonable price?

All you manage to do with your indignant comments is to sound like a whiney f*nb*y. If it's got 128-bit then it's got 128-bit and WHO CARES, really? Like with dicks, it's not size that matters (and trust me, I know this from personal experience. ).


*G*

 

[Megadrive1988]

I never buy the first iteration of a new technology anyway, so my next jump in graphcis cards will be the R350 or NV35 anyway. Though I HOPE NV30 has a 256-bit bus, because that guaruntees that NV35 will as well. because if NV30 has 128-bit bus, that doesn't necessaraly mean NV35 would get one.


right now I think the odds of Nv30 being 256 bit is 50/50 at best. Of course, I am extremely interested in seeing its IQ and texture mapping/pixel shading ability, as it will give good indication on what to expect from the entire line (NV31, NV35, etc)

 

[Guest]

Who believed Ati could do 110million trans on 0.15?

who would believed year ago what Matrox had on it's sleeves?
ATI doubled their transistor count, matrox more than tripled.

You should know nappe1 that I knew what they had up their sleeves you forgot that already

mmm guess I should cut it down more that pic still to big.....