Graphics Card Reviews

[Willzzz]

I am getting quite annoyed that despite the vast array of different sites offering graphics card reviews, none of them seem to answer the actual questions I have about the cards. There are several flaws that I see repeated again and again.

1) A reliance on average FPS figures
Personally I am much more interested in minimum FPS, that is surely the best indication of whether a game is playable at certain settings? Many sites do provide this, but why isn't it standard practice?

2) Poor comparison with other cards
Too many sites insist on comparing a say a custom GTX 660 with a reference 7870, why don't they compare like with like? Especially when they overclock, saying that an overclocked custom card beats a reference card isn't very interesting, I want to see how it does vs a rival custom OC card.

3) Poor or very limited choice of game benchmarks
This is pretty self explanatory, but even when sites do it right they often let themselves down on overclocked benchmarks. Techpowerup have nearly 20 benchmarks, but when they look at overclocking it's BF3 or nothing.

4) Running games at inappropriate settings
Too often we see games being run at too low settings and thus every card is hitting 90+ FPS which is useless, or we see too high and every card below 30 FPS. Either that or we see benchmarks that are based on less demanding areas of the game.

I was quite impressed by Xbit labs, but still some annoying ommisions:
http://www.xbitlabs.com/articles/gra...gtx-660_9.html

They have a good selection of games, are comparing with a decent 7870, and showing overclocked numbers for every benchmark, with both minimum and average FPS. So far so good.
But why do they jump from no AA to 8xMSAA especially at 1600p, where is the much more sensible 4xMSAA? Where is the comparison with an overclocked 7850?

I know from Techpowerup that even with a hefty 20% overclock the 7850 can't beat the 660 in bang4buck on BF3, but that's BF3, it's not exactly a neutral title. Can any site tell me how a 7850 OC performs vs a GTX 660 OC?

Sadly Hexus seem pretty poor in this regard:
http://hexus.net/tech/reviews/graphi...ex-oc/?page=14

What does this really tell us? Does the 660 still beat a 7850 when both are overclocked, or does the 7850 leapfrog it? Why is it only on one title and an AMD favoured one at that?

Which sites run the best review formats in your opinion?

[ladcrooks]

' Which sites run the best review formats in your opinion? ' - its a minefield! You also got to remember each site has a different test bed! And by going over a couple of review sites, you should be able to sum things up. I do know where your coming from. I myself, wish they would keep overclocking as a separate article for all cards when comparing!

[Fjo]

I'm quite partial to HardwareHeaven. They don't tend to mix stock clock and OC'd cards and their standard settings are 1080p and 4x MSAA (with multi-monitor resolutions and FXAA etc. included quite often too). The reviews aren't usually as extensive and frequent as the usual big sites but i always check with them to see if they've looked at what i'm interested in. They also give you images of how the games look with the hardware in review, rather than others that just show you a vid of the game section on undisclosed kit.

Example (a few pages in):
http://www.hardwareheaven.com/review...lefield-3.html

[Zak33]

Wilzzz.... I promise you, every Hardware site worth it's salt frets over all of your points, every single time.

Choosing HOW to Benchmark is hard.... choosing which rigs to use is harder as they get rebuiilt and remade to keep up with mobo's/cpus/ram upgrades etc, but hardest is working out how to score things.

Then.,... then we have actual phyical issues : such as...."do we have ALL the cards that are relevent RIGHT now AND all the cards that WERE relevent 3 months ago... all in one place?"

Cos these cards dont stay put. They go back to the hardware company, or get used in a rig testing sopmething else is a different town/city etc.

Choosingt the games is hard because in many cases, the Hardware review site makes their own Time Demo file .. ie they don't use the built in test. CoD was a great example of this. Everyone had their own benchmark TimeDemo recorded and , if we're all into this as we prolly are, we all know that even the MAP chosen to record it on makes a difference, as does the time recorded for. Record for 5 minutes and play back as fast as possible on 10 different card at 3 different resolution/IQ settings, and do each one 5 times for an average... and see how you get on. Then try one that's longer. Fresh Win7 install, fresh driver install etc.... it's a HUGE job. Getting it RIGHT and believing your own eyes on weird results is a challenge in itself.

Even the remperature of the labs offices could effect a card that Auto Overclocks... imagine a cold winters day, windows open cos the PC's are grinding away... and nice cool breeze coming into the place. Then, 4 months later and new card, in a different office complex with aircon.... or no windows and just hot blokes with lots of Sprite to keep them calm and cool

One day, the perfect set of results from the perfect set of Benchmarks will occur... and everyone will say "that works... we'll all do it that way..." and then a new hardware mod, or OS or something will come out and blitz it all. A new driver will arrive that "pushes" the scores on that set of benchmarks that everone uses... and it'll all start again.

It's a challenge.. its a GREAT challenge.. but man.. it's a chjallenge.

NOTE: I don't review for HEXUS. But I see and meet the soldiers who do... and man.. they work hard!

[Willzzz]

Yeah, some nice touches on HardwareHeaven, although I find some of those graphs (particularly the BF3 one) a bit hard to read.

I think ladcrooks is right, you just have to read a good selection of reviews and combine the information given to get a good feeling for the merits of each card.

There are lots of sites that I like, but none provide a good enough review on their own.

I like some of the features on Tech Report:
http://techreport.com/review/23527/n...raphics-card/6
In particular the frame latencies and the customisable graphs. (side note: I find it fascinating that the regular 660 has better 95th percentile+ frames than the OC version)

It's hard to ignore techpowerup for the sheer volume of data on the 32 page report.
Anandtech and Hardwarecanucks both provide a very good general picture.

[Willzzz]

Oooh, thanks for the reply Zak

I should clarify that I do of course enjoy the Hexus reviews, I wouldn't be posting here if I didn't enjoy your articles

And yes I can see that it isn't easy, after all not a single site that I could see answered the question of "does the 7850s superior over-clocking enable it to compete with the GTX 660?"
It seemed to be a fairly common question on forums, with many people guessing an answer, so I was surprised no site tried to answer it directly.

[c12038]

This is a polish site for GFX cards try it out

http://www.benchmark.pl/komparator-gpu.html

give benchmark comparisions

also try this one too

http://www.gpureview.com/

Both are good websites

happy reading