RADEON 4870X2 broken dogtracker?

[y eye]

pcgh
"Onboard Crossfire with R700
PCGH proves micro stuttering on the Radeon HD 4870 X2
Jul 28, 2008 16:26 - In our benchmark review of the Radeon HD 4870 X2 we already mentioned the problem: the micro stuttering did not vanish. Now we deliver the proof.
Radeon HD 4870 X2: micro stuttering is not gone.In February we wrote: Micro stuttering may destroy the performance gains from current multi GPU technologies. Since the beginning of the year hardware forums all over the world discussed what it's all about the infamous micro stuttering. People called for a "Frametime Harmonizer”, a device that emits the frames of multiple GPUs at regular intervals. The drivers were said to fix the problem. Many people expected a hardware solution with the Radeon HD 4870 X2. But the problem was not fixed (properly).

Test system and methods
After having explained the problem in the X2 review, we now deliver the appropriate pictures. The data below was gathered with the same card, which we unfortunately had to send back already. The Catalyst driver is the very same, too. It is a Beta state piece of software with the internal serial number is 8.52 (Catalyst 8.7: v8.512). To keep the data flowing we used a Core 2 Duo E8500 @ 3.6 GHz (400 x 9).

The frametimes below are taken from Half-Life 2: Episode 2, which we upgraded with the Fakefactory Cinematic Mod 8. In order to receive low framerates, we activated 24x CFAA (8x MSAA plus CFAA and Edge Detection) and 16:1 AF. The charts show the first thirty frames (not seconds!), that we recorded with Fraps. As it turned out that the tool doesn't always record what is displayed on the monitor, we convinced ourselves that the stuttering was there, before we let Fraps do the monitoring and got these results. This is the best way to prove visible stuttering with hard numbers. But not everyone notices the problem. A lot of people don't even recognize the micro stuttering and therefore don't see any differences to normal situations. The trick is not to see that the game stutters but how it does. "

The full article with reference to a previous review on Stuttering RADEONS as well as a Video of the 4870X2 Stutter may be found at: pcgameshardware.com. A question remains regarding the posibility that the reference card used in the above review may have been defective. So I bellieve that it is best to ask the experts. What do you think about the results of this 4870X2 test?

[brentpresley]

I think it will be a HUGE debate until we see if the microstutter happens on cards from multiple manufacturers, under multiple driver revisions.

If that occurs, then I believe it is a true design problem.

But, honestly, only time will tell on this one.

[3lfk1ng]

Here

I agree, with the new XFIRE controller... even if there is microstuttering, I doubt that it is as noticeable.


Goodbye GT280

They said it was "more noticable" in games when the framerate dropped below 40fps.... if your playing a game at anything lower than 60fps, you have issues.

[Tark]

Originally Posted by 3lfk1ng

Here

I agree, with the new XFIRE controller... even if there is microstuttering, I doubt that it is as noticeable.


Goodbye GT280

They said it was "more noticable" in games when the framerate dropped below 40fps.... if your playing a game at anything lower than 60fps, you have issues.

That chart is a great reason not to base everything off one chart, clearly its no where near the overall results.

[3lfk1ng]

Of course not, but we already know the card has put the GTX280 to shame, so it's not going to do any further damage.

The 4870x2 does good in almost all the results.

[YMAA]

The only way those results are plausible is if they didn't install the driver for the nVidia cards. There's no way an HD3870 could have beaten a 9800GX2 by such a wide margin.

[Eurasianman]

Originally Posted by YMAA

The only way those results are plausible is if they didn't install the driver for the nVidia cards. There's no way an HD3870 could have beaten a 9800GX2 by such a wide margin.

ATi optimization code? *rolleyes*

[YMAA]

nVidia results aside, I don't think an HD3870 can do that well at that resolution at that level of antialiasing with that texture pack installed. I use it on my Oblivion installation and it does add some overhead to the GPU, textures take up a lot more video memory with it in use. I call bull on those results, except for maybe on the HD4870 series because I don't know much about them.

[Eurasianman]

Sorry, I forgot this: /joking

[Viperabyss]

I'm with YMAA on this one. The benchmark results seemed too fishy to be true.

Oblivion at 2560 x 1600


Notice how a much wider 512bit memory bus GTX200s being outperformed by a 256bit memory bus card?

The exact same test done by Anand:


Now, I'm not sure if Anand used Cat 8.6, but I really don't think a simple driver update will suddenly give RV770 cards a significant jump in performance

[YMAA]

Was Anand'a test done with the high res texture pack?

[Dean]

Originally Posted by YMAA

Was Anand'a test done with the high res texture pack?

No, but it also wasn't done with the same moron pack that the other testing was done with.

[Clue69Less]

Seal watches gay porn?

[Dean]

Originally Posted by Clue69Less

Seal watches gay porn?

You should ask him that.

[YMAA]

Hence our ads.

[randomizer]

There is no way an 8800 Ultra woulld come last, Oblivion at these settings is where that card excels (relative to the others).