Saturday Commentary

[BrianG]

Originally Posted by saturnotaku

I mean, what good is AA if you can't see it in the first place. Hehe.

[RuneSR]

HalcYoN wrote: "As for the Squab, I don't like them, so I don't link them."

Intersting since Firingsquad was good enough for MikeC to include in this survey:

http://www.nvnews.net/vbulletin/show...0&page=1&pp=15

So MikeC finds the research quality at Firingsquad satisfactory, but HalcYoN does not?

Of course NVNews can decide whether or not it wishes to link to Firingsquad, but I do find it quite strange and academically unsatisfacory that interesting results are being ignored due to personal preferences - especially since results from the VST using MAX settings indicate that for the first time 6800U is loosing to X800Pro in 1600x1200 4X/8X and 2048x1536 4X/8X:

http://www.firingsquad.com/hardware/..._vst/page4.asp

Furthermore there are other VST studies showing less pronounced effect, but we do not know if these studies also used MAX details:

http://www.xbitlabs.com/articles/vid...ke-source.html

http://www.gamespot.com/features/6105370/index.html

Regarding HalcYoN's comment about Doom3, it is well known that Doom3 is performing quite better on 6800U than X800XT PE, but you can find results, where the difference is more or less pronounced - try comparing:

http://www.digit-life.com/articles2/doom3/index.html

and

http://www.xbitlabs.com/articles/vid...m-3-tests.html

Digit-Life found that 6800U using High Detail 1600x1200 (with or without 4X/8X) was 18 - 20% faster in Doom3 than X800XT PE (without the Humus tweak), but using XBitLabs 6800U is up to 50% faster. My point is that academically we have to include both results, whether we like Digit-Life and XBitLabs or not, otherwize you end up getting biased results, which I and probably other readers would not consider an improvement regarding the news posting at NVNews. Also, I am not an "enemy" of NVNews: actually NVNEWS is my current homepage when I start my Internet browser - and has been that during the last 4 years.

[Woodelf]

Originally Posted by BrianG

Thanks for getting back on topic.

I honestly believe that AA and AF are the new 32-bit color. They are taken for granted and are part of the gaming experience now. I see most game engines and game designers accounting for performance with AA and AF enabled. Hence, the games now control settings like AA and AF. To coin the ad slogan, "The way it's meant to be played."

That said, Doom3 does look great with out AA, but I think that is mainly due to the darkness of the environment.

Oh I believe they are the new 32 bit color too, but remember the trouble 32 bit had getting accepted over 16bit (thanks to voodoo2/3 fans). I'm just saying that (IMHO) the struggle is probably not over, and that they are still luxuries. Remember, both game engines (HL2 and D3) are/were well over due. What setting's do we anticipate for Stalker, or unreal 3, or even a D3/HL2 engine game that some developer decides to really graphically impress with. I anticipate low AF, AA and shadows, just to keep a decent resolution. All I'm really saying is that this has happened before. New card's come out that allow current games AA/AF, then the new game (far cry for instance) comes along that put's and end to it. AA/AF have been long standing issues (especially AA) that end up on the "first to turn off list". If this weren’t true, it wouldn't still be an issue. I really believe that, had id wanted to, they could have really gone all out on the graphic's detail. One thing they should have done was to enhance the gray looking skin on some of the characters. Sure it would be nice if we could all way's look forward to
full AA/AF, but history says different.

[saturnotaku]

Originally Posted by RuneSR

Digit-Life found that 6800U using High Detail 1600x1200 (with or without 4X/8X) was 18 - 20% faster in Doom3 than X800XT PE (without the Humus tweak), but using XBitLabs 6800U is up to 50% faster. My point is just that academically we have to include both results, whether we like Digit-Life and XBitLabs or not, otherwize you end up getting biased results, which I and probably other readers would not consider an improvement regarding the news posting at NVNews. Also, I am not an "enemy" of NVNews: actually NVNEWS is my current homepage - and has been my homepage during the last 4 years.

OK, what part about this being discussed to death in the forums are you not getting? We've seen the tests. And "academically" nV News can post whatever it wants, however it wants. If you don't like it, go elsewhere it's that simple.

And what's "biased" about the results? In just about every single test across the board, ATI wins. But the margins are completely different and that's why you have to take into account the fact that 1) this is a beta game, 2) test sites used different driver versions on the NVIDIA cards (61.77 and 65.62).

I will say to you once again, read the thread I linked to. There is plenty of discussion there where people are drawing their own conclusions.

Also, I am not an "enemy" of NVNews: actually NVNEWS is my current homepage when I start my Internet browser - and has been that during the last 4 years.

I just love when people reach into the barrel and pull this out.

[Woodelf]

Can some one tell me what Nvidia has (been able) done to their drivers (tested) in preperation for this stress test?. Or are we just sweating ATI propaganda.

This remark is alittle disturbing:

"So what did we learn with our benchmarks using Valve's Source Engine Stress Test? Well simply the fact that ATI's R420 generation of hardware outperforms NVIDIA's NV40,"
http://www.vr-zone.com/?i=1193&s=6

Like it's "end of story", or something.
That's why I say, is this driver, or really hardware
related.

[Spotch]

Originally Posted by BrianG

What do you have to say?

Q3A at 1600 x 1200 or even 2048x1536 but does your sky look like this unacceptable mess? Its even worse on Team areana maps with moving clouds that cast shadow effects on the ground.

http://webpages.charter.net/Spotch/q...1-52-22-90.JPG

Can someone please help me with this? Latest Q3a version using 61.72 on GF6800GT no compressed textures and max details. Win XP pro patched with all official service packs and DX9b.

[Woodelf]

Originally Posted by Spotch

Q3A at 1600 x 1200 yeah but does your sky look like this?

http://webpages.charter.net/Spotch/q...1-52-22-90.JPG

Can someone please help me with this? Latest Q3a version using 61.72 on GF6800GT no compressed textures and max details. Win XP pro patched with all official service packs and DX9b.

Man thats and old one, wasn't that a DXTC problem. I don't play it anymore so I don't know if it ever got fixed. Look's like 16bit.

[RuneSR]

Saturnotaku wrote: "But the margins are completely different and that's why you have to take into account the fact that 1) this is a beta game, 2) test sites used different driver versions on the NVIDIA cards (61.77 and 65.62)."

I believe you are misunderstanding what I am writing. Let me repeat a sentence from an earlier reply: "Furthermore this is not a discussion about getting the 6800U or X800XT PE to look best (who wins or not is not the primary issue), it is only a matter of showing all the relevant data".

My concern is not 1) if video card A or B is best, 2) whether or not VST correlates with full HL2 performance or 3) whether or not 65.62 will represent upcoming WHQL-driver performance. My concern was that NVNews did not provide links to certain articles that may provide interesting information about the VST test.

NVNews is normally doing a fine job and have often linked to articles where ATi was doing especially well - and this is what is making NVNews a great site - namely being unbiased. And being unbiased by posting all relevant news regarding a subject (be it VST, Doom3 etc.) is the subject of my postings in this thread. If links to all relevant data has been published, intelligent readers should have little trouble making their own conclusions. I do not normally visit the forums: thus my concern is only associated to the (lack of) material published on the front page of NVNews.

[mezkal]

Originally Posted by Woodelf

Oh I believe they are the new 32 bit color too, but remember the trouble 32 bit had getting accepted over 16bit (thanks to voodoo2/3 fans). I'm just saying that (IMHO) the struggle is probably not over, and that they are still luxuries. Remember, both game engines (HL2 and D3) are/were well over due. What setting's do we anticipate for Stalker, or unreal 3, or even a D3/HL2 engine game that some developer decides to really graphically impress with. I anticipate low AF, AA and shadows, just to keep a decent resolution. All I'm really saying is that this has happened before. New card's come out that allow current games AA/AF, then the new game (far cry for instance) comes along that put's and end to it. AA/AF have been long standing issues (especially AA) that end up on the "first to turn off list". If this weren’t true, it wouldn't still be an issue. I really believe that, had id wanted to, they could have really gone all out on the graphic's detail. One thing they should have done was to enhance the gray looking skin on some of the characters. Sure it would be nice if we could all way's look forward to
full AA/AF, but history says different.

Frankly I think consumer graphics hardware needs a whole new approach to to AA. If you look at professional level hardware like the Wildcat and Wildcat Realizm (sic) boards they have specific RAM used for FRAME BUFFER alone. Current architecture of consumer graphics hardware is just too RAM intensive for decent AA at maxed out resolution and texture details. A hardware FRAME BUFFER could overcome that limitation.

Of course I recommend going the other direction entirely.

UMA.

I think there needs to be a complete rethink of graphics hardware at the consumer level.

GPUs should be interchangeable. Just think about each time you upgrade your graphics card. You pay for VGA/DVI connectors, AGP/PCI interface and a whole slew of other components over and over each time.

I think it's time that UMA was properly investigated.

Personally, it makes MUCH more sense to have really fast RAM on your mobo and have all the connectors neccesary (DUAL DVI/SVIDEO VIVO) on the same board. Then just purchase your GPU as required. It would save money for consumers and probably would allow Motherboard manufacturers to raise their overall prices without too much consumer lashback. It would also give RAM manufacturers cause to celebrate as they could sell more of thier product too consumers rather than bulk discounting it towards OEMs and IHVs.

I know the XBOX is older tech but it does have a very workable UMA implementation based around Nforce/Nforce2.

I can't see why PC's shouldn't follow suit. Technologies like AGP/PCI-e really only get in the way (DIME certainly is being avoided as much as possible by game devs as it arrests the CPU while large chunks of texture/model etc data get shifted/accessed from motherboard hosted RAM), as data is served across the bus from HDD to CPU to VRAM. I think it would be far more efficient to serve from HDD to CPU to RAM and let decent MEMORY PROTECTION and Bounds Checking in the Host OS do its damn job.

But who knows right? It's probably on the cards *somewhere* down the line.

[BrianG]

Seems I touched a nerve with that Doom3 jab. Relax. Rune, honestly, if you use nV News as your homepage, it must be to keep up with the competition. Have you been collecting links to VST comparisons? I wonder why anyone would do such a thing. You are taking an oppurtunity to use an editorial feedback link on the mainpage to take stand on ATI vs. NVIDIA in the VST when the point of the editorial was to take the numbers with a grain of salt and wait for the game to be released. Forgive me if I am a bit suspicious, but a reader that has nV News as their homepage for four years FINALLY registers for the forum when it comes time to link as many VST comparisons as possible? Odd.

You may find the results absolutely stupendous. I don't. Again, it is a synthetic preliminary benchmark. I have linked to many of the comparisons. I have discussed the advantage ATI has in this BENCHMARK. What am I dodging other than another site with the same information, yet again? I could post link after link of water-cooling projects, but I don't. Do you find it academically unsatisfactory that I slight water-cooling. Probably not. This is an ATI site as much it is a water-cooling site.

Please get back on topic and discuss the editorial. You missed the point of my comments and you won't address anything but synthetic benchmark results, so you are no longer needed in this thread. You can join one of the many threads (now that you are a forum meber) about the VST and discuss it until you are blue in the face.

[saturnotaku]

Originally Posted by Spotch

Can someone please help me with this? Latest Q3a version using 61.72 on GF6800GT no compressed textures and max details. Win XP pro patched with all official service packs and DX9b.

Be happy to help. Let me install the game and test it. I have a 6800 Ultra with the 61.77 drivers.

[vampireuk]

Shall we all buy some tin foil hats?

Also I'm waiting for a guy called berg to show up, if you so much as stick a hint of opinion in a news post he is on you