Gabe Newell speaks [Archive]

Rob R.

09-11-2003, 10:18 PM

This is a bunch of hogwash. I'll wait until a reliable source of information comes out from a third party tester.

If anyone who calls them self a gamer doesn't understand that drivers are made to include all the latest games, you must have a fundamental misunderstanding of how the market works. If I was nVidia, I would have also been quite surprised in Valve's obvious decision to "fix" the numbers. I personally find it totally immature on their part. Is this company run by 10 year olds?

I lose more and more respect for ATi everyday. Following their moves is like watching another political smear campaign or a soap opera. What are even worse are ATi fans who scour the internet looking for this stuff so they can parade it around in some cheesy victory march to prove once and for all, they're on the winning team. Well yay for you! Neat! Personally, I don't think I want to be a part of any of that.

nVidia has something that has stabbed right into ATi's market, the top-of-the-heap market. Now ATi is scrambling around doing stuff like this at every corner. If they want the fastest card again, they should make one instead of making one up.

Rugor

09-11-2003, 10:32 PM

OMG

I had to read that to believe it.

Now that a game bench is giving the same results as all the synthetics, namely poor shader performance in the CineFX architecture you're actually losing respect for ATI?

Should ATI have inserted clip planes in 3DMark03? Should they have made sure that a lower quality setting overrides what the end user and the game designer intended? Should they have focused all their efforts on improving performance in games that are used as benchmarks?

Funny how in real world unbiased tests--- ATI is beating Nvidia much more often than not, and the more DX9 features a game uses the better ATI looks because it has stronger shaders. Nvidia handles shadows better, but current games use shaders more than shadows.

Rob R.

09-11-2003, 10:35 PM

Rugor, you can't be serious when you said "unbiased". This is the appitimy of biased; this is the very definition of biased.

Rugor

09-11-2003, 10:54 PM

I wasn't referring to the Half-Life numbers when I said 'unbiased,' but rather to most of the benchmarks I have seen on the net from sites that try not to be biased.

However, I do find it interesting that both Gabe Newell and John Carmack are on record as saying they had to program NV3x specific paths in order to get playable performance out of the architecture. When using the OpenGL standard ARB-2 path the R300 core is about twice as fast in Doom III as NV30/35. Nvidia is faster when using the NV specific path than ATI is on the standard path. Now in HL2 ATI is much faster using the standard DX9 path than Nvidia is, so Valve had to code an NV Specific path. In this case the NV specific path is still slower on NV hardware than the standard path on ATI. However, this does make sense when you consider HL2 is very shader heavy (and shaders are NV's weak point) and D3 is very shadow heavy (and shadows are NV's strong point.)

Nvidia uses a lot of hardware specific extensions, while ATI tends to work to the standards. It's not surprising that Nvidia hardware works better on optimized applications and ATI doesn't need the same optimizations.

ComPooTer

09-12-2003, 12:23 AM

dont be mad cuz my radeon 9600 pro smokes a geforce fx5900 ultra on half life 2 :p

haha just kidding.............im proud of my little ol buddy here for 200 bucks

WAY TO GO ATI RADEON 9600 PRO! :D

hey that almost rhymed there

Bigjakkstaffa

09-12-2003, 04:00 AM

To be fair, i have seen several DX9 specific benchmarks floating about and in all cases the 5900FX gets beaten by a good margin, and in the odd case eaten alive by the 9800 Pro. This is just another benchmark to add to that list, regardless of the actual numbers, as it stands its impossible to say Nvidia's current DX9 performance is equal to, nevermind better than Ati's

--Jakk:t

Bigjakkstaffa

09-12-2003, 04:12 AM

http://www.amdmb.com/article-display.php?ArticleID=257

Well theres some early benches of the magical 50.xx det's and though there are no DX9 specific tests, the performance jump dont really look to be on course for that which Nvidia is promising

--Jakk:t

Someone Stupid

09-12-2003, 03:06 PM

Rugor beat me to the point, both Newell and Carmack have said the NV3x is essentially garbage with all the workarounds they have to do. Carmack did the driver optimizations nvidia was talking about so the driver wouldn't do it, as Carmack knew that most likely nvidia wasn't going to release a driver set to do it properly anyhow if they would at all. Going by Newell's comments he didn't use the new drivers nvidia released to Valve as Valve had gotten many before and those changes were never made, it was only there to inflate benches. I give Gabe two thumbs up for telling nvidia to put the driver out, then I'll use it, so to speak than to play a game of inflating benches. You know who takes the heat from that, the game developers, not the card maker. So I agree completely with Gabe here. I was an nvidia fan, but once it went downhill, well I went with the best product, nvidia no longer offered it, or close to it, so ATI was the natural choice seeing as they have their act together. More money on RnD pays off in the long run in an even market than marketting hype, and this is where nvidia is getting killed now as they marketted these cards as wolves and they are sheep.

People pointed out to nvidia the problems with the FX core long before it's release, the original 128 pipeline was the most obvious. Developers like Carmack had pointed things out they'd like to have seen changed. Nvidia just stuck it's thumb up it's **** thinking they knew best and they were shown not to know a bloody thing when they released the blowdryer. It had a horrible launch and they still can't get the core to be truly competitive as yet ANOTHER benchmark illustrates. So now they just are going to full force marketing of a crappy product instead of heavy marketting of a good product.

I go with what hardware is best for the cost, ATI has them beat there in every single way as nvidia doesn't have a D9 solution in my book. 500 bucks for card to run D9 games isn't a solution, it's a ripoff.

Rob R.

09-12-2003, 06:23 PM

Maybe it's time I join the ATi crowd and buy one of these dang things. I suppose this could be a peak at what's going to happen more and more often with the newest games coming out.

Rugor

09-12-2003, 06:28 PM

If you are looking at a new card to play new and upcoming games I would definitely go with ATI.

The R3xx series is a better value and stronger overall architecture than the NV3x series. This doesn't make all ATI products better than all Nvidia, but when it comes down to those two architectures ATI got it right.

My last card was Nvidia, my current is ATI, and I'll see what's out before deciding what to get next, but that won't be for a while.

Rob R.

09-12-2003, 06:38 PM

I guess I'll start putting my peanuts back for a Sapphire 9800 non pro and do a BigJakk mod on that bad boy.

Bigjakkstaffa

09-12-2003, 06:46 PM

:cool:

--Jakk:t

Bigjakkstaffa

09-13-2003, 07:16 AM

Another aticle here:

http://www.tech-report.com/etc/2003q3/hl2bench/index.x?pg=1

more interesting is this:

http://www.guru3d.com/forum/showthread.php?&threadid=63181

a test of the 50.xx's in DX9 game Aquanox 3. There is a noticable performance gain, but its only about 1fps. Also, the cards are all pretty close in performance, however this in itself doesnt match with other direct X 9 tests ive seen with the likes of the latest Tomb Raider and Halo Beta... which raises the question.. could Aquanox be optimised for Nvidia :p:x

Comspiracy theories abound :x

--Jakk:t

Qvark

09-13-2003, 09:44 AM

I just loooooooove my 9800 more and more :)

bob05

09-13-2003, 10:27 AM

Originally posted by Rob R.
This is a bunch of hogwash. I'll wait until a reliable source of information comes out from a third party tester.

If anyone who calls them self a gamer doesn't understand that drivers are made to include all the latest games, you must have a fundamental misunderstanding of how the market works. If I was nVidia, I would have also been quite surprised in Valve's obvious decision to "fix" the numbers. I personally find it totally immature on their part. Is this company run by 10 year olds?

I lose more and more respect for ATi everyday. Following their moves is like watching another political smear campaign or a soap opera. What are even worse are ATi fans who scour the internet looking for this stuff so they can parade it around in some cheesy victory march to prove once and for all, they're on the winning team. Well yay for you! Neat! Personally, I don't think I want to be a part of any of that.

nVidia has something that has stabbed right into ATi's market, the top-of-the-heap market. Now ATi is scrambling around doing stuff like this at every corner. If they want the fastest card again, they should make one instead of making one up.

I agree with some of what Rob R. said, in that I, like him, find that it is highly suspicious that a 9600 Pro could beat a 5900 Ultra. It is possible that HL2 has been more optimized for the 9x00 than a 5x00, that's something only Valve can answer. I mean come on guys, you all know as well as I do that the 5900 Ultra is a better card than the 9600 Pro. The difference is like comparing a Ti 4200 to a Radeon 8500. Maybe it is a driver issue, in that Nvidia's current drivers don't do some DX 9 games very well. We've all seen other DX 9 benches, and we've all seen that a 5900 Ultra can equal and occasionally out perform a 9800 Pro, and most certianly a 9600 Pro. That's why we can even despute the fact that the 5900 Ultra is the king, or that the 9800 Pro is the king. And as you all know, the NV 3x's performance on Half Life 2 has been questioned from the beginning, since the report of fragments/artifacts using certiant AA/AF options or what not. So I am pretty supsious of the HL2 perfromnace on NV 3x's anyway.

P.S. I'm not trying to defend Nvida or attack ATI, I'm simply saying that something is seriously wrong when a 9600 Pro can beat a 5900 Ultra. Period.

tasty danish

09-13-2003, 12:08 PM

i read an article somewhere (trying to find it now) talking about why the 5x00 cards sucked. there is something about the architecture of the card, how the data gets across of something, but anyway, it was basicly an explaination of why the current nvidia cards will never run dx9 good. i believe that is why a 9600 can beat a 5900. so yes, i 5900 is a much better card than a 9600 in dx8 and below, however the 5x00's are uncapable of running dx9 decently.

Rugor

09-13-2003, 12:25 PM

Bob05

Yes there is something seriously wrong when a 9600 Pro can beat a 5900 Ultra. I agree completely. The question that immediately leaps to mind in my case is what is it? The evidence available to me says the problem lies in the FX5900 Ultra's shader performance.

Every single test and application, synthetic or actual game, I have seen has shown that ATI has MUCH stronger shaders than Nvidia in the current generation. We know that HL2 is a very shader heavy game. The 9600 Pro has four pipes at 400MHz with strong shaders, the 5900 Ultra has four pipes at 450MHz with weak shaders. Given they are running the benches at a fairly low resolution (1024x768) it is likely the game is not bandwidth limited, so the 5900 Ultra's greater memory bandwidth doesn't help. If, as seems likely, it's using single-texturing with shaders then the 5900 Ultra's extra TMU is dropped from the equation. That puts all the work on the shaders. And in a shader-limited situation I would expect ATI's shaders even on the 9600 Pro to be more than enough to make up the 50MHz difference in core speed.

It does sound odd, but if you've been following what's been going on it does make sense. The FX series may be a DX9 part but the performance emphasis in the design was on running DX8 class games quickly rather than DX9 class games.

bob05

09-13-2003, 12:36 PM

Originally posted by Rugor
Bob05

We know that HL2 is a very shader heavy game. The 9600 Pro has four pipes at 400MHz with strong shaders, the 5900 Ultra has four pipes at 450MHz with weak shaders. Given they are running the benches at a fairly low resolution (1024x768) it is likely the game is not bandwidth limited, so the 5900 Ultra's greater memory bandwidth doesn't help. If, as seems likely, it's using single-texturing with shaders then the 5900 Ultra's extra TMU is dropped from the equation. That puts all the work on the shaders. And in a shader-limited situation I would expect ATI's shaders even on the 9600 Pro to be more than enough to make up the 50MHz difference in core speed.

Yeah, that is probably very true about that shader issue. But the 50.xx Detonator's should improve the the shaders, that link that Jakk had on it showed 50.xx improved Pixel Shader 2.0 scores by 25% and vertex shader by 12%. I'll hold my breath on the whole issue untill the 50.xx Dentonators and Half Life 2 are finally released. ;)

bob05

09-13-2003, 12:52 PM

Originally posted by Bigjakkstaffa
http://www.hardocp.com/article.html?art=NTIw

Here's Nvidia's excuse, erm, i mean response

--Jakk:t

Originally posted by Rugor
For those who don't follow links the excuse translates to:

"Don't use the drivers that are out now, use the special ones we only released to the reviewers that the people who make the game don't want you to use. We told Valve not to use any but our special drivers and look what happened."

"NVIDIA does have a new driver release on the horizon, the Detonator 50 series of drivers, however Valve instructed us not to use these drivers as they do not render fog in Half-Life 2. "

http://www.anandtech.com/video/showdoc.html?i=1863&p=5

So now can we all agree on no more Det. 50 bashing untill they are released?
;)

Rugor

09-13-2003, 12:53 PM

The new drivers will increase shader performance, whether through instruction rescheduling or other means but they will only serve to get the most possible out of the hardware, they won't improve the hardware itself.

Valve spent a lot of time wringing extra performance out of the NV3x architecture but it wasn't enough. I doubt any allowable driver trick can pull more than another 10% from that architecture. Anything more than that will probably require borderline cheats or worse rather than aggressive optimizations.

For Half Life 2 you want ATI.

Rob R.

09-13-2003, 01:11 PM

THG was recently given the opportunity to test the game as well and Valve's stance on how they were allowed to test is still suspect to me.

"Valve did not allow us to also run test with a new Beta version of the upcoming NVIDIA Detonator 50 driver. Valve explained that they only want to see scores of already released drivers. We're still waiting for a more detailed explanation from Valve what exactly they don't like on that Beta driver."

"What we did not like was the very limited time of only 90 minutes for the test runs. Some cards needed more than 20 minutes for a single configuration. This means there was no time for verification of the scores. In comparison, we had more than a whole day from 8:00 AM until midnight for the Doom III tests at NVIDIA early this year."

Things that make you go...hmmmmm...

**I am still going to buy a 9800 non pro because it represents an excellent and undeniably good value, but this whole mess is leaving a very bad taste in my mouth.

Bigjakkstaffa

09-13-2003, 01:20 PM

Valve are IMO, one of the few honest and caring games developers in the industry today, an in comparison to most best selling teams actually look after their customers. As such i myself cant really see Valve having any hidden evil motive for anything, and heck if they were doign under the tabel deals, why woudl they be doign it with Ati, arguably the poor relative of Nvidia who they coudl squeeze more money out of. Also, commercialyl it would make no sense for Valve to cripple performance on Nvidia cards as Nvidia have a stranglehold on the GPU market, and for Valve to do such a thing would surely have a massive impact on their sales due to the number of gamers with Nvidia based solutions.

However im certain there is more to come from the FX cards in performance, whether it be in the form of the 50.xx's, or in the form of blatant cheating, which is what i think Newell and co. are worrying about... afterall theyve done it befre, whats to say they wont do it again. However everyone seems to be focusing on HL2 as the big assault on the FX series and its DX9 capablilities, in reality its not alone. All of the DX9 tests i have seen thus far have run at best slower on Nvidia cards than Ati alternatives, and at worst Ati has been far and away in the lead. Hl2 seesm to have gotten everyone ariated because its the first really highly anticipated DX9 release and secondly the OEM deal with Ati has allowed for conspiracy theorists to shift the blame for Nvidia's poor DX9 performance, something previously confirmed in other benchmarks, to Valve being in the sack with Ati.

Personally i think that DX9 performance on the FX will get better, but i wouldnt be at all suprised if this improvment doesnt come at the cost of quality and in all honesty i cant see the det 50.xx's making up the huge ground Nvidia seem to be promising, its gonan take time for them to get on a level footign at least and i dont think we'll see it in a single driver update

--Jakk:t

maje87c

09-13-2003, 01:39 PM

Originally posted by Rob R.
I guess I'll start putting my peanuts back for a Sapphire 9800 non pro and do a BigJakk mod on that bad boy.

What exactly does the 'BigJakk mod' consist of? Every news article seems to make my geforce 3 more and more inadequate. Not to mention, I need something better than a GF3 for all the upcoming games.

bob05

09-13-2003, 01:40 PM

Originally posted by Bigjakkstaffa

.................or in the form of blatant cheating, which is what i think Newell and co. are worrying about... afterall theyve done it befre, whats to say they wont do it again................
--Jakk:t

Hey ATI did it also, they are capable of doing it also. ;) Althought I seriously doubt either side will cheat, seeing as benches will be under the microscope for quite a while. Cheating would be the stupidest move Nvidia (or ATI for that matter) could make. I do agree with Bigjakkstaffa when he says that one driver will not be the magic bullet. But I think there is more to the NV3x yet to be unleashed, just hope Nvidia can unlease it soon enough. :(

bob05

09-13-2003, 01:42 PM

Originally posted by maje87c

What exactly does the 'BigJakk mod' consist of? Every news article seems to make my geforce 3 more and more inadequate. Not to mention, I need something better than a GF3 for all the upcoming games.

I'll put this question to rest so as it won't hijack the thread. Here:

http://www.sysopt.com/forum/showthread.php?s=&threadid=145610

http://www.sysopt.com/forum/showthread.php?s=&threadid=143644&highlight=Flashrom

Enjoy. :t

Rob R.

09-13-2003, 01:51 PM

Originally posted by Bigjakkstaffa
Hl2 seesm to have gotten everyone ariated because its the first really highly anticipated DX9 release and secondly the OEM deal with Ati has allowed for conspiracy theorists to shift the blame for Nvidia's poor DX9 performance, something previously confirmed in other benchmarks, to Valve being in the sack with Ati.
--Jakk:t

LMAO!

Ok I'll admit it, I have bought into this conspiracy theory. None of this makes sense to me. Here's the questions of my concern.

--- Why have a "shader day" about how video cards will run on your game at all?

This is for other people to do like THG, HardOCP, and the like. The way it was presented was extremely obvious to be pro ATi as well.

--- Why work with nVidia on a driver release for your game and then not use their work in progress on a day meant to show how your game will run on their cards?

Seems like an odd decision if they're trying to work closely with them.

--- Why be secretive of the reason why you do not want to use the drivers that will be available on game release?

You either have a good reason or you don't, and they didn't have a good reason.

--- Why allow a group of third party testers in and deny them both the time and freedom to test the game as they please?

Maybe they have something to hide. Actually, I can't even think of another single reason for it.

Rugor

09-13-2003, 01:51 PM

Yes there is more performance that can be eked out of the NV3x architecture in many applications, provided you have the time and resources to hand-optimize the rendering path for it.

However, every single unbiased benchmark and test I have seen shows that the CineFX architecture is weaker in DX9 level performance than ATI's current architectures. ATI even runs Dawn faster, and that's through a wrapper.

You can't get blood from a stone, and eventually people will realize the GfFX series is not a great solution for DX9.

Rob R.

09-13-2003, 02:04 PM

Originally posted by Rugor
You can't get blood from a stone, and eventually people will realize the GfFX series is not a great solution for DX9.

Rugor, It is pretty obvious to me that you know quite a bit about video cards. I have been dumbfounded by a couple of your posts in this thread and spent an hour researching everything you said just to understand what it was you were telling me. I tend to believe what you are saying here. I think ATi is the better card and I am willing to put my money where my mouth is now (I cant wait to get it too!) I just think this whole situation was handled extremely poorly and I feel like some wool has been pulled over my eyes.

They said: "Here's our results, as you can clearly see, nVidia sucks!" then "If you don't believe us, you can test it for yourself as long as it's the EXACT same way we did it and don't even think about doing anything else"

Doesn't seem right to me.

Rugor

09-13-2003, 02:09 PM

Rob let me try to answer your points in order:

--- Why have a "shader day" about how video cards will run on your game at all?

It was an event sponsored by ATI and even they were surprised by Gabe Newell's topic.

--- Why work with nVidia on a driver release for your game and then not use their work in progress on a day meant to show how your game will run on their cards?

There are two reasons why they did not use the Det 50's for this test. First, those drivers are NOT publically available, and may never ship to the public in their current form and so may not be representative of the performance people will see. Second, the new drivers do not render fog in the game and whether this is a bug, an optimization, or a cheat it is not the experience Valve wants to give its customers and so they refused to use them. There may be other "optimizations" in the driver but not rendering fog at all is enough to take them off the table.

--- Why be secretive of the reason why you do not want to use the drivers that will be available on game release?

First, because since these are BETA drivers they probably will not be the ones that will be available on game release. They are similar but there's no way to guarantee they are going to be identical. Every reputable reviewer I know of will only use publically available drivers for comparing hardware in public.

This is where I have to go into some speculation, but if you look at the slide where Gabe showed some of the "optimizations" he had already found in someone's drivers, that included artificially boosting IQ for screenshots it could well be that he thinks these drivers may have those or other "optimizations" and that invalidates them, but he did not want to attack Nvidia directly by name because they are working to improve performance.

--- Why allow a group of third party testers in and deny them both the time and freedom to test the game as they please?

Because that is standard for pre-release benchmarks. I don't know how many times I've seen that done. It's normal, especially when there is a limited time and a number of third party testers that you want to show it to.

The conspiracy fails Occam's Razor, it's simpler to accept they're telling the truth as they know it.

Rugor

09-13-2003, 02:20 PM

The results do seem right to me, because they are right in step with all the other Pixel Shader based results I've seen for the GfFX series.

On 3DMarks2001SE using PS1.4 the GfFX 5600 Ultra has about 39% of the performance of a Radeon 9500 Pro in the Advanced Shader test given the same CPU. Nvidia's shaders are weak and the game is shader heavy.

Rob R.

09-13-2003, 02:22 PM

Ok, I can accept most of that but I need one point clarified. Now when I think of fog, I think of the fog that reduces draw distance which is a way of improving the performance of a video card in massive environments. I have to assume you are not referring to this type of fog but rather a fog that is a special effect and would appear all around you?

Bigjakkstaffa

09-13-2003, 02:37 PM

Hey ATI did it also, they are capable of doing it also

To be fair, the changes ATi made were more optimisations for the features used in the tests, wheras with Nvidia it was a case of degrading quality when the 3dmark2003.exe was run. As such, i dont think its really fair to call what Ati did 'Cheating' in the face of Nvidia's actions.

Now when I think of fog, I think of the fog that reduces draw distance which is a way of improving the performance of a video card in massive environments. I have to assume you are not referring to this type of fog but rather a fog that is a special effect and would appear all around you?

In this sesne, i may be wrong, but im thinking of fog as in 'CS Smoke Grenade' fog, the sort used for atmosphere and effect that can really reduce frame rate, not that sort used to obscure distant vision and so reduce draw distance.

--Jakk:t