ATI Radeon 5870 vs 5770 in FCP X: Big difference?

Barefeats (http://www.barefeats.com/wst10g12.html) did a comparison of the two cards in the past couple of days. Of course it was a day after I bought a 5870 to put into my Early 2008 Mac Pro. The cards are equivalent until the software takes better control of the graphics using OpenCL. That is rumored with Lion. My card runs great, and in theory will give me a longer life, however, right now, today, there does not seem to be much difference.

My card is running great. The biggest difference in the 5770 and the 5870 is in the GPU cores. There are double the GPU cores in the 5870. If you have the extra bucks, it is a tougher call. If you do not, then I would go with the 5770 and leave it at that. I am told that the 5870 is better for gaming, however, I do not play games.

Good luck!

Oh, one other thing, the 5870 comes with two power cables. They have to go into the motherboard at the upper left margin of the fan assembly. Be careful, use the smaller connectors on the cable going into the mother board. The 5870 is a bit tight, but will work fine. Use caution when you push it down into the slot and if you push the tail end of the card, it will pull unseat the card. Once you have it securely in the slot, then bolt it down with the PCI card retention bar inside the Mac Pro. It will take about 15 minutes to install. No special software issues. I reset the PRAM after installing.

Again, good luck!

You are welcome!

I also have the 2008 Mac Pro with radeon 2600.

I could not buy FCP X on it but did on my 2010 Mac Mini. Transferred the app over to the Mac Pro and it works.

Problem is we don't know if it will remain functional in future updates or in Lion.

Irrespective of this I'm also considering the 5770 or 5870 to add a bit of longevity - these cards will not be around forever and no guarantee any replacements will be supported even unofficially in future OS releases on the 2008 machines.

AC

So I've just upgraded my 8-core Mac Pro 3,1 (early 2008) from the Nvidia 8800 to the ATI 5770 to the 5870 in the hope of speeding up my FCPX import/transcode/render/export workflow and can confirm that;

1). the 5770 and 5870 all love the 3,1 platform

2). you can run the 5770 and 8800 side-by-side (great for BOINC where I had the 8800 GPU earning credits while the 5770 drove my screens)

3). there is no measurable difference in the performance of the 8800, 5770 or 5870 on import and transcode

4). there is a 20% performance improvement using the 5870 vs 5770 in Motion type configurations (in this case sclaing a video stream 70% and warping the top and bottom left +/-100) - I did not have time to confirm the result for the 8800

5). I could find NO discernable difference in export speed between Nvdia's 8800, the ATI 5770 or 5870 cards despite Compressor in theory being OpenCL/GPU accelerated

There are two wrinkles that may have affected the above result;

a). I have installed the CUDA drivers from Nvidia

b). I was "forced" to install Quicktime 7 by another app and I am not sure if this impacted the use of teh GPU on export

All of the above tests were run on an 8-core early 2008 (3,1) Mac Pro with 32GB RAM, SSD boot drive, 7200RPM 3Gbps Seagate media drive running Lion 10.7.3, FCPX 10.0.4

I hope that helps - I'll write up the details and post to my blog when I get the time...

ptm (@xthestreams)

http://blog.xthestreams.com

Here is the test that will show the difference in the graphics cards in FCP-X.

Export a project timeline that has not been rendered. This is about the only time in FCP X where the graphics card is really being taxed.

While I didn't do that, I did test the render time for a 50 second clip of 25Mbps 1080p video, the times were roughly 20% improved for the 5870 vs the 5770 (per #4 above), unfortunately I didn't perform the same test for the 8800, I might go back and do so when my machine isnt locked up with real work (i had to render/export out a 40 minute movie last night which took 17 hours, hence my interest in rendering/export times.

Let me know fi you've got any benchmarks you'd like me to run, happy to sdee if I can help get some science behind what appears to be witchcraft...

I would be interested in seeing the differnence exporting an unrendered project timeline with two 5770's installed. In theory it could be twice as fast as just one and perhaps 75% faster than one 5870.

I *did* try a configuration with the 5770 and the 8800 both installed (ie: one Nvidia and one ATI), there was no performance increase at all, in fact the 8800 (whcih I had running "headless") was stone cold when i removed it form the chassis after teh benchmarks - in benchmarks where it was teh only GPU, it had gotten extremely hot.

This leads me to believe that FCPX doesnt take advantage of multiple (different) GPUs, just the "fastest"?

Does anyone know the innards of FCPX well enough to explain what's actuallu going on and how video (semi) pros can speed up day to day tasks (short of buying an iMac).

Maybe FCP-X and Motion would use two 5770's, would be interesting to try out just in case. If that was the case it would be more cost effective than buying just a 5870.

Davinci Resolve will use the more GPUs you throw at it as well as REDCINE-X and perhaps premier CS6.