(08-20-2011, 05:57 AM)NaturalViolence Wrote: [ -> ]It's probably incorrect (my numbers that is). I'm still tweaking it and those numbers were probably interpreted from the tomshardware benchmark charts. Unfortunately I don't think I have enough data from the dolphin gpu benchmark project to complete the list.
Edit: Looking at the benchmark results now. It seems to put a 4870 around 4.0MP and a 5770 at 5.0MP. I would expect a 5750 to fall in between somewhere.
4870 was better than the 5770 quite awhile ago, drivers might've lessened the gap or reversed the scales. It should be better than a 5750 no question though.
I don't know about pc games (nor do I care quite frankly) but as far as dolphin is concerned the 5770 is significantly faster. The gpu benchmark project showed the 5770 performing about 1/3 higher than a 4870.
*Looks up card specs to compare*
5770
Memory bandwidth: 76.8 GB/s
Shader throughput (SP multiple + add): 1360 GLOP/s
4870
Memory bandwidth: 115.2
Shader throughput (SP multiple + add): 1200 GLOP/s
Interesting. You're right the 4870 should be significantly faster. If I can't trust these benchmark results then I'm going to have to bs this thing even more. I noticed a serious inconsistency in the benchmark between the with and without SSAA results. Vdram bandwidth appears to be the primary performance determining variable with every card in the regular 1x, 2x, 3x tests. Cards with twice the memory bandwidth perform twice as well, this makes sense since efb emulation is extremely demanding on vdram bandwidth and that scales with framebuffer resolution. However the SSAA results don't make any sense at all. Cards perform much better than they should compared to the no SSAA results and memory bandwidth no longer determines performance. In addition radeon cards gain about a 20% advantage over comparable nvidia cards. Ideally what I should see is a linear relationship where doubling internal resolution cuts performance in half (since the gpu should be the bottleneck), yet I do not see this. I expect the low IR (internal resolution) results to show a non-linear relationship because the cpu begins to bottleneck the gpu at such low IR. However the super high IR benchmarks should show a linear relationship, and they don't. And you see wild inconsistencies as well with the SSAA results.
I'll begin working on the guide again tomorrow when I have some time, I might even finish it.
Not really. I need more data about what dolphin needs (how much memory bandwidth/shader throughput is needed at certain settings and conditions) which I can only get by interpreting benchmark results taken using dolphin. But I appreciate the thought.
hello
i bought the really fast HD 6950 with 2gb to replace my good old HD 5450.
crysis 2 with hires-texture-pack is running really nice.
but in super mario galaxy (lowest setting) i got just an improvement about 5fps.
HD 5450 (512MB DDR3) -> 35fps
HD 6950 (2026MB GDDR5) -> 40fps
on higher settings (internal resolution & aa) the difference is bigger but of course the fps decrease below 30.
i would love to play sm galaxy smooth and constant. that seems to be not possible with my cpu (AMD Phenom II X6 1100T, 3.3ghz).
is the actually top-AMD-cpu too slow for sm galaxy?
who of you guys can run sm galaxy really smooth at 60fps? on which hardware?
thanks for your replies
pudelmuetze
Quote:is the actually top-AMD-cpu too slow for sm galaxy?
Yes.
Quote:who of you guys can run sm galaxy really smooth at 60fps? on which hardware?
Only an overclocked i5 2500K or higher can handle SMG2 at fullspeed with LLE audio.
I am running the following system:
* Tyan S5512 server based motherboard, Sandy Bridge
* NVIDIA GTX 460 (768MB, DDR5)
* 8 GB ECC RAM
Running Linux and the OpenGL backend...
What is the minimum to be able to run most Wii stuff at an acceptable framerate? I am having difficulties even with 1x internal resolution, for some reason.
. .My average speed is roughly 60%.
-Thom
sorry, I forgot to mention that I am using a Xeon E3-1240 CPU. (4 cores, 3.30GHz, HT).
Which revision are you using? The latest revision currently has a bug which severely reduces performance on certain games, you should be using 3.0
If you are using 3.0 then post screenshots of your settings and I'll take a look.
(10-18-2011, 06:30 AM)tschak909 Wrote: [ -> ]I am running the following system:
* Tyan S5512 server based motherboard, Sandy Bridge
* NVIDIA GTX 460 (768MB, DDR5)
* 8 GB ECC RAM
Running Linux and the OpenGL backend...
What is the minimum to be able to run most Wii stuff at an acceptable framerate? I am having difficulties even with 1x internal resolution, for some reason. . .My average speed is roughly 60%.
-Thom
sorry, I forgot to mention that I am using a Xeon E3-1240 CPU. (4 cores, 3.30GHz, HT).
(10-18-2011, 04:43 PM)NaturalViolence Wrote: [ -> ]Which revision are you using? The latest revision currently has a bug which severely reduces performance on certain games, you should be using 3.0
If you are using 3.0 then post screenshots of your settings and I'll take a look.
oh ok, I pulled the latest rev from SVN. Thanks.
-Thom
LOL, you guys don't want to see what i'm playing on..
anyways, I meant to ask, APU's are out of the question huh? (even Overclocked?)
I heard a APU-DOES HAVE intergrated graphics BUT its supposed to be on par with a 64xx AMD GFX card.