The official Fermi benchmark thread

[j.o.s.h.1408]

Here we go:






Anymore else you guys find, post it here

[Perfectionist]

What's the source tho? If no way to check if it's legit not worth looking at really.

[TooNice]

Just wait for more trusted sources tomorrow. Just had a glance at the tables, and looks like it's got it's moment, but not consistent enough. Also, those who wants to 'play' 3DMark won't be pleased.

[Perfectionist]

I don't trust any benchmark programs anymore due to the whole practice of drivers detecting them running and changing stuff around to make sure they get a better result for the card... An actual game test is the only thing worth taking notice of I reckon, and NOT ones that have endorsements or branding by either graphics companies.

[Odc]

I don't trust any benchmark programs anymore due to the whole practice of drivers detecting them running and changing stuff around to make sure they get a better result for the card... An actual game test is the only thing worth taking notice of I reckon, and NOT ones that have endorsements or branding by either graphics companies.

I must have missed this about the drivers detecting benchmarking programs. Any links?

[Perfectionist]

eh I've read about it all over, I thought it was common knowledge?

just google benchmark detecting or driver detect benchmark etc...

pcworld.com/article/111012/nvidias_benchmark_tactics_reassessed.html

geek.com/articles/games/futuremark-confirms-nvidia-is-cheating-in-benchmark-20030523

zdnetasia.com/benchmark-maker-nvidia-is-wrong-39133935.htm

Futuremark.com Audit Report: Alleged NVIDIA Driver Cheating on 3DMark03

not just them either:
Intel graphics drivers employ questionable 3DMark Vantage optimizations - 3DMurk all over again?, October 12, 2009

[jim]

Google nvidia detecting 3dmark

It's the first link

EDIT: Or the fourth link in the above post

[Perfectionist]

Just wait for more trusted sources tomorrow.

Yeah. I think it's pure clever timing for "leaks" to occur at this time before the real, impartial reviews in the hopes that they will distort/eclipse them and people will make judgements upon seeing "the specs" (which of course could be totally fabricated, deniably) without waiting and making an informed decision.

[sadbuttrue]

I don't trust any benchmark programs anymore due to the whole practice of drivers detecting them running and changing stuff around to make sure they get a better result for the card... An actual game test is the only thing worth taking notice of I reckon, and NOT ones that have endorsements or branding by either graphics companies.

Even then one cannot be 100% condifdent in the results. Both companies tweak their drivers for specific games, and I wouldn't be surprised if those tweaks were aimed solely at the games internal benchmarks. The most trustworthy results come from random timedemo maps imo.

[Perfectionist]

Yeah that's what I mean really. Play a map or whatever of the game with demo recording enabled, not a standard routine something that's completely individual to whoever is reviewing, then play the demo file back as if you were playing it exactly the same but for different cards. Seems the best way. As long as you kept far away from the developers that have obviously been bought off with "the way it's meant to be played" branding all over etc

[TooNice]

I am not sure if there are any reason to trust game benchmarks any more than benchmarking programs. One of the earliest 'scandal' of the sort that I recall is the 'quack3.exe' stunt ATI pulled back when Quake 3 was a key component of any CPU/GFX card benchmark suite.

I do value gaming benchmark more that 3DMark, but only because you can't play the later. Yet to be honest, I would not say that it's completely useless. More often than not, it does give a fairly accurate guideline of the 'class' of the graphic card.

Still, anything that's popular may end up 'optimised'. To be honest, I don't mind a game specific subroutine that improves performance. That's fair optimisation. What I do mind however is if the subroutine trade off IQ for performance without telling the user.

[jim]

Having said that, if they can tweak the games to work well with their graphics card, then that's fair cop really. That's why the vast majority of consumers are buying graphics cards, so that's what they should be judged on.

So I don't think you can't really "fake" game results, unless you found some way of cheating the system to force it to output poorer quality visuals - and even then you'd get caught out fairly quickly by an eagle-eyed reviewer. Obviously you could convince the designers to hamper users of the other brand, but again that's going to get found out fairly rapidly... and unless you do it for all games it'll be obvious.

[kalniel]

So I don't think you can't really "fake" game results, unless you found some way of cheating the system to force it to output poorer quality visuals - and even then you'd get caught out fairly quickly by an eagle-eyed reviewer. Obviously you could convince the designers to hamper users of the other brand, but again that's going to get found out fairly rapidly... and unless you do it for all games it'll be obvious.

Like disabling dx10.1 for Assassin's creed etc?

OP, in what way is this official? Someone splashing a dodgy 'nvidia confidential' across it makes me think it's anything but!

[DanceswithUnix]

Do those strike anyone as rather high AA & resolution settings?

I would have thought that would turn it into more of a ram bandwidth test.

[matty-hodgson]

Do those strike anyone as rather high AA & resolution settings?

I would have thought that would turn it into more of a ram bandwidth test.

When you're splashing £300 - £400 on a graphics card you're gonna want to be playing at at-least 1080p with a decent helping of AA.

[kalniel]

When you're splashing £300 - £400 on a graphics card you're gonna want to be playing at at-least 1080p with a decent helping of AA.

You'll probably want about 6 monitors as well, right?