nMP vs CompressHD

[pcconvert]

I am using Matrox CompressHD (H.264 encoder PCI card) on my 3.84 quad hackintosh with ATI Radeon HD6870 and Mavericks. I was thinking about getting nMP and was wondering what would be the encoding speed difference. With latest FCPX and Compressor I see some improvement in the software only encoding (OpenCL?) but it's nowhere near as fast as CompressHD. Does anyone have real life comparison experience between the two - nMP and CompressHD? How efficient is the OpenCL power of those n500 cards?

 

[linuxcooldude]

I am using Matrox CompressHD (H.264 encoder PCI card) on my 3.84 quad hackintosh with ATI Radeon HD6870 and Mavericks. I was thinking about getting nMP and was wondering what would be the encoding speed difference. With latest FCPX and Compressor I see some improvement in the software only encoding (OpenCL?) but it's nowhere near as fast as CompressHD. Does anyone have real life comparison experience between the two - nMP and CompressHD? How efficient is the OpenCL power of those n500 cards?

OpenCl is not software encoding. It is using the GPU's for hardware encoding. I use a compressHD card too in a Mac Pro 2009.

I find they tend to be slow in new drivers when upgrading the operating system. When they did finally upgrade the drivers it caused my system to constantly reboot until I removed them.

And factor in you have it on a hackentosh probably makes matters worse.

CompressHD only does H264, so other formats are not supported.

 

[Stephent]

I am using Matrox CompressHD (H.264 encoder PCI card) on my 3.84 quad hackintosh with ATI Radeon HD6870 and Mavericks. I was thinking about getting nMP and was wondering what would be the encoding speed difference. With latest FCPX and Compressor I see some improvement in the software only encoding (OpenCL?) but it's nowhere near as fast as CompressHD. Does anyone have real life comparison experience between the two - nMP and CompressHD? How efficient is the OpenCL power of those n500 cards?

Many of the h.264s I've been outputting on my nMP have been near real time. maybe 10% slower. Would have to take a closer look. 6-core d700 64GB RAM

 

[pcconvert]

I thought FCPX is using OpenCL to use GPU's power for encoding (?). Yes Matrox is kinda slower in drivers but they did release Mavericks and my hack works fine, no rebooting, sleeps well etc. I am more interested in the performance comparison - CompressHD vs. nMP...

----------

Many of the h.264s I've been outputting on my nMP have been near real time. maybe 10% slower. Would have to take a closer look. 6-core d700 64GB RAM

Thanks! This was what I've been looking for - could you post some example real number? (say 5 min clip 1080p@5mbit took x minutes to render? One pass/two pass etc.) This would be great!

 

[handsome pete]

I thought FCPX is using OpenCL to use GPU's power for encoding (?). Yes Matrox is kinda slower in drivers but they did release Mavericks and my hack works fine, no rebooting, sleeps well etc. I am more interested in the performance comparison - CompressHD vs. nMP...

----------



Thanks! This was what I've been looking for - could you post some example real number? (say 5 min clip 1080p@5mbit took x minutes to render? One pass/two pass etc.) This would be great!

There's really no solid answer on this. It all comes down to the specific output settings. If you're looking to do a single pass encoding, then I'm afraid the Mac Pro is going to fall short of many other options out there, even the iMacs. The Xeons don't feature quicksync, which allow the consumer version Intel chips the encoding acceleration they're capable of in this regard. However, that becomes irrelevant with dual pass encodings, so the Mac Pro should still shine there.

 

[Stephent]

I thought FCPX is using OpenCL to use GPU's power for encoding (?). Yes Matrox is kinda slower in drivers but they did release Mavericks and my hack works fine, no rebooting, sleeps well etc. I am more interested in the performance comparison - CompressHD vs. nMP...

----------



Thanks! This was what I've been looking for - could you post some example real number? (say 5 min clip 1080p@5mbit took x minutes to render? One pass/two pass etc.) This would be great!

Have a few outputs to do tomorrow will let you know