-
Most NVidia cards now work on the mac
-
hi juan ,
which on one is the non classified edition ?ps: any particular cable adapt needed ?
thank!
g
Davinci Resolve Control Surface
MacPro
Cubix desktop 4
2 Red Rockets
GTX470+GTX470+GTX285
24GB RAM
HP Dreamcolor
Panasonic 58PF Plasma -
There is an over clocked version of the card called “classified edition” which requires more power and will not work. The regular version will.
You need a 6pin to 8pin adapter.
-
As you guys can see my EVGA GTX 590 running on my mac pro 2009 this is happening with the new driver from nvidia. I plugged 2 cables mini 6pin to pci-e 6pin in logic board to the GTX for powering. The first problem I found here is the second cuda core, is only 512mb and show as a generic NVIDIA. Any suggestion to fix this ? Testing here and I’ll give news. I’ll test the DaVinci and the premiere CS5.5. It seems that 3d software, use only one core. For those 3d softwares I think a quadro will be better choice. Nvidia guys, THANK YOU, awesome.
https://postimage.org/image/b49h9ppb5/
https://www.nvidia.com/object/macosx-270.00.00f01-driver.html -
The power supply in the Mac pro can’t support that 590.
As soon as the second gpu kicks in the system will kernel panic.
I was able to see both cards as properly listed gpus in my config and was even able to test it with davinci. But as soon as I stressed the gpu enough to kick in the second gpu… It was done.
-
I tried the 680 (MP 4.1, 10.7.3 etc) but no dice 🙁
It seems I’ve to remain modest, ordering a 570 now … -
I can confirm this too. 680 does not work. For now. Seems only Fermi architecture is supported in drivers. May change in the future.
-
Hi Juan,
Very brave testing with the 680 there. Have you tried it on a PC with resolve. Also have you heard or seen anything positive about use of the Geforce 590? The 590 has significantly larger bandwidth capability due to it’s dual pipes and higher bit architecture. It’s a bit of a gamble spending those $’s on uncertified gear isn’t it!
Best,
M.
-
Yes, I will try 680 in windows…but have not yet. Also have tested 590 in MacPro. Drivers work but built in power supply, can’t support both gpus working. Kernel panic when the second gpu kicks in. 590 are not officially supported in resolve, you get a warning dialog. But it does seem to work. Haven’t been able to check full capabilities because of power issues.
I recommend the safest config option is the 580.
-
Hi,
Many thanks, it seems odd that the 680 doesn’t work. I’m pretty sure that the kernel panic from the 590 on the Mac is down to something like the EFI strings in the graphics config. I may be wrong. When you say kicks in, do you mean when the machine boots? How is the 590 on a PC?
Also… sorry, what about performance? Is the 590 significantly better? I’m thinking specifically about the memory bandwidth of this card vs the 680. In theory the wider bit bus should give the 590 the edge over the 680. I presume though that the 680 would perform better in a PCI gen 3 slot…
Sorry to ask, I just don’t have the resources to test this at the mo.
Best wishes,
Matt.
-
[Matt Walsh] ” it seems odd that the 680 doesn’t work.”
Not at all actually the 680s are a whole new architecture (Kepler instead of Fermi) it makes perfect sense that they don’t work. I was just hoping that NVidia would have been forward looking when they released their new drivers.
[Matt Walsh] “‘m pretty sure that the kernel panic from the 590 on the Mac is down to something like the EFI strings in the graphics config.”
No, this was using the new NVidia drivers… which natively support the card.
[Matt Walsh] “When you say kicks in, do you mean when the machine boots? How is the 590 on a PC?”
When I say kicks in, I mean when I put on enough nodes or turn on the NR. Do a search on the forum for 590 and you’ll see me previous posts on this.
[Matt Walsh] “what about performance? Is the 590 significantly better?”
From what I’ve seen… No. Like I said before, right now the best card is the 580.
[Matt Walsh] “wider bit bus should give the 590 the edge over the 680.”
There are apparently some inherent differences in the kepler architecture that may never make the 680 as CUDA capable as the 580… specifically they switched from a hardware task controller to a software task controller. When you’ve got tens of thousands of CUDA tasks, that takes a performance toll.
Reply to this Discussion! Login or Sign Up