Hardware Sapphire ATI Radeon HD 4870 X2

[Tim S]

http://www.bit-tech.net/hardware/200...n-hd-4870-x2/1

After claiming the single card performance crown, the 4870 X2 is proving immensely popular amongst high end enthusiasts. We take a look at Sapphire's version of this graphics monster, and try our hand at overclocking it - oven mitts on standby!

[wuyanxu]

nothing is going to change the fact it's multi-GPU and have potential driver problem and it may not perform as well as it should be.

great product for those who are willing to go multi-GPU for the average FPS performance, but as article said, the value of it really isn't very good.

[Tyrmot]

But then of course, the value of any new bit of kit is never very good at launch is it? It's just to tempt those with the cash to burn and don't care what they pay... On the plus side, people like that are subsidising it for the rest of us! (thanks!)

[Kris]

Hey,

just a nice review, however I don't really see the point of testing a card like 4870 X2, or even GTX 280/260 in a game like Half Life 2 - it's so sure that they'll deliver absolutely stellar performance anyway... I'd really go for something diffeent, that's more suitable to cards of this calibre. This would also give the reivew some other value like "how will this card run t<insert game here>"?
Also, you keep on mentioning "maximum settings", in the Crysis performance tabs, yet say that you've tested with everything set to "high". Everyone knows, that Crysis also has a Very High setting That's of course true only for DX10 testing...

But cheers for the review ^^

[Baz]

We're actually accessing which new titles to test graphics cards on right now. There are so many graphically demanding titles coming up, if you have any suggestions of benchmarks you'd like to see just let us know.

[badders]

GRID's quite hard on my 8800GT - I get about 34FPS at 1920x1200 on 4xAA - it slows a little when there's lots of smoke, but I'm sure that would be a nice benchmark.

[The boy 4rm oz]

I can max GRID out at 1600x1200 res and still get 65-80FPS constant with my 8800GTX.

[tech9]

hi all, i was wondering why this card wasn't water cooled or even had an better aftermarket cooling instead ? as i understand it better cooling would give u a bit more head room and could improve overall performance.

woo finally made me an account here love this site

[wuyanxu]

Originally Posted by The boy 4rm oz

I can max GRID out at 1600x1200 res and still get 65-80FPS constant with my 8800GTX.

same, 1680x1050, CS16xQAA, not sure about FPS, but always silky smooth.

i'd like to see more Dx10 tests, World in Conflict Dx10 at max is a must, Devil May Cry 4 Dx10 at max?

[mrb_no1]

Originally Posted by Kris

Hey,

just a nice review, however I don't really see the point of testing a card like 4870 X2, or even GTX 280/260 in a game like Half Life 2 - it's so sure that they'll deliver absolutely stellar performance anyway... I'd really go for something diffeent, that's more suitable to cards of this calibre. This would also give the reivew some other value like "how will this card run t<insert game here>"?
Also, you keep on mentioning "maximum settings", in the Crysis performance tabs, yet say that you've tested with everything set to "high". Everyone knows, that Crysis also has a Very High setting That's of course true only for DX10 testing...

But cheers for the review ^^

Firstly, I'm not the tester but my impression was that half life 2 was included for 2 reasons;
1: It is a common title that many geeks/games enthusiasts will have played so can relate to it.
2: It is one of the few games that is optimised for ATI cards. Remember the "Nvidia, the way its meant to be played" that seems to appear on all games these days. Well i dont know how much effect it does have when hard and software developers get into bed with one another but maybe its chosen because its one of the few games that is for ATI cards..."ATI, because our drivers are better than Nvidia!"

Secondly, my guess on the crysis question is probably because when testing lower end/older cards the bit-tech guru's werent able to run the 3870 and the like at very high settings, thus high settings was the best they could do and for the purpose of having benchmarks that might be useful to people they kept using 'high settings' rather than switching between the two, confusing everyone and removing any usefulness of their benchmarks.

Arguably you might ask Harry, Tim or Rich to include some benchmarks at very high on the top end cards for those of you wanting to know how large your e-peen might really be if you purchased the card. Other than that, there is no point imo.

On a different note, i dont know why Quake Total war, or whatever the latest quake is as the games are soo poor that your benchmarks often tally over 100fps, at which point you arent looking for more fps as its pretty much pointless. Saying that i'm not entirely sure what i would use in their place. I'm a rts fan so sup com/forged alliance or company of heroes until starcraft 2 comes out maybe (note:did not like that World in Conflict so much but most rts seem to fall short since playing sup com.) Maybe use the new unreal tournament or something, i dont know there, i'm stuck playing bf2 and bf 2142 because i love 'em atm. Maybe take a poll and get a feeling for what people might want to see in benchmarks just to give you an idea, or a group that arent being catered for when you review a card, just a thought.

peace

fatman

[Tim S]

It raises an interesting point - we want to keep the games relevant to you, the readers, so if you have any suggestions please don't hesitate to make them! We'll have a look into DMC4 - there's a copy lying around somewhere.

[Bauul]

As overpriced and dual-card-unreliable as it is, it's nice to ATI kicking ass again, especially after they had said they were submitting the higher end products to nVidia.

[[USRF]Obiwan]

Despite the stellar performance the card has. On pure logical basis, I will probably go for the GTX280 when globally comparing: price, power consumption (both idle and load) and performance.

[Goty]

Quick question: Why the different driver revisions between the preview 4870X2 and the Sapphire board?

[frontline]

Nice review. I think the games used as benchmarks are generally ok, i'm more of a first person shooter nut, so prefer it if most of the games used are of this genre. Half Life 2 Ep 2 is worth having in to see how cards perform on the OB engine. I'd also like to see some dedicated multiplayer benchmarks for games like COD 4, TF2 and Day of Defeat Source, as these often differ significantly from the Single player games using the same engine (especially when you factor in 16-20 players + on a server).

Forthcoming releases look like they will be good for benchmarking, Fallout 3, Far Cry 2, Stalker Clear Sky etc.

[Tim S]

Originally Posted by Goty

Quick question: Why the different driver revisions between the preview 4870X2 and the Sapphire board?

They're run at the same clock speeds and it was more interesting to see how 8.8 worked out than to retest two cards and get exactly the same results for both.

[wuyanxu]

Originally Posted by Star*Dagger

Anyone considering a lesser card should think about where the state of games will be in a year or 2.

i assure you, in less than 1 year time, a faster, cooler graphics card will come out.

4870 is based on 3870, it's not a very large improvement. so is gtx280 based on G92 which is a tweaked+smaller G80.
only when they take the risk and develop really new architecture we can see a large improvement.

[Tim S]

Originally Posted by wuyanxu

4870 is based on 3870, it's not a very large improvement.

It's a massive improvement and there are many things that changed - it's easier to list what didn't change.

The only thing that didn't really change was the shader core - the only update there was the local data store and re-association of texture units with shader units. 3870's texture units were completely decoupled from the shader units and just about all of the cache was global.

[mrb_no1]

Originally Posted by wuyanxu

i assure you, in less than 1 year time, a faster, cooler graphics card will come out.

personally, thats the wrong attitude for someone with an interest in the current technology market. Its like me saying i wont go quad core because within 12 months they'll have doubled the core# and halfed the wattage, or something along those lines....of course they will have done something like that!

Personally, when you keep uptodate with techs, you buy to have the best for buck, or the best period! Not saying i'll wait 12 months for something better. When i got into Pc's 6 years ago, i might aswell as bitten the bullet and said i'll wait 6 years, and then buy something really awesome :P Apologies, i know thats slightly obtuse of me and not particularly funny, but the original quote struck a nerve in my brain

peace

fatman

[wuyanxu]

Originally Posted by mrb_no1

personally, thats the wrong attitude for someone with an interest in the current technology market.
... [snip]..
peace

that is correct in the sense you mentioned it. but the point im trying to get across is that multi-GPU is power-inefficient, hot and doesn't perform as well as single GPU.

GT200 55nm will be out later, even if it's a bit slower than 4870x2, its better idle power consumption and performance will make that a better purchase. how many people went and bought GTX280 as soon as it wasreleased for the fact that it's the best? most people waited and went for the sensible high-end of 4870 for its price/performance, or gtx280's price drop. now 4870x2 is released, and IMHO it's better to wait for nVidia's response before buying the best of the best.

look at 3870x2, and see how 4870 trumps it easily. multi-GPU is not there. yet.