Final Words & Conclusion
Final Words & Conclusion
It's always good to see an actual retail sample of a graphics card arrive. And that gave us the opportunity to do a little extra. As stated in the introduction, the 2nd card we received is a 100% reference model (we are not allowed to disclose just yet who the vendor is), but they have the same clock frequencies, the same memory amount, the same PCB, same components and the very same cooling. That's not a bad thing my Guru's, contrary.
The right processor ?
There's a couple of things I like to discuss. First off, I have seen questions as to why we use a Core i7 965 Extreme processor in our VGA reviews. Well, honestly you are not going to see much difference compared to say a Core i7 2600K. Have a look at our CPU scaling article.
And sure, inCrossfire that dynamic changes a bit, once once you dip below 1600x1200 that's where you'll see some performance differences due to the processor compared to say the new Core i7 3960X. Realistically, even then the performance offset will not be a extensive. CPU limitation is a little overrated (as long as you have a modern age quad-core processor). We can not imagine that if you have the money to invest in two R7970 and then connect it to a 17" 1280x1024 monitor.
So the test platforms are still well suited to test this gear. And sure with the DX9 titles like Far Cry 2 you will notice a little CPU limitation here and there, but honestly once you passes 60 FPS on average .. you're good regardless. It's the same story with x8 / X16 / Gen 2 / Gen 3, the measurable performance difference are just exceedingly small.
Crossfire Stability
The past year if you have read up in our forums, AMD's Radeon team has been a step too slow with driver support. While there are monthly updates and hotfixes titles like Rage and Skyrim have been plagued by driver bugs, especially in Crossfire modes. It took AMD weeks to fix. When you drop 60 EUR on game you want to be able to play it on release day, period.
When we relate that to our test suite with CrossfireX in mind AMD did not return with a homerun either. We had some issues, COD MW2 had a negative scaling issue, Anno 1404 refused to scale properly below 1920x1200 (though that one might have been the one title severely CPU limited) and Dirt 2 showed massive graphics corruption. We inserted Anno 1404 but didn't even bother with the other two titles hence I did not include them in the benchmark session.
The flipside of the coin however is that all other titles behaved well and performed beautifully, really. However the high-level 0-day game driver support needs to improve. On a recent AMD event we talked about this with AMD Radeon developers, and we have been assured that starting 2012 more people are added to the driver development team, to be able to guarantee you much better game support when the game is actually released. Good news is that the latest drivers also support manual creation of multi-GPU game profiles, allowing you to enable Crossfire yourself, if AMD has not fixed it. But well, this somewhat sore topic remains open and we'll follow it closely. It is however the risk you take with multi-GPU gaming, that goes for GeForce cards versus drivers as well albeit it seems to be better managed in camp NVIDIA.
Crossfire Scaling
Obviously this topic is closely related towards the previous paragraph, but once Crossfire is properly supported, well it just doesn't disappoint. The games scale well with the new Tahiti GPU architecture. And depending on your monitor resolutions you'll see relative performance scaling go upwards with your monitor resolution. The modern GPU stringent titles scale anywhere from 1.6x towards a staggering 2x performance, and that's just impressive. Playing Crysis 2 at 2560x1600 in ultra quality mode with the HQ texture pack in DX11 at 70 FPS on average certainly brings a big smile to my face.
Power and Heat
Much like the reference review has shown, the cooling is certainly sufficient enough for two cards inside a chassis and during gaming temperatures will stay at roughly 80~85 Degrees C. The IDLE noise levels are close to NIL really, and when stressed massively two cards will be audible, you will hear quite a bit of airflow. It however remains at acceptable levels.
Initially we where impressed by the board the power consumption of this product. A single card we rated at roughly 200 Watt TDP, that means when you completely stress it that's the power consumption. Two cards indeed calculate towards roughly a 400 Watt power draw when gaming. In total our PC during hefty gaming consumed well under 600 Watt on X58 and a slightly overclocked Core i7 965 which is not the most energy efficient platform. With the PC in Idle with a second card we measured roughly the same (163~166W) power consumption, meaning that the low IDLE power consumption of the cards is just very low.
ASUS Radeon HD 7970
Overall obviously you'll need a very well ventilated chassis, always keep cold air coming at the graphics cards.
With the Crossfire overview we also have been able to get glimpse into the future, The pending Radeon HD 7990 won't be far away from the performance levels you have seen today.
As impressed as we where with the reference product we are similar impressed with the ASUS R7970. Impressive as well is obviously Crossfire scaling. We do hope that AMD puts their money where their mouth is and get their game driver support at higher levels.
With the Crossfire overview we also have been able to get glimpse into the future, The pending Radeon HD 7990 won't be far away from the performance levels you have seen today.
Should you have missed out on it, check out our Radeon HD 7970 CPU scaling performance article right here as well as a new R7970 overclock guide right here.
- Leave/read comments on this product
- Sign up to receive a notice when we publish a new article
- Or go back to Guru3D's front page
