Skip to main content

Radeon HD 4870 X2: Four Cards Compared

Performance Gain From CPU And Graphics Driver

In the following tables, we compare some combinations and test cards directly against each other. After the chip name and the memory size of the card you’ll find the driver version. This is followed by the abbreviation for the CPU: X68 = Core 2 Duo X6800 Extreme Edition at 2.93 GHz (comparable to E6750 and E8200) or 3.47 GHz; E85 = Core 2 Duo E8500 at 3.8 GHz; QX = Core 2 Quad QX6850 at 3.67 GHz. Overclocked graphics cards are labeled with an OC.

If you’ve wondered why only the two X2 cards from HIS and Sapphire were overclocked by us, not the ones from MSI and Asus, you’ll get the answer with the following results. Overall, the overclocking results in a performance gain of no more than 1%, and for such an insignificant performance boost, you should not risk losing the warranty on your $500 card.

Performance Gain HD 4870 X2fpsPercent
Sapphire HD 4870 X2 (2x1024 MB) OC 8.10 QX@3.674752.1100.0
HIS HD 4870 X2 (2x1024 MB) 8.10 QX@3.674783.4100.7
MSI HD 4870 X2 (2x1024 MB) OC 8.10 QX@3.674814.5101.3
Asus HD 4870 X2 (2x1024 MB) Top OC 8.10 QX@3.674823.8101.5
HIS HD 4870 X2 (2x1024 MB) OC 8.10 QX@3.674828.8101.6
Radeon HD 4870 X2 4CF (4x1024 MB) 8.10 QX@3.675033.3105.9

The next table shows the single-chip cards, tested with both the standard CPU speeds and the overclocked quad core CPU. The X6800 Extreme Edition at 2.93 GHz corresponds to an E6750 or E8200. When replaced by an overclocked quad-core QX6850, you’ll get a frame rate 9% to 16% higher.

Performance gain, single-chip cardsfpspercent
GeForce GTX 280 (1024 MB) 177.39 X68@2.933765.6100.0
GeForce GTX 280 (1024 MB) 180.42 QX@3.674367.3116.0
GeForce GTX 260 (896 MB) 177.39 X68@2.933464.8100.0
GeForce GTX 260 (896 MB) 180.42 QX@3.673967.0114.5
GeForce 9800 GTX+ (512 MB) 180.42 X68@2.933246.6100.0
GeForce 9800 GTX+ (512 MB) 180.42 QX@3.673578.6110.2
Radeon HD 4870 (512 MB) 8.6 X68@2.933355.7100.0
Radeon HD 4870 (512 MB) 8.10 QX@3.673779.7112.6
Radeon HD 4850 (512 MB) 8.6 X68@2.932965.6100.0
Radeon HD 4850 (512 MB) 8.10 QX@3.673240.2109.3

Here SLI and CrossFire combinations are improved with more CPU power—the gain in frame rate is significantly higher, ranging from 31% to 38%. When using dual-chip combinations, you can not only increase the relative performance gains of the faster CPU, you can even double performance.

Performance Gain SLI and CrossFirefpsPercent
GeForce GTX 280 SLI (1024 MB) 177.39 X68@2.933522.0100.0
GeForce GTX 280 SLI (1024 MB) 180.42 QX@3.674632.0131.5
GeForce GTX 260 SLI (896 MB) 177.39 X68@2.933481.1100.0
GeForce GTX 260 SLI (896 MB) 180.42 QX@3.674569.0131.3
GeForce 9800 GX2 (2x512 MB) 175.16 X68@2.933351.4100.0
GeForce 9800 GX2 (2x512 MB) 180.42 QX@3.674616.1137.7
Radeon HD 4870 CF (512 MB) 8.6 X68@2.933482.9100.0
Radeon HD 4870 CF (512 MB) 8.10 QX@3.674825.7138.6
Radeon HD 4850 CF (512 MB) 8.6 X68@2.933241.4100.0
Radeon HD 4850 CF (512 MB) 8.10 QX@3.674433.0136.8
Gain from single-card to CrossFirefpspercent
Radeon HD 4870 (512 MB) 8.6 X68@2.933355.7100.0
Radeon HD 4870 CF (512 MB) 8.6 X68@2.933482.9103.8
Radeon HD 4870 (512 MB) 8.10 QX@3.673779.7100.0
Radeon HD 4870 CF (512 MB) 8.10 QX@3.674825.7127.7
  • Not only do we have four super-fast Radeon HD 4870 X2s to test, but also a list of 31 other graphics configurations including CrossFire and SLI setups. If you're in the market for AMD's fastest card available, you'll want to see this.

    Radeon HD 4870 X2: Four Cards Compared : Read more
    Reply
  • neiroatopelcc
    "Because of accessories and price, Sapphire is our best-buy recommendation."
    One slight warning about sapphire though. If you have problems, don't expect their support team to help you before you've solved the problem yourself!

    I made a ticket regarding some issues with my 4870 on august 7th, and received a reply on the 26th of september! That's 46 days to address an error they simply stated would go away with a bios upgrade from their homepage!

    As for the article, I actually liked the detailed driver errors they encountered. Not that I liked the errors themselves, but I liked them being explained. Usually you just read 'after spending some hours resolving driver errors ....' without getting any wiser.
    Reply
  • Pei-chen
    Wow, AMD cards consume power like a Detroit SUV. I like Nvidia GTX 2xx series’ Toyota Prius like efficiency at idle.
    Reply
  • ilovebarny
    Why didnt they use the GTX260 Core 216? its like way better than the regular GTX260. And i just read yesterday that Nvidia was only going to make GTX260 Core 216 now. http://www.fudzilla.com/index.php?option=com_content&task=view&id=10497&Itemid=1
    Reply
  • ilovebarny
    And i wish they had SLI'ed the 9800GTX+.
    Reply
  • Chizops
    Why didn't they try using core i7 (extreme maybe)
    Reply
  • enforcer22
    Pei-chenWow, AMD cards consume power like a Detroit SUV. I like Nvidia GTX 2xx series’ Toyota Prius like efficiency at idle.

    Hmm your right. Power house vs crippled mouse.. yeah your analagy sucked im sure mine did to but all i saw from what you typed was i like weak stuff dont give me more power.
    Reply
  • bdollar
    seems to me if you are going to be comparing the highest end cards and even crossfire them for 4x you would have the highest resolution as one of the options. i would think people considering going x2 in crossfire would consider a 30" screen.

    don't get me wrong, i liked the article but would have liked to have seen the resolution spectrum hit the top.
    Reply
  • It's nice to see the 9800GX2 included in the tests. I was considering the 4870x2 due to all the rave reviews but they never had the comparison like this against my current 9800GX2. I won't be getting new card anytime soon it seems. Thanks.
    Reply
  • cleeve
    Wow, tons of info there, Tino. Nicely done!
    Reply