Sidebar Sidebar
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

Tesla C2075 6GB working in Mac Pro 5,1

C

CompNeuro

macrumors newbie
Original poster
Aug 16, 2012
6
1
Just plugged in using existing power cables (only requires 2 x 6 pin).

Haven't run any benchmarks yet but CUDA-Z screen shots attached.
 

Attachments

  • Screen Shot 2012-08-16 at 15.29.45.png
    Screen Shot 2012-08-16 at 15.29.45.png
    82.1 KB · Views: 6,225
  • Screen Shot 2012-08-16 at 15.33.50.png
    Screen Shot 2012-08-16 at 15.33.50.png
    60.9 KB · Views: 778
  • Screen Shot 2012-08-16 at 15.34.15.png
    Screen Shot 2012-08-16 at 15.34.15.png
    73.2 KB · Views: 658
  • Like
Reactions: Petri Krohn
Benchmark attached.
 

Attachments

  • Screen Shot 2012-08-16 at 16.05.32.png
    Screen Shot 2012-08-16 at 16.05.32.png
    124.5 KB · Views: 732
Well, it is just the same chip as in GF100 aka Geforce GTX 480 :)

The Tesla C2075 will actually be slower if frame buffer isn't an issue compared to the Geforce GTX 480.
 
Actually it's a GF110. The c2070 uses a GF100 and has a higher power requirement.

The point of a Tesla card is the increased double precision speed and available memory.
 
CompNeuro said:
Actually it's a GF110. The c2070 uses a GF100 and has a higher power requirement.

The point of a Tesla card is the increased double precision speed and available memory.
Click to expand...

Ah yes, thanks for correcting me :)

Although, you could get the same performance in Mac OS X just be installing a Geforce GTX 580 then, as they use the exact same driver.
 
Pressure said:
Although, you could get the same performance in Mac OS X just be installing a Geforce GTX 580 then, as they use the exact same driver.
Click to expand...

They do? The teslas are built for computation workflows, thus the large amount of ram. Out of curiosity, what is the point of one under OSX? Also I thought that one was headless, and you have gaming benchmarks one it:p.
 
Pressure said:
Ah yes, thanks for correcting me :)

Although, you could get the same performance in Mac OS X just be installing a Geforce GTX 580 then, as they use the exact same driver.
Click to expand...

AFAIK the double precision speed is capped in the firmware/hardware. Are you saying that you can get full speed DP on the GTX 580 under OSX? :confused:

In any case, you can't get the same amount of memory, and if your workload is memory bound, that is an issue.

----------
thekev said:
Out of curiosity, what is the point of one under OSX?
Click to expand...

I was running under Windows/Linux. My lab just got 8 new Mac Pros (I bagged one :D) so giving OSX a spin. All my tools\software run under OSX, just checking out the performance of the card


thekev said:
Also I thought that one was headless, and you have gaming benchmarks one it:p.
Click to expand...

Nope it has a single DVI out.
 
CompNeuro said:
AFAIK the double precision speed is capped in the firmware/hardware. Are you saying that you can get full speed DP on the GTX 580 under OSX? :confused:

In any case, you can't get the same amount of memory, and if your workload is memory bound, that is an issue.
Click to expand...

Ah yeah, had forgotten that they were restricted on the hardware level...I'll crawl back in my hole now heh :eek:
 
CompNeuro said:
I was running under Windows/Linux. My lab just got 8 new Mac Pros (I bagged one :D) so giving OSX a spin. All my tools\software run under OSX, just checking out the performance of the card




Nope it has a single DVI out.
Click to expand...

That's very cool, but I didn't know they had dvi.
 
is very much like a Quadro 6000, but without the DisplayPorts

CUDA-Z is showing 4096 MB, you will find that openCl can see all 6GB.

We have done a C2050 and were able to add boot screen and run at PCIE 2.0 as well. But the ECC RAM is a bit of a bother, sometimes it turns on, sometimes off.

Will be nice when the 6GB shows up everywhere.
 
CompNeuro -- Do you mind posting your experience using the Tesla card on a MacPro? Are you having issues with the ECC settings (as described by MacVidCards)? Can your CUDA applications only access 4GB RAM? Information on using a Tesla on a MacPro is very limited. Any information you can provide would be helpful. I am considering putting a C2075 in my machine.
 
If anyone in Hollywood has a C2075 and wants to bring it by for rechipping, I am 99% certain I could give it boot screens and PCIE 2.0.

First one is the Guinea Pig, I'll do it for a 12 Pack of Stella.
 
Mac version of Tesla C2070

We have introduced a Mac EFI Tesla C2070.

All 6Gb is available in GPGPU apps, though as noted by OP, Mac OSX can't see it all. Simple cosmetic issue, the RAM is there.
 
CompNeuro said:
Benchmark attached.
Click to expand...

The GTX 670 seems to be equal in performance (at least by Heaven). But the Tesla is one awesome card!

Just ran a test on my MP with the same settings.
 

Attachments

  • Screen Shot 2013-02-09 at 12.25.46 PM.png
    Screen Shot 2013-02-09 at 12.25.46 PM.png
    172.7 KB · Views: 391
Did Teslas get built in support with the drivers? That's a pretty big development.
 
They have had support of one sort or another since 10.6.8 Quadro Drivers but it wasn't useful until 10.7.3 or so, especially for the cards with more than 4GB of RAM.
 
You must log in or register to reply here.
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.
Top Bottom
Back