Is 2026 terribly slow because of the new database structure, or is something else going on?

I’ve been working on a long term project, and upgraded to 2026 a few months ago. Is it me, or does take MUCH longer to Save a Batch, or a Workspace? It’s fairly large - 100ish media clips, 25 Batches, 40 sequences, but the Batches aren’t terribly complicated - maybe 30-40 noes per Batch, but the saves are super slow… I never timed it before, but I’d guess 10 seconds? Now it’s 65 seconds every auto-save of the workspace!

I’ve seen some other threads about removing all old versions of Flame - is that recommended? I’m already “over the falls” on this one, so there’s no going back to 2025, but I’m also not super Terminal friendly, so I don’t relish copy-pasting commands I don’t understand…

1 Like

I have not had this experience, although I have removed all traces of 2025 from the system. One thing I have noticed is that my 2026 archives open and close much faster than the 2025 archives. And by that, I mean the 2025 archives when I was still running 2025.

I would also say never upgrade mid-project

1 Like

100% Agree, most of the time. I was thrilled to have a new type tool, LOL. But seriously, there’s a lot of animated text, so working in that old Text tool was a major pain point for me. WAY worse than having slow saves.

where is your project directory?

For me when I put it on my nas its incredibly slow to save even a single batch (5s)

2 Likes

for 2026, I’m using my RAID5 for both media and projects.

@davejahns - hardware RAID5 is normally set to a fixed block size.
writing many small files into large blocks is inefficient.
when you get round to it, consider using a mirrored z-raid with unspecified block size.
do not use hardware raid controllers.
also, read about sync and async and RAID5 write holes.

3 Likes

Having this same issue in 2026.1. Project gets to a certain size and the project becomes unusable. Loading workspace and autosave takes 3-5 minutes plus. Entering BFX the same.
I created a new project and wired the most need media across, but same issue persists.
Since the introduction of BFX I have used BFX and connected segments as my default workflow.
I’m on my second big show in 2026 and this problem has now made Flame unusable at this point.
I have extracted all BFX setups and converted to BatchGroups, but at this point it seems like the database is still the issue.
I’ll gather further info and post this to a new thread and post a ticket with Autodesk as well. (Funny how these things happens on the weekend and day of final delivery)

What storage are your projects stored on now? Same as stone+wire was before or different?

There are multiple paths in the project window. Media and project files are separate. They’re not super intuitive for their defaults, so worth making sure everything is on suitable hardware.

Storage is on the same Stone+Wire hardware and set manually to the correct path for that storage. Setups are on the default internal SSD of the MacStudio. I suspect that this issue is also tied to the BFX workflow getting to heavy at some point.

1 Like

In some very rare instances, this might be a related issue, although it’s prevalence is in Windows environments, not macOS or Linux:
ZSTD-22 is 30 times slower than it should be

1 Like

Anecdotally, I was able to restore someone’s unarchived project from a backup .tzst this morning with zero hassle.

I like 2026+

3 Likes

We are seeing this horrid performance also.
I’ve been working with ADSK for about 6 months on a different critical failure bug in 2026.x that just got solved. I’ll start on them with this issue shortly. We may have to stick on 2025 for quite a bit longer.

1 Like

Maybe a given but thought I still mentioned. Up until v2025.x, if I carried multiple batch iterations I’d experience minor to massive slowdowns while performing general lib/housekeeping tasks (saves, loads, moving/copying things around etc). The more the iterations, the more the slowdowns. Having minimum to no iterations brought things back to normal.

1 Like

@AamirQ - especially slow if batches contained actions which contained embedded geometries which originated in fbx or alembic format.

sigh - life is so short.

i lament that this amazing toolset of flame 2026+ did not exist 30 years ago.

we could have all been rich and patient and reasonably cheerful.

1 Like

Currently just creating a new BatchGroup is incredibly slow as well. Same waiting period as loading workspace and autosaves (v2026.1)

This might still be a result of the bloated (contaminated) database of the current project I’m battling with.

Hi ALan,

would this be your recommended way forward (reverting back to 2025)? I just had a very challenging client session on final delivery day of the above mentioned problematic project.

Please keep me posted on your findings. I would like to contribute to this process where possible. I just have not had time to gather more data on this current issue, while desperately trying to deliver this current project.

I’ll start gathering more information on the issues at hand and start building a case file.

@eddiead have you opened a ticket with out Support team? if so, please share the ticket number so I can follow-up.

You might also want to upgrade to 2026.1.1 so you you the latest bug fixes (Help). Also, the 2026.2 Update is available.

To Finn’s point make sure projects are stored locally on a nvme or at bare minimum a ssd. Also make sure you aren’t running low on your boot drive.

3 Likes