Originally Posted by Kuivamaa
Matching richland iGPU clocks is kinda pointless since it uses an older architecture.
What I'm trying to find out with all the suggestions in my post is:
By how much is the built-in GPU on the A10-7850K bottlenecked by the DDR3 memory interface ? Surely the Anand test system with the HD 6750 is using an Intel CPU, but shouldn't the difference in GPU performance offset that ?
My theory is that AMD is using more GCN cores clocked lower in order to try make their product better than it really is, because I suspect that a 384 core GCN part clocked at Richland's 844 Mhz instead of 720 Mhz is going to yield the same kind of performance as the 512 cores running at 720 Mhz. From the benchmarks it's clearly visible in some cases that the fps hit a wall, no matter if it's the Kaveri A8 or the A10. I suspect it's a RAM bottleneck, which means it's debatable whether it's cost-effective to use 512 GCN cores at 720 Mhz over 384 cores at 844 Mhz.