I have a question. I saw someone on Reddit claiming that the maximum allocatable memory for a GPU on a DGX Spark is 96GB. I’m curious if this is true, and if so, why is this limit imposed on the unified memory? Is it for system stability or something?
s0ne
1
Not true, there’s no such limit anywhere
Related topics
| Topic | Replies | Views | Activity | |
|---|---|---|---|---|
| Dgx spark can only use 119GB of memory running ComfyUI, so it is killed, isn't there 128GB of memory, why can't it be used? | 3 | 551 | December 4, 2025 | |
| The DGX system itself takes up 20GB memory? | 20 | 1561 | November 23, 2025 | |
| DGX Spark becomes unresponsive (“zombie”) instead of throwing CUDA OOM | 16 | 1696 | April 10, 2026 | |
| Memory Creep on DGX Spark: Where Your 128 GB Actually Goes (And How to Stop It) | 2 | 974 | March 30, 2026 | |
| Training transformer on DGX Spark | 3 | 1714 | October 16, 2025 | |
| Is there a way to reduce baseline memory usage on DGX Spark systems? | 2 | 369 | February 3, 2026 | |
| Difference in Total VRAM available for different Sparks | 18 | 1274 | June 4, 2026 | |
| Any advice on capping memory usage when running DL programs? | 1 | 69 | November 18, 2025 | |
| My DGX Spark Hangs ... is this normal? | 4 | 432 | April 13, 2026 | |
| DGX Spark stability / out of RAM / overheating | 27 | 1548 | June 11, 2026 |