By using this site, you agree to our Privacy Policy and our Terms of Use. Close
DM235 said:


FLOPS are a metric for measuring graphics, but they are not the only metric.  The only reason the 5870 has such a high FLOPS value is because it has 1600 shaders (compared to 1280 shaders in the 7870). 

The other important metric is the core clock.  The 5870 only runs at 850 MHz while the 7870 runs at 1000 MHz. 

So the 5870 has more, but slower, shaders.  If you could utilize all of those shaders, you would be better off.

Given that console hardware is fixed for a long period of time, I think it is quite likely that late in the generation developers would have SDKs that would let them harness that extra power.


Not only that, but the Radeon 5870 was based on the VLIW5 architecture. (Very-Long-Instruction-Word)
It was an Architecture designed for optimal Direct X 9 rendering performance as it was merely an evolution of the Radeon 2000 series, which back then... Games were still pretty much exclusively utilising the Direct X 9 rendering path, so it made sense.

As time went on, AMD introduced the VLIW4 architecture with the Radeon 69xx and 68xx series as their data indicated that the 5th unit in the shader array  was pretty much going un-used in newer games. - Which were starting to drop Direct X 9.
The result was that the 69xx series had less overall shaders, but more shader arrays that will become saturated, resulting in more performance.

The GCN architecture however is more forward-looking, it's a compute heavy design which shows in allot of compute scenarios such as Bitcoin mining and Folding@Home, however each shader is vastly more programmable and powerfull than any other AMD architecture that came before it.

So in the end, whilst the 5000 series has more theoretical performance than it's more modern brothers, it is much harder to extract the performance in any meaningfull way.




www.youtube.com/@Pemalite