• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

The 4870X2 vs 5870 shootout!

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

Neural Net

Member
Joined
Jul 3, 2006
Location
UK
Hi guys,

Sandy and I took the opportunity to compare the performance of the 4870 X2 and 5870, both using the 9.11 drivers. The results were taken from two systems, not identical in parts but running at the same speeds. Specs are as follows:

Systems:

Sandy's:
I7 920 D0 @ 3.78GHz 1.37v XSPC Delta V3
Gigabyte x58-UD4
6GB Corsair DDR3 @1440MHz C8 timings
HD4870x2 cooled with an Accelero
1KW Enermax Galaxy
500GB Samsung Spinpoint F1
Windows 7 Home Premium

Neural Net's:
I7 920 D0 @ 3.78GHz 1.24v Coolermaster V8
Gigabyte x58-UD3R
6GB Patriot Viper DDR3 @1440MHz C8 timings
5870
750w Antec Truepower
500GB Seagate Barracuda
Windows 7 Home Premium

While the systems are comprised of different parts, the biggest difference (apart from the graphics cards) are the motherboards, which at one step up from the UD3R, the UD4's only real benefit is being able to hold more RAM and the cooling solutions are identical. We ran through a series of benchmarks which I'm sure you guys are familiar with. We focused mainly on more demanding titles to really push both cards to their limits. We haven't had an opportunity to test everything under the sun but there is enough to produce some meaningful results - sorry, no Far Cry 2 results!

Noise & Heat:

There is no doubt that a stock 4870X2 is a loud card, the fan speeds up considerably when running games so it is easily heard over the rest of a typical system. It runs hot in stock form too, with temperatures easily approaching the far side of 90°C under load.

Equipped with the Accelero cooler however, the card is practically silent as its 3 fans spin at a far lower RPM, even under load. The trade off is that the card now becomes a 2.5 slot monster, but in real-world terms that only really matters for the select few that want to run two of these in crossfire. Otherwise the cooler does a great job, much better than stock which struggles in high ambient temps and cases with low airflow.

Somewhat surprisingly, the 5870 is a very quiet card, part helped by the more efficient 40nm process. Although we can't provide any figures for this, the 5870 never becomes intrusive and you will only just hear it over 120mm fans running at 800rpm.

Benchmarks:

Now let's get on with the benchmarks! Unless otherwise stated in the charts, all benchmarks were run at 1920x1200 with either 0 or 8x Anti-Aliasing, the reason being that most gamers out there will buy one of these cards to run at this resolution preferably with all the eye candy turned up to its absolute max.

Crysis 1.21
crysisresults.jpg


All benchmarks were run at Very High settings. As we can see here, the 5870 scores a significant victory with average frame rates at playable levels across all benchmarks; the 4870 X2 puts up a good fight but simply can't keep up. Surprisingly, switching on 8x AA resulted a very small performance hit for both cards with the X2 performing particularly well in this regard, albeit at unplayable frame rates.


GTA IV Patch 10.0.0.4
gtaivresults.jpg


GTA IV is another heavy hitter with some great visuals. Arguably the Core i7 is doing a lot of the work here as despite the 5870 scoring another victory both are easily running it at playable frame rates, but only the 5870 manages to break the 60fps barrier. If we were using a Dual core CPU or even a Core 2 Quad at less than 3Ghz, the game would be unplayable at these settings, which are detailed below for anyone that's interested:

Video Mode: 1920 x 1200 (60 Hz)
Texture Quality: High
Texture Filter Quality: Very High
View Distance: 32
Detail Distance: 100
With everything else on Very High/100 and No AA (as GTA IV doesn't support it).


Street Fighter 4 Benchmark
sfivresults.jpg


Results for Street Fighter 4 were close but mixed. As you can see for less demanding games the difference between the two cards is pretty meaningless, you're never going to run into performance issues with either. The X2 manages win one victory here but it is slower overall.


Resident Evil 5 Benchmark
re5results.jpg


Resident Evil 5 can't be seen as anything but a victory for the X2, winning both tests by a noticeable margin, particularly with no AA enabled. As was said earlier however, being a less demanding game both cards easily run it at playable frame rates, but the X2 is certainly more efficient in this instance.


X3: Terran Conflict Benchmark
x3results.jpg


While it may seem easy to run in places, X3 can be very demanding as shown by the minimum fps both these cards achieve. As this is a sandbox game, player actions can have a significant impact on performance, meaning that while it is able to run quite happily a lot of the time, it will begin to stutter when a lot is going on. Both cards are pretty much even here, on average the X2 wins but this seems irrelevant when this victory doesn't come with a higher minimum frame rate. Neither card can keep it above 30 fps to have it somewhat enjoyable when the action gets hectic.


3DMark06
3dmark06results.jpg


Now we move onto benchmark applications, which while they have arguably become less relevant over the years can still shed light on a card's strengths and weaknesses. Here we can see that both cards are CPU bottlenecked at the default resolution, which is no surprise as neither of them were built with 1280x1024 in mind. Pushing up the res and AA however and we see that the 5870 isn't as optimised as the X2 for DX9, let's see how it fares in Vantage.


3DMark Vantage
3dmarkvantageresults.jpg


Here the 5870 is indeed faster that the X2, even at what should be the CPU limited resolution for the Performance benchmark. What we see is nothing we haven't seen already - the 5870 is undoubtedly the faster card, but it is not miles ahead.


Unigine Heaven Benchmark (DX10)
unigineheaven.jpg


Finally we have the Heaven Benchmark, which next to Crysis is perhaps the most demanding test we've put these cards through. With 8x AA the 5870 manages to pull ahead, but not so much that you would notice watching it.



Conclusion:

It's safe to say that the 5870 is definitely the faster card out of the two but it only really comes into its own with the more demanding titles, which bodes well for it being a decent DX11 card when more DX11 games come out over the next 12 months.

Perhaps the most disappointing aspect of the 5870 is the lack of overclocking headroom provided through CCC. You can flash the BIOS and use MSI Afterburner, but the highest CCC can push is 900Mhz, a paltry increase over the stock speed of 850Mhz and not everyone will be comfortable with flashing their new bit of kit.

Before the X2 was discontinued it was reaching prices as low as £240. If both were out today the 5870 would demand a £60-80 premium, but there are additional benefits to owning this card. Obviously it supports DX11, it also runs cooler and draws far less power, the fact that on early drivers it manages to beat an X2 is also impressive: Harking back to the first month of the X2's release, its initial drivers were buggy, inconsistent and clearly lacking in performance. The 5870 has been a much smoother ride.
 
Interesting review, but a few questions:

What setting did you use for Catalyst AI? While this handles driver optimizations, it also handles certain aspects of Crossfire which would obviously come into play with the 4870x2 but not the 5870.

Also, what filter did you use for the 8xAA tests, Box or Edge-Detect? Edge-detect is really the main perk of these cards but performance has always been vastly different than using the Box filter.

I still think it would have been better to swap the cards and make sure that there really aren't any other differences being caused by the rest of the system. I know they are very similar but there are countless things that could impact performance.
 
Interesting review, but a few questions:

What setting did you use for Catalyst AI? While this handles driver optimizations, it also handles certain aspects of Crossfire which would obviously come into play with the 4870x2 but not the 5870.

Also, what filter did you use for the 8xAA tests, Box or Edge-Detect? Edge-detect is really the main perk of these cards but performance has always been vastly different than using the Box filter.

I still think it would have been better to swap the cards and make sure that there really aren't any other differences being caused by the rest of the system. I know they are very similar but there are countless things that could impact performance.

The idea of these tests were to be realistic, not overly technical as the results can be misleading. Sorry I did not state; the AA is the default of what the game uses - this is the fairest way to test as it's what most gamers will use.

Swapping the cards into the same system wasn't an option as Sandy and I live more than 500 miles apart. As for countless things impacting performance, that's quite an exaggeration, this test was done to show a realistic indication of performance you can expect.

very nice review man

Thanks! Hopefully Sandy will pop up in the thread because he deserves some props too. :D

very nice
the 5870 looks like it was indeed driver limited, rather than bandwidth limited

That's definitely how it appears. I may also do some more tests with the 9.12 drivers to see if they yield any increase in performance.

Wow. great writeup! I've been curious myself now for a little while. Guess this pretty much settles it, eh?

Thanks! :D I was keen to test this since a few reviews have had the X2 beating the 5870, which seemed at odds with my observations as I upgraded from one myself (I couldn't test both myself unfortunately as I sold the X2 a few weeks before my 5870).
 
Last edited:
The idea of these tests were to be realistic, not overly technical as the results can be misleading. Sorry I did not state; the AA is the default of what the game uses - this is the fairest way to test as it's what most gamers will use.

Using the edge-detect filter does not mean manually specifying AA in the control panel. You can use the filter and still allow AA levels to be set in-game. The Difference is that when you use Edge-Detect it uses stream processors on the card for some of the AA calculations and 4x in-game AA would become 12x and 8x in-game AA would become 24x. It's a nice option given that both cards have 1600 stream processors.

Swapping the cards into the same system wasn't an option as Sandy and I live more than 500 miles apart. As for countless things impacting performance, that's quite an exaggeration, this test was done to show a realistic indication of performance you can expect.

I'm not trying to say you're results aren't relevant to a certain degree, but there isn't a single review site on the planet that would benchmark cards across different systems even if they are mostly similar. Yes there are countless things that can impact performance, perhaps not that would account for a huge difference, but then again there isn't a huge difference between the cards in most of the benchmarks. In many cases they are only a few fps apart and that is certainly well within the the range of what could be caused by minor system variables. TBH the fact that you are 500 miles apart having different people run the benchmarks only exacerbates the issue compared to if the systems were built at the same time, by the same person, running side by side (for example), which is what I thought at first.
 
Great review...! Its a Granny Smith apple to a Red Delicious, but hey, its still apples (and MUCH better than some testing I have seen around here lately). I cant imagine switching the cards between system wouldnt yield anymore than a negligable difference and across most games the difference between these cards is always more than negligable from the reviews I read.

Im hoping the GPU Content Edititor sees this and talks to you guys about a Front Page article..........!
 
Thanks for the review, I love the charts!
Im really bumming about the 900mhz limit, I dont want to flash a $400 card when I get it! I wonder if GPU tool will allow you to go past that limit? I know it does with my 4890 1GHz barrier.
 
Using the edge-detect filter does not mean manually specifying AA in the control panel. You can use the filter and still allow AA levels to be set in-game. The Difference is that when you use Edge-Detect it uses stream processors on the card for some of the AA calculations and 4x in-game AA would become 12x and 8x in-game AA would become 24x. It's a nice option given that both cards have 1600 stream processors.



I'm not trying to say you're results aren't relevant to a certain degree, but there isn't a single review site on the planet that would benchmark cards across different systems even if they are mostly similar. Yes there are countless things that can impact performance, perhaps not that would account for a huge difference, but then again there isn't a huge difference between the cards in most of the benchmarks. In many cases they are only a few fps apart and that is certainly well within the the range of what could be caused by minor system variables. TBH the fact that you are 500 miles apart having different people run the benchmarks only exacerbates the issue compared to if the systems were built at the same time, by the same person, running side by side (for example), which is what I thought at first.

Interesting stuff about the AA, thanks for the info, I'll look into that, although anything over 8x is pointless.

I understand your problem with the differences, but this was never meant to be a test to rival that of Anandtech, Hardocp or whatever. This is intended as a "real-world" test for users who aren't big into the minor tweaks that yield negligible performance differences, can see can an idea of what to expect. This is also the reason why we kept the i7s below 4Ghz, even a large difference such as one lower CAS latency timings on the Ram will make little to no difference on speed, as demonstrated by Anandtech here.

Great review...! Its a Granny Smith apple to a Red Delicious, but hey, its still apples (and MUCH better than some testing I have seen around here lately). I cant imagine switching the cards between system wouldnt yield anymore than a negligable difference and across most games the difference between these cards is always more than negligable from the reviews I read.

Im hoping the GPU Content Edititor sees this and talks to you guys about a Front Page article..........!

Thanks Earthdog, it's certainly Granny Smith vs Red Delicious, but I think the Crysis results speak loud enough on how much horsepower the 5870 has over the X2 (at least when certain games are in a position to exploit it).

Thanks for the review, I love the charts!
Im really bumming about the 900mhz limit, I dont want to flash a $400 card when I get it! I wonder if GPU tool will allow you to go past that limit? I know it does with my 4890 1GHz barrier.

Exactly the way I see it. It would be great if CCC had a limit of 1Ghz, that would probably be just about right for most users that want to overclock their card. Any further seems to require better-than-stock cooling. :)
 
Don't worry Jokers... i'm never far away... :D

Sorry been working a lot recently since its the run up to christmas!!

Most of the credit goes to Neural as he is the one who did the final write up... i just ran some benchmarks!!

Cheers for all the comments guys... as for AA effects i left them default in CCC (let application decide) i never bothered tweaking it becuase everything looks perfectly fine to me....

Thanks for all the positive feedback... its been a lot of fun doing it, we were hoping to get some card overclocking results in too but the x2 isn't a great clocker and im scared about bricking it if i try and flash the bios as im poor at the moment and dont fancy buying another graphics card just now!!

If anyone has any requests for any benchmarks or specific settings we would be more than happy to run them...

:D
 
Don't worry Jokers... i'm never far away... :D

Sorry been working a lot recently since its the run up to christmas!!

Most of the credit goes to Neural as he is the one who did the final write up... i just ran some benchmarks!!

Cheers for all the comments guys... as for AA effects i left them default in CCC (let application decide) i never bothered tweaking it becuase everything looks perfectly fine to me....

Thanks for all the positive feedback... its been a lot of fun doing it, we were hoping to get some card overclocking results in too but the x2 isn't a great clocker and im scared about bricking it if i try and flash the bios as im poor at the moment and dont fancy buying another graphics card just now!!

If anyone has any requests for any benchmarks or specific settings we would be more than happy to run them...

:D

Definitely, Dirt 2 could be interesting? :)
 
RE: Dirt2 I would like to see the difference in in fps as far as DX9 vs DX11 for the 5870...

To test that would I run the game at High settings to run in DX9 and Ultra to run in DX11?

Thanks to both of you for the review. Good work.

Smells like a front page article to me :)

Nice job guys :)

Thanks guys! :bday:

oh yeh i forgot about dirt2!! Iv got the demo on steam so i will run it at some point... :D

Good stuff, I'll try to run it in DX9 and DX11, see how it performs.
 
I would imagine, yes.

Some screenshots of the differences would be pretty swell as well (especially considering the discussion in that other thread).
 
I would imagine, yes.

Some screenshots of the differences would be pretty swell as well (especially considering the discussion in that other thread).

I'll try to, but screenshots don't really to it justice for some reason. I'll see if I can use FRAPS to get a decent video of both.
 
Back