This is what I get for having an older GPU. Hand break is nowhere near that speed for me. More like 70% of real time when doing a single pass on an MKV from a Blu-ray for example. (Maxed out on the CPU though...)
BruceX on FCPX 10.1 and 10.9.1 (ProRes) -- 42 seconds
Dual Radeon 5770 1GB VRAM (Slots 1 &2)
BruceX on FCPX 10.1 and 10.9.1 (ProRes) -- 33 seconds
I was expecting better than that given the comment from Barefeats (assuming 7970 is ~ 20% faster than 7950)
By my calculation using the single 7970 vs. 7950 factor I'd guess a dual 7970 on this system would be about 24 second, so something "slightly slower" than that for dual 5770 could be anything from 25 seconds up... maybe as much as 28 seconds, but 33 seconds is quite a bit slower than "slightly slower".
Have had difficulty physically installing 2nd 5770 into Slot 4 to save a pcie slot for another Accelsior card.
It would seem that even with the 2nd GPU in slot 4 that there should be enough room to fit low profile 2.5" drives in the bays above -- at least for bay 2.
72 second rendering to ProRes LT on my main Hackintosh.
It's a i7 3770 Quad Core 3.4Ghz, 32GB RAM, GTX670.
Scores 14108 on Geekbench3 64 Bit and 631 on LuxMark Sala.
So that sounds about right, but it still unclear whether 2 or more GPUs will work on a Hack.
I guess you could add more GPUs via thunderbolt right?
The entry level nMP does look like great value for money, but I hope the £2k I spent on the entire suite, RAID, IO and LED monitors will last for a few more years...
FCPX 10.1 OSX 10.9.1 -> fail
FCPX 10.1 OSX 10.9.1 -> 20sec, prores
FCPX 10.1 OSX 10.9.1 -> 20sec with project set to 4K, H.264
Seeing how well dual 5770 are doing (especially as their combined luxmark scores are lower than the 7970) I'd be interested to know how well two 7750 would do. They match or outperform the 5770 at openCL tasks and are more energy efficient, drawing all the power they need from the pcie slot.
I'm even more interested in whether cards need to be matched. We're not making use of crossfire here so I don't see why they'd need to be. It would be great to just add a 7750 to say a 7970 or 7950 for a little boost for £70. Or potentially sizeable boost given the 5770 performance. For mac pros this would seem ideal given that no additional power cable is required if the combined 7950 and 7750 could outperform 2 5770s.
For those with mixed cards to test, try putting your fastest card in PCI slot 1. then put your slower card whereever. Hook your display up to your slower card. From what I saw in Alex4d tags regarding GFX cards and some reading on Anandtech, I think the OS will use the display card as a raster and the second card as an Auxillary. If the Auxillary card is your faster card, I think that will bebnnift the most. I don't hae two cards so I can't test myself. But I'd be interested to see if anyone can see a difference.
Given the problem with h.264 the more recent numbers being reported are coming in for Prores, but I am seeing some saying Prores LT. There is also other Prores settings. I think some of this variation is making comparisons difficult. Can anyone clarify what we are supposed to be comparing?
FWIW I did get my 2nd 5770 moved from slot 2 to slot 4 & the results was a slow down from a time of 33seconds using slots 1 & 2 to a time of 36.5 seconds using slots 1 & 4. That shouldn't be a surprise really.
However, I did try connecting the monitor to the slower GPU card (slot 4) as someone suggested: this made no difference in performance.
The other point I found is that although the instructions say to use the fastest storage location possible, for this test I think that is unnecessary. In fact, since my fastest storage is on the PCIe that's what I was using. Turns out I get slightly fractionally faster times when using slower storage... even external usb thumb drives!
>So in my case seems like faster conencted storage will improve export times although an earlier poster seemed to have the opposite experience.
I suspect the reason why I see faster times when using slower storage is because my fastest storage is located on PCI express so I wonder if the high activity being pumped by both GPU cards is in some way competing/conflicting with the writing to the Accelsior card.
That's all I can think of when I see (marginally but consistently) faster times when saving the file to a SATA hdd or FW HDD, or USB flash.
Noticed you have access to multiple 5770s, a 5870, and a 7950. Have you tried BruceX combining the 7950 and the 5770?
I'm not an expert on power distribution, but for what it's worth I tested BruceX with a 5770 with a GTX 285 and had no power problems. The GTX 285 has a max draw of 204 Watts compared to the 7950's 200 Watt max. Granted these are internet-enabled numbers so double check of course (
Won an eBay auction for the 5770. The flashed version are on average 50 bucks cheaper but I decided to look for apple one at the end since I am using a Cinema Display having the 2nd video card with mini display port may come in handy in emergency cases..
Once I receive it I will do the test once again to see how the dual 5770 system work for me
I don't want to give up my internal hard drive bays, so I'm going to put in slot 3 configured as a 4 lane? (x4)
It looks like I only give up 5-6% performance of the card using that speed slot, so it is fine.