# Force purge gpu memory **URL:** \ **Category:** Flame Questions\ **Created:** [April 29, 2021, 9:12am UTC](https://forum.logik.tv/t/force-purge-gpu-memory/2975 "2021-04-29T09:12:55Z")\ **Posts on this page:** 13\ **Page:** 1
**Author:** ![philobedo](https://avatars.discourse-cdn.com/v4/letter/p/b19c9b/32.png) [@philobedo](https://forum.logik.tv/u/philobedo)\ **Post date:** [April 29, 2021, 9:12am UTC](https://forum.logik.tv/t/force-purge-gpu-memory/2975/1 "2021-04-29T09:12:55Z")
Hi. Is there a way to simply force purge GPU ram? Like a button somewhere in the interface? (If not, it would be super cool to have one!!!) I have a problem where, while working on 6k images, each time I press render, flame tells me that render time will be 25 minutes for 100 frames. If I close flame, relaunch and press render again, render time falls to 5 minutes… ☹ arghh I’m on a Linux box with poor P4000, on 2021.1 flame version. ---
**Author:** ![philobedo](https://avatars.discourse-cdn.com/v4/letter/p/b19c9b/32.png) [@philobedo](https://forum.logik.tv/u/philobedo)\ **Post date:** [April 29, 2021, 9:19am UTC](https://forum.logik.tv/t/force-purge-gpu-memory/2975/2 "2021-04-29T09:19:11Z")
Just one more clue. On a render that says 25 minutes, if I render the batch in 2 parts. First half of the setup then having the result of that render as a source for the second part, the render times are like 1 minute + 5 minutes… I know the hardware P4000 is quite limited for 6k work but still, there is a problem with how Ram is filled and flushed during renders isn’t there? ---
**Author:** ![johnag](https://sea2.discourse-cdn.com/flex020/user_avatar/forum.logik.tv/johnag/32/573_2.png) [@johnag](https://forum.logik.tv/u/johnag)\ **Post date:** [April 29, 2021, 9:41am UTC](https://forum.logik.tv/t/force-purge-gpu-memory/2975/3 "2021-04-29T09:41:31Z")
@philobedo. Just wondering that when it says it will take 25mins, does it actually take 25mins? I had a situation similar where it said one thing ad did another. ie the render got to half way then boom it completed. My situation was also tlfx rendering not batch. might be very different(edited) ---
**Author:** ![philobedo](https://avatars.discourse-cdn.com/v4/letter/p/b19c9b/32.png) [@philobedo](https://forum.logik.tv/u/philobedo)\ **Post date:** [April 29, 2021, 10:22am UTC](https://forum.logik.tv/t/force-purge-gpu-memory/2975/4 "2021-04-29T10:22:03Z")
I have to say that I didn’t let the render go to the end of 25 minutes. But the average time for the 40 first frame was consistant with the complete estimation… and I was already at 10 minutes for these 40 first frames… ---
**Author:** ![richardamckeand](https://avatars.discourse-cdn.com/v4/letter/r/858c86/32.png) [@richardamckeand](https://forum.logik.tv/u/richardamckeand)\ **Post date:** [April 29, 2021, 4:56pm UTC](https://forum.logik.tv/t/force-purge-gpu-memory/2975/5 "2021-04-29T16:56:09Z")
Did you try eciting the application and rerunning the render? It will clear the vram but you might fill it again once you look at the result view once you reliad your batch. It would be great if we could toggle enable vram compression (as I think macs do) and get a lot more performance out of the available vram - it’s the most underspecced resource in the flames they give lowly freelancers like me 🙂 +1 for a simple button to flush the vram though ---
**Author:** ![philobedo](https://avatars.discourse-cdn.com/v4/letter/p/b19c9b/32.png) [@philobedo](https://forum.logik.tv/u/philobedo)\ **Post date:** [April 29, 2021, 4:58pm UTC](https://forum.logik.tv/t/force-purge-gpu-memory/2975/6 "2021-04-29T16:58:20Z")
Yes, exiting flame is my workaround for now. But it’s obviously not ideal… ---
**Author:** ![PlaceYourBetts](https://sea2.discourse-cdn.com/flex020/user_avatar/forum.logik.tv/placeyourbetts/32/6852_2.png) [@PlaceYourBetts](https://forum.logik.tv/u/PlaceYourBetts)\ **Post date:** [April 29, 2021, 8:24pm UTC](https://forum.logik.tv/t/force-purge-gpu-memory/2975/7 "2021-04-29T20:24:59Z")
Yeah. That would be a good feature request to put out there. I sometimes get a suspicious render time and I’m sure it doesn’t normally take that long. I like to think I am quite efficient and I usually have a 20 min threshold. A quick exit and relaunch will fix it most of the time but I’ll vote up any feature that avoid that step of the process. ---
**Author:** ![philobedo](https://avatars.discourse-cdn.com/v4/letter/p/b19c9b/32.png) [@philobedo](https://forum.logik.tv/u/philobedo)\ **Post date:** [April 29, 2021, 9:27pm UTC](https://forum.logik.tv/t/force-purge-gpu-memory/2975/8 "2021-04-29T21:27:04Z")
Request posted. 😉 ---
**Author:** ![PlaceYourBetts](https://sea2.discourse-cdn.com/flex020/user_avatar/forum.logik.tv/placeyourbetts/32/6852_2.png) [@PlaceYourBetts](https://forum.logik.tv/u/PlaceYourBetts)\ **Post date:** [April 29, 2021, 9:30pm UTC](https://forum.logik.tv/t/force-purge-gpu-memory/2975/9 "2021-04-29T21:30:03Z")
I’ll go and up vote. Do you have a number? ---
**Author:** ![philobedo](https://avatars.discourse-cdn.com/v4/letter/p/b19c9b/32.png) [@philobedo](https://forum.logik.tv/u/philobedo)\ **Post date:** [April 29, 2021, 9:33pm UTC](https://forum.logik.tv/t/force-purge-gpu-memory/2975/10 "2021-04-29T21:33:04Z")
Fl-02573 ---
**Author:** ![PlaceYourBetts](https://sea2.discourse-cdn.com/flex020/user_avatar/forum.logik.tv/placeyourbetts/32/6852_2.png) [@PlaceYourBetts](https://forum.logik.tv/u/PlaceYourBetts)\ **Post date:** [April 29, 2021, 9:34pm UTC](https://forum.logik.tv/t/force-purge-gpu-memory/2975/11 "2021-04-29T21:34:29Z")
Addition of Vram flush button - Up voted ---
**Author:** ![philobedo](https://avatars.discourse-cdn.com/v4/letter/p/b19c9b/32.png) [@philobedo](https://forum.logik.tv/u/philobedo)\ **Post date:** [April 29, 2021, 9:37pm UTC](https://forum.logik.tv/t/force-purge-gpu-memory/2975/12 "2021-04-29T21:37:13Z")
Thank you! ---
**Author:** ![philobedo](https://avatars.discourse-cdn.com/v4/letter/p/b19c9b/32.png) [@philobedo](https://forum.logik.tv/u/philobedo)\ **Post date:** [April 30, 2021, 12:16pm UTC](https://forum.logik.tv/t/force-purge-gpu-memory/2975/13 "2021-04-30T12:16:20Z")
So to keep everyone up to date I had a feedback from Autodesk on that question. To make it short, they’ve work on that already in the last without success so, only real solution is and will still be to restart flame when needed. Second thing, Stephane reminded me that P4000 doesn’t meet anymore the minimum requirement needed for flame 2021, which is a 12GB GPU at least. I’ll tell the facility I work in they have to put a little money on a RTX 8000… 😍