eGPU isn't fully used during rendering
Hello, I switched from a 10.14.6 MBP to a 10.15.1 Mac Mini, both using a Vega 56 in a Asus XG Station Pro enclosure.
When I render the pictures (jpg export), the GPU activity monitor is showing 100 % (with the Macbook Pro) and the Mac Mini uses aprox 50-55 %. Export times are simular, but the score of the Macbook Pro is higher than of the Mac Mini. Not sure if it's a MacOS problem (does storage affect the score?, otherwise both Macs have similar specs with 6-core CPU and 32 GB RAM). Both setups only use one thunderbolt 3 connection, so no other device is interfering. In both MacOS versions C1 is forced to prefer the eGPU.
I have the same trouble with C1 v20, rendering is even slower than with v12.
When I render the pictures (jpg export), the GPU activity monitor is showing 100 % (with the Macbook Pro) and the Mac Mini uses aprox 50-55 %. Export times are simular, but the score of the Macbook Pro is higher than of the Mac Mini. Not sure if it's a MacOS problem (does storage affect the score?, otherwise both Macs have similar specs with 6-core CPU and 32 GB RAM). Both setups only use one thunderbolt 3 connection, so no other device is interfering. In both MacOS versions C1 is forced to prefer the eGPU.
I have the same trouble with C1 v20, rendering is even slower than with v12.
0
-
Please provide CPU model numbers, RAM and drive configuration.
Go to https://www.cpubenchmark.net/ and type in your CPU model numbers - PASSMARK will provide you with a synthetic score that you can use to gauge the potential amount of work each CPU can do irregardless of hyper threading or not. ( I am thinking your OS/Software combination is getting the HT wrong.)
Performance is also dependent on System Memory (RAM) and how well it is used by the software.
For large files the SSD/HD speed is likely a big issue as well.0 -
Mac Mini: Core i7 8700H, 32 GB RAM, 256 GB SSD
Macbook Pro: Core i7-8750H, 32 GB RAM, 1 TB SSD
eGPU: Vega 56
The CPU of the Mac Mini is faster. I cannot imagine that the SSD is the bottleneck.0
Post ist für Kommentare geschlossen.
Kommentare
2 Kommentare