How to spec a Windows PC for Affinity Photo?

[PLShutterbug]

Apologies for this duplicate post (first as response to one on building a PC to learn) but since it has been 24 hours with no replies I wonder if I should have posted a new topic all along.

I have been using Affinity Photo on a Surface Pro 3 with Core i7 @1.7GHz, 8GB RAM, 256GB SSD, integrated graphics, and Windows 10/64 bit. Things like focus stacking and HDR merge are unbearably slow, especially during adjustments once the initial merge operation is complete.

Today I bought a new PC - a laptop with Core i7-7700, 16GB RAM, 256GB SSD, NVIDIA GeForce 1060 with 6GB RAM, and Windows 10/64 bit.

After setting the new system up I did a focus merge with 10 24Mpxl images.

Surface Pro: 8 minutes 31 seconds to complete the operation.

New PC: 3:23.

I'm pleased that the new PC is faster, but still I'm surprised at how long the operation takes (I had never timed this before so I didn't really know how long I've been waiting ...). I also note a significant (several seconds) lag if I choose one of the source images to clean up the merge ... its performance looks nothing like what I've been seeing in the tutorial videos (are Macs really that much faster?).

I can return this PC and exchange it for a desktop version with Core i7-8700, 16GB RAM + 16GB Optane memory, 7200rpm spinning HDD, NVIDIA GeForce 1050Ti with 4GB RAM, and again Windows 10/64 bit. Looks like that machine’s i7 processor has 50% more cores and threads, and has about 15% faster clock speed.

Alternatively I can spend more and get an even faster system. However it looks like the next step beyond the Core i7/8700 is a Core i9, which will add about $2,000 to the PC cost.

Any suggestions for getting the most PC for the money, for use with Affinity? Photo editing will be the primary and overwhelming use case for this PC. I don't game, or do anything else that stresses even my Surface.

Suggestions?

Thanks, Karl

[stokerg]

Hi Karl, 

As a Mac user, i can't really comment when it comes to PC specs.  I see you already posted on this thread and i'd suggest reading all the pages as they do contain a lot of information on what Affinity uses on a PC.

As for merging, it can be slow if merging straight from RAW files. Developing the RAW files and then merging may give quicker results.

 

[PLShutterbug]

@stokerg, thanks for responding. Regarding RAW vs. develop: you seem to be implying that I can develop the files and then save them as some other format prior to performing merge operations. However I don't see any option to do that from the Develop persona. It only allows me to cancel my work or hit "Develop" which opens the developed image in the Photo persona. Since "New HDR Merge" or "New Focus Merge" requires files as input, somehow I need to save the developed files.

Are you suggesting a batch operation with macros on stack or HDR filesets that ends up saving the files as another file format that Photo can more easily use? What format would that be?

Second question: once RAW files are merged, should manipulating the different source files in the stack be slower if they originally were RAW than if originally they were that intermediate format, or does Photo do some magic during the merge that makes the source file list now work the same regardless of original file format?

Thanks.

 

[John Rostron]

The disadvantage of loading raw files directly into a stack is that you get the default development process. This may be fine in the great majority of cases, but not always. If you have saved jpegs as well as your raw, this may give you a guide as to whether the raws will need processing. If the jpegs look good, you could stack those instead!

 

17 hours ago, PLShutterbug said:

you seem to be implying that I can develop the files and then save them as some other format prior to performing merge operations. However I don't see any option to do that from the Develop persona. It only allows me to cancel my work or hit "Develop" which opens the developed image in the Photo persona. Since "New HDR Merge" or "New Focus Merge" requires files as input, somehow I need to save the developed file.

Yes you do need to save the developed file. You can save it as afphoto, tiff, png or even jpg. I always save as high quality jpgs. This adds an extra step for each file. This may or may not be quicker than loading the raw files into the stack. It is a matter of try it and see.

John

[PLShutterbug]

OK, more results. I dropped by Costco yesterday and they made me a screaming deal on a display-model Dell XPS 8930, Core i7-8700, 16GB DDR4/2666 RAM, Win10/64 and NVidia 1050ti graphics card.

Loading the same ten images as above for a focus merge, this beast took only 2:11. Cleaning up the image after the merge was significantly faster as well.

I picked up a 500GB Samsung 860 EVO SSD. I'll try that next to see what other performance gains I might get.

Is this interesting to the crowd?

[SrPx]

Yes, it is interesting. I don't do photography, but a lot of people around here do. And that satisfying result is only with the 8700, and seems without overclocking (I neither overclock my machines). So, a stock 8700k might be quite faster, even ( and an overclocked one, way more). As I mentioned in other threads, besides I don't ever work with RAWs, I don't have much of a clue on  if RAW processing and most often photography related filters and other operations, do use heavily single core, or multiple cores. I'm very interested in this fact about Affinity applications, specially under MS Windows 7 and Windows 10. I'd be to suspect that higher clocks will always benefit, no matter what (ie, 8700k stock way faster than 8700, and Ryzen 2700X being way way faster than Ryzen 2700 stock (without overclocking). IF Affinity Photo uses multiple cores heavily (or maybe even if slightly) there could a large difference in favor of the Ryzen 2700X compared to a 8700K. But the latter is a beast, anyway.  I have said this before, but you don't go really wrong with a 8700, 2700, 8700K or 2700X. But if you are not overclocking (and even if you, are, still is "technically" true) the two latest ones are the fastest. I have  my preference on the Ryzen 2700X, of all these. But any of these is a great machine. My point is, if working with these huge images, makes sense to go for the faster solution. If is for a hobby, and is not so critical, I believe the intel's 8700 and the ryzen 2700, very specially this latest one, are quite more energy efficient.

And if one looks into more extreme solutions, both in cost and number of cores, there's the Threadripper platform and intel's X-Series, the core i9. BUT.... first, these cpus are really expensive, to my taste (a bit above 300 bucks of the main stream of the above paragraph vs 800 -1800 here), and those are really watts eaters. Threadripper 2nd gen is way much more energy efficient than its first appearance, but still. Even so, if one would be rendering a lot, I think the AMD Threadripper 2990 WX is the best way to go. But is clocked at 4.2 Ghz, and these machines when overclocked really eat a lot more watts. Also, is harder to get significant gain when overclocking solutions with many cores. Intel's top solution is much better in performance than the top AMD's in single core operations (for its IPC and other matters) but is 1800 bucks for just the CPU! and with way fewer cores. Boards are really expensive in intel's X-Series, and TR4 boards are more expensive but a bit cheaper than intel's extreme branch, but don't quote me on that... A great solution tho, for someone rendering a lot, but also making use of applications making use of single core, high speed clocks, is Threadripper 2950 X. Still 16 cores and 32 threads (is not 32 threads and 64 cores like 2990 WX, but still, A LOT) but at 4.4 GHZ, and stays pretty near the intel's very top offer in benchmarks results. While costing 900 USD instead of intel's top solution, which is around 1800. (half the price!).

Thing is, all these beasts, in tests using the typical applications we use, have almost identical performance to an AMD Ryzen 2700x (remember, this one has a lot more performance in multithreaded apps, while keeping ion very good shape in single core, when compared to a 8700k from intel). So, only if Affinity is heavy in using multiple cores and multiple threads, instead of, like many applications yet, using heavily just single core and high speed clocks, in case is multithreaded it might end in a significant gain when doing those resource consuming 2D, photography operations. For these users so much wishing higher speeds with huge images (besides fast RAM and a SSD disc) then it might be something interesting to dig. If all in Affinity is mostly single core, and throwing there the most GHz possible, IMO a 2700x or an intel 8700K is the way to go. 2700 overclcoked, if want to be very energy efficient. But after seeing the power usage of Threadripper and intel core i9  in general, I don't think anymore 2700x uses really much power....At least in comparison.  If having more cores do get strong benefit in Affinity, the champion, the golden purchase is imo a Ryzen 2700 or 2700X.