Performance Windows vs Mac

Announcement of new releases, bugs, support, suggestions
Post Reply
strawbale
Posts: 8
Joined: 26.03.2020 14:56

Performance Windows vs Mac

Post by strawbale »

I'm a newbie starting with the trial version on a DPReview forum member's Panasonic G80 & G9 files (full size focus bracketing jpegs and raws, as well as post focus 4k & 6k video files) to test these camera models as potential upgrade (from my small sensor compact, that doesn't do focus bracketing nor post focus video) and to (then) purchase HF.

As a side effect of doing the same stacks on my Mac Mini (2018, i5-8500B 6 core, 8GB RAM, Intel UHD 630, SSD, 10.14.6) as the DPR forum member on his Windows PC (Win 10, i7-8700K 6-core & hyperthreading, 32GB RAM, GeForce RTX 2080, SSD) I/we noticed the considerable performance difference:

Windows vs Mac:
Jpeg (43 20Mp files, 5MB) stacked (method B, R26, S3): 3 sec vs 20 sec
Raw (44 20Mp RW2 files, 24MB) stacked (method B, R26, S3): 37 sec vs 4min10sec

Hence the Windows stacking was 6.7x faster. Is this to be expectected given the hardware differences or is HF on a Mac inherently slower?

During the stacking I noticed that all cores were used; apart from when the RW2 raw files were converted one by one with Adobe DNG converter, when only 1 core was used (hence that part took 3min45sec out of the 4min10sec total) - is that normal/to be expected?

Would a similar 6-core CPU with hyperthreading (e.g. Mac Mini's i7-8700B) be considerably faster (than my non-hyperthreading 6-core)?

The iGPU didn't seem to be used, so a (more powerful) video card doesn't make any/much difference?

The 8GB RAM seemed sufficient (for just one stack) as RAM was not swapped. Repeating several stacks forced RAM swapping though.

Thanks a lot in advance for any response!

PS: the slower performance on my current Mac Mini won't stop me from usinf HF - I'll just wait a bit longer :wink: (and maybe upgrade to 16GB RAM).
User avatar
Catherine
Posts: 1157
Joined: 29.04.2019 22:38

Re: Performance Windows vs Mac

Post by Catherine »

There are a few things that need to be checked before we can call this a fair comparison:
1. From your description ("raw files were converted one by one ..., only 1 core was used") I conclude that "Parallel image loading" is disabled in your Helicon Focus settings (it is by default). If it's enabled on the Windows system, it accounts for a large portion of the difference. Since you have an SSD, you should definitely enable this (Helicon Focus menu - Preferences - Performance) and enjoy the performance improvement.
2. Does the Windows system have OpenCL acceleration enabled? That RTX 2080 card will make everything much faster.
3. For raw stacks, the first time a stack is processed it will be slower than the second and the next times. You should make sure to compare either the first runs or the second runs, but not first to second. For JPEG and TIFF source images this is much less noticeable, but there still may be a slight difference between the first and the next runs.
4. Make sure to compare the same focus stacking methods with the same parameters. Auto-adjustment settings may also affect performance if they're different.

Additionally, the built-in benchmark will help you compare the base performance of your corresponding systems as it will not be influenced by many of the factors I mentioned above. But it does use OpenCL if enabled, and the second run will be a bit faster than the first so keep that in mind.

I would expect to see about 20-25% performance increase from hyperthreading alone in the built-in benchmark, but less in normal operation. Also keep in mind that 8700K has a higher base frequency and boost frequency, and it can be overclocked.

So to answer your last question - no, I don't think you would see different performance of Windows and Mac version of Helicon Focus on the same hardware. The differences you see come from different hardware, and likely different Helicon Focus settings.
strawbale
Posts: 8
Joined: 26.03.2020 14:56

Re: Performance Windows vs Mac

Post by strawbale »

Catherine wrote: 05.05.2020 15:57 Thanks a lot for your detailed response - much appreciated!

There are a few things that need to be checked before we can call this a fair comparison:
1. From your description ("raw files were converted one by one ..., only 1 core was used") I conclude that "Parallel image loading" is disabled in your Helicon Focus settings (it is by default). If it's enabled on the Windows system, it accounts for a large portion of the difference. Since you have an SSD, you should definitely enable this (Helicon Focus menu - Preferences - Performance) and enjoy the performance improvement.
OK, that box was empty, set to 'enabled' now.

2. Does the Windows system have OpenCL acceleration enabled? That RTX 2080 card will make everything much faster.
Don't know, would have to ask the user.

3. For raw stacks, the first time a stack is processed it will be slower than the second and the next times. You should make sure to compare either the first runs or the second runs, but not first to second.
That's what I did.
For JPEG and TIFF source images this is much less noticeable, but there still may be a slight difference between the first and the next runs.

4. Make sure to compare the same focus stacking methods with the same parameters.
We used the same settings (B,R26,S3)
Auto-adjustment settings may also affect performance if they're different.
Will check that with him.

Additionally, the built-in benchmark will help you compare the base performance of your corresponding systems as it will not be influenced by many of the factors I mentioned above. But it does use OpenCL if enabled, and the second run will be a bit faster than the first so keep that in mind.

I would expect to see about 20-25% performance increase from hyperthreading alone in the built-in benchmark, but less in normal operation. Also keep in mind that 8700K has a higher base frequency and boost frequency, and it can be overclocked.
Thanks for reminding. Fastest CPU for a MacMini would be an i7-8700B (https://ark.intel.com/content/www/us/en ... 0-ghz.html). That wouldn't make it a lot faster, I suspect?

So to answer your last question - no, I don't think you would see different performance of Windows and Mac version of Helicon Focus on the same hardware. The differences you see come from different hardware, and likely different Helicon Focus settings.
Good to know.
strawbale
Posts: 8
Joined: 26.03.2020 14:56

Re: Performance Windows vs Mac

Post by strawbale »

Enabling Parallel image loading reduced the Raw stacking from 4min10sec to 1min40sec !
Second run (with same settings): 29 sec
User avatar
Catherine
Posts: 1157
Joined: 29.04.2019 22:38

Re: Performance Windows vs Mac

Post by Catherine »

strawbale wrote: 05.05.2020 17:04 Thanks for reminding. Fastest CPU for a MacMini would be an i7-8700B (https://ark.intel.com/content/www/us/en ... 0-ghz.html). That wouldn't make it a lot faster, I suspect?
No, definitely not a lot. The "mini" form-factor isn't suited for powerful CPUs (at least not without liquid cooling).
strawbale
Posts: 8
Joined: 26.03.2020 14:56

Re: Performance Windows vs Mac

Post by strawbale »

Catherine wrote: 05.05.2020 17:19
strawbale wrote: 05.05.2020 17:04 Thanks for reminding. Fastest CPU for a MacMini would be an i7-8700B (https://ark.intel.com/content/www/us/en ... 0-ghz.html). That wouldn't make it a lot faster, I suspect?
No, definitely not a lot. The "mini" form-factor isn't suited for powerful CPUs (at least not without liquid cooling).
Indeed, I read quite a few throttling issues with the i7. I think my i5 non-hyperthreading 65W TDP 6-core is the sweet spot for this Mac Mini.
Post Reply