HUGE 3dfx interview w/Gary Tarolli

[ElFenix]

i think the point about what would you really like, a card that is wild or a card that is stable, is best illustrated by the same turning 360 degrees in 1 second that we do with everthing else. when you're at 150fps, you turn 2.4 degrees per frame, so at 75 you're turning 4.8. would you rather have it going from 2.4 to 4.8 and back, or staying at 2.6+/-0.1 the whole time?

 

[AlabamaMan]

The point though is that when NVIDIA does release NV20, they are going to hype HSR and claim probably full overdraw removal. And you know what, I'm betting they'll have a tech demo to show it and I bet I've got it mostly figured out how they'll do it. And if they do it, it will be NOTHING like a game situation.

My point about NV20 being so expensive is that they don't have a mid or low cost DX8 product

hot damn! only one month at 3dfx and psychic already. using his magical powers, dave can predict that nvidia is going to use hsr, claim full removal, make a demo and the demo will be nothing like real game plus nv20 will be very expensive.

freaking amazing, for an arrogant (edited). if your shameless self-promoting was not diluted by 3dfx pimping and nvidia bashing, i would probably get sick from it. "i wrote the paper, i run the site, i work for the industry, (are you truing to intimidate people by throwing that in their faces?) you are all ignorant". don't you have anything better to do to pimp 3dfx for a whole day? "geforce t&l sucks, v5 fsaa rocks, nvidia is not an innovator": all sounds dandy until you realize that all of it is coming from 3dfx employee, who is obviously not designing chips or he would not be educating "ignorant peasants" on a message board for hours.

i think it is really pathetic one people working for one company spit at one while promoting their own in public. all pr people (including your dear Bubba, the $4/hr men) are scum, and right now you are just like them. wonder why no engineers from other companies are doing this (besides that they actually *might* have something to do)? did you for one second think that this might be considered unethical?

one more thing: ever see you post again at fool bbs, i am sending sec a letter. the thin veil of presumed objectivity might be enough to cover you promoting of company on which you are (only!) for dependent for salary and stock option while talking down their competitors, but doing that on the financial board is too much.

aman.

ps. i am waiting for the slew of "i am objective, i only work for 3dfx" and "i know everything and you are an ignorant worm" replies ,or my personal favorite ?your opinion is wrong? go ahead, dave, do you worst.

 

[Ben]

Whew!

I just read the whole thread from start to finish.

Now my brain hurts.

 

[IBMer]

Well with that said.

Just because I work for IBM doesn't mean that I only support product developed by IBM.

You assume this.

Before he worked for 3DFX he had the same Ideals.

Your just trying to discredit him because you don't agree with him. It is easy to say that he isn't really telling the truth because he works for 3DFX now, but if you had read some of the articles on his site you would know that some of the things he has brought up in this thread, he has stated in articles before.

You shouldn't be so rude and judgemental to someone that you don't know, and you don't associate with.

How dare you.

 

[ElFenix]

oh, and i remember when carmack's computer was a mac g3 with an ati rage in it (he confessed this when asked why quake 3test or demo was out on mac before it was on pc or something along that line). and we all know how good those are. the fact that he has a geforce in there now isn't really all that great.

alabamaman, i'd take a bet that nvidia actually does something along those lines. if hsr is in their chip like scuttlebutt says it is, and that is the major hardware upgrade, they will hype it to all h3ll, and release demos that use it to the fullest extent possible, while for whatever reason, they run like crap on the competition. this is how products are advertised and sold. i think i liked them better when they were underdog and 3dfx was king, i don't remember reading about how bad 3dfx or nvidia pr people were back then.

 

[dawheat]

The only reason I don't like the reasoning behind the V5 steadier fps explanation is that it sounds like having a lower fps is actually a good thing which it is not.

Someone with a GTS if necessary, can always cap fps to a level lower than the high peaks, but still higher than a V5, and get the same fps smoothness. However the reverse for the V5 is not true. I just have a hard time stomaching the idea that lower fps is a good thing. If someone is playing a clean 1on1, then I say- let the fps stay super high. If you're playing a heavy deathmatch, then you can always capfps and still average a steady fps higher than the V5.

I agree that the V5 has some superb qualities- but this is NOT one of them.

 

[Michael]

Alabama Man - Are you an expert on SEC matters? I'm just curious why you think the SEC wouldn't just toss your letter into a circular file? Is there some statute that Dave is violating? He sure isn't an officer of the company and would have very limited inside knowledge.

The normal round of personal attacks to discredit someone has started. I hope Dave doesn't respond to them and that he tones done the way he words his posts so he's not an easy target.

Michael

 

[dawheat]

<<The point though is that when NVIDIA does release NV20, they are going to hype HSR and claim probably full overdraw removal. And you know what, I'm betting they'll have a tech demo to show it and I bet I've got it mostly figured out how they'll do it. And if they do it, it will be NOTHING like a game situation.

My point about NV20 being so expensive is that they don't have a mid or low cost DX8 product>>

PR is PR- whether from Nvidia or 3dfx. Both will do what they can to stretch the boundaries of truth, but as long as neither cross the line, no biggie.

The problem with criticising the NV20 regarding price is that Nvidia has to have direct competition at the very top to push prices down. Who's job is that? If ATI and 3dfx can't come out with a product that can compete vs the NV20 at the very high end, then Nvidia has no reason to lower its price. There is no way the NV20 will have full overdraw- even though it will not, it may still spank every other card out there. That's why I think we all want ATI and 3dfx to come out with their A-game.

Also as for low cost DX8 products, I don't know about that. What I DO know is that value gamers never had it so good with the GF2 MX and the ATI Raedon 32MB- both cards can be found for around $120-140 and offer excellent performance, while the V4 doesn't quite give the same performace/price value.

 

[BenSkywalker]

Dave-

&quot;Actually, this is not true at all. The GTS has a set amount of lighting capability. This lighting capability is very weak combared to it's transform unit. However, because a CPU has a bigger cache, it is able to fetch the lighting data much faster and so with each successive light, the performance hit is not nearly as drastic on a CPU.&quot;

Every bench agrees with me. I'm sure you have access to some of the most powerful, or yet to be available, hardware now Dave, fire up say MDK2 and back it up with some numbers.

It doesn't matter that the GF's T&amp;L has superior transformation performance, it still has significnantly faster lighting performance then a CPU. Want a list of benches to try?

&quot;But when you put NVIDIA's T&amp;L engine on the scale of things, it does basically suck. Its functionality is very limited and it has horrible performance, especially with lighting.&quot;

The T&amp;L engine is able to blow away O2 workstations of not too long ago(before SGI switched to the &quot;horrible&quot; nVidia T&amp;L hardware), and can even put IR2s to shame in some cases. It may be CAD, but the T&amp;L engine certainly has exceptional performance, despite what you have been saying for the last year.

FSAA superior? There has been a few lengthy discussions on this over at B3D and from a mathematical basis higher res can be proven to be superior. RGSS still reduces accuracy moreso then OGSS which reduces it moreso then higher resolution. A lot of people will pick no FSAA until monitors are the limitation, I know I will. Very few games make it worth while and as geometry complexity increases the amount of noticeable aliasing is reduced significantly(assuming that aat that point in time all boards will have high tap anisotropic to handle texturing issues). Looking at it from a &quot;scientific&quot; perspective higher res is bettet, it is a matter of when we exceed monitors capabilities and &quot;need&quot; to use FSAA due to lacking the option of going higher with resolution.

If you want to argue that 1024x768 4x is better then 1600x1200 I certainly disagree with you for the overwhelming majority of situations particularly in gaming(outside of sims). 1600x1200 offers more detail then 10x7 4x.

FSAA isn't that big of a deal in most games to most people(sim fans being an exception), and none of the current boards are that good at it. The V5 still looks like he!! compared to a software based render engine, just slightly better then the GF2 or Radeon(Yes, slightly. Try out some of the better render engines and compare their FSAA to the hardware options available now, it isn't close).

 

[AlabamaMan]

ibmer: How dare you.

how dare i do what? form an opinion and, god forbid, voice it? i am not trying to discredit him because i don't agree with him. i am just posting my opinion. and in my opinion, what he is going is unethical, unprofessional and in a very arrogant manner to boot.

ibmer: Just because I work for IBM doesn't mean that I only support product developed by IBM.

imagine that, i work for a company myself, and as an employee i have a certain loyalty toward my company and animosity towards our competitors. think about last time you saw someone on tv promoting his/her company putting down their competition. every time i see it, it makes my stomach turn and i don't see how what dave is doing here is any different. (wait, that's not true: he is much more arrogant than your average promoter.) idk about you, but i would sure as hell never buy a product from a company who's employee call people &quot;ignorant fools&quot; in public because they have a disagreement.


Michael: Are you an expert on SEC matters? I'm just curious why you think the SEC wouldn't just toss your letter into a circular file? Is there some statuate that Dave is violating?

no, i not an expert on sec meters, and for what i know they could very well laugh at my letter and shred it after reading the first sentence. but i am pretty sure there are rules about employees making public statements. either way, i am not going to sit idly when something what i consider unethical (here is that word again) is taking place.

The normal round of personal attacks to discredit someone has started. The normal round of personal attacks to discredit someone has started.

hardly. afaiac, dave?s statue and tone discredit him. i am just pointing out why I think that. you are welcome to disagree. Some of my comments might have been too harsh, but dave?s demonstration of psychic powers just ticked me off.

 

[AlabamaMan]

in most games, 1024x768 with 4x fsaa will be much slower then 1600x1200. while you can play some games at 1600x1200 on v5 (q3a), you can't do so at 1024x768 with 4x fsaa.

 

[ElFenix]

as far as i've heard sports games really benefit from fsaa. nothing worse than having a puck pixel pop. of course, the geforce's fsaa doesn't fix that.

 

[Sephiroth_IX]

I have to agree with DawHeat on this one.

 

[Michael]

Alabama Man - Please explain what is unethical? Dave identified himself as a 3dfx employee over and over again. Where is the stock hype? Is he being inconsistent with what he has written in the past?

Heck, who are you and why should we all be concerned about what you consider &quot;unethical&quot;?

Michael

 

[DaveB3D]

Michael,

You saved Alabama. I had this HUGE post and totally ripped into him worse than you know, but I saw your post and it rather softened me. Thank you.


Ben,

I was a bit confused about what you were speaking of. I understand now. I was speaking overall efficiency on a clock-per-clock basis, dealing entirely with transformations and lighting. My mistake. In games, yes the T&amp;L unit is a good deal faster, however hardware lighting is not common place right now and I wouldn't expect it to be so for a good 6-months to a year. Light maps simply look better right now, and I'm sure you understand how both ways work (if not let me know) so I won't go into it.

In CAD apps, it blows things away. If you note benchmarks, you'll note that there is like a 10x performance increase with GF boards over software T&amp;L. But also take note that the 10x performance increase is not present in games.

FSAA superior? There has been a few lengthy discussions on this over at B3D and from a mathematical basis higher res can be proven to be superior. RGSS still reduces accuracy moreso then OGSS which reduces it moreso then higher resolution. A lot of people will pick no FSAA until monitors are the limitation, I know I will. Very few games make it worth while and as geometry complexity increases the amount of noticeable aliasing is reduced significantly(assuming that aat that point in time all boards will have high tap anisotropic to handle texturing issues). Looking at it from a &quot;scientific&quot; perspective higher res is bettet, it is a matter of when we exceed monitors capabilities and &quot;need&quot; to use FSAA due to lacking the option of going higher with resolution.

Actually, on B3D there wasn't ever really a conclusion on this. High resolution provides more detail, while anti-aliasing provides more accuracy. I'm not sure where you got the idea that either reduces accuracy and that RGSS does it more than OGSS. This is simply not the case. There is no way to lose accuracy when dealing with 4x the original data. (unless you do something screwy)



----

There is just one thing I will say to Alabama. Consider that maybe my insight into NVIDIA products comes from my time at Beyond 3D and so I therefor know a good deal about it. No, I certainly don't know everything but I know enough.

 

[ElFenix]

cheap gamers had it good when both the v3 and tnt2, the top of the line cards, were under $200 last summer. the geforce mx has performance like what came out a year ago (geforce sdr). the v3s and tnt2s both had better (double) performance than what had been out a year earlier, and at cheaper prices( v2 sli was $500+/-). so in comparison cheap gamers are getting screwed right now.

 

[DaveB3D]

Dufusyte,


exactly!

 

[ICyourNipple]

Dave (or Ben),

honestly, with no bias whatsoever, what CPU is a GeForce 2 T&amp;L engine equivalent to in a real game situation assuming the game is heavily optimized for the T&amp;L engine?

 

[DaveB3D]

Hrm.. Tough Question.

I know in total raw performance, a GTS is equal to like a 2 GHz P4 (or something like that anyway).

In a game though, one factor would be hardware lighting of course. But honestly, I don't consider that a overly large factor.

I would say clock for clock, probably a 1.5 GHz CPU if the game is heavily optimized.

 

[KarlHungus]

Dave -

I do believe that Ben was illustrating the case where the same amount information is provided for all methods, i.e. 1600x1200 vs. 800x600 with 4x FSAA. When looked at in this fashion FSAA could be consider a form of lossy compression, thus incurring a loss in accuracy relative to the original.

Also, one other question: are you positive that the GeForce2 has no fillrate headroom at 640x480? It seems to me that even with the offloading of the transformation calculations to the T&amp;L unit of the GeForce2 the bottleneck could still remain with the CPU. The end result is that the GeForce2 could have approximately the same fillrate headroom as the V5 5500 at this resolution.

Just a thought.

 

[AlabamaMan]

micheal - i don't know how i can explain what is ethical and what is not. it up to a person, really. and if you don't care, for what i think, well, i am sorry to hear that.

dave - Consider that maybe my insight into NVIDIA products comes from my time at Beyond 3D and so I therefor know a good deal about it.

let me get this straight, you *know* that nvidia is going to hype it as 4x effective, you *know* make completely unrealistic demos and you *know* that it will be too expensive? sorry, no dice.

and btw, don't do me any favors.

 

[DaveB3D]

I want to share something.. something that has cross my mind. As I've been pondering over this thread, looking back some of my posts haven't been the best and yes, I've posted a bit more as a 3dfx employee than I have as myself. I regret doing this and so from here on in (in all the threads here) I'm going to think about what I'm saying more. Think if I really feel certain ways and to what extreme. I'm going to post as myself. In doing that, I'll share what I know to the limites that I can. There are obviously MANY things I can't talk about. In my doing this I hope to offer insight into the technology, but I have to ask something of you guys. Do question me about things 3dfx has in the works. You know I cannot comment on those things and so please don't ask them. Just think of me as Dave from Beyond 3D and that is who I will be and the type of posts I will deliver.


I do believe that Ben was illustrating the case where the same amount information is provided for all methods, i.e. 1600x1200 vs. 800x600 with 4x FSAA. When looked at in this fashion FSAA could be consider a form of lossy compression, thus incurring a loss in accuracy relative to the original.

Also, one other question: are you positive that the GeForce2 has no fillrate headroom at 640x480? It seems to me that even with the offloading of the transformation calculations to the T&amp;L unit of the GeForce2 the bottleneck could still remain with the CPU. The end result is that the GeForce2 could have approximately the same fillrate headroom as the V5 5500 at this resolution.


I see what you mean. However, it isn't really giving a loss as the image it is producing with each single pixel has a more accurate color value than the pixel of the original image. Signal theory plays into this some, but I'm not going into that tonight. I actually think I need to go refresh myself a bit on signal theory and aliasing a bit. However, with a high-resolution image you will find more detail per-pixel than a lower resolution anti-aliased. Really, with current hardware there is no set setting that works for all games. Some games I play high res, some I play low res with little settings, others I play at 10x7 with 2x AA and others at the same setting with 4x AA. It just really depends.


As for the GTS capping. No, I'm not entirely certian, however I'm confident it is close. Play a game of Q3 and watch the numbers.

 

[DaveB3D]

I think my own words defend me.


The point though is that when NVIDIA does release NV20, they are going to hype HSR and claim probably full overdraw removal. And you know what, I'm betting they'll have a tech demo to show it and I bet I've got it mostly figured out how they'll do it. And if they do it, it will be NOTHING like a game situation.


Note

&quot;probably&quot;
&quot;betting&quot;
&quot;if&quot;

Key words. If you really are going to flame someone, it isn't a good idea to put words into their mouth. As for the price, I know what they were expecting. I actually adjusted it (in their favor) given the way memory prices are going.

 

[DaveB3D]

Also.. might I add about the demos. They make completely unrealistic ones now, so I don't see why that would change in the future.

 

[IBMer]

<<in most games, 1024x768 with 4x fsaa will be much slower then 1600x1200. while you can play some games at 1600x1200 on v5 (q3a), you can't do so at 1024x768 with 4x fsaa.>>

I say that 1024x768 with 2XFSAA looks better than 1600x1200 in Q3 and I will tell you why.

There is no extra detail to be gained by going to 1600x1200. The textures in Q3 are at most 512x512(not counting the S3TC Special Levels) all going to 1600x1200 does is cause the textures to stretched and cause them to look blurrier. This isn't what I call detail.