Final Words

The two new additions to the test suite, Age of Conan and Race Driver GRID certainly beef up the portfolio of advantages AMD has in the current generation. Not only that, but we got quite an interesting surprise with GRID. Yes, even with R700 the menu screens were slow, jerky and painful at 2560x1600, but gameplay with R700 was much improved. We will still see the upper limit at twice the performance of a single RV770, but the fact that R700 looks to offer at least the potential for much better scaling at the high end (where the card will be marketed) than two card solutions is an advantage, even if it only comes from the additional RAM. We will need to look at CrossFire again when the 1GB 4870 is available to get a more apples to apples comparison here.

We don't have any solid evidence in this article of the GPU to GPU communication hardware making a real difference in scaling, but we will be interested to see if that changes by the time R700 hits store shelves. We can say that, for those who want to game at the extremely high end, 4870 X2 with it's 2GB of on board RAM will be a more consistent solution than 2x 512MB 4870 cards in CrossFire, as evidenced by our Race Driver GRID test.

But as we've said many times before, the success of CrossFire is in the consistency of it's performance. We absolutely need to see AMD put everything they have into making all games past, present, and future scale well with CrossFire. There needs to not be even the inkling of a question that CrossFire might not improve the performance of a game. Until then it is very risky for AMD to put all of its hope for the high end business into a multi-GPU solution.

Maybe they've done it. Maybe their driver team, by the time 4870 X2 launches will have improved driver support to where it needs to be. Maybe the changes to the R700 hardware will be enough to fill in the gaps and bring performance up closer to the theoretical limit for many more games. The information we have here shows the incredible potential R700 has, but it will be absolutely necessary to wait until launch day to see if the execution behind the hardware has been enough to realize this potential across the board.

This isn't a case where the quality of the hardware, but rather the quality of the driver and programming style of modern game developers will be key in delivering value. For the launch of this card we will be looking at as large a cross section of games as we can, as the importance of broad testing has never been more clear than with AMD's first part after their declared strategy of using single card multi-GPU solutions to compete in the high end space. We will need to see more examples of improved single card multi-GPU performance over CrossFire, as well as improvements in CrossFire performance in general if AMD is to be taken seriously going forward in the high end space.

The 4870 X2 will be AMD's proving ground. This preview shows what might be, what could be ... but we must wait for final hardware and final drivers before we can honestly evaluate the card for what it is. Let us hope AMD knows how important having pervasive compatibility really is for this launch.

Power Heat and Noise
Comments Locked

55 Comments

View All Comments

  • ZootyGray - Monday, July 14, 2008 - link

    I look forward to Anandtech testing. Other sites - some good, some biased, some illiterate, etc.

    I describe Anandtech testing as thorough, fearless, accurate, brutal, relentless, and uncompromising. And few typos/language issues. And this results in far fewer fanboy/junk comments as well.

    If I want the real truth, I come here, and I read it slowly - it's like a feast.

    Thanks Anand and company.
  • jamesbond007 - Monday, July 14, 2008 - link

    On page 7:

    It is likely that the extra 512MB of RAM available to each GPU has significantly impacted perforamnce since we are testing with all the options cranked up and 4xAA.

    Anyways, I can't wait to see tests with 2x 4780X2 going! I also would have liked to see the 4780 (single) in all of the tests, but I figured the card would get roughly half of the performance of the new X2. No big deal.

    The power consumption diagrams are making me realize why the 1000W+ PSUs exist. :) Good gravy, boy! A guy would have to pick up an extra shift or job just to pay the electric bill for the new cards coming out these days. Then again, most guys who buy this card likely live in their parent's basement. :p

    Just joking, of course!

    Game on, ATi! Nicely done.
    ~Travis
  • hooflung - Monday, July 14, 2008 - link

    This can't be a 1G version Derek. This has to be the 2G version everyone else has gotten. Can you confirm that?

    Nvidia... meet face. Age of Conan on this card is simply amazing. I think I'll trade in my 3850 512 for the X2 once it comes out. Something I said I wouldn't do for a while.
  • yacoub - Monday, July 14, 2008 - link

    Kind of annoying you didn't include the ATI Radeon HD 4870 card in every test, as that's the most expensive one most people will care about. The $500+ stuff is ridiculous. The $299 card is a bit more worth reading about.

    The other thing is how you have to test at 2560x now just to show the biggest differences. Kinda shows that something like an 8800GT is still fine for 1680x and even 1920x for most games.
  • 7Enigma - Monday, July 14, 2008 - link

    Let's see this is a preview about AMD's new high-end card that will directly compete with the 280. Of course most of us will never buy it, it doesn't mean it isn't important. There is already a review about the $300 card, it came out already with an indepth review.

    Every comparison has a nice graph and chart at the bottom showing performance at several different resolutions. If you are too lazy to THINK instead of just looking at the pretty color bars and seeing which one is longer, that's on you.

    Really these complaints are just getting rediculous. If they failed to review the cheaper cards or only showed a single resolution/setting (cough...cough..(H)) then I could see your point. Fact of the matter is people shelling out >$500 for a graphics card probably DO have the screens with the resolutions in the bar graphs.
  • Alexstarfire - Tuesday, July 15, 2008 - link

    No offense to you, but it would seem to me that they were just a bit lazy on making the extra lines in the graph. I know that at Guru3D that they include different resolutions in the graphs.

    To me, it's not that I'm lazy, but it's a lot more difficult to compare different numbers across a range of cards and settings than it is to see which one is above and/or below others on a graph.
  • FXi - Monday, July 14, 2008 - link

    Put 2x 4870x2's in CF on an Intel X48 or X58 and you can kiss BOTH Nvidia's chipsets AND GPU's goodbye.

    Advantages:

    Crossfire is fully supported on Intel chipsets with NO bridge chip required.

    Crossfire won't be denied to function on Intel chipsets - the current bridge chip on Skulltrail has had 280 SLI denied by Nvidia, because they don't "feel" like enabling it.

    Dual screens on Crossfire? No problem. Crossfire doesn't artificially limit dual screen use to just their workstation cards like Quadro.

    4870x2 is already managing to beat SLI 280's in some places. Driver improvements will only make it a stronger beating in the future.

    Don't you want a company that is committed to working WITH Intel, where it makes sense, rather than fighting them?

    Support for DX 10.1 "just in case" it should end up enabled in some games.



    Can you possibly imagine what the R800 is going to do? People should be considering the 4870x2 over the 280 and Crossfire of 280 SLI without a doubt. High res gaming has never looked so good.

  • DigitalFreak - Monday, July 14, 2008 - link

    Fanboy much?
  • piroroadkill - Monday, July 14, 2008 - link

    Uh, multiple monitors are really important to a lot of people, I know it is to me, rendering SLI pointless
  • Griswold - Monday, July 14, 2008 - link

    So, AMD needs to bring their drivers to the point where all games of the past benefit from multi-GPU? Are you sure we need to play 5-10 years old games (assuming they even run on todays systems) at 5000fps just to prove a point? :P

Log in

Don't have an account? Sign up now