I know there have been a few threads in the past (had a good look through!) on GPU for the still popular Mac Pro 5.1, and I am reading lots about the Radeon RX580 as being a good option in a 4;1 or 5,1 Mac pro.
I’ve been looking at the 8Gb Sapphire Pulse, and from what I see of CB scores on the PC side, this card offers somewhere around the 125 fps mark. I don’t expect the same on the Mac, but I wondered if anybody had any experience with this card on Mac OS, and if their general viewport capabilities felt faster/smoother, as well as for use with Pro Render? - something that I as yet, have not been able to try out!
Your viewport will not be faster, maybe just a little bit, the problem with your machine is the limited single thread performance, so you will likely see half the CB score comparing to PC or recent Mac. Computing will be a lot faster though, it will be a nice upgrade for everything running OpenCL(and maybe Metal if supported on older machine). One of my students upgraded his PC to the 580 from a medium range Quadro card and he is quite happy, also for the price.
This is what has stalled me from trying out a different GPU previously, because the differences may not be great enough to justify the spend.
In the case of the 580 at around £300, it’s still not as huge an investment as some Nvidia options i’ve considered previously, but regardless, if theres no great advantage or speedup in the main software I use, (C4D, then it may be folly to spend anything that could ultimately be used for a newer machine at some point. The reality is however, that a machine that is significantly quicker in all these aspects is still going to be a huge outlay.
Does anybody else have any experience with this or other cards added to an older Mac Pro and Cinema usage?
I have a 5,1 with an RX480. Open GL CB score is identical to the 5870
I use it to test ProRender, and other Open CL workflows, in which it performs nicely, but unless you have at least £300 worth of OpenCL projects coming your way, I’d say hold off.
I tested Octane on my PC with a Nvidia 1050Ti and although large projects are out of the question, it seems to work with it, and Redshift as well.
Having said that, I have no idea how good, or bad for that matter, the Nvidia drivers are on the Mac Pro.
I’m sure I read somewhere recently that driver support for the Nvidia cards had diminished, but that it had started up again with news of the release of a new MP somewhere down the line, which would be advantageous for any of us interested in using the established GPU renderers on the newer machines. Assuming any of us mortals can justify the costs!
Aside from the Cinebench scores, would there be any other advantages in the GPU RAM increase from 1Gb to 8Gb in C4D? - I know this should be beneficial in some Adobe CC uses, but not sure with Cinema!?
More GPU RAM means that more textures, geometry and so forth can be loaded directly into GPU memory for - sometimes, depending on the application - faster OpenGL/DirectX performance. If you are rendering with a GPU renderer, it also means that 8GB of scene data can be rendered, not 1GB. 1 GB is very low in 2018. My 2017 model work laptop has 6GB GPU RAM.
Also, Google the GFLOPS rating for your old GPU and for the new one you want to buy. That is the rough compute power of each card for GPU accelerated applications.
In any application that is GPU accelerated, whether through CUDA, OpenCL, DirectX or OpenGL or anything else, the GFLOPS rating of the GPU will give you a rough idea of how much faster you can expect the new GPU to perform.
So if your old card does 3,000 GFLOPS max and the new one is rated at 6,000 GFLOPS max, you are roughly looking at 2 X faster GPU computing with the new card, provided that the application being accelerated makes efficient use of those GFLOPS. The fastest, most expensive GPUs are currently at around 13,000 to 15,000 GFLOPs.
GFLOP which is short for GigaFLOP means “Billion Floating Point Operations Per Second”. It is a rough indicator of how many floating point math operations the GPU can perform in 1 second on average.
GFLOPs may not have huge bearing on the OpenGL viewport in C4D aside from some extra FPS. I don’t know exactly how the C4D viewport uses the GPU.
But in any scenario where the GPU is doing the calculating - such as running a GPU renderer, or GPU accelerated video encoding - GFLOPS matters a lot.
Thanks for the further info guys, I appreciate your input.
I think that upgrading the processors looks to be a worthy option all round then, since a 33% increase in CPU render speed is significant enough for about £250, and will also have at least some benefit on the viewport based on the info above.
Maybe Apple will surprise us with the new Mac Pro, and it wont be priced completely out of the ball park… :]
Going from a 6 core 2.66 to a 12 core 3.46 is like buying a “brand new” MacPro, with speeds approaching the newest trashcan Macs. I went from a 12 core 2.66 to the 12 core 3.46 and my render times dropped by about 25%. Coming from a 6 core, you will have an even bigger bump in speed, and the whole Mac will feel snappier! Well worth the money. Make sure you get as much RAM as you can afford as well.
I’ll actually be going from the same machine specs as you - 12 core 2.66 to 12 core 3.46! - it’s currently running 48Gb of RAM as well, so I would imagine thats enough! - although, I do have 2 free slots, so I guess I could take it up to 64Gb!?
Did you do the CPU upgrade yourself? - I’ve been looking it up online, and it looks a lot easier than I thought it might be!
The snappiness of your system is affected mostly by the single thread of the CPU and going from a dual 2.66 to a 3.46 will boost your performance about 20%, if this worth your effort and money it’s your choice, otherwise you can wait for a new machine or choose right now an iMacPro, even the base model will be easily twice as fast concerning system and viewport snappiness.
Agreed. I have now bought a pair of X5690 3.46Ghz for £210, and will upgrade my existing MP for now. It’s the most economical choice.
Tbh, I can’t currently imagine justifying £5k at the bottom end on an iMac Pro, regardless of the doubled speeds in some areas.
I understand Adam, my point was that since you are looking for snappiness a 20% increase is very little(that’s why I’ve never upgraded anything beside RAM on Mac or PC), a 100% increase on the other end is quite a bit and it’s almost as far as you can go for now (no matter if you are on Mac or PC). There’s a good chance that next Mac Pro will be slightly slower than iMacPro in this regard due to the dual CPU choice so be aware of that ad well;)
Understood! - as much as I crave a truly snappier system (only really C4D viewport really), I will have to wait until I can invest in a newer machine with a higher single core speed. With this old work horse though, I should be looking at a 30% render speed increase with the new chips, and on the multi-thousand frame renders i’m doing a lot of lately, this can only be a good thing!
And when that faster desktop machine becomes a reality, the MP will be an even more useful Team Render node.
I did not do the upgrade myself. I was put in touch with a guy in LA who will either take your MacPro tray and upgrade it, or send an upgraded tray that you swap out and send back to him. It does seem pretty easy to do the CPU upgrade yourself, but this guy does hundreds of upgrades and guarantees the work. I only have one workstation, and could not afford any down time or problems. It’s been about a year since I upgraded and all is well!
One note: I got the upgraded tray with 64Gb RAM and initially my MacPro wasn’t recognizing all of the RAM SIMMS. Zapping the PRAM fixed that.
I also upgraded my video card to a Radeon R9 280X with 3Gb RAM. Not sure that did much to speed things up, but it’s fine for now. If I decide to try Octane 4 when it is released for MacOS, I may re-think the video card.
I contemplated that kind of service, but its about 3x the cost, and until I saw the procedure, I wasn’t aware of how relatively simple it should be. It would certainly make more sense going from a single CPU tray to a dual though.
Another possible upgrade you may consider, if you haven’t done it already is to get an SSD as a boot drive. It has made my workflow very much faster, and you can use it in any future build. I did spend a good chunk and got a 1TB SSD 3 years ago but haven’t looked back since.
Yes, I did put a 240Gb in as my boot drive some years ago, which was all I could afford at the time. Its been enough to hold the system and apps, but my user folder resides on a standard SATA HDD, so Im definitely not feeling the full benefit of the speed, other than in booting and app launch.
Ive been looking at replacing it with a larger volume now that prices are far more favourable, as my Maxon folder alone is pretty enormous!
Definitely a recommended upgrade though!