Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
MP 7,1 Two AMD Radeon Pro Vega II Duo
- Thread starter kittiyut
- Start date
- Sort by reaction score
I’m also interested in the performance gain or rather efficiency of them. Two Duos mean 4 GPU cores right? wondering if the performance is really quadrupled, or very near to, by having those two duos compared to a single Vega II.
Judging off of Apples own performance charts on the main Mac Pro page, Two duos are not worth it for most uses. If you program can utilize multiple GPU's by themselves (without needing CPU) then the performance is scaled very well for example the results for 3D animation rendering.
For other tasks like video editing, Apple only used two radeon vega ii's and not two duos which tells me there is no benefit having 4 graphics cards. If there was they would use them since having better gains over the previous Mac Pro's is a good thing for them to show off.
They also only used 384GB of RAM in all of their testing which may say something as well. In my testing with 192GB of ram I've never seen more than 70GB being used for video/photo editing and its usually around 40-50GB.
[automerge]1578158014[/automerge]
As far as two GPU's FCX will definitely use both but you better have a ton of GPU work to be done. In most of my tests a single Vega II GPU is used between 20-50%. The only time I've seen it maxed out is when I was playing back 14 streams of 4K ProRES RAW in a 16K project with Afterburner and color grading on that as well. The GPU was majorly bottlenecking the CPU so for that crazy kind of workload (same thing as 4 streams of 8K RAW) you definitely want two Vega II's or a duo.
[automerge]1578158026[/automerge]
For other tasks like video editing, Apple only used two radeon vega ii's and not two duos which tells me there is no benefit having 4 graphics cards. If there was they would use them since having better gains over the previous Mac Pro's is a good thing for them to show off.
They also only used 384GB of RAM in all of their testing which may say something as well. In my testing with 192GB of ram I've never seen more than 70GB being used for video/photo editing and its usually around 40-50GB.
[automerge]1578158014[/automerge]
As far as two GPU's FCX will definitely use both but you better have a ton of GPU work to be done. In most of my tests a single Vega II GPU is used between 20-50%. The only time I've seen it maxed out is when I was playing back 14 streams of 4K ProRES RAW in a 16K project with Afterburner and color grading on that as well. The GPU was majorly bottlenecking the CPU so for that crazy kind of workload (same thing as 4 streams of 8K RAW) you definitely want two Vega II's or a duo.
[automerge]1578158026[/automerge]
I can see dual Vega II Duo fit really well in a 3D workflow (once Octane X and Redshift are ported and stable) Most GPU Rendering engines will scale rather well with additional GPU and with the high speed connections between the cores and improvements in Metal (memory pooling to come) large scenes will quickly load on all cores in synchronies allowing for some impressive viewport performance that just now starting to being seen on the Nvidia side with Octane recently supporting NV Link.
The question is - the studios buying these MacPro workstations will probably have a windows or linux based render farm with more GPU power than one of these towers can ever produce. While having fast local hardware is good for the artists will they need TWO Vega II Duo? Sure if they can swing it otherwise I see a single Vega II Duo working really well.
The question is - the studios buying these MacPro workstations will probably have a windows or linux based render farm with more GPU power than one of these towers can ever produce. While having fast local hardware is good for the artists will they need TWO Vega II Duo? Sure if they can swing it otherwise I see a single Vega II Duo working really well.
One good use would be AI calculations and simulations. There's a lot of work going on with GPGPU stuff and AI; mostly from our friends in Team Green, but the same could be done with AMDs.
"Could be", but isn't. Team Red doesn't have the ML/AI libraries that Team Green has.
One of my PhD AI researchers recently commented that "There hasn't been a Nvidia GPU in a Mac for the longest time, nobody in the ML community uses Mac with a Nvidia card anyway (I'd say <1%, disregarding Hackintoshes)".
My AI/ML researchers use SSH into our Linux GPGPU servers to run AI/ML. The MacBooks are for email and the web - not research.
Apple burned the ML/AI bridges by actively killing CUDA on the Apple OSX platform.
I just wonder how multiple GPU cards intend to work in the 7,1. Do I get doubled performance when I add two cards or do they all work independently?
Da Vinci resolve candle benchmark which is a gpu test shows double the performance with 2x vega II compared with 1x. But it’s going to depend on the app used as to whether performance will double with other applications.
Hm ok, let's say you have a 580 and a Vega II in the MP, which GPU will be the main GPU for macOS? Can you select it? Is this related to the connected display?
In Da Vinci resolve you can choose how the gpu’s are used eg one for display etc and the other for GPU processing. Again this is application dependent.
I would assume it depends on what gpu you plug the display into. If you plug the screen into your Vega II, that one will be the “primary” one. It’s just great that we get to chose this again, even giving the option to power two screens by two GPUs, unlike in the trashcan days.
I have the dual Vega II Duo and it's a beast. I use it in Resolve. It shows up as 4 cards in Resolve.
attached are some images.
attached are some images.
No ~
Not immediately. Its possible that demand for the Mac Pro 7.1 is keeping all of these in the current ordering pipeline.