Mac Pro Handbrake Anomaly

[pl1984]

There is a thread in the Mini forum where some discussion was being had on which processor to choose (i5 or i7). In that discussion was a Handbrake benchmark one of the members asked people to run:

https://forums.anandtech.com/threads/benchmark-your-computer-4k-with-handbrake-1-1-and-h265.2544492/

Interested in how it performed I ran the benchmark on my 2010 and 2013 Mac Pros. Surprisingly the 2010 model easily bested the 2013 model despite having lower processor specifications in every way:

The 2010 averaged 6.5 fps whereas the 2013 was only able to achieve 2.5 fps. Despite having more cores and a higher per core clock speed the 2013 was 62% slower than the 2010. See the following post of mine in that forum for details on the configurations of the two systems:

https://forums.macrumors.com/threads/i5-or-i7.2153452/page-9#post-26825854

Is there anyone here that has a 2010 Mac Pro who would be willing to perform this benchmark and report the fps? I also performed this on my Z440 system (4c/8t 3.5GHz, 16GB RAM, Quadro NVS 310 GPU, 1TB NVMe SSD, Windows 10 1803 which achieved 2.2 fps. The Z440 and 2013 Mac Pro are similar in speed making the 2010 the anomaly. I'd like to see what others can achieve with theirs. Ideally the configuration would be similar to my Mac Pro but some variation may be acceptable.

Anyone?

 

[Ludacrisvp]

I don't have a 2010 ... However I do some video encoding from time to time and had planned to test the new machine I've got in comparison to the old one I have. This seemed like a good excuse to test.

But I do have a 2008 with 2x X5482 (8 cores total @3.2) and a hack with Haswell E5-1650 v3 (6 core / 12 thread @3.5) Note that neither machine is running windows right now so these are 10.13.6 results... i don't believe either machine supports QuickSync.
Using their file and presets using Handbrake 1.1.2 I get pretty low perf on both systems with unique SSD for source and target media in both machines.

Started both encodes at the same time, haven't completed by the time of this posting...
2008 - avg ~1.03fps when it was 25% encoded
hack - avg ~3.22fps when it was 82% encoded

 

[pl1984]

I don't have a 2010 ... However I do some video encoding from time to time and had planned to test the new machine I've got in comparison to the old one I have. This seemed like a good excuse to test.

But I do have a 2008 with 2x X5482 (8 cores total @3.2) and a hack with Haswell E5-1650 v3 (6 core / 12 thread @3.5) Note that neither machine is running windows right now so these are 10.13.6 results... i don't believe either machine supports QuickSync.
Using their file and presets using Handbrake 1.1.2 I get pretty low perf on both systems with unique SSD for source and target media in both machines.

Started both encodes at the same time, haven't completed by the time of this posting...
2008 - avg ~1.03fps when it was 25% encoded
hack - avg ~3.22fps when it was 82% encoded

Thanks for testing this. I have a 2008 2 x quad core 2.8GHz Mac Pro which returned a rate of 1.5 fps which will likely be in line with what yours will score when it completes. I've re-run the benchmark on my 2010 and it continues to score in the 6's every time.

 

[Slash-2CPU]

Mac Pro 2009 flashed to 2010 firmware. 3.46GHz 6-core W3690. 24GB(8x3GB) DDR3-1333:
encoded 1806 frames in 911.77s (1.98 fps), 11821.39 kb/s, Avg QP:29.08

My Win7 5930K 6-core water-cooled and clocked at 4.2GHz doesn't even break 6.5FPS. There's no way that a 2.8GHz quad-core is beating a 3.46GHz 6-core 2010 Mac Pro and an overclocked, water-cooled 5930K.

Working drives on both systems are 960 Pro 512GB SSD's.

 

[tommy chen]

remember: a xeon CPU has no IGPU for quicksync

HEVC aka H265 need it !

 

[pl1984]

Mac Pro 2009 flashed to 2010 firmware. 3.46GHz 6-core W3690. 24GB(8x3GB) DDR3-1333:
encoded 1806 frames in 911.77s (1.98 fps), 11821.39 kb/s, Avg QP:29.08

My Win7 5930K 6-core water-cooled and clocked at 4.2GHz doesn't even break 6.5FPS. There's no way that a 2.8GHz quad-core is beating a 3.46GHz 6-core 2010 Mac Pro and an overclocked, water-cooled 5930K.

Working drives on both systems are 960 Pro 512GB SSD's.

I agree. The numbers I was obtaining were definitely unusual. As mentioned in the other thread I re-ran the benchmark this morning and achieved the following result:

work: average encoding speed for job is 1.154117 fps​

This is inline with what I expected to see yesterday morning. I cannot account for why it is different today compared to yesterday. I can only assume yesterdays result was operator error. It's the only thing that makes sense.
[doublepost=1542810422][/doublepost]

remember: a xeon CPU has no IGPU for quicksync

HEVC aka H265 need it !

That is correct. However this benchmark specifically has the operator disable QuickSync so that only the CPU is tested.

 

[tommy chen]

That is correct. However this benchmark specifically has the operator disable QuickSync so that only the CPU is tested.

ok, I had overlooked that

 

[h9826790]

remember: a xeon CPU has no IGPU for quicksync

HEVC aka H265 need it !

And H265 does NOT need QuickSync, it can speed up encoding process, but not a requirement.

 

[Ludacrisvp]

And H265 does NOT need QuickSync, it can speed up encoding process, but not a requirement.

Indeed. I’ve done lots of h265 encoding on my 3,1, 1080p video takes a while but provides a great file size on BD backups vs h264 if comparable quality.