New graphic card for Photo editing?

[AT.HA]

Hi,

I will choose a new graphics card for my Mac Pro 5,1 (mid 2010) because my own report a problem and I want to upgrade later for macOS Mojave.

Apple  suggest some compatible cards: https://support.apple.com/en-us/HT208898

I not a gamer and I edit video rarely. 

My card will work for Raw Photo Editing with Affinity Photo, graphic design with Affinty Designer and DTP with future Publisher.

Between these two cards (Apple suggestion; and I know they will not work with Mac Boot Screen; no problem for me) which one should I choose?

Sapphire Radeon RX 580 Pulse 8GB GD5

and 

MSI Gaming Radeon RX 560 128-bit 4GB GDRR5

 

For photo editing/graphic design/Dtp is important 4GB VRam vs 8GB VRam? Makes real difference?

What other specs should I pay attention to?

Thanks in advance for your help.

Regards.

 

 

[Dan C]

Hi AT.HA

Provided the graphics card is Metal-capable, Affinity will be able to use GPU acceleration to provide a better application experience. Other than this, it is unfortunately company policy that we don't recommend specific hardware, as it can cause liability issues. My apologies I can't help further here, however other users on the forums may be able to offer their own personal advice!

[AT.HA]

Hi Dan,

thank you for your help. Both cards are Metal capable.

I understand company policy.

But, you can please confirm that for still images work (photography) I will take advantage of more VRAM (8GB instead of 4GB? Just to be sure that I fully understood your words " Affinity will be able to use GPU acceleration to provide a better application experience.".

My current problematic and very old card (2010) is 1GB VRAm and it is slow to process several tasks, like open files, export, etc. Batch export is a nightmare; as the worst example.

VRam is not only important for moving images; right?

Thanks .)

[SrPx]

EDIT: Went for lunch, and I had left my post without hitting "submit", so, I hope I am not repeating some of what Dan C just replied. In any overlap, whatever he said has preference... I can indeed speak about brands, not tied by any legal matter or NDA, so, there you have it below, a bloated and useless arguable opinion-full take at it, from me.   

Almost nobody can give you an accurate answer about these other than Affinity developers themselves, or, people having very particular info (not me), and also, tech savvy enough, quite deeply, into that specific matter and Affinity apps. So , take all what I am telling here (and anyone not included in the mentioned groups) with, not just a grain, but a huge truck load of salt. Both are good cards, even if I have always been an nVidia fan and user, but I have to say AMD's offerings tend to be very nice on our pockets, and they are finally delivering true performance, both in CPUs and GPUs, even if in graphic cards they're keeping a somewhat slower pace (u know, one can't do all perfect at a time...). And that's even unfair, as Vega's cards are good, just expensive, and in the integrated cards, they clearly have beaten in performance all Intel's integrated ones (AMD's 2200g and 2400g CPUs integrated cards do beat intel's latest integrated, in benchmarks) 

Among those two cards you mention, there's a large difference in "power" (in processing capability, haven't looked into power consumption) in favor of the 580. But it tends to be as well in price.  I've heard 580 is pretty much similar in performance (I believe 1060 gives slightly better average frame rate in games played at 1080p (higher resolutions, and there the the 580 is better), but I don't care about games) to one of my fav cards ever in price/performance, lately, the nVidia 1060 (which is an amazing purchase, btw). Already that gen is a bit of old news (nvidia has launched its new thing, RTX platform), but is still impressive. Now, If I, as someone who edits video, renders in 3D a lot.... I would opt by the RX 580 ...IF...noooone of my apps are of the kind that depend only on Cuda, or where Open CL performance is sadly poor, which has happened n some instances. Blender (Cycles rendering) was the case, but heard/read from  several sources that the devs have reverted that. Gotta check latest render benchmarks, tho. Being proved in benchmarks that it (RX 580) is so similar to the nvidia's 1060 in performance (and that at newegg.com, you can find a RX 580 at around 180 -220 US $, while a nvidia 1060, around 240 US $, and up (300, etc)....Only considering the 6GB version! As also happens with 1050 2gb and 1050 ti 4gb...the 1060 3g is less powerful, not just has less memory) , so, am not buying a crappy chip with crappy capability just having 8gb, but a similar-to-1060, and so, the 8 GB then do make a large difference in some apps. Blender rendering will not crash if the scene was larger in total than 6gb (I tried the other day to render a huge scene with my crappy 2gb...lol), for example. You can fit more stuff there. Digital painting software (from other brands!!!) can have larger brushes, etc.  If in the future, or current betas, Affinity developers start doing more stuff in the card (which I don't know if would be good or bad...), IMO, memory will certainly help. Specially in large operations. But has been told that this is not the trend in these apps, due to (hardware reasons) slow transfer to and out of the card (explained terribly by me, lol). So, in any case, probably should not bring you too much of a headache. Focus more in the other components for now, IMO.

Of vital importance is knowing if ANY of your essential apps, in its accelerated by GPU department, uses CUDA cores (nVidia take at it), or Open CL (AMD's) (Edit: forgot to mention that nVidia cards tend to suport BOTH cuda and open CL) . It'd be outstandingly silly to purchase an AMD card if all you need is cuda cores, and your app(s) do not have an option to set both CUDA or Open CL, or, just as bad, having the option for both, it performs really bad in Open CL (or CUDA! There's some scenarios where it is already happening that, and important ones). If one of these scenarios happen, there you would have a total decision making clue.

I don't know you, but my machine works in many different matters (surely because I tell it to do so)... Affinity is only one part of it. More memory, and more capability, more cores (essential for 3D rendering), more processing units, more turbo and base speed, faster memory (despite both RX you listed being DDR5) more everything that the 580 has over the 560 (remember there's a very interesting middle point not mentioned, the 570....which Apple sets there in the "perhaps" category...yeah, too much of a risk...) , I'd opt for the 580, easily (again, the cuda/opencl in the apps being a major factor). Mostly for a singular estelar reason: If like me, you end up handling OTHER apps where memory in the GPU is crucial, you'll find this 580 a total bargain, as 8 GBs of GPU memory is going to the roof in price in nVidia brand. 6gb is the last "affordable" (!?) step (the 1060). While not the case in AMD with its RX series. You get usually a bunch of memory in cheap AMD cards. Which is great for rendering. Now, that said, not ALWAYS the case, attending to rendering benchmarks. Some low end RX are beaten badly by nvidia low end, even being sold as "even lower" than those AMD offerings, because the processing power, despite having more memory, is really poor in the very low entry AMD cards. As far as I know, not the case of the 580 (middle-high end) : A brilliant card with 8gb. I'd opt for that one, for sure.

And we are not talking about the 590, is quite more high end (still, quite a nice pricing at newegg.com, 280 bucks  is a very nice price for that card, nVidia gets really pricier at this level), and some Blender users are talking wonders about it, since quite some time. For blender rendering ( I always mean Cycles rendering inside Blender), the amount of memory is crucial. But only if you consider, imo, the middle-high range, where you don't find any card with excessively poor processing capability. I find, in nvidia, the 1070 (it's middle-high range) , and now the 2070, very sweet spots for someone doing graphic content seriously. And having quite deeper pockets, hehe (around 340 to 400 US $ and +). Those are your "cheap" entry to the 8GB wagon, in nVidia, lol. An amazing card, tho. The new version of that, the 2070, goes easily farther the 500 $ barrier. But right now the price/performance ratio of a RX 570 and 580 is really amazing. I only want to buy these things locally, otherwise, I'd have opted by a 570/580. My distributor didn't sell it at the time, and 590 was a bit too much (price-wise). I'm not tied to any brand, as you see, despite liking hardware a lot... The one thing that really worries me about AMD's cards, tho, is the non Cuda, but Open CL matter... reason why I tend to recommend the combo of AMD Ryzen CPU + nVidia card... But nVidia keeps going with very high prices, so...

Now... a 180 $ good deal at Newegg , which is easy to find, for the 8gb RX 580, is to me, sth quite reasonable for having a very solid card for graphic content creation. Now, a 2070 (and I know, quite more poweful!) is 500 bucks. Too much for the 8gb barrier. With that I can buy an entire new Ryzen machine, in certain great local shop...

BTW, both cards have DDR 5 memory, the ones you list, 560 and 580. (there's a typo in what you listed). But realize the 560 has "only" 4GBs of memory, while the 580 is 8gb. Too much of a difference. Even more, is quite a faster GPU memory in the latter. As all specs are, much more powerful than with the 560. 

My ultimate advice, if you sort out if what you need is CUDA cores or Open CL, (and is the case of the latter, if anything) and you find at newegg, or ( my preference) a local shop you trust and know, a RX 580 for around 190 - 220 us$, I'd take that. Is just IMO more future proof as software keeps demanding more and more, in general. The card is quite much better in every department than a 560. If they had listed a 570 as a 'fully sure' supported by Metal, well, then I'd have my doubts...But at the price am seeing now the 580 is, quite cheaper than a 1060,  a 1060 from nvidia in performance, but with 2 more GB than the nvidia model...  Of course, price related, YMMV, as local shops have sometimes very rare pricing, often just the case of having easier certain distributors, shipping issues, stock matters, etc. IMO, the 580 is a no-brainer if is compatible with your apps, great choice. Look, I've just bought a 1050 2gb, which is WAAAY worse despite being nvidia, and I had to pay localy around 150 euros (basically, the price now at newegg for a RX 580 !!! )

Quote

For photo editing/graphic design/Dtp is important 4GB VRam vs 8GB VRam

Being super honest here, nope, I don't think is gonna make much of a difference for a big while (specifically in that!). But... dunno.. they use the card for some things... like some panning, visualization, etc. I don't know if I'd be very wrong to think that the much more powerful chip the 580 has will add at least something in performance there. This little bit can turn to more in certain scenarios.

There are digital painting apps, a few, that as their function is more limited and specific, can afford to go all GPU, and then, if you shall use any of those besides Affinity, you will be extremely happy to have purchased a 8gb powerful card again, (dig those apps' docs for the cuda/opencl matter!), and way faster card, than a much slower, and 4gb one.

For that what you mentioned, specially RAWs, I'd make the wild guess that the major thing to worry about is CPU clock speed, more cores, more CPU capabilities, be it a modern cpu, and system memory: RAM (perhaps even ram speed and latency! In any case, great performance price ratio : 2666MHZ in speed and CL 16 or CL 17 in latency. Gain over that is minimal in most cases, but price grows exponentially. A ryzen machine can see more improvement than intel's, if you overclock the ram in the bios (or AMD's software for that), so, needing to purchase a well for that the ram at 3200 ( in Ryzen mother boards, till 2666 is supported by default). IMO, extremely more important to purchase 16 GB (or more) at 2666 than 8 at 3200 or more.). Starting from the base of using a decent average card so that it does not bottleneck a powerful system, then all what will matter above that -in bucks spent- is cpu, ram and disk. Meaning, a 560 should do quite well, in its department. But I would definitely jump into a 580 if finding a good price. (a pity that for Metal the 570 is not fully validated...after all, is 8gb, too. Has many processing units, has nice speed, etc)

Of course, far more important for the tasks you mentioned are imo the CPU, RAM, and maybe an hybrid system, having the OS and apps in an SSD disk, and store/save files in HD, a mechanical drive. Maybe for longer life of the SSD, if u know how to (Mac OS is bit of a mystery in that for me), set all cache systems in the HD.  A typical HD with 7200 rpm. (ie, a Seagate Barracuda. Cheap as a disk can get for 1 terabyte (40 or 50 bucks) and frikin' reliable (mine is almost 10 years now, of crazily intensive usage)). I know how to set browser's cache, OS temp folder/files, OS swap file, etc, in the HD, despite being other stuff in the SSD for faster load, but under Windows. No freaking idea on if/how is that possible in Mac OS. But surely is 100% possible.

There are people way more tech savvy than me around here. And that speaking generally, as, about Affinity performance, no one does know more than the devs.

 

[SrPx]

14 minutes ago, AT.HA said:

My current problematic and very old card (2010) is 1GB VRAm and it is slow to process several tasks, like open files, export, etc. Batch export is a nightmare; as the worst example.

Yep, well, I doubt that's the card's fault....

Open files.... --- > Disk speed, CPU and RAM speed, probably even RAM amount... probably multi core powerful processors help there, too (if only because any OS is doing always several things at a time, so might have some more threads for everything, more overal performance).

Export or batch---> I believe pretty much the same than above.

The card would be the culprit ( in these apps) maybe in graphic glitches (or in a 3D app, getting stuck redrawing the 3D wireframes), slow canvas panning or the like. And even then, could be not the card, but the cpu , ram or disk  unable to move well all the data. As far as I know, the card here is mostly used for visualization. So, I'd worry way way more about the other components to get more speed in those actions.

 

[AT.HA]

Thanks a lot SrPx. 

The 580 option is more capable and more future proof; as you said.

I am upgrading memory (16GB to 32 GB) too and an SSD for the system, hopping to improve the whole performance of this old Mac.

The machine is the 3,33 GHz 6-Core Intel Xeon version/Mac Pro 5,1

[SrPx]

Intel Xeons ( ideal for servers and render farms) are wonderful machines, generally. But the fact is that these are processors from 2010.

I quite believe you will see a large improvement with the SSD added, and RAM increase, in the tasks that you mentioned.

6 core is quite nice, and well, 3.3GHz, not horrible.  A today's Ryzen 7 reaches 4.2 or 4.3 in its top gamma on the main stream, soon to reach 4.5 or more in next offerings.  The intel's 8700K reaches with overclocking 5 GHz. And latest 9900k, quite expensive, but easily goes further the 5 ghz barrier. Add on top the newer CPU features to help in the tasks, many improvements in these CPUs since 2010, plus improvements in the mother board, BUS, etc, etc. 

So, you have a "kind of" weaker point there. But a SSD and duplicating the RAM tends to provide a whole new world of experience in a machine, so, I'd be quite optimistic.  If later on in the year you see a good moment for another upgrade, maybe a last one on that 2010 machine, well, maybe a slightly (I think you can get till 3.6, not sure if I'm wrong) faster CPU, if there's by then yet support for that. Or... If really the cpus you could get for this platform wont deserve the bucks, just let it be with this one for the time being, see how well it does some more years, then jump to a globally new thing. I do think it is a machine that could live quite longer, yet, with some optimization in workflow, system, etc. 

I have to say, with 2.8 GHz ( but i7,...2009, even older), non overclocked, am doing just fine (Windows, even) with heavy tasks , but RAWs editing (I don't do that) is a really taxing operation on any system, and obviously, batch processing these or any large dimension images, are always very heavy tasks for any system or app, actually...

[AT.HA]

Thanks .-)

I hope to keep using this machine.

[Fixx]

SSD will help a lot with your mac, and doubling RAM is always good. At the moment though display card does not do much work with Affinity software. Versions 1.7 will have much improved GPU usage, but at the moment there are no tests which cards are best for 1.7 suite.

[AT.HA]

Thanks a lot SrPx and Fixx.

The 580 option is more capable and more future proof; as you said.

I am upgrading memory (16GB to 32 GB) too and an SSD for the system, hopping to improve the whole performance of this old Mac.

The machine is the 3,33 GHz 6-Core Intel Xeon version/Mac Pro 5,1