AMD Execution Thread [2023]

Status
Not open for further replies.
With how slow the 7040 rollout is... I wonder if iln the end zit will get released widely together with Intel meteor lake.

I guess, more competition means cheaper laptop then.
 
They're not drastically different these days. It used to be a bigger gap, but as things like async compute became completely normalized in games and all, the 1070's lead over the 580 aint what it used to be. Maybe like a 25% difference in general raster performance. Similar memory bandwidth and quantity and all. Neither are DX12U-capable. I'd say they're in pretty much the same boat by modern standards.

But yes, that also means they're still very usable GPU's, so long as you dont have high expectations or a need to play the most cutting edge games at 60fps or something.
These days a 1070 is often only 10% faster than a 580.
 
These days a 1070 is often only 10% faster than a 580.
Whenever someone benchmarks it 580 is at about the same as it was in comparison to 1060 (a bit faster) and 1070 (a bit slower) these days.

 
Last edited:
Whenever someone benchmarks it 580 is at about the same as it was in comparison to 1060 (a bit faster) and 1070 (a bit slower) these days.

There is a big gulf between a 1060 and a 1070 in relative terms. Mid range versus upper midrange.

A 580 is clearly more performant in modern games than a 1060 than it used to be. This isn't disputable, no matter how many hours you spend trying to desperately find cherry-picked examples showing otherwise.

And I say this as a 1070 owner, to be clear. I dont have this weird need to try and defend the brand of GPU I buy against any and all criticism like some do.
 
Did you even look at the links or you've just had this "weird urge to try and defend the brand of GPU"?
I was going to post something snarky, but I looked at your links (couldn't read them, but graphs are easy) and you got a fair point. The benchies all line up with my own 580 experiences and the nVidia numbers all look about right.

Fair argument, very well documented too. The point goes to DegustatoR.
 
Finally, a public release of HIP/ROCm SDK for Windows:

 
Finally, a public release of HIP/ROCm SDK for Windows:

Its only HIP plus math. No AI .
 
Not in the initial release, though MIOpen is being ported to Windows.

 
Again, you were citing very outdated twitt. Since then, Moritz implemented multi GPU support, so currently 4 MI250 is faster than 8 RTX A6000 in his simulations.
The comparison would be a bit more relevant if he used A100 since it has twice the memory bandwidth than the A6000 (~ 1500GB/s vs 800GB/s), then the comparison could be made using only 4 gpus for each vendor. He even states all that matters is memory bandwidth and capacity when examining individual gpu results using his fluid simulation app.
 
Status
Not open for further replies.
Back
Top