***The OcUK Graphics Card Performance & Review Thread***

[Malazan]

Why overclock some cards and not others?

Strange when you were hoping to show direct comparisons.

 

[FrenchTart]

Better to not bother at all than to post something misleading or incorrect as these numbers appear to be. Benchmarks from a company trying to sell the hardware in question should always be looked upon with suspicion.

The post was made, and feedback was given. This is how forums work.

Exactly.

 

[icey_haj]

Why overclock some cards and not others?

Why OC the 4870 but not the 260? and why use the older 192- 260 instead of the new 216-260?

Unless my eyes are deceiving me, then from what I can see, he did use the '216' 260.

 

[Malazan]

Unless my eyes are deceiving me, then from what I can see, he did use the '216' 260.

Yes, you're right aopologies to the OP.

Still doesn't answer why some cards got OC'ed and others didn't.

 

[icey_haj]

Yes, you're right aopologies to the OP.

Still doesn't answer why some cards got OC'ed and others didn't.

Apologise to me as well.

To be fair, the guy did overclock most of the cards, so you do get a general feel for how the cards perform.

 

[Rroff]

I think they are quite a good representation of the overall picture with these cards...

If I was doing it I'd have gone for 1920x res, 4x FSAA (too much driver "cheating" goes on at x8 to give a consistant picture) and thrown in GRID and COD4.

 

[arc@css]

Close between the 295gtx and 4870X2. Glad I've got one of them

 

[Malazan]

Apologise to me as well.

To be fair, the guy did overclock most of the cards, so you do get a general feel for how the cards perform.

Yep, a good general feel. As you can see from this thread many just want their precise requirements tested, and aren't happy with a general picture - sadly

 

[death_looms]

Thanks rjkoneill, between this and your build log its pretty good stuff

 

[FrenchTart]

Yep, a good general feel. As you can see from this thread many just want their precise requirements tested, and aren't happy with a general picture - sadly

I'm happy with the idea of 4870 512 < 4870 1GB & GTX260 216 < GTX280. I paid £140 for my 4870 1GB a while back so I'm more than happy with the purchase, even if a GTX260 216 turns out to be slightly faster.

What I don't agree with is benchmarks that don't really give a real world performance analysis. No AA is not a situation that I would find myself in anytime I play a modern game, aside from a few *very* rare cases. Not to mention the rumours that fly around about the ATI cards having slightly better performance with AA. They might do, or they might not - why not put the question to rest by giving us benchmarks that use a setting that all of us will use (and have a large impact on performance on BOTH sets of cards)?

 

[psychas]

nice benches 260gtx 55 nm 216 neck in neck with 4870 1 gig well at least in those games

 

[londonbairn]

For me it iterates that the 295 is not worth the outlay over a X2, and as many have shown in situations the X2 is faster.

If the 295 was worth the extra I wouldn't hesitate, but it simply isnt

 

[w3bbo]

Nice little round up m8, appreciate the efforts.

Gotta say though I agree with some of the comments here, the benchies are a little misleading. I can't see the point of benching a 4870x2 and GTX295 at any res below 1900x1200, especially with no AA. The cards are simply not aimed at the low end of the market. Sorry if that offends anyone who has less than a 24" screen but tbh, wtf were you thinking buying a ultra high end card for a medium size screen for?

Those cards at least were made for High resolutions and AA. Run them on a 30" with at least 4xAA/AF and then you will see why the x2 and 295 cost so much. They need room to stretch their 'legs' as it were.

 

[rjk]

Include some overclocked GTX260-216 results if you want to be a fair comparision, unless you don't want users at OCUK you know the GTX260 is a greater overclocker.

And for god sake add some AA in Far Cry 2.

i benched all the cards i had access to - obviously i could only use the test/sample cards we had available

cant be ripping stock out of the warehouse now!!!

Thanks for the review OcUK.
I think that a Vantage graph with Physx turned off would be interesting though.

i really did consider turning physx off but the tech team discussed it an all agreed that physx is a feature of the card and should therefore be used

Thanks for this.. However:

No AA?

Why would anyone with one of these cards play without AA? Obviously Crysis tends to get dogged down by AA but then again, that is just one of the many reasons to not use Crysis as a benchmarking tool.

4ghz i7 is a "generic setup that a lot of customers tend to use"?

I think a 4ghz i7 setup is perhaps a *little* more powerful than most people's systems. Even those who have overclocked i7 rigs will likely have them running a tad bit slower than 4ghz. Not to mention all the people on 3.2-3.6ghz Kentsfield and below.


I'm trying not to be overly-negative here but the reviewing methodology is rather questionable.

ok - perhaps i was being a little brash

the reason i sued 1680*1050 was because the ocuk forum benchmarks use that res
it is a common res that everyone can relate to

the whole point of the results is to give newcomers aplace to check/ask questions on what the best card for them is

the results, whilst not being performed to everyones requirements - give a clear indication of the performance of each card
i had limited time to bench each card - that meant each card got one run at each bench

i would have personally liked to include another benchmark
perhaps world in conflict or perhaps a time demo of cod4/waw [although i am not aware of any standardised benchmark tools for those games]

i could have run the test rig at stock speed

but that would just be...

boring

i have tried my best to be unbiased and that meant giving each card the best possible platform to run on

i would have liked to do crossfire/sli results - but it meant i would need duplicates of each card and we had one sample of each

 

[rjk]

Nice little round up m8, appreciate the efforts.

Gotta say though I agree with some of the comments here, the benchies are a little misleading. I can't see the point of benching a 4870x2 and GTX295 at any res below 1900x1200, especially with no AA. The cards are simply not aimed at the low end of the market. Sorry if that offends anyone who has less than a 24" screen but tbh, wtf were you thinking buying a ultra high end card for a medium size screen for?

Those cards at least were made for High resolutions and AA. Run them on a 30" with at least 4xAA/AF and then you will see why the x2 and 295 cost so much. They need room to stretch their 'legs' as it were.

understood
however they do outperform all the other cards so you get the idea
i will do a 1900*1200 bench on the 295 and X2 on my next saturday when i have access to the large screen

 

[gary996]

understood
however they do outperform all the other cards so you get the idea
i will do a 1900*1200 bench on the 295 and X2 on my next saturday when i have access to the large screen

That would be nice and maybe some 3 core crossfire and 4 core for comparason

Would like to see if its worth buying another 4870 (either single core 1gig or another x2)

 

[Metalface Mark]

I would like to see a 4850 1gb in that line up to see how it performs against the bigger boys (as this is what i have just now, and have been considering upgrading it).

 

[layte]

i really did consider turning physx off but the tech team discussed it an all agreed that physx is a feature of the card and should therefore be used

Are PhysX hardware enabled results allowed by Futuremark?

From what I understand of the license that covers the professional version of the software, you are not allowed to publish results that use non sanctioned drivers or performance enhancements (like PhysX).

 

[londonbairn]

If you re-run it, turn PhysX off, it's not exactly being used in any decent games

 

[w3bbo]

Are PhysX hardware enabled results allowed by Futuremark?

From what I understand of the license that covers the professional version of the software, you are not allowed to publish results that use non sanctioned drivers or performance enhancements (like PhysX).

HWbot does not allow.