Hey Gavin, could you share couple of A7RV’s MRAWs and SRAWs?
Uncompressed and lossless compressed RAWs seem to be working fine.
I have been trying to bulk edit like I used to. It will do the first photo then goes to the second and just stays there and doesn’t complete and move to the next photo like it used to. I just edited everything one by one which took FOREVER. HELP!!!
1 Like
Sure, I currently only shoot in MRAW but these couple were taken on my camera on Saturday. Just confirmed they instantly crash Photo AI 3.0.0 with no error message. Also you can see my thread in the Bugs and Issues forum here where a person from the company confirms the issue.
MRAWs (Lossless Comp (M))
_DSC4072.ARW (39.8 MB)
_DSC4071.ARW (40.1 MB)
I also just took one in SRAW (Lossless Comp (S)) to check and it shows the same issue:
_DSC4135.ARW (37.7 MB)
2 Likes
I agree, these are features which I suggested be made.
However, as long as there is such a huge memory leak when running batches of images, none of these improvements help my workflow that much.
1 Like
If one runs frame level video processing on thousands of images, having a presets is very helpful. Alas, 3.0.0 bleeds RAM like 2.4.x did, making presents not that useful at the moment.
1 Like
Thanks Michael! Will take a look
Same here. I miss the DNG option… please allow the user to choose what format they want
Also observed the memory leak issue. Watched as my 32GB on my notebook and the 64GB on my tower maxed out at 99% & 100% while trying to process a few hundred images.
3 Likes
same here. apart from the TIFF output format (before was DNG) colours temperature or saturation or both change completely… A pain to restore original ones
Easy to call it a memory leak as if it’s an error in the code, but it looks more like an intentional way to maximize the utilization of available memory resources. Much in the way that Windows manages the Standby List, they use it all, and then release the oldest segments as they need to allocate memory for newer items. The slowdown occurs during the cycles while the old memory allocation is being released to make room available for the new work. Leaving a little more residual working space to cover the reallocation period and optimizing the memory release routines should make the slowdowns less noticeable.
100% incorrect. I have a background in IT and programming and this is clearly a memory leak. What makes you qualified to make such statements?
2 Likes
This is odd, TPAI refuses to save the output image:
“Skipped: no edits applied” - regardless of changes made in TPAI and/or export dialog
I experienced and reported this as well for either the current PAI or GP (I can’t find the post in either thread…).
Hi Gavin, exact my issue. What is crazy. Some files in the same series (same day, time, RAW format and location) i can open and some not. Other software didn´t have a problem to open all files.
Memory management is one thing, but I doubt that letting the application crash and burn when it runs out of memory is an intentional way of releasing.
I ran my first batch in a long time this weekend (I usually just run bigger batches after shooting an event) and fully expected to experience the slowdowns but decided to try even so as there were only a bit over 100 photos to process and I was too lazy to go back to an earlier version.
I however did not have slowdowns - for me PAI just crashes if I try to do more than 30 photos (it dies somewhere around 34-35 processed photos with my 32GB of RAM). I actually loaded the task manager and watched memory consumption going up with every photo processed, just for PAI to die running out of memory. However - if I start PAI and load 30 photos, process them, close down and restart PAI and load 30 more it works perfectly.
1 Like
I think everyone here knows Photo AI is in full development and, thus, that some new stuff won’t work optimally at first or that certain older routines are negatively affected. The Topaz team are aware and will address the memory leak issue – sooner or later. If you need a totally reliable app, stick with a version that meets that requirement. At this point that seems to be version 2.3.2.
2 Likes
Yes, definitely unforgivably poor memory management if they let it overflow into a crash versus just bottlenecking into a slowdown.
If they’ve actually acknowledged that this is a memory leak already then so be it, but good business practice dictates that they either immediately issue a patch to fix it, or withdraw the software from public access to avoid litigation.
p.s. I’ve been writing code since 1975.
3 Likes
Apart from the colour change issue, even mild sharpening gives an unnatural look to the image. Still not upto DxO Photolab in terms of IQ.
Don’t remember where I read it (if it was in this thread or the previous version) but I do believe someone quoted a confirmation of the leak and that they’re dealing with it…
I’m decently pleased with the latest updates and really like the possibility of stacking filters, but I do however feel like the line between alpha/beta/release is very fluid at times considering what’s being released…
2 Likes
The following are snippets out of eMails I received from Topaz Labs Support to my report of a “Apparent memory leak in Photo AI”, which was submitted on March 12th, 2024:
Sent: Wednesday, March 13, 2024 11:21 AM
Thanks for sending that info. My team believes they’ve found the cause of the memory leak and our Head of AI is creating a potential fix for this. We appreciate your patience and understanding while we get this sorted for you.
Sent: Thursday, March 14, 2024 2:13 PM
We’ll keep working on that! If we have test builds, we can also send them to you to try out.
Sent: Friday, March 15, 2024 1:13 PM
The Upscale enhancement is particularly resource-intensive so if you’re running that on hundreds of images, let alone the memory leak, it could cause the computer to freeze up. We’re still investigating the memory leak on our side. If you could, try processing smaller batches (~100 images) and see if that works better.
Sent: Monday, March 18, 2024 12:45 PM
Thanks again for the testing. Our AI engine team is looking into this as a top priority.
Sent: Thursday, March 21, 2024 11:11 AM
We’re still working on it! I don’t have a solution for you just yet. Thanks for understanding!
Sent: Friday, March 29, 2024 10:58 AM
Our development team is running some tests for the memory leak now so it wasn’t included in yesterday’s release but it’ll come to a future update soon.
Sent: Thursday, April 4, 2024 11:37 AM
We are still working on it. The backend team has a lot on their plate and has not gotten around to fixing it yet.
At this point, I stopped sending Support eMails as it seems, now five weeks in, that this issue will be fixed when it gets fixed. Until that happens, my only option, if I want to run large batches is to go back to a version which does not have the memory leak.
As far as testing goes and I have done a lot and sent my test plan & results to support. The memory leak appears to be linear and maybe impart caused by a recursive function being called when doing batch processing.
I have not monitored memory usage for the application while processing single very different images over time. Which is to say keeping Photo AI open for an extended period while working on individual photos. This could indicate the leak is not just related to batch processing. Generally, that is not my workflow.
My Windows 11 system has 256GB of RAM (it can take up to 2TB, not that I could afford that!), and so I can process well over 100 images in batch before I need to restart Photo AI.
As others have reported, the issue is not related to operating system memory management, either in MacOS or Windows.
If one’s computer has 32GB of RAM, then if Photo AI is consuming 20GB of RAM, as it did in my last batch run of 100 images, the operating system’s memory management will start thrashing virtual memory pages to disk, slowing down everything, potentially crashing the application and possible the operating system as well.
At the beginning of this thread, it was stated by Dakota Wixom concerning Presets, “This will speed up the process of editing large batches of hundreds or thousands of photos at a time”, for me this would be great as I process tens of thousands of video frames… that is if it worked.
The product description, see Topaz Photo AI - Maximize Image Quality with AI, in the Easy workflow section states, “Use the standalone to batch process thousands of images in sequence…”, at the present time this is not possible.
As I have been a programmer since the early 1980’s, a systems administrator and network engineer for decades now, I do have sympathy for those in product development.
This said, the other day I started a 30 day evaluation of a network/systems heath tool called EventSentry by NETIKUS (https://www.eventsentry.com). I was having difficultly with collecting the data I wanted and producing a report at the detail I wished. When I contacted technical support, I was blown away at the degree of information which was sent back to me (i.e. screen shots, written instructions, etc.) which addressed my very specific question. In my opinion this is how a software company should respond.
10 Likes
