Login to Account Create an Account
Controversial Those Serious CrossFire Microstuttering and Runt Frame Issues - AMD Responds
Somewhat unsurprisingly, Nalasco downplays the significance of these issues, claiming that judging rendering smoothness using frame rating which is an NVIDIA developed test methodology, is actually no better than using the long established software-based FRAPS frame rate testing tool, since the effect of rendering issues are actually exaggerated and do not reflect a user's real world experience:
It is important to understand that regular frame pacing does not necessarily translate into gameplay smoothness. Other issues, including uneven game simulation updates or long latencies from user input to displayed frames, can often have a more significant perceived impact. So while we believe that FCAT does provide some interesting information, on its own we do not see it as being any more useful than FRAPS for helping to identify or solve smoothness issues. On the contrary, the results from tools like these can be misleading since they can exaggerate issues that have little noticeable effect, while completely missing issues that are much easier to notice while playing a game.
While there is certainly truth to input latencies, game engine and other factors affecting game rendering smoothness, his claim still isn't quite true. PC Per, which published the first in depth frame rating article linked to above reported that the poor frame rating results were actually corroborated by subjective gameplay testing. Other sites also report hitching and jerkiness when using CrossFire, even though the reported frame rate is high.
Nalasco then goes on to say that a tool to properly measure gameplay smoothness and responsiveness does not yet exist and that testers should instead use subjective gameplay testing to get a feel for how smoothly a game is playing:
While we appreciate the desire for users and reviewers to have a simple tool that clearly and simply evaluates gameplay smoothness and responsiveness, we do not believe that such a tool exists yet. Many game developers (as well as our own driver engineers) use tools like Microsoft’s GPUView to better understand and solve these issues, but these tools are not particularly user friendly or easy to interpret. It may be possible to develop better measurement tools in the future, but for now we recommend that reviewers supplement existing tools with subjective, un-biased gameplay evaluation to provide the most meaningful results.
I don't buy it. This looks to me like AMD has been caught with their pants down and are desperately trying to downplay this serious problem as a damage limitation exercise. So, if microstuttering and runt frames are so "insignificant", then why have AMD previously acknowledged this problem, saying that a driver fix will be eventually released in a couple of months that fixes it? In fact, AMD have now issued a (very) prerelease version of the new driver which significantly improves the runt frame problem. The results can be seen in the PC Per HD 7990 review. It's still somewhat worse than NVIDIA at this stage, though:
This driver is very much not ready for primetime however, as explained by PC Per's Ryan Shrout:
There is a catch to all of the data presented here today - AMD at the last minute provided me with a prototype driver with a completely from the ground up frame pacing algorithm in an attempt to fix the runt frame issues we have been seeing with CrossFire for nearly a year. However, this driver is not nearly close to end-user availability so we were not comfortable reviewing a product with it. I don't expect this driver until the late-June or July time frame, but the early results look promising.
So, AMD are improving the situation by using frame pacing, similar to what NVIDIA has been doing for a long time now? Better late than never, I guess.
All in all, it looks like Nalasco was spinning PR speak in a bid to downplay the seriousness of this problem as a damage limitation exercise, rather than honestly face up to it. The interview covers other aspects of multi GPU rendering such as Alternate Frame Rendering (AFR) and is well worth reading.
PC Per made this damning conclusion of the HD 7990 in their review (our emphasis):
I knew this review was going to turn out like this after publishing our faux-HD 7990 performance results last month. With its performance completely dependent on CrossFire technology, the HD 7990 as a $1000 graphics card has a very hard time justifying its price. With our early testing of the Catalyst prototype driver showing positive results though, there is yet hope for CrossFire to be fixed in this generation, at least for single monitor users! But until that driver is perfected, is bug free and is presented to buyers as a made-for-primetime solution, I just cannot recommend an investment this large on the Radeon HD 7990.
The HD 7900 series cards have now been out for about a year and a half and are heading for replacement by the 8000 series cards later this year, so it beggars belief that AMD are only now getting to grips with this problem, now that the merde has truly hit the fan. This does not give one confidence in the company to deliver a decent graphics solution and is reminiscent of their significant performance lag behind Intel in the CPU market. They certainly shouldn't be charging top dollar for products with such serious flaws. This also extends to single card configurations too, since the new $1000 HD 7990 rather incredibly exhibits severe and very annoying coil whine when in 3D mode!
Would you spend $1000 on a card with such massive flaws? Tell us in the forum.
Also, I'll have an editorial up about this very, very soon...
Sources: ComputerBase PC Per