nVidia has confirmed nv30-125m transistors

[Wombat]

That's not so true these days, as images get pre-processed and send back to RAM. Otherwise people wouldn't have been so excited about eDRAM.

[PowerHungry]

Maybe instead of spending time finding out useless facts like "If converted to US Dollar bills and placed end-to-end, would stretch from New York to Hong Kong.", they could research some better image quality techniques.

[isochar]

If it delivers the 2x performance over a GF4, then it'll hit its target market perfectly - crappy 2d won't matter. (Think of the bragging rights of 22k+ on 3dmark at stock speeds)

[WyWyWyWy]

Originally posted by isochar
If it delivers the 2x performance over a GF4, then it'll hit its target market perfectly - crappy 2d won't matter. (Think of the bragging rights of 22k+ on 3dmark at stock speeds)

Actually that might be a good thing. It would push the competitors (ATI & Matrox) so at the end of the day it is all our benefit.

[superfly]

Originally posted by isochar
If it delivers the 2x performance over a GF4, then it'll hit its target market perfectly - crappy 2d won't matter. (Think of the bragging rights of 22k+ on 3dmark at stock speeds)

Doubtfull that'll happen...Current cards like the 9700 are already pretty much cpu limited in 3d mark,at least when the benchmark is run at it's default resolution....


2X the performance is probably at higher resolutions and with AA/aniso filtering enabled,much like what happends with a 9700 pro over a GF4 4600...The biggest differences happen at those settings....

[isochar]

Doubtful, perhaps. However, I wouldn't put it past nVidia to pull off something like that.

I personally am expecting quite an impressive card with hardware/software engineers from nVidia/3dfx/gigapixel (plus ATI & Matrox) working on their first project from the ground-up. And with the initial impressions that a few lucky souls have expressed, it seems that the nv30 will introducing some very interesting technology.

[Novdid]

That score in 3dmark is all about the CPU you have in the box. Hence my conclusion is that they are calculating that they will have the chip out when the top of the line CPU's are Clawhammers and Prescotts...

[isochar]

No, it's not all about the processor. Throw a Parhelia and a 9700 Pro in the same box and see almost a 2x difference at the standard benchmark.

[Novdid]

Yes, but in the Parhelia example it's definitly not the CPU that is the bottleneck...

[superfly]

Pretty soon,all video cards will be cpu limited even at resolutions as high as 1600*1200 32bit/4X anti-aliasing and 16x anisotropic filtering,if the current pace is maintained,and i believe that it'll happen within the next 18 to 24 months,considering there's already talk of upcoming video cards having nearly 60 gig/sec worth of memory bandwith by that time...

We'll be lucky if we see prescott p4's running over 4.5 ghz by then,which at best,will be perhaps 60% faster on average,compared to the fastest chip you can get your hands on today...

Cpu's double in overall performance every 18 to 24 months on average,while video cards in that same time frame,get 5~6 times faster....If we look back just 2 short years ago,the highest clocked cpu was a 1.5 ghz P4 and the fastest video card at the time at least as far as gaming goes was a GF2 ultra..

[Wombat]

&lt<!>sigh</!>&gt

[Pace]

PowerHungry: Add to that their figure about French and German speakers. I can't be bothered working it out, but I think it'd be easy to prove a little white lie in that statement.

Wombat: The "knowledge" returns

P.

[Wombat]

Knowledge? Just memory.

The "advancement" of video cards has almost no tie-in to CPU speed. The processor needs of the video card are design-dependent, and don't have any relation to its overall performance.

<I>A selective abridged history:</I>
Video cards used to almost always be CPU limited. Then things jostled around a bit, and almost any CPU would provide more than say a TNT relied upon it. When the G400 came out, it relied a lot on the CPU to do some of the work that the WARP core couldn't do. Even Matrox didn't know where the ceiling was. Most people had around 600MHz Athlon/P3s at the time, and it wasn't until Greebe got his G400 paired with his unlocked Athlon (which ran at 850) did we see the G400 peak with around a 750-800MHz processor (Quake benchmarks, some apps required more or less). Then the "revolutionary" GeForce came out, and nVidia hyped the shit out of it, saying that the T&L engine was an "advancement that would free up the CPU to do AI calculations for today's demanding games," yada yada. Now we have the 9700, an entirely different architecture, that seems to fall back on the CPU much more. Superfly missed the mark completely - it has nothing to do with the "current pace," just the design o' the day.

If you want to make any prediction at all about it, maybe in the long run video cards will be forced to do more work on their own. There's nothing that design improvements or process changes can do to change the fact that using the processor requires the video card to run electrons all the way through the card, an interface slot, over the motherboard, and through the processor's interface. The speed of electrons through a conductor is only so fast, and travelling that entire distance will become a hindering factor.

[WyWyWyWy]

Originally posted by superfly
Cpu's double in overall performance every 18 to 24 months on average,while video cards in that same time frame,get 5~6 times faster....If we look back just 2 short years ago,the highest clocked cpu was a 1.5 ghz P4 and the fastest video card at the time at least as far as gaming goes was a GF2 ultra..

Actually, why?
If graphics cards are more profitable, I can understand. But now CPU and graphics cards profits seem to be on the same level, then why ain't CPUs advancing as quickly as graphics cards? Competition? Hmm....

[Greebe]

Pull your head out of the AMD and sniff the Intels