As a user and developer: This is unfortunate. With proper optimizations speed can be increased and/or power consumption reduced. Think of a game which needs a lot memory bandwidth, but low ALU performance. Reduce ALU clock and voltage, save power.
Originally Posted by Ericg
There was as recent paper about this to do this automatically.
On the spaghetti code I somewhat disagree: We have written drivers and software and optimized stuff for every hardware generation and card. No spaghetti code, nice architecture. On the other hand, we had only 3-4 cards per generation to support.
I've got a bit of a funky system, it's a Sony Vaio Z with an external Radeon GPU (6650M) that is bridged to PCIe over Sony's proprietary implementation of Thunderbolt (aka Light Peak). The external Radeon GPU in the setup actually works but only with the open source driver and not Catalyst so I'm at the mercy of the Radeon driver's performance. Unfortunately the performance on this setup is very poor and even the Gnome Shell DE feels a bit sluggish in day to day use, could this be related to the issue Phoronix has with the 6570? I believe the 6650M and 6570 are pretty closely related which makes me question this, has Phoronix ever determined what the issue is with respect to their 6570 setup?
For a point of comparison, the Intel 4000 HD in the Vaio Z feels noticeably faster then the Radeon both in the desktop and for games. When I run glmark2, the Intel 4000 gives me a score of 1800 whereas the Radeon only gives me 1100 with the 3.11 kernel and mesa 9.2 on Ubuntu 13.04. I was really hoping to see at least parity between the two on my setup.
[QUOTE=kwahoo;359124]No, it hasn't. I posted my 6670 results while ago.
Is there any differences if the brand is ATI or AMD or Sapphire?
Thanks for your time
Will Mantle help bridge the performance gap down the road?
Perhaps AMD's new Mantle will bring performance gains to linux gaming eventually? (Steam OS first?)
Originally Posted by DarkFoss
- it comes to Linux
- You have a GPU with GCN architecture
First of all. Hi to everyone.
I am visiting phorinix.com for some time now and it's really a great site!
Now the cause for finally registring for the forums:
What am i doing wrong with my gentoo machine. Because when testing nexiuz and xonotic the fps don't come near the results of the phorinix benchmarks.
But when booting Fedora 20 Alpha everything changes for the good. Even for tf2 and dota2.
Currently i have mesa 9.2 and xf86-video-ati-7.2.0 running in both distros.
Where is the bottleneck?
Shure, i could switch over to fedora or whatever, but i like the gentoo way more.
What is your hardware (CPU and GPU)?
Originally Posted by ThoM
Originally Posted by Asariati
GPU: Radeon HD5870
I do not know what exactly was wrong, but since i updated mesa (on gentoo 9.2.0-r1) and the latest emul-linux-x86 (20131008) everything just runs as it is supposed to.