By using this site, you agree to our Privacy Policy and our Terms of Use. Close
secpierre34 on 21 June 2013
Edit Thread
Viper1 said:
Turkish said:
Viper1 said:

I guess you don't realize that it is being paired up with the HD 7760D inside that AMD A10 APU?

Yes, and its not even close to being on par with PS4.

Actually, it's not too far.  At worst you've have to drop the AA, soft shadows or volumetic fog in games but after that, you'd not really see much difference.

HoloDust said:
Viper1 said:
HoloDust said:
Viper1 said:

I guess you don't realize that it is being paired up with the HD 7760D inside that AMD A10 APU?

You are aware, of course, that A10 + 6670/7670 combo will perform at best at level of 6770/7750, and that even vanilla 7850 (let alone PS4's GPU) performs at 2x 7750?

HD 6670 - 768 GFLOPS
HD 7660D - 614 GFLOPS

Total - 1,382 GFLOPS.  That's much more than the HD 7750 but equal to a single HD 6770

Or if the APU will let you crossfire with an HD 6770 for $30 extra, there's your more than equal performance (614 + 1360 = 1,974 GFLOPS).

Oh boy, here we go again with FLOPSing....

6770 is rebranded 5770 http://www.rage3d.com/articles/amd_radeon_hd6770-6750_rebranding/

and this is how it compares to 7750:

http://www.anandtech.com/bench/Product/538?vs=535

Now, 7850 vs 7750:

http://www.anandtech.com/bench/Product/549?vs=535

and 7850 vs 5770:

http://www.anandtech.com/bench/Product/549?vs=538

So why are you only comparing the performance of 1 of the GPU's when the system I posted has 2?

And yes, I used GFLOPS as it the closest means of measuring comparable performance across architecturally different products.

single point performance of gflops is the worst measure of performance across architecturally different products. You have it backwards. The Gflops difference between two gpu of the same architecture is the best way to go. However across different architecture it's a dead end. The reason being that one architecture might be suited for optimal performance per cycle, whereas the other one might perform few calculations per cycle but a lot more cycles. Now over the course of the same time they will perform the same amount of calculation know as flops. However if the task at hand deals with more calculation per cyle than they one best suited in that area will will win by a wide margin even though they can compute the same amount.

So flops comparison is no good unless we are dealing with the same architecture. Need further proof think of what a cpu does at around 200-400 gflops against a gpu computing physics at 2tflops. In fact gpu only started to compute physics only three years ago.