Another Richard Huddy Interview

[Diplo]

The whole problem with shader replacement, when the game is used as a benchmark, is that benchmark becomes only an indication about performance of the card on that game and even more on that game version.

Which is why you never buy a video card based on just one benchmark (or, indeed, one review). I think that's pretty much common sense.

However, if you had been foolish and bought an NV30 class card, isn't it the least Nvidia can do to at least try and give you decent performance? Don't the end users deserve that? I find it a little hard to swallow that ATI fans want to deny gamers large performance increases based on the fact that maybe the game will be used for benchmarks (especially the same ATI users who vocally defended ATI when they did exactly the same thing with Quake 3).

I don't think there's anything wrong with doing per-game optimisations as long as there is A) no discernible effect on IQ and B) there is a degree of openness about the fact that it's being done. Is it any different, really, than Valve working closely with ATI coders to make sure HL2 runs well on their hardware? Or Nvidia working directly with developers to help them write the most efficient shaders possible?

All I'd ask is that if Nvidia are doing optimisations for one particular game then they should say so openly (same goes for ATI). However, if those optimisations are more generic then that's their business. Again, though, no one (AFAIK) has yet to prove Nvidia have done anything.

 

[kyleb]

how can you say his comments are fud when you just said you don't know if he has ever been caught lying?

You don't have to lie to spread FUD. And marketing crap is usually not lies, just one version of the truth

fud is "fucked up disinformation", and disinformation is synonymous with lie; so sorry, but you do have to lie to spead fud.

 

[kyleb]

I already showed a case of selective quoting to show how huddy was being misleading.

but i am looking for a rational example.

 

[Dave Baumann]

No, its "Fear, Uncertainty and Doubt" - you can create those elements without lying.

 

[Xmas]

In a literal translation, it does.

Looks like English and German grammar rules are way more different than I thought.

fud is "fucked up disinformation", and disinformation is synonymous with lie; so sorry, but you do have to lie to spead fud.

I thought there is a consensus that FUD means fear, uncertainty and doubt.

 

[kyleb]

doh, sorry about that; the acronym finder agrees with you. it also has the definition i gave but "fear, uncertainty and doubt" is the bolded one. i suppose i just got them confused because most of the fud around these days falls into both categories.

 

[WaltC]

I already showed a case of selective quoting to show how huddy was being misleading.

Furthermore, why don't you just wait till the mods are out and see if the NV40 cards performance plummets (in comparison to the r4xx cards of course) as someone suggested they would. Guess what it ain't going to I promise you that. The NV30 is another matter as we all know.

OK, but the context of Carmack's remarks was the R4x0/nV4x on-site benches he conducted for D3. As well, Carmack didn't say "only nV3x," he said "especially nV3x", which obviously adds nV3x to nV4x in Carmack's "heavily optimized" category. To say that his case for nV3x was worse than for nV4x does not eliminate the intent to include nV4x. Isn't it already a matter of public record that while Humus' optimization improves performance for Rx0 it at the same time degrades nVx performance? Since the change Humus made is exactly like the kind of things modders will do, I see no reason to have to wait on those mods, as the only difference will likely be one of degree, with nV4x "falling off the fast track" to a more or lesser degree.

 

[nmyeti]

So does this mean JC "loves" the idea of "driver programmers analyzing texture data and changing driver parameters," then...? Is his beef with automatically switched optimization versus manually switched optimization? Must be, since he never mentioned nVidia's trilinear optimizations at the same time he saw fit to mention ATi's, which I found really strange...

I don't care what JC loves or hates and I don’t pretend to understand his feelings about things that he hasn’t commented on. You are taking this personal. The quote I was responding to was

Also fyi, Carmack was referring to the trilinear optimizations when he commented on the colored mipmaps issue. It also looks like Huddy isn't the only one taking Carmacks comments extremely out of context because he never said anything to the tune of disdaining that particular optimization.

Clearly he does disdain those optimizations. I suspect he would hate them equally as well from either Nvidia or ATI, but again that’s not really in the scope of my response. I was simply responding to the quoted material that attempted to establish JC’s feelings on the subject of such optimizations as being indifferent when he took the time to address them and used the word hate to describe his feelings about said optimizations.

Again I don’t give a rat’s ass about what Carmack loves or hates. Just as I don’t really give a rat’s ass which video card company makes the car that happens to be in my system, so don’t take my comments as a fanboi using the great JC to justify his video card purchase.

-Nathan

 

[WaltC]

Again I don’t give a rat’s ass about what Carmack loves or hates. Just as I don’t really give a rat’s ass which video card company makes the car that happens to be in my system, so don’t take my comments as a fanboi using the great JC to justify his video card purchase.

-Nathan

Uh...OK...

To me the issue was in pinning down what he said, especially in the context of what he knew about but didn't bother saying. To me, complaining specifically about ATi doing tri opts, but nobody else, when both companies are doing tri opts, was a very blatant, bizarre oversight for someone of his experience to have made. It certainly could not have been anything other than a deliberate oversight, which causes me to wonder that JC might have thought nobody would pick it up...

 

[2senile]

http://www.3dvelocity.com/articles/atiinterview/rhuddy.htm

& yet another Huddy interview. Amusing read in places but I didn't think B3D needed a "Yet another Richard Huddy interview" thread so I put it here.

 

[FUDie]

http://www.3dvelocity.com/articles/atiinterview/rhuddy.htm
& yet another Huddy interview. Amusing read in places but I didn't think B3D needed a "Yet another Richard Huddy interview" thread so I put it here.

Did anyone else notice "Today's date is Tuesday, September 31, 2004."? It's an interview from the future! Odd... my calendar shows that September has only 30 days and that the 31st (if there was one) would fall on a Friday. Maybe George Bush thinks he is Caesar and has to change the calendar? Will November become Bushember?

Regarding the discussion comparing TV to a PC monitor. One big difference is that most TV programs are recorded on film or video. This means that motion blur is built in. You don't have that advantage in a PC game. Later they discuss motion blur, but it wasn't mentioned in the earlier discussion regarding televisions. TVs also have a low pass filter which helps remove aliasing, IIRC.

One problem with the suggestion of changing quality on the fly is that it can cause framerate fluctuations and quality issues. For example, say you are just above the threshold for enabling feature X, but when you enable feature X, performance drops too much. You could end up ping-ponging back and forth which would be very annoying. Of course, you could try to catch situations like this, but it's not easy to cover every possible scenario.

-FUDie

 

[2senile]

FUDie; It's like "The Twilight Zone" over there which could explain the date.

The interviewer has been on about changing stuff on the fly for a couple of years. Seems to be his pet dream.