DT uses 8.8 GB RAM and triggers the “Out of application memory… Force Quit” in just a few minutes

DEVONthink
1

Hello DEVONians. I am come again with another heartache.

I’ve seen similar posts about this problem and know what Jim will ask so here’s the info:

  • System & Dir Info:
    MBP 16” 2021 (M1) • macOS 12.7.6
    RAM: 16 GB
    HDD: 994 GB
    DEVONthink version: 3.9.14
    DB size: 45.96 GB
    DB file location: ~/Databases/New.dtBase2
  • Indexing:
    • I have placed the dozen Finder folders I index in a Group called INDEXED FILES, which is 2 TB as you can see in the images. The other the two other 2-TB Groups I have are 2 TB because each contains a Replicant of INDEXED FILES.
    • These indexed folders are —
      • All in my internal HDD home folder.
      • The include the Attachments folder from ~/Documents/Bookends plus 21 others. (Useful fact: Each Finder folder that is indexed starts with the string DT_IDX [dir name].)
      • I also have indexed some large Finder folders on external HDDs, as you can see.

Image of the DB’s TOP LEVEL —

top_level
top_level2084×1548 393 KB

(Oh, I see above that the Tags Group on top is huge. Or is it? It’s just a bunch of Replicants, right? But maybe it’s still RAM-taxing? Should I kill it (assuming it’s easy to re-create o demand)?

Image of inside the Group that holds the INDEXED FILES Group:

inside_FINAL
inside_FINAL2218×1684 368 KB

Here are my two questions —

Q1: DT uses 8.8 GB RAM and I need to Force Quit

I can’t use DT anymore because its RAM usage is so big I have to Force Quit within a few minutes. When I launch DT, the used RAM is immediately 6 GB. After a few clicks, 8.8 GB. Then I get a warning, and have to Force Quit.

What’s the best way to stop this? Kill the huge HDD indexings?

If you say, “Split your DT DB into sub-DBs,” then I’m worried. There’s lots of Replicants spread throughout.

But if you insist on splitting, what would be your strategy? Is there a KIND of record that you’d move off-shore? Would it be the indexed ones?

Has anyone else had the Force Quit problem before, and solved it? What did you do?

Here is a troublesome related question:

Q2: Organization anxiety

My DT world is like another Finder. But I also have a Finder. Half of my total content is in DT; the other half, in Finder. Has anyone here thought, “It’s time for me to get my Finder organized. I’ll implement Johnny Decimal or something like that and then the Finder will be useful again?”

I mention this because I’m the guy who has been using DT as an outliner for 15 years. If I try to stop the Force Quit problem by offloading as much of my DT stuff into a finally-organized Finder, I’m in for some serious problems, since most of the value-added I produce is in Unsorted lists of Groups. These won’t make it across.

My two problems—(a) preventing the Force Quit and (b) finally being an adult about my Finder—are related but pretty different. The second one is deep and so maybe should be ignored. I only mention it because I’m guessing that someone will say, “Reduce DB size by offloading as many records as possible into the Finder, specifically the [by kind? by size? by logic? by workflow?] records.”

Thanks in advance.

2

Crike! I see that I asked a very similar question 3 years ago and never responded to Jim’s reply! (Medical emergency is the reason, I dropped everything at the time.) I’ll paste the Help > Report Bug the next time it happens.

3

I would start out by not indexing 4 TB of data into DT. And updating the OS.

OTOH, 8 GB doesn’t seem very much, but you didn’t say how much memory you have in your machine. Nor how many files you index.

4

8 GB is nothing. My about 600 GB of databases fires the memory usage to 24 TB only for simple searches.

2 Likes
5
  • What is shown in the File > Database Properties for the database?
  • How much RAM and free hard drive space do you have on this Mac?
6

There is a total of 2 TB of data indexed, put into one Group. But this Group is replicated twice; two of them are visible in the image.

Crike! I forgot to include my RAM and internal HDD storage. I’ll edit those now.

  • RAM: 16 GB
  • HDD: 994 GB
7

You have … 24 TB of RAM!? Is that correct?

8

RAM and free hard drive space:

  • RAM: 16 GB
  • Free HDD space: 11 GB

File > Database Properties:

Screen Shot 2025-10-01 at 15.09.30
Screen Shot 2025-10-01 at 15.09.301184×1448 185 KB

9

I’ll address Q2 as I’m sure you’ll get a lot of help on Q1.

FWIW, I began a reorganization of my files as well not too long ago. I had files everywhere in the finder, NAS, different application’s storage, DT, you name it. Some of it semi-organized through good intentions and most not. The segregation by application just complicated things when I needed to aggregate material across different applications for an objective. What did I do in a mind map somewhere, where is the powerpoint, that PDF, the work I’m writing, what I wrote long ago, those old genealogy files my mom left me when she passed away, things the kids did, etc. Yes search is your friend and DT’s built in AI helps but isn’t enough for someone predispositioned to be scattered like me.

I went through the Johnny Decimal system and ended up modifying it for my needs, but it really helped me organizationally and influenced how I use my tools and DT. It is centered on the analogy of the Library and the Dewey Decimal system. I’ve even adopted that paradigm further in how I think of indexing and terms of a “card catalog” and “call numbers”.

The advantage of it is it helps you with an organizational structure for everything. With some forethought you can organize how you prefer to live and work that makes sense to you. You create a simple numbering scheme that aligns. You can put the simple numbering at a level that makes sense to you wherever you work then (finder folder, your mind maps, DT, etc). That number (its at most “2 digits . 2 digits” long) helps you know “where” you organized it. And then something that’s crucial that I tried to avoid in the beginning is the heart of it all - creating an “index” or “card catalog” as I call it that tells you where things are so you can find things (besides search).

Basically you figure out everything about your life and things you care about and come up with rough divisions that actually are pretty separable and make sense for you, even when there is cross-talk between those areas. For me those areas became:

0 - ZK (my obsidian zettelkasten), 1 - Life (personal), 2 - Library (references), 3 - Business (company), 4 - IT&Media, and 5 - R&D

Each of those became further divided into how I think about each area:

  1. Life - 11. Family, 12. Finance, 13. Insurance, 14. Legal, 15. Hobbies/Community, 16. Home

And then within them you typically just go one level deeper. I now know organizationally where all the info is relative to individuals in my family, etc. The card catalog just indicates what those names and numbers are and anything I want to note about them generally and “where” things generally are. When I go “there” to put stuff or get stuff, I know I can go to the iCloud folder with the specific number or the DT DB and group with the specific number etc. If I’m perusing something in finder, I pretty much now at least know how I’ve organized it. And still the power of search, links and tags if you want cross folder access.

Anyway, where I am today is 6 DT databases. Three of them are indexed over shared iCloud folders with the J.D. hierarchy that is shared with others ones. One of the nonindexed ones is my general Library that has evolved into all sorts of things I keep and have organized but also now hosts research papers and references organized within this system based on a combination of arXiv, ACM, and IEEE taxonomies that make sense (to me). Similarly for the other areas. The card catalog keeps me honest and I can record if something is in DT, the NAS, a USB drive or whatever. I assure you I don’t go overboard, I just make a general referral for a whole area of material as to where it is. But I can pretty much easily decide where something now goes organizationally, e.g. 23.06 CL (Computation and Language) - the 2 is Library, the 3 is computer science, the 06 is the number I gave based on the arXiv for CL in its taxonomy. 23.06 is all I need to know which because numbers are short you tend to remember overtime. And using your reference manager (I use Zotero), papers are already coded with the same taxonomies. I put the “CL” just as a memory tickler in the group title just to help. Additionally this in turn helps the DT AI as you’ve now organized the types of items how you like to organize them. Then the power of linking and search helps it farther along.

Don’t know if this is the type of reply you were looking for, but it’s made a big difference for me. Seems daunting I know but it’s actually easy and you have to have some sort of plan. You can start big picture and work your way through things. I use Obsidian and templates which have been great for quickly adding, editing, and using the “card catalog”. Since the 0-ZK is indexed in DT, I can use DT or Obsidian whichever I prefer in the moment.

Good luck!

1 Like
10

The database contains a lot of words and a really huge number of unique words (due to poor OCR results or lots of languages?), therefore the RAM usage is not unusual.

11

No, DT uses 24 GB of RAM with a search in my scenario. My mini M2 Pro has 32 GB and my Studio, 36 GB. Currently DT is at about over 27 GB due accumulated searches and/or other stuff, with some memory pressure in my mini.

You have to keep in mind that DT has to load a lot of search information in memory. Personally, that DT uses 24-30 GB of RAM to perform operations on about 600 GB of data, less than 5% of the total, seems to me something very optimized.

12

Thanks so much for taking the time to write this thoughtful answer. A few days each year, the reality of my chaos hits me, and I … Well, instead of organizing, I begin fantasizing about which system I ought to use to do it. No progress made in a decade of fretting about this problem.

13

OK, thanks.

If you wanted to reduce the RAM needed, is there a certain class of record that you’d offload?

14

Well, similarly to you, the reality is that I lived with it until I retired. Wish I had done it long ago. It would have helped me both at home and at work. I think you know what to do. Find the time, pick something (anything), work toward it incrementally and adapt to your own needs as you learn. I don’t think you’ll regret it. Again, good luck!

15

The kind doesn’t matter, therefore basically those items with the highest word count.

16

I hope I don’t run in the same type of troubles. Currently I have a large database with mostly reference pdfs, accounting for 35 million words and 0.5 million unique words, which is well below the statistics of the case that initiated this thread. My M2 Mac with 32 Gb RAM reports a memory usage of about 1.25 Gb attributed to Devonthink (with two databases open), and 0 swaps to disc used. However, I’m in the process of transferring my lifelong references database from another program (TheBrain), which will at least triplicate the total number of references (pdfs, mostly medical and scientific, so likely lots of unique words). The other program (TheBrain), which also has almost instantaneous search within the database, is also loaded and my mac attributes about the same memory usage to it (1.27 Gb). I understand that there is no linear relationship between the size of the database and the memory usage, but I tend to think that I should be okay. I’m reluctant to split the databases because everything is interconnected. I would appreciate any thoughts, particularly if you think I’m heading for trouble.

17

Nothing sounds concerning about the info you’ve provided.