V5 in 2K Follow-up

hmm, found a dead link and an article that had been pulled at NVIDIA\'s request so take that as you will....

As for believing 3dfx more than Nvidia because they have always been truthful, I don\'t know. I think their problem with the full OpenGL ICD is about as bad as Nvidia\'s statement regarding their clock speeds. How long has it been since they said they would do it? A year? From what I understand, it STILL isn\'t stable.

I don\'t trust the marketing hype from EITHER company. Show me benchmarks and screenshots.

magila: definately not a fully-educated guess, especially coming from a competitor. However, I would trust 3dfx on this a lot more than I would trust nVidia saying \"NV10 will definately be faster than Voodoo4\", simply because 3dfx has delivered on its promises in the past (IMO) more than nVidia. Of course I\'m not 100% convinced that Voodoo4 will be faster, not even close, but I just think if 3dfx estimates that it will be *on that particular benchmark*, then I think they have good reason to say so.

I\'m looking forward to the NV10 as much as anybody though, I own a TNT and I know that nVidia can get great drivers out there with excellent Quake3 support. 3dfx has yet to show this, so it\'s not like I\'m discounting nVidia just because of something 3dfx said. I\'ll wait and see, just like everyone else.

And please no more \"you are all stupid to argue about cards that haven\'t been released yet\" posts, it\'s getting really annoying and we don\'t need that here.

hey aleks, if you ask me I still think saying that V4 will be unequivocally faster then NV10 when we still have not seen the final clock rates for either chip is just a semi-educated guess

Traxx: we are free to make educated guesses all we want, given the information we have. You don\'t need to come here and tell us to basically \"shut up\". Nobody here is fighting, just making interesting guesses, nothing more nothing less.

I also thought this part was very juicy:

Patrick Grinsell: LetÂ’s just take Quake 3 at 1024x768, just throw that out there.
Jim Hopkins: Unequivocally our product will be faster than the NV10.

I find this pretty interesting because 3dfx has never really lied before. They never said that Banshe would be faster than Voodoo2 for example, at a time when everyone wanted to hear them say so. If they know everything we know, and that is that NV10 will be 2-4x faster than TNT2, then an unequivocal \"yes\" makes me think that Voodoo4 will at least be as fast as NV10 in Quake3. Of course we\'ll have to wait an see, but at least he didn\'t say something like \"we will position the next product as a mid-range, but wait till you see the NEXT thing\", that would have meant \"sorry, NV10 will be faster\".

uh..
traxx
i dont see any of the sencless bragging that was going on b4..
this is some real good arguments for both cards....
yes we know that they arent out..
but pro\'s and con\'s arent that bad :)

Guys. Come on. No card is better than the other at this point. Read: THEY ARE NOT OUT YET. Everything is subject to change... we all remember the TNT hype, V3 model changes, that first Galze3D product.. the NV10 does not \"own\" the V4/V5. The V4/V5 does not \"own\" the NV10. Chill. We can clearly see that nvida is going for better quality of the game, while 3dfx strives for faster play; however there is no clear choice for what is better at this point.

well i only chose 100fps cuz thats the fps cap in quake3 1.08 heh and i\'m used to seeing it when i play in 640x480
hehe

oops sorry for the anon post (sliped on the mouse)

First off about fill rate, fill rate is dependent on two things:

1. the core clock of the chip itself
2. the number of procesors runing in parallel
(memory clock can also have an impact but only if it\'s clock is much lower than all modern graphics cards)

the reason fill rate is purely dependent of clock rate is because, no matter what kind of tricks you try to do otherwise, you can only procces 1x[the number of parallel procesors] pixels per clock cycle.

now with that said the fact that the NV10 and the V4 will probably be made at the same dye size (.18 microns) they will also probably be clock comparably, meaning thier fill rates will be comparable.

now about T&L, this is were NV10 gets a very large advantage in fact that it will be implemented in hardware which is inherently much faster than doing it on the CPU (think 2 to 4x and beond), although many of you have raised a valid point about the fill rate being the bottle neck, with the massivly more powerful T&L engine we will see games which utilize upwards of 10,000 polys in an average scene which will be a great increase in realism.

and by the way 100+fps dosen\'t do shit since most people have a monitor refresh of ~75 Hz (which is plenty), fill rate IMO is close to critical mass because until some new amazing display tech becomes viable as a replacement for CRTs (not likely to happen in the near future) these new cards will be able to max out every reasonably priced (>= 1000$) display in existance.

Interesting that they give an unequivical statement that V4 will beat NV10 at 1024x768. From an investing/legal standpoint, this is a potential shareholder suit similar to the S3 Virge deal if they don\'t meet their promises.

If, as they admit, they don\'t know all about the NV10, this is a stupid statement to make, especially considering their recent performance in the stock market. Unhappy shareholders in today\'s litigious world go straight to court.

Of course, I wish the interviewer had stated more specifics, such as on an XXX speed processor at a certain color depth, but oh well. Interesting none the less.

I thought this was very interesting: It seems that whenever the topic of T&L discused, the specs of the Playstation2 come up. \"The PSX2 has (T&L)and thats why it can push 10m polys, and make those game screenshots look so sweet.\"

When comparing a video subsystem designed for a console with that of a PC, One should\'nt forget that the consoles output to a TV a therfore the target resolutions are much much lower that what we expect (need) from PC hardware. If beleive the resolution for a typical TV is only 320x240 with higher res only available on HDTV systems.

This being the case, I can understand why Hopkins is much more concerned about the fillrate bottle neck. Fillrate is much less of an issue for a console, while most gamers will refuse to play games lower than 800x600

That said, I\'m sure T&L will eventualy be nessasary to have. But, will nVidia be able to deliver T&L + the Fillrate to match 3DFX\'s next chip while still remaining competitively priced?

Jim Hopkins: The presentation from engineering is that if you have a game that is aware of a hardware based geometry engine, if youÂ’re running at 640x480 you will actually, and you have what we know of NV10Â’s fill rate, then the geometry engine will help you quite a bit. If you were above 640x480 in resolution, you will become fill rate limited before the geometry engine becomes a problem. So, as you get higher resolution, above 640x480, fill rate becomes a bigger bottleneck than having hardware geometry does. IÂ’m not saying that Napalm does or doesnÂ’t have the various and sundry things that are included in a geometry engine. What Napalm, if thatÂ’s the name, indeed does have is the ability to provide fill rate thatÂ’s faster than anything else out there and at resolutions, particularly 1024x768, our engineering guys have calculated that fill rate is the ultimate bottleneck that keeps you from being able to run faster on benchmarks.

#15, no one said they didn\'t

Should have read your link b4 asking :)
Guess nvidia will win round... 3? or is it 4?
i\'m really hoping the V4 or V5 will be an answer to the nv10..
cuz the nv10 will own the market for a long while..
and whats gona happen when cards are pushing 100fps in quake3
i think they should devote more attention to the actuall look of the games in the future..
once ya max out monitor refresh rates... THEN WHAT? where will we go when we hit the video card maximum polys per sec?
video cards are gona be comming to an end...
i hope there will be a way to accelerate voxels/voxtels (what ever) in the future..
they (voxels) look promicing on games like DF2..
and i\'d love to see voxels in tribes 3 (hehe)
but a full voxtel game is a way off.... (one that LOOKS nice)
just my 2 and a half cents again

is it confermed that the nv10 WILL have t&l?

yes, it is

OMG!
this is a first steve...
no 3Dfx vs NVidia comments.
ie.. 3DFX rules nvidia sucks or vice versa.. :)
is it confermed that the nv10 WILL have t&l?
i dont remember hearing offically nv10 will have t&l
its gona be a back and forth battle until one finally smacks the other one with a death blow...
it\'ll probably happen some time in the next year or so..
i\'m still debating on which card will be better..
they both look promicing..
My guess is as good as steve\'s and any one elses...

oops I messed up the auto-linking there, go to http://www.riva3d.com/nv10/nv10.html for the report

\"The NV10 also takes advantage of the .18 micron die...\" taken strait from the Riva 3D NV10 report (http://www.riva3d.com/nv10/nv10.html)

I tend to agree with #1 but I can\'t help but take issue with their statments regarding the speed of the V4 vs NV10. Although they do admit they don\'t know a whole lot about the NV10, I can almost guaranty you they don\'t know enough to make such difinative statements as \"Yes\" to the question of which will be faster. Also I think they also do something which 3DFX has been known to do in past (no one needs 32bit color and large texture support, 3DFX talking about V3 vs TNT2), and that is downplaying potentialy revolutionary features which they happen not to have (in this case it is hardware T&L). Refer to my #33 post under \"Voodoo5 in 1st Half of 2000?\" and the resulting conversation. And with the very high clock rates the NV10 will acheive with a .18 dye size it is doubltful that V4 will be able to beat the NV10 in terms of fillrate by enough to overcome the NV10\'s onboard T&L.

BAH! \"company\'s\" not \"companies\"

sheesh...

On page 10 (yes, I read that far, hard as it was) I was totaly dumbfounded that both Ballard and Hopkins were under the impression that a full OpenGL ICD had already been released. Ballard even tried to direct the interviewer to the files on the 3dfx site!

Jim Hopkins: Thanks for the heads up. WeÂ’ll push on that now that weÂ’ve heard about it.
(After Grinsell relates that lots of people are complaining about the lack of a full ICD.)

Like I said... I can\'t believe that these guys are SO out of touch with things. It makes me fear for the companies future. Now why did I buy that stock again? Someone please remind me...