Welcome to our community

Be a part of something great, join today!

  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

FCPX 10.2 is out

Have a 2009 MP Tower and a GTX9804GB SC card sitting on my desk. I was about to install it, when I came across a note that the 980 defaults to PCIe 1.0 on OSX. According to MacVidCards you need to send it in for a $180 "flash" to get PCIe 2.0 support on Mac. Does anyone know the dealio? I suppose I could just install it and see how it goes… FYI, current GPU is Quadro4000 and I have a RR-X in slot 2.

Cheers - #19
 
(snip) Additional notes on the 980
[/B]- The 5870 is still "snappier" in UI related things in a way that makes the GTX feel like there is a ways to go on the driver front. However, now that the new plugin is contributing to Rendering/exporting the R3Ds I will probably make "my final decision" on rolling the dice on better driver support and keeping the 980 instead of going to an r9 290x.
- The GTX 980 still gets completely Spanked in the BruceX test (the 5870 is 2x as fast on my machine) - But rendering those R3ds is a different story. (snip)

I'm wondering if this is related to the 980 being seen as a PCIe 1.0 device? If so, perhaps the throughput is too constricted to manage RT playback - but the rendering engine is able to shine since data throughput is less of an issue than computational muscle for such tasks. Just curious…

Cheers - #19
 
I had thought the 09 and 2010 ran PCI2, but the 08 was PCI1 (2.5). And the need to send it in was to get PCI 2 in windows.

(I could be wrong on this)

With the 980 I saw 16x reported in both models in slot 1, but oddly enough the GTX 970 reported 8x when in my 08 machine. Even more odd, I felt like the 970 was faster in redcine.

However I haven't checked the actual PCI link speed in system profiler... But according to Cuda Z - Host to Device was around 2500 on the 08 machine and double that on the 2010. So I think the 08 is falsely reporting PCI2 (16x) speed.

I'd be curious what you find out comparing it to your Q4000, it's definitely worth a test.
 
Last edited:
My understanding was that the 980 card could not properly negotiate the PCIe 2.0 protocols running on OSX. The hardware is certainly capable of PCIe 2.0 on the tower and PCIe 3.0 possible on the GPU itself, but inaccessible without a proper host slot/low level driver/???. If I understand correctly, the nVidia web drivers will allow you to use the Maxwell nVidia cards on OSX but they are NOT able to overcome the board being "stuck" in the PCIe 1.0 spec. I theorize this might be the reason that using GPU optimized decoding schemes gets a big boost from all that Maxwell compute power, but still can't get data down the pipe fast enough to support smooth playback.

I'd sure love to know what the $180 flash does. Described here: http://www.macvidcards.com/what-are-the-benefits-of-buying-a-flashed-card.html

Cheers - #19
 
Hi Blair -

It is my understanding that stock 970 and 980 (maxwell cards) run out of the box PCI 2 in 09 and 2010 mac pros with the nvidia driver.

The EFI firmware flash is needed to get a boot screen and run pci2 in windows.

For ATI cards it's a totally different story.
 
I'm such a newbie to FCP X and am only operating with a field lap top mac book pro and 16GB of RAM.

Will an external GPU (like Sonnet or what?) help me out more with transcoding speeds from Dragon files with these new updates?
 
I'm such a newbie to FCP X and am only operating with a field lap top mac book pro and 16GB of RAM.

Will an external GPU (like Sonnet or what?) help me out more with transcoding speeds from Dragon files with these new updates?

No external GPU support yet for Macs I'm afraid. Maybe after Skylane and Thunderbolt 3 in new Macs, hopefully end of the year or early next year...
 
FCP X: POOR R3D playback performance using NVIDIA GPUs
http://www.reduser.net/forum/showth...OR-R3D-playback-performance-using-NVIDIA-GPUs

This was mentioned by others (Matthew B) on this thread.

I'm hoping to see if Apple and RED can get better R3D performance with NVIDIA's GPUs particularly the GTX 970.
I left Apple some feedback on this as well. The 970 is too powerful for it to be performing less than AMD 6870.
It's kind of ridiculous.

Brian Timmons
BRITIM/MEDIA
 
Hey Brian -

Nvidia released a new web driver today and it has made a world of difference for my gtx 980 on my 2010 mac pro - I replied on your other thread for more info,

But I'm posting here on this thread too, since I gave a rundown earlier of the need to switch between v5 and v6 of the red workflow plugin to optimize for editing or exporting. With this new web driver from nvidia it seems like I no longer need to keep v5 around.
 
Regarding the 750M - The nvidia web driver as of today allows you to install it to laptops and imacs without patching, and from a quick scan in a timeline it is running 5k R3Ds in fcpx significantly smoother. Haven't tested rendering yet.
 
Thanks Matthew B for emailing the news on the new NVIDIA drivers.
Happy to report that it does solve the playback issue for the GTX 970.
I commend NVIDIA for stepping up their driver support and APPLE and RED for
taking advantage of these NVIDIA cards.

The 750M on the Macbook Pro also plays back better though it seems to fluctuate between perfect playback to slightly choppy. Scrubbing is fine.
I hope further tweaks can solve this as it's so close to perfect playback.

The BRUCE X test performance is still bad on my 970 but this may be due to my card running in PCI 1 bus speed. My MBP actually beats out the 970 in my Mac Pro.
One would think this has to do with the CPU speed but the BRUCE X test uses the export function which almost wholly relies on the GPU.
May look into some Rom hacking solutions or perhaps paying $180 for MAC VIDEO CARDs to hack the ROM.

Glad to see this update.

Brian Timmons
BRITIM/MEDIA
 
Back
Top