RadeonSI Lands Zen CCX/Thread Optimization For GLThread Mode

[phoronix]

Phoronix: RadeonSI Lands Zen CCX/Thread Optimization For GLThread Mode

After the AMD Zen tuning for RadeonSI was dropped earlier this month, a revised implementation has now landed in Mesa 19.0-devel...

http://www.phoronix.com/scan.php?page=news_item&px=RadeonSI-GLThread-Zen-Land

[schmidtbag]

Obviously, seeing benchmarks of this is welcome, but I personally would be interested to see before and after benchmarks of GLThread Mode when compared to something like a Ryzen 2300X (or something similar) vs a 2700X (preferably of the same clock speeds).

[tuxd3v]

Obviously, seeing benchmarks of this is welcome, but I personally would be interested to see before and after benchmarks of GLThread Mode when compared to something like a Ryzen 2300X (or something similar) vs a 2700X (preferably of the same clock speeds).

Yes the benchmarks, I agree.

I agree with L3 Shared cache pinning behaviour,off course
IMO, this would have even a bigger impact in a Thread-ripper for example..

But I also love to see, a comparison like that you mention.

I haven't understood the RadeonSI subject completely...about random pinning.