What don't I understand?

[jpine]

Ok, guys. Help me understand what's going on here. I bought a used 5,1 with a flashed 7970 (3GB) and 64GB of RAM that I intend to use with the Oculus Rift. This rig is for research and not for developing commercial apps. Just for fun, I compared the OpenGL capabilities (I don't care about OpenCL) of the MacPro rig to that of a brand new i5 iMac with the basic 1GB mobile GPU. I expected the MacPro to squash the iMac. As you can see from the screen shots of the Valley OpenGL test, that did not happen.

One other thing: The 7970 was not flashed by Macvidcards, so I assume the wattage of the card is stock. Does this card run too hot for the MP to safely handle? I'm open for suggestions for a better GPU that does not require an EPS. I do have a 2.5GB GTX 570 that was flashed by MVC laying around so I'll test the MP with it and see if the results change any.

BTW, the resolution for the test was 1280x720. That is the resolution for the Oculus Rift.

TIA!

JP

 

[h9826790]

I just ran that test with my 7950 (OC to 850MHz), and get almost identical result as yours. Even though I think the 7970 can do a little bit better than that, but should not be too much.

 

[666sheep]

Double minimum FPS could be considered as "squash".

What you're missing is the fact that Unigine scores are CPU clock/single thread performance dependent.
Resolution matters as well. Difference would be bigger in higher res.

 

[mikeboss]

maybe the 7970 isn't properly modded and the PCIe Link Speed is running at only 2.5 GT/s instead of 5.0 GT/s.

 

[jpine]

maybe the 7970 isn't properly modded and the PCIe Link Speed is running at only 2.5 GT/s instead of 5.0 GT/s.

Link speed is part of the modding process under OSX? I thought that was an issue pertaining to Windows under Bootcamp.

 

[h9826790]

Yesterday, I ran the test when remote controlling the computer, it seems the result is not very accurate.

So, I just did it again. This time I was sitting in from of the computer, no active remote control section. The computer still running as a server, have few background tasks, and not intentionally optimise to run the benchmark.

And this time, I have a better result.

(This was using the basic preset. I don't know why it says custom with medium quality.)

I expect a 7970 can do better than this. So, there may be something wrong in your system. Or as 666sheep said, your score was limited by something else (e.g. your single core performance).

 

[N19h7m4r3]

At the resolution you're running the tests at your CPU is taking the lime light. Crank it put to the extreme setting to single out the GPU more.

Also note that AMD's OpenGL drivers have always been behind NVIDIA's, and NV has always held the lead in the OpenGL performance.

 

[mikeboss]

Link speed is part of the modding process under OSX? I thought that was an issue pertaining to Windows under Bootcamp.

nope. you need to do this modification on the card itself ->
https://forums.macrumors.com/showthread.php?p=17183586

did you check in the System Profiler if the link speed is 2.5 or 5.0 GT/s?

 

[skinned66]

Also note that AMD's OpenGL drivers have always been behind NVIDIA's, and NV has always held the lead in the OpenGL performance.

Co-sign!

 

[gochi]

another reason why i got the 680gtx instead f the 7970 radeon

 

[mikeboss]

results with my (tiny little bit undervolted & underclocked) 7970:

 

[jpine]

nope. you need to do this modification on the card itself ->
https://forums.macrumors.com/showthread.php?p=17183586

did you check in the System Profiler if the link speed is 2.5 or 5.0 GT/s?

Thanks!!! When the previous owner flashed the card, he didn't remove the resistor. The card is indeed running at 2.5 GT/s. The bad news is that there is no way in hell I'm risking turning this card into a paper weight by trying to remove the resistor myself. I think I'll just put it on eBay or here on the forum and use the money to buy a MacVidCard.

Thanks guys!!!!!

 

[jpine]

Well, I went ahead and popped the R17 resistor off the 7970 and am now running at 5GT/s. At 720P, it makes no difference. The difference is noticeable if I crank up the res setting in Valley. The "min" fps is doubled running at 5GT/s.

 

[mikeboss]

received a fresh 7970 this morning and benchmarked it before removing the resistor:

FPS 64.8
Score 2712

 

[666sheep]

If you care about Unigine scores, you gotta get highest clocked CPUs and overclock your GPU. That way 7950 can beat stock clocked 7970s easily.
But it's only a benchmark