Has Bare Feats helped you? How about donating to Bare Feats?


BARE FEATS LAB - real world Mac speed tests

Do GPU upgrades always speed up Pro Apps?

January 17th, 2019 by rob-ART morgan, mad scientist
February 7th - Removed dual NVIDIAs from graph -- see note in
RED

Before you upgrade your GPU or double up on GPUs, it helps to know if it will make a difference no matter what pro app you run. We used three pro apps to illustrate this point.

GRAPH LEGEND

GTX 1080Ti = single NVIDIA GTX 1080Ti running High Sierra
GTX 1080 = single NVIDIA GTX 1080 running High Sierra
RX 580*2 = dual AMD Radeon RX 580s running Mojave
Vega Frontier = single AMD Radeon Vega Frontier running Mojave
RX Vega 56 = single AMD Radeon RX Vega 56 running Mojave
RX 580 = single AMD Radeon RX 580 running Mojave
HD 7950 = single AMD Radoen HD 7950 running Mojave

Test Mule was a 2010 Mac Pro 12-Core 3.33GHz Xeon X5680, 96GB of 1333MHz DDR3 memory. NOTE: Results for the GTX 1080 paired with GTX 1080 Ti were removed when I discovered an efi error in the unflashed 1080 Ti that hampered performance. The dual NVIDIA GPUs should have produced a looping playback as fast or faster than the dual RX 580s.

DaVinci Resolve

This is a looping playback of a 1080p clip (Candle project) with 3x noise reduction added and rendering on the fly. (HIGHER frames per second = FASTER)

Using the DELIVER function, we transcoded our R3D 5K clip to ProRes 4444 XQ 5K. (LOWER Time in Seconds = FASTER)

Motion

We rendered a RAM Preview of the 600 frame Atmospheric sample project. (LOWER Time in Seconds = FASTER)

Final Cut Pro

We used the BruceX sample project to create a ProRes 4444 XQ 5120x2700 Master File. (LOWER Time in Seconds = FASTER)

WHAT WE LEARNED

In the case of DaVinci Resolve, faster GPUs (and multiple GPUs) do not speed up every process. If a process is CPU heavy and ignores or only makes limited use of the GPU, there will be little or no boost from a faster GPU. One example of this is graph #2 showing Resolve using the Deliver panel to transcode the 5K RED clip to 5K ProRes 4444XQ.

NOTE: Results for the GTX 1080 paired with GTX 1080 Ti were removed when I discovered an efi error in the unflashed 1080 Ti that hampered performance. The dual NVIDIA GPUs should have produced a looping playback as fast or faster than the dual RX 580s.

In the case of Motion, faster (and multiple) GPUs make a significant difference in completing the render of RAM Preview. And, again, NVIDIA GPUs are at a disadvantage running both Motion and Final Cut Pro.

We featured the 2010 Mac Pro tower in this testing session. When we tested the same combination of AMD GPUs connected to the iMac Pro there were more surprises. Let's just say that the 'obsolete' 2010 Mac Pro is still useful for running pro apps.

Comments? Suggestions? Feel free to email me,
Follow me on Twitter @barefeats

WHERE TO BUY APPLE products:
USA readers can help us earn a commission by clicking THIS TEXT LINK or on the Apple display ads at the top and on the sides this page. Then search for the product you seek.

You can also order Apple products from B&H Photo Video.

OUTSIDE USA?

Readers in Australia can support us by clicking our link to the Apple Store AU.

.

WHERE TO BUY THUNDERBOLT 3 EGPU BOXES
The Sonnet eGFX Breakaway Boxes are available at B&H Photo Video.

The AKiTiO Node and Node Pro are available from B&H Photo Video.

WHERE TO BUY AMD GPUS
B&H Photo has the AMD Radeon RX 580 and other AMD GPUs.

We found the AMD Radeon RX Vega 56 on NewEgg. Check eBay for all models of Vegas new and used.

Has Bare Feats helped you? How about helping Bare Feats?

copyright 2019 Rob Art Morgan
"BARE facts on Macintosh speed FEATS"
FEEDBACK or QUESTIONS?