Zerene & LR execution on HDD vs. SSD vs. M2 vs. RAM

Have questions about the equipment used for macro- or micro- photography? Post those questions in this forum.

Moderators: Pau, rjlittlefield, ChrisR, Chris S.

Adalbert
Posts: 1162
Joined: Mon Nov 30, 2015 1:09 pm

Zerene & LR execution on HDD vs. SSD vs. M2 vs. RAM

Post by Adalbert »

Hello everybody,

I have just executed a test with all my disks for LR and Zerene but I really don’t know how to explain the results :-(
Image
So, the differences are hardly noticeable and LR was faster on HDD than on SSD :-)

For this test I used 140 CR2 files each ca. 29MB.

For LR I only changed the appropriate input and output directories.

For Zerene I changed the temp-directory too (set in preferences and restart Zerene)

BTW, the RAM-disk has been created by the ImDisk

BR, ADi

Pau
Site Admin
Posts: 5163
Joined: Wed Jan 20, 2010 8:57 am
Location: Valencia, Spain

Post by Pau »

I'm not at all expert in computers but to me it's clear that being all the times very similar the limiting factor is not the drive, could it be the processor?
Pau

Macro_Cosmos
Posts: 932
Joined: Mon Jan 15, 2018 9:23 pm
Location: Sydney

Post by Macro_Cosmos »

Agreed, seems like you're being bottle-necked by the CPU. It's a 3900 from AMD right?

Adalbert
Posts: 1162
Joined: Mon Nov 30, 2015 1:09 pm

Post by Adalbert »

Hi Pau, hi Macro_Cosmos,
Yes, the CPU is Ryzen 9 3900x but it was used in max 40%
BTW, “Overlap I/O” was activated and “Use all available cores” set in Zerene (24)
BR, ADi

JKT
Posts: 169
Joined: Fri Oct 28, 2011 9:29 am
Location: Finland
Contact:

Post by JKT »

I'd suspect thermal throttling by the CPU. The Ryzens seem to run very hot. You did have liquid cooling, so it shouldn't do that, but you never know.

Get a program that shows temp core-by-core and see how high it goes. I'd also suggest checking the performance core-by-core, but with your CPU there is going to be a lot of small windows. :)

lonepal
Posts: 320
Joined: Sat Jan 28, 2017 12:26 pm
Location: Turkey

Post by lonepal »

Hi;

I think you need to test again after setting the minimum CPU status to %100 from power settings otherwise CPU may behave different at each step according to multi core usage algorithyms.
Regards.
Omer

rjlittlefield
Site Admin
Posts: 21027
Joined: Tue Aug 01, 2006 8:34 am
Location: Richland, Washington State, USA
Contact:

Post by rjlittlefield »

The numbers definitely look CPU-bound.

I agree with JKT's suggestion to check for throttling due to CPU temp.

I seem to recall that Windows 10 factors the clock speed into its overall % report, so an overheating CPU looks poorly utilized.

Another issue, specific to Zerene Stacker, is that ZS has trouble utilizing large numbers of processors on a single stack, even with overlap I/O enabled. You might try processing two or three stacks at the same time, and see if the processor utilization goes up much in that case.

--Rik

Adalbert
Posts: 1162
Joined: Mon Nov 30, 2015 1:09 pm

Post by Adalbert »

Hello JKT,
I'd suspect thermal throttling by the CPU. The Ryzens seem to run very hot.
I'm afraid I can't confirm that. The temperature was about 50 degree Celsius.
The fans of the “Corsair Hydro Series iCUE H115i RGB Pro XT” have hardly turned at all :-)

But the Intel I7 in my old computer can be really hot :-(

BR, ADi

Adalbert
Posts: 1162
Joined: Mon Nov 30, 2015 1:09 pm

Post by Adalbert »

Hello Omer,
I think you need to test again after setting the minimum CPU status to %100
good idea but how can I do it?

BR, ADi

Adalbert
Posts: 1162
Joined: Mon Nov 30, 2015 1:09 pm

Post by Adalbert »

Hello Rik,

My guess is that Windows 10 manages the I/O operations via huge buffer in RAM so that I cannot see any difference.

Actually in this test I wanted to compare the performance of the disks
but Zerene running in many sessions could decrease the entire time of the execution but independent of the disk used, I think.

BR, ADi

JKT
Posts: 169
Joined: Fri Oct 28, 2011 9:29 am
Location: Finland
Contact:

Post by JKT »

Adalbert wrote:Hello JKT,
I'd suspect thermal throttling by the CPU. The Ryzens seem to run very hot.
I'm afraid I can't confirm that. The temperature was about 50 degree Celsius.
That liquid cooler really does its job! I have Ryzen 3800 and the temp is pretty fast in 70 C. At least the Noctua works silently unlike the original boxed cooler.

Have you checked your power saving profile? (Control Panel / Power Options)

Adalbert
Posts: 1162
Joined: Mon Nov 30, 2015 1:09 pm

Post by Adalbert »

Hello JKT,
power saving profile
I have just set it to the max and started the test on the RAM disk again :-)

The execution time was exactly the same but
CPU usage till 70% during PMax
CPU usage till 50% during DMap
Max temperature about 65 degrees Celsius
GPU usage low but sometimes till 40% (Radeon RX 5700)

BR, ADi

Adalbert
Posts: 1162
Joined: Mon Nov 30, 2015 1:09 pm

Post by Adalbert »

Hello everybody,

This time I have split the stack into 2 sub-stacks.

So, firstly I have created the project and aligned all frames.

And secondly I have created 2 batches with the Bugslabber as follows:
Image

And finally I have started the both batches at the same time:
Image

But one sub-stack has failured with the following error-message:
Image


BTW, the temp was on M2 this time. For the test of performance will be defined on the RAM disk

BR, ADi

rjlittlefield
Site Admin
Posts: 21027
Joined: Tue Aug 01, 2006 8:34 am
Location: Richland, Washington State, USA
Contact:

Post by rjlittlefield »

From EXIF propagator error stream: Error renaming temporary file to R:/MultimediaFoto/LightRoom/TIFF/ZSP/SavedImages/Slab 1 PMax.tif
I don't see anything in the file path/name that appears to be unique to the project or to the instance of ZS.

So, I'm wondering if both scripts tried to access the same file at the same time, and one of them got this error as a result.

--Rik

Adalbert
Posts: 1162
Joined: Mon Nov 30, 2015 1:09 pm

Post by Adalbert »

Hello Rik,
Yes, the both scrips are located in the same directory (of the Zerene-project) and produce the output-files with the same name :-(
So, I will have to edit the scripts generated by the Bugslabber in order to get the unique names.
Do you know any better method how to split the stack into some sub-stacks which can be executed at the same time?
BR, ADi

Post Reply Previous topicNext topic