XFX’s Radeon HD 7970 Black Edition Double Dissipation: The First Semi-Custom 7970
by Ryan Smith on January 9, 2012 6:00 AM ESTCompute Performance
Compute tasks, in spite of the name, are not always purely GPU bound. Depending on the task memory bandwidth can also play a significant part – which is why memory bandwidth was the single biggest increase on the 7970 over the 6970 – and our compute benchmark end up reflecting this.
Among our tests only the DX11 Compute Shader Fluid sample fully benefits from the increased clockspeed of XFX’s factory overclock, gaining 9% over the reference 7970. Our Civilization V and SmallLuxGPU benchmarks meanwhile only gain 4-5%, and our AES benchmark only gains 2%, the latter likely due to the fact that the setup time for the program’s dataset does not decrease, only the execution time does.
93 Comments
View All Comments
wifiwolf - Tuesday, January 10, 2012 - link
I would assume it's a nice fit for you too as you tend to persist.Morg. - Tuesday, January 10, 2012 - link
Precisely.Information : good
Information + Information about the Information : better
The content presented here is not worthless, one just has to know what it is and how it is limited (i.e. anandtech needs funding, they can't do all benchmarks themselves, etc.)
AssBall - Wednesday, January 11, 2012 - link
Trolls like you?Not informative.
Not factual.
Not worth reading.
MrBunny - Monday, January 9, 2012 - link
point made. formulation could be more along the lines that this cooling solution(though being louder at idle but cooler) is nicely executed being the card is overclocked and there by beating the reference design cooler easily in temps and noise.the only thing that they need to fix is the the idle fan pwm so it can be silent at idle aswell.
@Njoy i think he read it just right.
Morg. - Tuesday, January 10, 2012 - link
Just edit the bios manually when tools are available and you can change the curve from the original (which XFX didn't bother to modify for some reason .. they simply had to lower the first point in the curve to 15% or something - unless as I said there was a minimum voltage for the fans to start -- )R3MF - Monday, January 9, 2012 - link
how is GCN an architecture targetted at compute tasks when it is no more capable of DP FP than the VLIW4, in that it is still only capable of doing DP tasks at 1/4 speed of SP?or, is the 1/4 only a function of crippled consumer drivers, whereas professional products will see perhaps 1/2 for DP FP?
Morg. - Monday, January 9, 2012 - link
Probably the latter.All in all, GCN is exactly like Fermi (which is also like an older design) and the performance characteristics should be very close in the end - where it matters (i.e. not gamer products).
R3MF - Monday, January 9, 2012 - link
would be a shame if true, especially when paying $549 for the hardware!Morg. - Tuesday, January 10, 2012 - link
Are you really doing GPU accelerated computing ??R3MF - Wednesday, January 11, 2012 - link
me? no.but it is going to become a very mainstream thing for performance hungry applications, and i always dislike buying artificially disabled products.