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

DustinB85

macrumors newbie
Original poster
Nov 24, 2011
26
0
OKAY SO, this is yet another should I update type thread. But this time it is a tad different since I've done a couple upgrades in the pasted and looking for a different set up this time.

1.) Has anyone used the R9 290X GPU in a 2009 Mac Pro with either Adobe Premiere Pro CC or Final Cut Pro??? I am looking for the best GPU setup for a relatively decent cost. I've tried both the GTX 670 and Sapphire 7950 Mac Edition in the past. I am looking forma tad more horsepower than either of those. So I am looking at the R9 290Xs, GTX 980 or possibly a dual card set up. I don't play games on the MP just strictly concerned about rendering performance/playback during editing and playing nice with Adobe CC and FCPX. All opinions and help greatly appreciated.

2.) I am going to upgrade from the quad core 2.66 to the 6 core 3.33. Any objections or better options from that cost standpoint? I would love to go to 8 core + but that requires an whole different tray and the cost jumps pretty heavy at that point.

Other than that I plan to upgrade to a 4k monitor setup since I am now shooting 4k on a few of the cameras. If any has any input other than those two items to get my 2009 performing as well or better than the cost of the new Mac Pros or even top self iMac let me know. Other specs are 32 GB ram, 256 GB SSD boot off PCI, Raid 0 internal HD (3*1 TB), fourth bay has 3TB HD. PCI eSata card with 4 esata ports 4 running external HDs in a RAID 5. 2nd CD tray used to hous another SSD as a scratch disk drive.
 
1. Yes people use that card. You need supplemental power supply or use Pixlas Mod to tap power direct from the PSU.

2. The 3.33ghz is fine. I tested Premiere, After Effects, Media Encoder and they barely use more than a few cores for rendering. Most people here (like me) who have 12 cores got the CPUs cheaply but those extra cores are mostly doing very little.

If you are going to configure two RAIDS you need to forget about El Capitan. There is no more RAID configuration assistance in Disk Utility.

To be honest, if you are only going to use your Mac Pro for editing video then you should leave the cMP alone. I tested the X5690 3.46ghz CPU in a 4.1 vs the E5-1650v2 3.5ghz CPU in the 6.1. In Media Encoder the newer machine was almost twice as fast at rendering 4K. You can also attach your desired RAIDS over Thunderbolt. If you use something like a Drobo then you won't have to worry about El Capitan's lack of RAID set up support.
 
From what I'm reading on Hackintosh boards, the Hawaii drivers (R9 290/390[X]) are still very buggy (freezes, black screen, no dual displays, ...).
Maybe @MacVidCards or @netkas can confirm if those issues still persist?
 
1. Yes people use that card. You need supplemental power supply or use Pixlas Mod to tap power direct from the PSU.

2. The 3.33ghz is fine. I tested Premiere, After Effects, Media Encoder and they barely use more than a few cores for rendering. Most people here (like me) who have 12 cores got the CPUs cheaply but those extra cores are mostly doing very little.

If you are going to configure two RAIDS you need to forget about El Capitan. There is no more RAID configuration assistance in Disk Utility.

To be honest, if you are only going to use your Mac Pro for editing video then you should leave the cMP alone. I tested the X5690 3.46ghz CPU in a 4.1 vs the E5-1650v2 3.5ghz CPU in the 6.1. In Media Encoder the newer machine was almost twice as fast at rendering 4K. You can also attach your desired RAIDS over Thunderbolt. If you use something like a Drobo then you won't have to worry about El Capitan's lack of RAID set up support.

Thank you for the reply. Couple follow up questions.

1.) What GPU were you running in the 4.1 when you compared it to the 6.1?

Other thoughts;

The internal RAID 0 is already setup prior to my install to El Capitan so that really isn't a big deal. Plus I have a raid card for my external drives. I would see no speed benefit moving over to a thunderbolt set up at this time since the bottleneck is the hard drives (non SSDs) so the PCI eSATA card has more than enough data transfer speed.

Moving to a 6 core 3.33 will only cost a couple hundred dollars so that is money well spent. I see no benefit ( and nobody ever has made a case for it to be worth it ) to the cost upgrade the tray so I can move to the 8 or 12 core setup. At that point it I'd be comparing the 4.1 to the 6.1 altogether.

I guess at this point my only real play is to figure out the best GPU set up for a 4.1 they will play nice with Adobe and Final Cut. Like a said I want to have a GPU card/setup that's a step up from the 7950 for video editing purposes.
 
Thank you for the reply. Couple follow up questions.

1.) What GPU were you running in the 4.1 when you compared it to the 6.1?

Other thoughts;

The internal RAID 0 is already setup prior to my install to El Capitan so that really isn't a big deal. Plus I have a raid card for my external drives. I would see no speed benefit moving over to a thunderbolt set up at this time since the bottleneck is the hard drives (non SSDs) so the PCI eSATA card has more than enough data transfer speed.

Moving to a 6 core 3.33 will only cost a couple hundred dollars so that is money well spent. I see no benefit ( and nobody ever has made a case for it to be worth it ) to the cost upgrade the tray so I can move to the 8 or 12 core setup. At that point it I'd be comparing the 4.1 to the 6.1 altogether.

I guess at this point my only real play is to figure out the best GPU set up for a 4.1 they will play nice with Adobe and Final Cut. Like a said I want to have a GPU card/setup that's a step up from the 7950 for video editing purposes.

I compared H.264 encoding between both machines, which at first I thought would be accerlated by the GPUs but it turned out that it was purely CPU. The 6.1 was about twice as fast as the 5.1. But if you want to see a massive speed up without buying a new system then install Boot Camp on your machine. H.264 encoding was up to four times faster in Windows than OSX with Media Encoder CC 2015 on the same machine.

It's fairly cheap to upgrade your current machine and you will obviously see a great increase per dollar spent. It will also help maintain your resale value so that if you are impressed by a future Mac Pro model at least you can sell your machine at a good price.
 
  • Like
Reactions: DustinB85
I compared H.264 encoding between both machines, which at first I thought would be accerlated by the GPUs but it turned out that it was purely CPU. The 6.1 was about twice as fast as the 5.1. But if you want to see a massive speed up without buying a new system then install Boot Camp on your machine. H.264 encoding was up to four times faster in Windows than OSX with Media Encoder CC 2015 on the same machine.

It's fairly cheap to upgrade your current machine and you will obviously see a great increase per dollar spent. It will also help maintain your resale value so that if you are impressed by a future Mac Pro model at least you can sell your machine at a good price.


Yeah the Adobe mercury playback engine is not compatible (or does not recognize all GPUs), there is a hack to make it do so but unless you do that it only allows you to use CPU only rendering which is the slowest thing on earth and I wouldn't wish that on my worst enemy. The reason Boot Camp is faster is because the windows version of Adobe recognizes more GPUs that's most likely allowing you to use the full Mercury engine playback without the hack you have to do on the OSX side.
 
Yeah the Adobe mercury playback engine is not compatible (or does not recognize all GPUs), there is a hack to make it do so but unless you do that it only allows you to use CPU only rendering which is the slowest thing on earth and I wouldn't wish that on my worst enemy. The reason Boot Camp is faster is because the windows version of Adobe recognizes more GPUs that's most likely allowing you to use the full Mercury engine playback without the hack you have to do on the OSX side.
Doubt it. We have a thread here where I ran the test thinking the GPUs accelerated H.264 rendering and encoding. It turns out CC2015 doesn't do that at all and you can read the technical reasons in detail in our conversation below...

https://forums.macrumors.com/threads/real-world-premiere-media-encoder-speed-test.1932390/
 
Doubt it. We have a thread here where I ran the test thinking the GPUs accelerated H.264 rendering and encoding. It turns out CC2015 doesn't do that at all and you can read the technical reasons in detail in our conversation below...

https://forums.macrumors.com/threads/real-world-premiere-media-encoder-speed-test.1932390/

That test was ran with a GTX 980 GPU. Adobe premiere Pro and media encoder will not allow you to enable the GPU mercury playback engine with that card on OSX. It forces you to use the mercury playback via CPU (software). It doesn't as a matter if you install all the Cuda drivers etc. I ran a GTX 670 prior to my 7950 which is the main reason I had to learn the hack from PP to recognize it as a "certified" card. The rendering was night and day difference. Now windows might be different on out it deals with not supported GPUs as long as you have the correct drivers maybe Premier Pro allows the Mercury Play Back to work correctly without a work around. I ran Windows 8 via Boot Camp on my mac pro for a long time and I didn't notice too much of a difference ever is far as performance. However, I do know though that some Adobe products have been known to run slightly faster on windows but nothing to write home about.

Here shows the list of cards Adobe recognizes for GPU (full Mercury engine payback)

https://helpx.adobe.com/premiere-pro/system-requirements.html
 
That test was ran with a GTX 980 GPU. Adobe premiere Pro and media encoder will not allow you to enable the GPU mercury playback engine with that card on OSX. It forces you to use the mercury playback via CPU (software). It doesn't as a matter if you install all the Cuda drivers etc. I ran a GTX 670 prior to my 7950 which is the main reason I had to learn the hack from PP to recognize it as a "certified" card. The rendering was night and day difference. Now windows might be different on out it deals with not supported GPUs as long as you have the correct drivers maybe Premier Pro allows the Mercury Play Back to work correctly without a work around. I ran Windows 8 via Boot Camp on my mac pro for a long time and I didn't notice too much of a difference ever is far as performance. However, I do know though that some Adobe products have been known to run slightly faster on windows but nothing to write home about.

Here shows the list of cards Adobe recognizes for GPU (full Mercury engine payback)

https://helpx.adobe.com/premiere-pro/system-requirements.html

Please read the discussion in detail. H.264 encoding cannot be done on GPU (there is a hackish plugin out there). The test results therefore were CPU only. This was proven when I disabled the GPUs and set to software rendering. The results were the same. If you are going to encode to another codec then that's another discussion because I haven't tested all of them.

So to get back to the point, if you want a cheap upgrade then by all means the six core Xeon upgrade will help until you are ready for a new machine. And using Windows will help you see even more speed increases if you are using Creative Suite.
 
Please read the discussion in detail. H.264 encoding cannot be done on GPU (there is a hackish plugin out there). The test results therefore were CPU only. This was proven when I disabled the GPUs and set to software rendering. The results were the same. If you are going to encode to another codec then that's another discussion because I haven't tested all of them.

So to get back to the point, if you want a cheap upgrade then by all means the six core Xeon upgrade will help until you are ready for a new machine. And using Windows will help you see even more speed increases if you are using Creative Suite.


Yeah need to go back and look at that in detail. I was unware/never noticed that when outputing to H.264 (which is obviously my main compression used for Youtube/Vimeo uploads).

I acutally been using FCP X for the last 10 projects or so because I'm only using the stock GT 120 until I pick a new GPU to try. Both are insanely slow obviously with that in there but final cut runs a tad better. Adobe PP just laughs at me if I try to use it right now and After Effects is really out of the question lol.

Thanks for the input, I'm definitely going buy new Mac Pro in the next year or two. Waiting to see the next refresh on it. I bought my mom the 27inch 5k iMac earlier this year and I used to edit one Video using Final Cut Pro/ DaVinci Resolve and Motion to see how it performed and outside of the nice display it didn't seem to blow away my Mac Pro all that much when I was still using the Radeon 7950 GPU. And I still shoot mainly on my Canons which are all 1080p so until my DSLRs are 4k my current Mac Pro still has years in it and can easily handle the GoPro Black's 4k and iPhone 6s 4k when I shoot with those.
 
In Premiere, if outputting to ProRes, does this utilize all 12 cores of the cpu? Wondering if a video editor, using Premiere and outputting to ProRes, would benefit from a 12c MP or a hex nMP?
 
Last edited:
In Premiere, if outputting to ProRes, does this utilize all 12 cores of the cpu? Wondering for a video editor using Premiere and outputting to ProRes would benefit from a 12c MP or a hex nMP?
I'll test in a moment for you (OSX only, no ProRes on Windows). CC2015 dropped the multi CPU configuration option in the preferences in favour of a new rendering engine which is a work in progress they hope to complete next year.
 
Last edited:

I used the following short 10 second clip (first download link) I found online to really stress the CPUs. It's a 4.5K RAW RED-ONE clip.

http://nonlinear.ru/index.php?/topic/176-скачать-red-45k/

I didn't apply any effects as transcoding it to ProRes422 will be sufficient enough to see what is going on in the CPU usage monitor. I did a software, OpenCL and CUDA render with the basic setting attached. Render times were:

Software : 55 seconds
OpenCL : 11 seconds
CUDA: 11 seconds

CPU usage was quite high, but not maxing out. It appears to be higher when rendering with the GPUs, that's because the CPUs also have to do a little extra work sending the data to the graphics card. Bear in mind each of these cores are almost half as powerful as a modern Skylake. So you could get these results today with half as many cores if a six core Skylake existed.
 

Attachments

  • image.png
    image.png
    101.4 KB · Views: 83
  • image.png
    image.png
    117.7 KB · Views: 88
  • image.png
    image.png
    143.9 KB · Views: 82
  • image.png
    image.png
    134.4 KB · Views: 76
Last edited:
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.