BLACK FRIDAY: 50% off my book π₯
A no-bullshit guide to growing on X. No tricks, just HARD, consistent work.
Code BLACKFRIDAY26 until Nov 30.
https://t.co/xVGjoqLFL6
Ordered 2x Seagate 28TB (ST28000NT000) hard drives and a UGREEN DXP4800 GT NAS. I'll be retiring my old 2-bay NAS and moving its 14TB drives into the new one.
Total 84 TB of storage!
@plotarmordev@m7jazin@ashxhart Add another spark there and it easily beats it though. Nevertheless, for image/video gen I actually recommend a real GPU π
sparkDash v2.91.0 is out β‘οΈ
π§Ή Remove Agent Junk! (NEW)
- Finds what AI coding agents left behind
- Old worktrees, /tmp folders, session logs
- Stale editor servers, scratch scripts, .bak files
- Smart pick ticks only the safe stuff
- Remove it all in one click, with safety checks
Also in Storage Manager
- Now works on Linux GPU hosts, not just Sparks
Also new
- ignis engines detected: live tok/s, KV use, TTFT
- Pico: decode rate + cached/uncached prefill
- macOS: physical memory + memory pressure
- Occupancy helper supports DSH 0.2.2
Faster & more efficiant
- GPU stats no longer start nvidia-smi on every poll
- One long-running reader instead of 4 calls per poll
- Less CPU on the host
Fixed
- Unreadable bandwidth shows "unknown", not 0 GB/s
- Fleet energy chart survives an offline node
- Energy per token labelled "lower is better"
- Pico decode rate no longer hidden
- Clearer prefill label (cached tokens excluded)
- macOS health works when pressure is unknown
- DSH token never sent over plain http
- Unsupported GPUs no longer crash-loop
Lots of changes here came straight from issues and PRs, thank you!
Get it here:
https://t.co/8H3h5wjPOh
sparkDash v2.91.0 is out β‘οΈ
π§Ή Remove Agent Junk! (NEW)
- Finds what AI coding agents left behind
- Old worktrees, /tmp folders, session logs
- Stale editor servers, scratch scripts, .bak files
- Smart pick ticks only the safe stuff
- Remove it all in one click, with safety checks
Also in Storage Manager
- Now works on Linux GPU hosts, not just Sparks
Also new
- ignis engines detected: live tok/s, KV use, TTFT
- Pico: decode rate + cached/uncached prefill
- macOS: physical memory + memory pressure
- Occupancy helper supports DSH 0.2.2
Faster & more efficiant
- GPU stats no longer start nvidia-smi on every poll
- One long-running reader instead of 4 calls per poll
- Less CPU on the host
Fixed
- Unreadable bandwidth shows "unknown", not 0 GB/s
- Fleet energy chart survives an offline node
- Energy per token labelled "lower is better"
- Pico decode rate no longer hidden
- Clearer prefill label (cached tokens excluded)
- macOS health works when pressure is unknown
- DSH token never sent over plain http
- Unsupported GPUs no longer crash-loop
Lots of changes here came straight from issues and PRs, thank you!
Get it here:
https://t.co/8H3h5wjPOh