[H] Benchmarking the Benchmarks

[Epsilon84]

http://enthusiast.hardocp.com/article.html?art=MTQ2MSwxLCxoZW50aHVzaWFzdA==


Timedemo benchmarking of video cards is broken. We have proven this on the preceding pages with today’s most graphically intensive gaming title. Many will argue that timedemo benchmarking is the only scientific approach to video card performance analysis that can be trusted. Why you would want to trust a performance metric that is in no way shape or form going to relate to your gaming experience is beyond me. There is also no doubt that there are some games out there that benchmark perfectly in relation to their real world gameplay. We just don’t know what they are, and quite frankly we don’t care. Today's Crysis benchmarks that in no way reflect real world gameplay are enough validation for us to keep on doing it “our way.” If you want someone’s idea of overall "relative performance" of a graphics card based on timedemo benchmarks, HardOCP.com is not for you. We are going to make sure that our video card evaluations give you a solid idea of the actual gaming performance you will experience at home when playing the game.

Some interesting comparisons between the 'canned' and 'real world' timedemos.

Discuss. :up:

[Syn.]

[H] provides a nice balance to review sites. I always try and look at both and see the best and worst cases with the hardware. Too many times have i seen a review showing much higher FPS numbers then the ones i get in the game.

Maybe with this article people will finally understand the [H] method.

[Omastar]

"HardOCP has been thoroughly mired in “real world” testing scenarios for years now, and there is no doubt that we have taken a lot of heat from various sources. Most of it is generated through ignorance and the human quality of being resistant to change."

No, you arrogant cretins, most of it was generated due to the fact that the logic BEHIND the arcane test methods was not thoroughly elucidated upon, and the metrics changed from review to review. It's so easy to play this 'holier than thou' card after the fact.

[Rock&Roll]

Lol. Omastar, one serious non-believer. I guess i don't understand why people get so belligerent about this subject. They have their test methods, which are valuable and trustworthy. Just my 2c. There are many other sites with value info as well.

I've been reading their site for years. When they shifted to this real world stuff, I was happily met with results that actually seemed on par to what I would get if I bought the card and stuck it in my rig. When I was younger, I always wondered why websites would get 20fps more in almost every game with the video card I just bought. What the [H] has done is bring the video card review down to the level of every other man who buys hardware and uses it for gaming instead of benching. This helps me decide whether I should fork out x dollars for an upgrade, or just hold on to what I've got.

[Omastar]

Never let one site or one source be the final resting place of your decision. When I'm looking for numbers on a video card, I go to about 5 different sites to get an overall trend, then go back and nitpick the numbers. To show your solidarity to one site's testing methodology, be it Anand or [Limp], is just to drink that site's Kool-Aid, exclusively. Fanboys, however, will never learn.

[M.Beier]

I think its a gray-area..,
Its impossible to catch same sequences and same amount of load during multple tests...
However, often drivers are specifically optimized to the darn fly-by's.... I think the 3Dmarks is efficient for "time demo"... :\

However, I hoped they'd spoke about the real issues... SiS Sandra, refreshing gains atleast 2500MB/s, SuperPi 1M being very inaccurate etc...

[happychappy]

No benchmark results can never be copied exactly, as Omastar said you shouldnt base your buying decisions off one site and if you dont trust a site then dont read their reviews, simple as that

[Syn.]

No benchmark results can never be copied exactly, as Omastar said you shouldnt base your buying decisions off one site and if you dont trust a site then dont read their reviews, simple as that

i dont think anyone is, the discussion is why people are dismissing the [H] reviews simply because they are different.

[munim]

Even though you can't recreate the exact series of events every time, I still trust them to get it pretty close, as is shown in their FPS graphs. Different video cards have the same dips and crests, and sometimes they don't occur at the same time but they are the same "crests" or whatever.

However, this is all a moot point because the graphs are there to back up their claim that X card delivers a better gaming experience than Y card. They have played most of the test game and deemed whatever setting to be the best setting that produces the best immersion for that card.

[v_rr]

[H] is a joke. They keep bashing HD 3870X2 in all the article, while 8800GTX happens the same thing and nothing say about it.
All conclusion is about HD 3870X2, and 8800GTX nothing they say.
Anyone reads [H]? Biased reviews, no thanks.....

The method they use is OK and its nice to see them doing it, but the way they do those test is very biased. They bash to much one card while the 8800 almost don´t say nothing.

[saaya]

kyle bennets method is interesting and a good addition to conventional videocard reviewing, but what i dont get is why he claims his methodology is the ultimate way to test videocards and keeps poking at anandtech and trying to make them look bad.

[munim]

[H] is a joke. They keep bashing HD 3870X2 in all the article, while 8800GTX happens the same thing and nothing say about it.
All conclusion is about HD 3870X2, and 8800GTX nothing they say.
Anyone reads [H]? Biased reviews, no thanks.....

The method they use is OK and its nice to see them doing it, but the way they do those test is very biased. They bash to much one card while the 8800 almost don´t say nothing.

The GTX has been out for ages, man. It's the 3870X2 that has something to prove. IIRC they didn't even bash it, they discussed the differences in the testing methodology they use compared to others and thus the different conclusions arrived at regarding the 3870X2.

[Rob Halford]

Anyone finishing high school has learned (and hopefully understood) the terms validity and reliability. As long as not everyone can reproduce their test with the same results, and we all know that that is impossible playing through a video game, then their results are not based on a scientific methodology. Time demos let us do exactly that, as long as you use the same equipment. And they do that over and over and over again. Limited as they are though...

:p:

[SKYMTL]

While I will not comment on the [H] article in particular, I WILL say that I am quite tired of this being brought up on an almost-weekly basis just to "prove" that certain methods are the "right" methods.

In all reality, the ONLY way to truly benchmark a game's performance is to benchmark the WHOLE game. That means every second of every level with repeatable tests. The fact of the matter is that there will always be highly varying performance even within the same game due to level design and the detail contained therein.

I'll give the example of Company of Heroes in DX9 at 1600x1200 resolution on any higher-end card. I play CoH quite a bit to take a break from reviewing so whatever GPU is in the system at the time is what I use...

In one of the first levels, you are forced to defend a bridge while under bombardment from German artillery through the later part of the mission. During the first 30 minutes frame rates are way up there because nothing much is happening. Then when the Germans let all manner of hell break loose, frame rates chug like crazy. The whole mission from start to end takes a good 45 minutes to complete.

Meanwhile, in another mission you get into a tank fight with a couple of your Allied POS Shermans against a couple of near-invincible Axis tank-killers. Frame rates stay high through the whole mission. BUT, there is one point where ATI cards suffer.

Finally, there is a mission near the end of the game where you are forced to defend a hilltop against what seems to be the whole German army....at night. Because of the complex lighting, may graphics cards suffer greatly and I have found that Nvidia cards in particular experience drastic slowdowns in this level.

SO, HERE IS THE QUESTION:

Which level do you use for an actual in-game benchmark and how to you ensure that the benchmark itself is completely repeatable?

In some levels the performance will be drastically different than with others while in some levels certain GPUs will perform better that others...

In the end, unless you bench the whole game (which is impossible given the timeframes) in-game benching of one or two levels just ends up equaling the same accuracy of benched with a timedemo.


I like the fact that [H] uses a different method than alot of other websites since it gives a bit different spin to results. However, is this metod the last word in benchmarking? Well, variety is the spice of life. Let's just leave it at that. ;)

[grimREEFER]

well, this only really affects hardocp cause they try and tell you the best settings to play games on, and with crysis, the bench tells you nothing lol. imo, just use the bench to compare the cards to each other, and let the people who buy the cards figure out what settings they want to play at.

[v_rr]

The GTX has been out for ages, man. It's the 3870X2 that has something to prove. IIRC they didn't even bash it, they discussed the differences in the testing methodology they use compared to others and thus the different conclusions arrived at regarding the 3870X2.

You are only giving reason to me. If 8800 is out there for ages it must have very good drivers.
I want to remember you that HD 3870X2 on that test is using CD drivers because Cat 7.2 final where not released and the crossfire-x driver (cat 7.3) only in March, so if the article want to be not biased it must bash 8800 GTX not a card launch 2 weeks ago with CD drivers and dual-gpu that put things harder in the driver team.

[H] is biased IMO. And this article prove that again. I remeber they make reviews of ATI cards with older drivers just to it look worse.

[grimREEFER]

You are only giving reason to me. If 8800 is out there for ages it must have very good drivers.
I want to remember you that HD 3870X2 on that test is using CD drivers because Cat 7.2 final where not released and the crossfire-x driver (cat 7.3) only in March, so if the article want to be not biased it must bash 8800 GTX not a card launch 2 weeks ago with CD drivers and dual-gpu that put things harder in the driver team.

[H] is biased IMO. And this article prove that again. I remeber they make reviews of ATI cards with older drivers just to it look worse.

the change in drivers wouldnt really make a difference. lol, this is like when there is a cpu review, and then someone goes "they used cas 5 instead of cas 4, this totally invalidates all the benchmarks!," as if the small memory latency difference is gonna make a significant difference on the cpu benches.

[munim]

You are only giving reason to me. If 8800 is out there for ages it must have very good drivers.
I want to remember you that HD 3870X2 on that test is using CD drivers because Cat 7.2 final where not released and the crossfire-x driver (cat 7.3) only in March, so if the article want to be not biased it must bash 8800 GTX not a card launch 2 weeks ago with CD drivers and dual-gpu that put things harder in the driver team.

[H] is biased IMO. And this article prove that again. I remeber they make reviews of ATI cards with older drivers just to it look worse.

They used Catalyst 8.1, the latest one available. Specifically, it is version 8.451.2 I don't know if it's a language barrier that is impeding your comprehension of what I'm saying, but here it is again. The point of conflict was regarding the 3870X2. Other sites were reporting that it was faster than the 8800 Ultra etc. HardOCP argued that this is not the case. They are bashing the way cards are tested by others leads to false conclusions about the card. In this case it was the 3870X2.

Anyone finishing high school has learned (and hopefully understood) the terms validity and reliability. As long as not everyone can reproduce their test with the same results, and we all know that that is impossible playing through a video game, then their results are not based on a scientific methodology. Time demos let us do exactly that, as long as you use the same equipment. And they do that over and over and over again. Limited as they are though...

:p:

Too bad high school didn't teach relevance.

[Mav451]

Omastar has a big point - while their reviews are more "real-world" than canned benchmarks, there's nothing else to compare it too!

OTOH, all other sites running the "canned" timedemo are running the very same demo, and that eliminates discrepancies in that aspect. Then you have:

Drivers
OS
Any essential h/w differences (cas on RAM, quad vs. dual, OCed vs. stock)...

good god. Just alot of differences on top of other differences. I guess my point is that, unless every single reviewing site had the exact same test bench, from drivers, os, cpu/ram/mobo down the last BIOS detail, then there's bound to be small to significant discrepancies.

And I agree that you can never just go to one site for reviews. Anandtech's "canned reviews" are nice, and THG's GPU charts, oddly enough, are great as a general measuring stick. H's is good for seeing what you can use in terms of AA, AF, and resolution.

[v_rr]

They used Catalyst 8.1, the latest one available. Specifically, it is version 8.451.2 I don't know if it's a language barrier that is impeding your comprehension of what I'm saying, but here it is again. The point of conflict was regarding the 3870X2. Other sites were reporting that it was faster than the 8800 Ultra etc. HardOCP argued that this is not the case. They are bashing the way cards are tested by others leads to false conclusions about the card. In this case it was the 3870X2.

Did you opened the link?
8800GTX as the same problem that HD 3870X2 have so.... and this is the main focus of the article. Not comparing performance because in that field they did almost nothing ;)

And again HD 3870X2 is the best single card in the market IMO.

[H] got a point, explored that point in 1 game and in a very restrict conditions (you don´t know witch part of the game they excluded from the benchmark that could give advantage to HD 3870X2, etc)

And if you read the title is Benchmarking the Benchmarks not bashing HD 3870X2 with CD drivers and ignore 8800GTX with more then 15 releases of drivers and have same problems. They are not reviewing HD 3870x2 vs 8800GTX. That can only be on your mind.

Your language barrier looks worst then mine.
If you look closer they only talk about HD 3870X2. Nothing more nothing less. And that´s the problem. It´s caled biased because 8800GTX showed exactly the same problems and they don´t say almost nothing about it.

If they really want to show their point they must talk about both cards and not only one. The conclusion is just patetic and only reserved to bash HD 3870X2. Not 1 word about 8800GTX.....

[grimREEFER]

3870x2 is good, but idk if u can definitively call it the best when you see how it plays crysis and quake wars.
this is straying off topic lol.

[WrigleyVillain]

kyle bennets method is interesting and a good addition to conventional videocard reviewing, but what i dont get is why he claims his methodology is the ultimate way to test videocards and keeps poking at anandtech and trying to make them look bad.

That's simple. Because he's an arrogant :banana::banana::banana::banana::banana:.

[Rob Halford]

Too bad high school didn't teach relevance.

Relevance is in the non-replicability of their tests, i.e. the two of us can’t reproduce their results with exactly the same equipment, and thus, they are not objective. Or another way to put it in English, they are BS from a testing point of view... SKYMTL’s post is addressing the same and other relevant implications of their methods. Like it or not.

[munim]

Did you opened the link?
8800GTX as the same problem that HD 3870X2 have so.... and this is the main focus of the article. Not comparing performance because in that field they did almost nothing ;)

And again HD 3870X2 is the best single card in the market IMO.

[H] got a point, explored that point in 1 game and in a very restrict conditions (you don´t know witch part of the game they excluded from the benchmark that could give advantage to HD 3870X2, etc)

And if you read the title is Benchmarking the Benchmarks not bashing HD 3870X2 with CD drivers and ignore 8800GTX with more then 15 releases of drivers and have same problems. They are not reviewing HD 3870x2 vs 8800GTX. That can only be on your mind.

Your language barrier looks worst then mine.
If you look closer they only talk about HD 3870X2. Nothing more nothing less. And that´s the problem. It´s caled biased because 8800GTX showed exactly the same problems and they don´t say almost nothing about it.

If they really want to show their point they must talk about both cards and not only one. The conclusion is just patetic and only reserved to bash HD 3870X2. Not 1 word about 8800GTX.....

Ah I see what you mean now! It is true, the GTX is also slower, but I think a big problem is that AMD is able to sort of cheat the regular fly-by benchmark. For instance, they released a driver or hotfix or whatever that enormously increased the FPS in the fly-by, yet in the game it only increased very little. I think what Kyle was trying to say is that using a fly-by can result in manufacturers misleading people.

Bottom line is this: Hypothetically, since I can't afford either card, I can buy the 3870X2, or I can buy the 8800 Ultra/GTX. I can only use the drivers that are available. Which one delivers the better performance in game? I don't care if one card has been out for 15 months and its drivers are refined.

Relevance is in the non-replicability of their tests, i.e. the two of us can’t reproduce their results with exactly the same equipment, and thus, they are not objective. Or another way to put it in English, they are BS from a testing point of view... SKYMTL’s post is addressing the same and other relevant implications of their methods. Like it or not.

Definitely, he acknowledges that I probably won't be able to repeat his "results". That FPS graph is not what he bases his conclusions from. He bases his conclusion from playing the whole game. His conclusion is not the mainstream, his conclusions are as follows, paraphrased " With these settings, you will have have the best possible gaming experience throughout the whole game". I guess if you don't find that relevant, then you should keep reading Anandtech etc.

[AbelJemka]

How can they prove their point using only one game Crysis? They can't they their methods are better than the others sites with only one game :shrug:

[Syn.]

You are only giving reason to me. If 8800 is out there for ages it must have very good drivers.
I want to remember you that HD 3870X2 on that test is using CD drivers because Cat 7.2 final where not released and the crossfire-x driver (cat 7.3) only in March, so if the article want to be not biased it must bash 8800 GTX not a card launch 2 weeks ago with CD drivers and dual-gpu that put things harder in the driver team.

[H] is biased IMO. And this article prove that again. I remeber they make reviews of ATI cards with older drivers just to it look worse.

They are using beta drivers that AMD sent them when they postponed the 3870X2 launch, Cat's 8.2 Beta (8.415.2). Does anyone really use drivers off cd's anymore? Oh and i am guessing you mean Cat 8.3 not 7.3.

They are not bashing 3870X2 they are only using it vs GTX because that is what people are spamming their forums about. If people actually read the article you will notice that they test GTX second which is why they just say:

With the 8800 GTX we see much of the same...

So what they are trying point out is that Timedemo benchmarking, no matter what card you are using, shows results that do not relate to actuall gameplay. Which means if you play the game you will not see the same frame rate, in most cases the actuall pefromance will be much worse.

kyle bennets method is interesting and a good addition to conventional videocard reviewing, but what i dont get is why he claims his methodology is the ultimate way to test videocards and keeps poking at anandtech and trying to make them look bad.

Kyle is using Anandtech as example because they are quite experienced review site which firmly promotes that they belive in Timedemo Benchmarking. Others may use it but they have not come out and said this way is the best.

SO, HERE IS THE QUESTION:

Which level do you use for an actual in-game benchmark and how to you ensure that the benchmark itself is completely repeatable?

In some levels the performance will be drastically different than with others while in some levels certain GPUs will perform better that others...

In the end, unless you bench the whole game (which is impossible given the timeframes) in-game benching of one or two levels just ends up equaling the same accuracy of benched with a timedemo.


I like the fact that [H] uses a different method than alot of other websites since it gives a bit different spin to results. However, is this metod the last word in benchmarking? Well, variety is the spice of life. Let's just leave it at that. ;)

Kyle has stated that they play through the whole game and find the settings which are most playable through out. The charts that they show in their reviews are only there as reference points. They only select few scenes from the game that show most general performance and show you results so that you can see for your self that they really did do these tests.

Their reviews, evaluations as they like to call them, are subjective because the end result is based on the experience the person testing has had. They dont hide it and are quite open about it. It is up to the reader to decide if they agree with that decision or not. To be honest i dont care about either these days much as i am boycotting Intel and Nvidia so that does not leave me with much choice. But i do like that they test thoroughly and come a conclusion that always seem to be more inline with what i get on my PC.

[Praz]

but what i dont get is why he claims his methodology is the ultimate way to test videocards and keeps poking at anandtech and trying to make them look bad.
Most likely because of the the technical level of the reviews at Anandtech lately. It seems their reviews are becoming a thorn in the side of quite a few other sites. They are being done at a level that requires knowledge that other sites apparently can't match.

[Rock&Roll]

i dont think anyone is, the discussion is why people are dismissing the [H] reviews simply because they are different.

yeah I agree. I don't think anyone reads just one site. But, all I'm saying is, there is room for Kyle's views and other's views in the world of computer hardware. Useful information comes in many forms. And, FWIW, to all of you who take the stuff so ridiculously serious--there is no review that can compare to having the card in your own box, and doing your own testing your own way. So, by that, I may say that NO review is good enough. I personally think that anyone throwing stones at a review site needs to STFU and start their own site if they think the've got a better method. That's what Kyle has done.

[SKYMTL]

They are being done at a level that requires knowledge that other sites apparently can't match.

:rofl: :rofl:

Thanks for making my day.

I don't think one GPU testing methodology requires that much more knowledge than any other. Canned vs """""real world""""" benchmarks are not a testament to a site's technical know-how.

[keiths]

Benching realworld if fine, their apples to oranges comparision is crap.

[technodanvan]

:rofl: :rofl:

Thanks for making my day.

I don't think one GPU testing methodology requires that much more knowledge than any other. Canned vs """""real world""""" benchmarks are not a testament to a site's technical know-how.

He isn't referring to just GPU testing, he's referring to articles explaining just how Card A works, what exactly makes it different from Card B, how it's been evolved from Card C, etc, etc. And I'm not just talking about them copying the specs from the manufacturer's website.

Not all of them are like that of course, but that additional content makes them stand a bit apart from other reviews. This doesn't necessarily make them better by any means, but some users are interested in that sort of thing and gives Anandtech a slight edge in the review department which other sites have to take note of.

[Praz]

:rofl: :rofl:

Thanks for making my day.

I don't think one GPU testing methodology requires that much more knowledge than any other. Canned vs """""real world""""" benchmarks are not a testament to a site's technical know-how.
but what i dont get is why he claims his methodology is the ultimate way to test videocards and keeps poking at anandtech and trying to make them look bad.
My response was to the bolded part above in general and not the technical expertise required for the video card review. The testing mythology itself I have no issue with as long as it's consistent.

[SKYMTL]

Kyle has stated that they play through the whole game and find the settings which are most playable through out. The charts that they show in their reviews are only there as reference points. They only select few scenes from the game that show most general performance and show you results so that you can see for your self that they really did do these tests.

Look, I happen to like their benching methodology but what you said above has come caveats. Take for example this quote from their BFG 8800GTS 512MB review regarding CoD4:

Our run-through consisted of playing the mission The Bog.

This means their results are just as accurate / inaccurate as results from these other sources.

They even have the benchmark results graphed out and all for the "run through". This is NOT based on their whole gameplay experience but rather one mission and it illustrates my points quite well. These are benchmarks based on one mission and do not take into account performance displayed throughout the game as a whole.

Personally, I think this pitting one website against another is just disgusting. I understand it is done for marketing / viewership reasons and standing by a benchmarking methodology is the right thing to do but there is a limit. Like I said before: it is good they benchmark this way since it gives readers a different perspective but this has become ridiculous already.

[SKYMTL]

My response was to the bolded part above in general and not the technical expertise required for the video card review. The testing mythology itself I have no issue with as long as it's consistent.


Understood. I stand corrected. :D

[Krizby87]

Benching realworld if fine, their apples to oranges comparision is crap.

I second that :D, their "playably framerate" methodology just ticks me off completely. Because each person's "playable framerate" is different from each other, and there are people rather play @ low resolution with AA/AF rather than play @ high resolution with no AA/AF. It would be much better if they just report what fps they are getting at which screen resolution with or without AA/AF and let the reader decide whether or not it is playable.

[Mav451]

I second that :D, their "playably framerate" methodology just ticks me off completely. Because each person's "playable framerate" is different from each other, and there are people rather play @ low resolution with AA/AF rather than play @ high resolution with no AA/AF. It would be much better if they just report what fps they are getting at which screen resolution with or without AA/AF and let the reader decide whether or not it is playable.

I agree on that. At high res, low AA is more than plenty, and I'd rather go with native resolution for my 24", then drop it 2 or more steps for massive AA.

On Anandtech, I don't feel their reviews are as consistent in quality as I'd like (especially to be a main review site) for my decision-making. I still end up going to other sites to augment what I've read there.

[zerazax]

While I like having a max playable settings feature, [H] really needs to do an apples to apples comparison too because that allows cards to be put at relative power to one another at a certain setting. As stated by many, some people prefer lowering AA / AF to get a bit more FPS. For instance, 30 FPS is technically playable for a shooter but some refuse to play at anything less than 60FPS so being playable at 30FPS has no bearing to them.

[Epsilon84]

While I like having a max playable settings feature, [H] really needs to do an apples to apples comparison too because that allows cards to be put at relative power to one another at a certain setting. As stated by many, some people prefer lowering AA / AF to get a bit more FPS. For instance, 30 FPS is technically playable for a shooter but some refuse to play at anything less than 60FPS so being playable at 30FPS has no bearing to them.

They actually do sometimes have an apples to apples comparison as well. Not sure why they don't do it all the time, it would certainly appease all the naysayers. :up:

[zerazax]

They actually do sometimes have an apples to apples comparison as well. Not sure why they don't do it all the time, it would certainly appease all the naysayers. :up:

Exactly, they don't do it often enough, and manage to make absolute conclusions on cards with 2 apples to orange game comparisons and not much else.

This ain't exactly basing a review on 3dMark06...

[RAMMAN]

allow me to list [H],s transgressions in recent times.over the years they have chosen games and graphical settings that bias the results towards nvidia,s gpu,s and amd,s cpu,s.they refused to benchmark the 8600gt with quake 4 which would have shown just how crappy this card really is.nor would he benchmark the 8600gt against the 8800gts 320mb.not to mention saying in their forum they were going to review the core 2 duo showing it not beating the a64.the next day we see they made it 100% gpu limited so both cpu,s performed the same.kyle tried to get away with it by calling the review something else and refused to review core 2 duo.gee i wonder why :rolleyes:
not to mention the 3.06ghz p4 fiasco and kyle banning anyone that disagrees with him in his forums.btw when was the last time kyle used a non twimtbp game for benchmarking? also kyle says he alway tries to find the best possible iq while still maintaining a playable framerate.than why does he insist on using medium settings with a high resolution in crysis when high settings on a lower res looks much better?

[Eastcoasthandle]

Didn't they say benchmarked time demos showed gains with Core 2 processors and considered them worthless (C2D review)?
Link (http://www.hardocp.com/article.html?art=MTEwOCwxMSwsaG5ld3M=)

[munim]

Didn't they say benchmarked time demos showed gains with Core 2 processors and considered them worthless (C2D review)?
Link (http://www.hardocp.com/article.html?art=MTEwOCwxMSwsaG5ld3M=)

From what I understood, the C2D made little appreciable difference in performance over the A64s at settings you would play at. I thought it is accepted that the limiting factor in games is the videocard.

[Eastcoasthandle]

From what I understood, the C2D made little appreciable difference in performance over the A64s at settings you would play at. I thought it is accepted that the limiting factor in games is the videocard.

Did I misunderstand the implication in that review implying that time demos favored C2D. Which mirrors what they are saying the video card review? :confused:

[Rock&Roll]

From what I understood, the C2D made little appreciable difference in performance over the A64s at settings you would play at. I thought it is accepted that the limiting factor in games is the videocard.

Such has been my experience, coming from a 2400mhz dual-core Athlon. However, I game at 1920x1200. If you game at say 1280x1024, you're going to get a few frames by the upgraded subsystem.

[munim]

Did I misunderstand the implication in that review implying that time demos favored C2D. Which mirrors what they are saying the video card review? :confused:

Haha yes. I am confused.

[Mav451]

allow me to list [H],s transgressions in recent times.over the years they have chosen games and graphical settings that bias the results towards nvidia,s gpu,s and amd,s cpu,s.they refused to benchmark the 8600gt with quake 4 which would have shown just how crappy this card really is.nor would he benchmark the 8600gt against the 8800gts 320mb.not to mention saying in their forum they were going to review the core 2 duo showing it not beating the a64.the next day we see they made it 100% gpu limited so both cpu,s performed the same.kyle tried to get away with it by calling the review something else and refused to review core 2 duo.gee i wonder why :rolleyes:
not to mention the 3.06ghz p4 fiasco and kyle banning anyone that disagrees with him in his forums.btw when was the last time kyle used a non twimtbp game for benchmarking? also kyle says he alway tries to find the best possible iq while still maintaining a playable framerate.than why does he insist on using medium settings with a high resolution in crysis when high settings on a lower res looks much better?

While that is really tough to read, there is some truth there:
Kyle definitely had (not has) an AMD slant. It's obviously gone now, especially with Phenom benches in the open. I think before those benches were available, he made it seem like it was something special.

His word alone, was enough for me to hesitate on my buying decision at that time. Of course once those #'s came out in September, it was obvious he was bsing. Which is a bit odd, b/c for the most part he had seemed truthful - until now. I wonder if AMD had an influence on his stance towards the Phenom?

[RAMMAN]

While that is really tough to read, there is some truth there:
Kyle definitely had (not has) an AMD slant. It's obviously gone now, especially with Phenom benches in the open. I think before those benches were available, he made it seem like it was something special.

His word alone, was enough for me to hesitate on my buying decision at that time. Of course once those #'s came out in September, it was obvious he was bsing. Which is a bit odd, b/c for the most part he had seemed truthful - until now. I wonder if AMD had an influence on his stance towards the Phenom?

i agree kyle is no longer biased towards amd anymore, since he now owns a core 2 setup.but its painfully obvious he has been biased towards nvidia,s cards since day 1 and still is :mad: .making the 8600gt look good by not benchmarking with quake 4 and not comparing it with the 8800gts 320 is a classic example.

[Krizby87]

i agree kyle is no longer biased towards amd anymore, since he now owns a core 2 setup.but its painfully obvious he has been biased towards nvidia,s cards since day 1 and still is :mad: .making the 8600gt look good by not benchmarking with quake 4 and not comparing it with the 8800gts 320 is a classic example.

lol, so he is Nvidia biased by not comparing one Nvidia card to another Nvidia card ? last time I checked Nvidia 8600 series is no worse than AMD's 2600 series, and they are both crap.

[Plywood99]

They are using beta drivers that AMD sent them when they postponed the 3870X2 launch, Cat's 8.2 Beta (8.415.2). Does anyone really use drivers off cd's anymore? Oh and i am guessing you mean Cat 8.3 not 7.3.

They are not bashing 3870X2 they are only using it vs GTX because that is what people are spamming their forums about. If people actually read the article you will notice that they test GTX second which is why they just say:



So what they are trying point out is that Timedemo benchmarking, no matter what card you are using, shows results that do not relate to actuall gameplay. Which means if you play the game you will not see the same frame rate, in most cases the actuall pefromance will be much worse.



Kyle is using Anandtech as example because they are quite experienced review site which firmly promotes that they belive in Timedemo Benchmarking. Others may use it but they have not come out and said this way is the best.



Kyle has stated that they play through the whole game and find the settings which are most playable through out. The charts that they show in their reviews are only there as reference points. They only select few scenes from the game that show most general performance and show you results so that you can see for your self that they really did do these tests.

Their reviews, evaluations as they like to call them, are subjective because the end result is based on the experience the person testing has had. They dont hide it and are quite open about it. It is up to the reader to decide if they agree with that decision or not. To be honest i dont care about either these days much as i am boycotting Intel and Nvidia so that does not leave me with much choice. But i do like that they test thoroughly and come a conclusion that always seem to be more inline with what i get on my PC.


Great post Syn. Sounds like you actually read ALL of the article.

And this time demos are best because of "scientific method" crap has got to stop. What's so scientific about running driver optimized time demo's? And what about games sponsored by ATI or Nvidia. Is it scientific then?

The only real way to evaluate something is to use it. Video card - use it. Car - drive it. I guess all car reviews are worthless since each person drives differently, no scientific method. When Jeff Gordon pulls into the pit stop and says the front of his car feels sloppy, does the pit chief say "what scientific method did you use to come up with that?" No. Because he knows the driver has experience.

Same with the [H] and their video card evaluations. Experience actually using the cards and doing this type of testing for years now means more than some "scientific" hooplah...


Ply

[JumpingJack]

http://enthusiast.hardocp.com/article.html?art=MTQ2MSwxLCxoZW50aHVzaWFzdA==



Some interesting comparisons between the 'canned' and 'real world' timedemos.

Discuss. :up:

Are you trying to draw me into an argument? :) :) :p:

[RAMMAN]

lol, so he is Nvidia biased by not comparing one Nvidia card to another Nvidia card ? last time I checked Nvidia 8600 series is no worse than AMD's 2600 series, and they are both crap.

my point is he tryed to make the 8600gt look good by showing best case scenarios when it clearly is crap, as you put it.unlike the other review sites he was trying to mislead his readers.i also remember reading somewhere on hardocp(i think it was a thread he made about his own setup)that he doesnt like ati cards in general.on the other hand he did do a negative review of the 8800ultra because of its poor price/performance ratio at the time.

[Eastcoasthandle]

my point is he tryed to make the 8600gt look good by showing best case scenarios when it clearly is crap, as you put it.unlike the other review sites he was trying to mislead his readers.i also remember reading somewhere on hardocp(i think it was a thread he made about his own setup)that he doesnt like ati cards in general.on the other hand he did do a negative review of the 8800ultra because of its poor price/performance ratio at the time.
can you find this post?

[munim]

@RAMMAN: I don't understand why he would compare the 8600GT to a 8800GTS 320MB.

[RAMMAN]

@RAMMAN: I don't understand why he would compare the 8600GT to a 8800GTS 320MB.

the obvious reason:to compare price/performance ratio.

[RAMMAN]

can you find this post?

i remember reading it in the thread "rebuilding the beast" i read through the whole thread just now but apparently it has no mention of ati or amd.btw i have no idea how he came to the conclusion the 8800gts 320mb is better than the hd2900xt.

[eric66]

'if you look at the amount of difference between the AMD and Intel CPUs, you will see that it isn’t enough to amount to anything. The only game that we saw any real-world difference in was Oblivion, and even that was tiny. A little overclocking would clear that difference up'

lol

[RAMMAN]

this is the closest thing to a smoking gun i could find:

http://forums.overclockers.com.au/showthread.php?s=&threadid=113066&pagenumber=1

http://www.xtremesystems.org/forums/showthread.php?s=&threadid=4828

if that doesnt discredit hardocp i dont know what will.

[initialised]

I would argue that the only valid 'real world' benchmark would be a to have a gameplay session recorded under double blind conditions on a system and settings that are not going to drop below playable frame rates.* However, 'canned' timedemos still give good relative performance data between different systems provided that they are not recorded under conditions that would favour one graphics/CPU/chipset architecture's area of expertise.

Where the 'best playable' methodology falls down is the fact that an end user's PC is likely to have an OS riddled with memory leaks and software that said user has forgotten about running in the background without the latest drivers or OS patches and without in game or nTune/Catalyst optimisations while the test bench of a review site or magazine will be offline and optimized for gaming/benchmark performance (I unplug the network and close down all but the bare minimum including explorer.exe for bench runs). So it is worth noting that 'best playable' on a bench/review rig may not be playable at end user level.

*Double blind means that neither the experimenter nor the subject of the experiment knows what to expect from the results so cannot apply (intentional or otherwise) bias to the outcome of the experiment. This is the cornerstone of modern medical trials. In this context it would probably be enough for timedemos to be recorded while the players are unaware that this is the purpose of the playtest at low settings on a high end system and for the the benchmark to consist of randomly selected chunks of gameplay.

[Omastar]

Aggregately, this is what it comes down to:

All other mainstream review sites: using canned demos that apparently aren't a fair representation for the game's actual performance. But, at least these demos can be replicated, and you can at least extract trends from these numbers, even if the numbers demos give are not what real-world playing and benching would yield.

[H]: let's do in game benchmarking that can never be 100% replicated among our competitors and call ourselves superior while we play the percentage differences of flawed benching methodologies.

Sounds like a great, big wash to me.

[G H Z]

kyle bennets method is interesting and a good addition to conventional videocard reviewing, but what i dont get is why he claims his methodology is the ultimate way to test videocards and keeps poking at anandtech and trying to make them look bad.

So they get noticed :up:

[initialised]

[H]: let's do in game benchmarking that can never be 100% replicated among our competitors and call ourselves superior while we play the percentage differences of flawed benching methodologies.Exactly, subjective methodology that cannot be replicated is unscientific and has no credibility (especially given the banner and side bar adds for Nvidia). If they are going to do 'play tests' through specific levels then have an independent person do the playing, record this and re run it on each test rig and have it available for download so the data can be verified.

The cynics amongst us could look at this as [H] being payed to play the latest games and write about how well they work on Nvidia hardware by Nvidia but that would be jumping to unsupportable conclusions.

My own experience has shown that the 3870 X2 is faster than the 8800 Ultra, the real question should be whether three Ultras in SLi are faster than two X2s in Crossfire.

[Omastar]

My own experience has shown that the 3870 X2 is faster than the 8800 Ultra, the real question should be whether three Ultras in SLi are faster than two X2s in Crossfire.

That's the 64 dollar question. It depends a good deal on how fast and how much AMD's CrossfireX drivers mature. I'd also like to see dual X2s versus dual 9800GX2s when they finally hit stores.

[initialised]

That's the 64 dollar question. It depends a good deal on how fast and how much AMD's CrossfireX drivers mature. I'd also like to see dual X2s versus dual 9800GX2s when they finally hit stores.In a way that doesn't matter as two X2s will be cheaper than two 9800s. On the other hand there's the halo effect which explains why far more units currently ship with GeForce than Radeon cards. But then what about motherboard support for multiGPU? Skulltrail does both but will all this extend to gaming cards, if it does what happens to Nvidia's and ATi's chipset business and AMD CPUs? HP can get Crossfire to work on Nvidia boards with hacked drivers so presumably ATi can but don't want to. Whatever happened to modularity?

Anyway, back on topic, [H] admit their method is unscientific, propose a sound scientific method using recorded demos (which appears to justify their conclusion that X2 ain't as good as it looks) then throw this away apparently because it was hard to do. I'd like to see another site do the same as a single set of data from [H] isn't enough. If ATi are tweaking performance (cheating) in a common benchmark to get better numbers then it is big news, ATi in Quake 3 and Nvidia in 3DMark03 come to mind.

[GreatWhiteHope]

this is the closest thing to a smoking gun i could find:

if that doesnt discredit hardocp i dont know what will.

I am not sure why a mistake 6 years ago would discredit them now and they did address it. They did F up the way they handled the forum issue though. That does not seem to be the case any more though as I see those guys handling both barrels from some members over there.

Friday November 15, 2002[H]ardNews

Honesty & Integrity:

If you think [H]ardOCP has been not been founded on honesty and integrity over the last five years, then you might as well stop reading now, as what I have to say will not pertain to you. We are an enthusiast site that existed before we were called "enthusiasts". We have watched the community grow and we have grown with it. Our readers are our peers and our number one concern, and never think any different.


There is a situation that we need to discuss that has been brewing since we posted the Intel 3GHz review on the 13th at 9pm CST. After two days of talking and listening to our readers and our peers that we have full trust in, it is evident that we have done something that is inexcusable.


After researching this fully for two days, I have been able to pull together 100% factual evidence that a single set of benchmarks in the 3GHz article were changed after publication of the article. The 3DMark benchmark was changed from "17829" to "17329". The fact of the matter is that we did not note this change. It is standard operating procedure that anytime a portion of an article is changed, outside spelling and grammar, it is noted and explained as to why the change was made. This was not done.


As I am Editor-in-Chief of HardOCP.com this is 100% fully my responsibility to make sure issues such as these do not occur. The change in the benchmark without notation is fully my responsibility and I did not fulfill my responsibility to you. For that I apologize.


Our detractors have been posting stories that have reached from one end of the spectrum to the other every since they smelled blood in the water. I assure you that we do not post lies, and we have never intentionally posted numbers or benchmarks that we have known to be wrong. Doing that is simply not in our character. We do not accept bribes and we do not play favorites. While this site has become my livelihood, it is never viewed as "for profit" business. Our first and foremost objective is to supply solid information to our readers. At this point, if you question our competency, I can understand as we have given you a reason for doing so. Please do not let this be a mark against our integrity and honesty.


I assure you that this will not happen again.


Kyle Bennett

Editor-in-Chief [H]ardOCP.com

I like the {H} benchmarking on video cards. It is great to have an alternative to the others. I am sad that bit-tech stopped their real world testing.

[THE JEW (RaVeN)]

If you wanted an account on how [H]'s data is a bit different than others, look at the whole "incident" fiasco. A good chunk of the original crew at GenMay went there after they tested the boards themselves. Results didn't line up with the [H] review. [H] banned a bunch of people and quietly "fixed" their review numbers.

I like their alternative POV, but big review sites seem too susceptible to big companies. When those sites or affiliates come out with articles stating that they don't take payola, that they're honest, or they edit old articles without stating an edit--some sort of flag should go up. Honestly, I find the specific results in forums my measuring stick (ie. niksub's wb testing).

Heck, even XS had a fun little "incident" over how the place was supposed to be run:

http://www.ocforums.com/showthread.php?t=282850

I have no comment on that as I wasn't here at the time. I think I was actually trying to join here then and registration was closed due to the brouhaha :lol:

A bunch of the "major" sites (heck, I still visit them) have had spats with other sites and/or themselves. Everybody gives into childish tantrum's eventually. I being just as guilty of that (probably more so) as anyone else.

[initialised]

CustomPC also do in game testing and they are giving it a similarly bad review. It's print only for now but will appear here http://www.custompc.co.uk/reviews soon

Hmmm... this is starting to smell like optimised drivers for specific games/benchmarks.

Can someone with access to both an Ultra or GTX and an X2 replicate the [H] data?

[RAMMAN]

I am not sure why a mistake 6 years ago would discredit them now and they did address it. They did F up the way they handled the forum issue though. That does not seem to be the case any more though as I see those guys handling both barrels from some members over there.



I like the {H} benchmarking on video cards. It is great to have an alternative to the others. I am sad that bit-tech stopped their real world testing.

that last paragraph from kyle is nothing but lies.

[Epsilon84]

CustomPC also do in game testing and they are giving it a similarly bad review. It's print only for now but will appear here http://www.custompc.co.uk/reviews soon

Hmmm... this is starting to smell like optimised drivers for specific games/benchmarks.

Can someone with access to both an Ultra or GTX and an X2 replicate the [H] data?

Extremetech also uses the FRAPS methodology and come to similar conclusions.
http://www.extremetech.com/article2/0,1697,2252547,00.asp

Hmmm... the plot thickens. :up:

[Eastcoasthandle]

Extremetech also uses the FRAPS methodology and come to similar conclusions.
http://www.extremetech.com/article2/0,1697,2252547,00.asp

Hmmm... the plot thickens. :up:

This review shows the 3870x2 on par with the 3870. They didn't do it right :rolleyes:

[Epsilon84]

This review shows the 3870x2 on par with the 3870. They didn't do it right :rolleyes:

The Crysis numbers do look a bit odd (driver issue perhaps?), but the other games show fairly healthy gains at higher resolutions.

[Eastcoasthandle]

The Crysis numbers do look a bit odd (driver issue perhaps?), but the other games show fairly healthy gains at higher resolutions.
All the numbers look odd ;)

[Epsilon84]

All the numbers look odd ;)

Well, I'd have to agree to disagree. ;)

In times like this, I really wish I had a 8800Ultra and 3870 X2 just to test things out myself!

[RAMMAN]

If you wanted an account on how [H]'s data is a bit different than others, look at the whole "incident" fiasco. A good chunk of the original crew at GenMay went there after they tested the boards themselves. Results didn't line up with the [H] review. [H] banned a bunch of people and quietly "fixed" their review numbers.

I like their alternative POV, but big review sites seem too susceptible to big companies. When those sites or affiliates come out with articles stating that they don't take payola, that they're honest, or they edit old articles without stating an edit--some sort of flag should go up. Honestly, I find the specific results in forums my measuring stick (ie. niksub's wb testing).

Heck, even XS had a fun little "incident" over how the place was supposed to be run:

http://www.ocforums.com/showthread.php?t=282850

I have no comment on that as I wasn't here at the time. I think I was actually trying to join here then and registration was closed due to the brouhaha :lol:

A bunch of the "major" sites (heck, I still visit them) have had spats with other sites and/or themselves. Everybody gives into childish tantrum's eventually. I being just as guilty of that (probably more so) as anyone else.

wow...really makes you look at fugger in a different light.

[KTE]

There used to be a time where the same creed/thoughts/lifestyles/beliefs/opinions were enforced on humans by some-one/people who believed they knew better, nearly always with extreme tyranny and oppression. The modern West named that period "The Dark Ages". Humans all differ, have always differed and will always differ. To try and eliminate this and to make one dogma, opinion and creed rule and override all others as the best because of your own opinion as nothing but another human with the same knowledge and a larger than life superiority complex, causes exactly what we know and have seen time and time again to cause in humans: bands/parties/sects/destruction/abuse/low acts-traits/wars/hatred/loss formed in every way, as happened through the Dark Ages, each touting his own until the most tyrant and powerful survives and eliminates the less tyrant and/or weaker using extremes, most often.

People should heed the last 10,000 year history and religious lessons for once. With any such fiercely contested dear topics humans are going to differ and have their own opinion and you can't do anything about that, you can't satisfy all, if you try to change opinions forcefully, you end up ruining everything being a severe extremist as in the Dark Ages. With most humans this covers opinions on everything, from clothes, cars, politics, cultures, songs, games, beauty, to companies, sport teams and what not. So the same sentiment applies here and is aroused in even this case where just judging "which review methodology is right" because each will declare his own. Hence, why site authors/partisans start getting personal and egotistical over their method compared to someone others. Whose to judge whose correct?

The point comes down to the same basics. Each knowledgeable reader decides for him/her-self based on experience and/or knowledge, simple as that. The vulnerable choose someone to follow instead, not facts. Each reason will be an opinion too and different minds accept and look to different techniques as valid for themselves. Different minds see and understand differently the same situation and sensory input (perspective) and this is of paramount importance where considering "user gameplay experience". It is best left for each to decide theirself after due research, and people should not try and eliminate differences as such. I for one appreciate all testing styles because each can and does portray what another may well miss. However, if you're trying to produce something accurate and real-life reflective for us and you tout you have the most complete, in-depth, comprehensive, accurate, full and real-life representative methodology then you need to look in the mirror at what you think you're doing and why, because you are obviously not the right-minded person to be authoring reviews for us. It is we who decide and judge how your methodology and editorial is, not you the author. A simple "We hope it is" or "Our aim is to" prefix infront of the salesman self-touting statement would make an immense difference in giving you some credibility for not being one-eye blind on your method pre-authoring. It is beneficial in all ways to keep an open mind.

If I was buying a GFX, I would look for a similar system setup to what I had in mind with the same games and settings I had in mind and see which does better overall in reviews; which are nothing but sneak previews, but mostly heed, what users who have them who I trust are showing/saying. Simple as that really. Sites are just a starting reference if you have nothing to go off for me, an indicator to base your uninformed inclining's on. They're relative to me, not absolutes. I use all unless I can find some major repeated discrepancy in one.

Then, I usually pick one [hardware config] I think is best and go out to test it myself behind curtains for myself, when I can, to decide if it holds up to what I understood from the reviews and feedback. There is nothing which can replace knowledgeable experience. I appreciate all efforts and all reviews unless they be deliberately compromised or flawed, they are all useful and serve for our benefit individually and combined. If one [setup] gives me too many problems compared to what reviews show, then I drop it until I have more time and pick something which does work adequately for my needs meanwhile. Fortunately for me, I can pretty much nick over any day down to the labs and test most equipment there not belonging or paid by me and it works out best for me overall in developing experienced opinions and deciding what to purchase. For most others, they have to go off reviews, so the more accurate, diverse and in-focus-depth they are, the better. :)

My brief opinions on review styles 1/2 way down: http://www.xtremesystems.org/forums/showpost.php?p=2756999&postcount=2

[Syn.]

[H]: let's do in game benchmarking that can never be 100% replicated among our competitors and call ourselves superior while we play the percentage differences of flawed benching methodologies.

Sounds like a great, big wash to me.

In [H] the FPS count is used to indicate weather the card has the performance for the graphics quality options set. If it dips under 25fps while playing and causes the game to slow down the quality settings are brought down. If a card is doing 80 fps on the quality setting currently set then they would raise the quality and test again until Maximum Playable Graphics Settings are found.

Other websites use set resolutions and quality settings and run them over and over again to tell you that Card A does 100 FPS in this Canned Benchmark and Card B does 90 FPS in this Canned Benchmark. What they dont tell you is that Card B can hold its FPS count at playable level on higher graphics level then Card A. From the FPS results you would conclude opposite, that Card A should preform better on higher graphics quality. But as such graphics level test are not done you only find that out if you actually play the game your self.

[H] will tell you how far the graphics card can go, others will tell you at what speed it will travel. Both combined make for the best conclusion so i would not exclude [H] tests simply because you do not understand them.

[schnulli]

If a card is doing 80 fps on the quality setting currently set then they would raise the quality and test again until Maximum Playable Graphics Settings are found.

and now you can ask yourself WHY? the [H] approach gets so much hassle!

it's funny thou while this isn't the first time we saw kyle getting top discussed number one in the interhurts; the real reson was anand and their ridiculous show of the amd/ati x2! and what shall i tell you? i saw it, i lost my breath, i went over to XS just to see some little kids moaning and whining WHY? their beloved card got messured with a "evil" nforce platform...r all here really so dumb?


you can argue what ever u want but one thing is for sure; anandtech did fail, not kyle!

[initialised]

Well, I'd have to agree to disagree. ;)

In times like this, I really wish I had a 8800Ultra and 3870 X2 just to test things out myself!I do at work but don't have the time to sort out testing there. Also to make it fair you would have to double up the test with both a good Xfire board (x38/790FX) and a good SLi board (780i) since nVidia cards tend to be on nVidia boards and ATi cards tend to ship on x3x or AMD 7X0 boards.

For comprehensive testing you would need to do:
3870 x2 vs 3870 CF vs 8800 GTS vs 8800 Ultra on Intel x38
3870 x2 vs 3870 8800 Ultra vs 8800 GTX vs 8800 GTS vs 8800 GT SLi on 780i
3870 x2 vs 3870 CF vs 8800 GTS vs 8800 Ultra on AMD 790FX
Finally 3870 x2 CF on x38 vs 8800 GTS/GTX/Ultra (2 & 3) SLi on 780i

For now I will stick with my 8800 GTS, if I was buying GFX next week I would probably get an x2 since I have a Crossfire board and performance is generally at GTX - Ultra beating levels. If I had a SLi board I'd wait for the G92 x2. Hopefully Skull Trail will level the playing field by offering SLi and xfire on a single platform.

Another thing to bear in mind is that AMD/ATi is shipping products before driver support is good while nVidia seems to be getting drivers sorted before shipping but ATi has a routine monthly update while nVidia is ad hoc.

Thats a lot of testing :eek: shame Tom's VGA charts don't keep up any more, they were good for this kind of comparison. Maybe metascore can fill the void by collating hardware review scores like they do for games and films.

[schnulli]

@ the one above me: sorry dude, but u have to learn alot ALOT more than u can investigate, as it seems...

[keiths]

For now I will stick with my 8800 GTS, if I was buying GFX next week I would probably get an x2 since I have a Crossfire board and performance is generally at GTX - Ultra beating levels. If I had a SLi board I'd wait for the G92 x2. Hopefully Skull Trail will level the playing field by offering SLi and xfire on a single platform.


Do you have a CRT monitor or run at settings in which framerate never breaks 60Hz? As LCDs are capped to 60 Hz, and even when set to 75 Hz, it gets down converted to 60 by the monitor.