Announcement
Collapse
No announcement yet.
Mesa 17.1-dev vs. AMDGPU-PRO 16.60 vs. NVIDIA 378 Linux Gaming Tests
Collapse
X
-
-
-
Originally posted by indepe View PostThat's gonna make the $100 GPU look really good....Last edited by dungeon; 31 January 2017, 02:39 AM.
Comment
-
Originally posted by dungeon View Post
Yup, and usable for reading only for driver developers of some sort... not for users who wanna buy that GPU hardware, as if they buy just GPU by reading this, they might be surprised how slower things are on their average CPU
Comment
-
Well, on PC market every mixages are possible There are no rules there of course, but equal or up of range of two is more likely to happen than else
So on average if someone pick $100 GPU i don't think he will get CPU of more than $200 or something like that.Last edited by dungeon; 31 January 2017, 02:56 AM.
Comment
-
Originally posted by dungeon View PostWell, on PC market every mixages are possible There are no rules there of course, but equal or up of range of two is more likely to happen than else
So on average if someone pick $100 GPU i don't think he will get CPU of more than $200 or something like that.
And if that turns out to be too expensive together, you go back to picking a less expensive GPU.Last edited by indepe; 31 January 2017, 03:09 AM.
Comment
-
Originally posted by smitty3268 View PostMy guess as to the real underlying cause is this: http://www.anandtech.com/show/10536/...ation-analysis which presumably lets them "reduce the memory bandwidth for rendering" which would lead to better performance, lower power usage, and in turn higher clocks (and more performance).
Here is a web demo of TBR via WebGL. You can switch to TBR hitting 6 on your keyboard and turn it off via 7. Of course that only changes FPS when you are under the cap so you have to increase the number of lights to achieve that first. The frame rate nearly doubles in my case.
https://github.com/tiansijie/Tile_Ba...DeferredShader
I don't know whether RadeonSI has a corresponding workaround OpenGL implementation to force TBR for immediate rendering but pretty likely the proprietary NV driver has.
So pointing that out was a good idea in my opinion and perhaps when someone has the time he might have a look at how the current implementation is. It might also be useful for the guys porting games to improve the performance. Implementing it directly is of course the best way.Last edited by oooverclocker; 31 January 2017, 03:12 AM.
Comment
-
Originally posted by indepe View Post
Unfortunately the best combination depends on the game as well... so I'd think you'd find a GPU using a test like this, and then look at some other tests to try to figure out which CPU you might need, for your favorite game, to support the same frame rate. Maybe.
And if that turns out to be too expensive together, you go back to picking a less expensive GPU.
Sounds irregular that Michael's Xeon or even that AMD's i7 6700K to be combined with RX 460 That might be only for marketing provocations reasons as that dGPU is faster than *any* iGPUs ... as whatever iGPU you have does not matter if CPU cost $200 or $2000, that RX 460 is still better so applies anywhereLast edited by dungeon; 31 January 2017, 03:39 AM.
Comment
Comment