Question My SSD is misbehaving and I'm not sure how to troubleshoot it.

[TehPenguin]

Hi,
One of my SSDs, the Crucial P3 Plus 4TB Gen4, does not behave how I'd expect an ssd to behave.
For example, when downloading a game over steam, it cannot sustain the full 600mbit download speed with the active time in the task manager going to 100% periodically.
https://paste.pics/R9RVU
When playing with LLMs that are all stored on it, loading a new model in slowed down to a crawl, matching HDD speeds. I believe I saw a similar speed slowdown when loading a game, but I couldn't reproduce it right now.
During the CrystalDiskMark, the response time got all the way up to 250.000+ ms in task manager at one point. The results are very inconsistent.
Here's an example: https://paste.pics/R9RY5
Weirdly enough, the last time I tried to see what's the problem, the crystal disk results were fine with the exception of the RND4KQ1T1 being rather low. Now the read and write speeds fluctuate between the advertised 5GB/s and 90Mb/s or so.
The temperature got up to 70°C during the test, there is a heatsink on the SSD. When downloading through steam, the temperature goes up to around 61°C. When idle its around 50°C

My other specs:
Win 10
7900x3D
Asus ROG Strix B650E-E
64GB DDR5 3000MHz CL30

How can I go about finding the culprit? Thanks!

 

[Bob.B]

Hi,
One of my SSDs, the Crucial P3 Plus 4TB Gen4, does not behave how I'd expect an ssd to behave.
For example, when downloading a game over steam, it cannot sustain the full 600mbit download speed with the active time in the task manager going to 100% periodically.
https://paste.pics/R9RVU
When playing with LLMs that are all stored on it, loading a new model in slowed down to a crawl, matching HDD speeds. I believe I saw a similar speed slowdown when loading a game, but I couldn't reproduce it right now.
During the CrystalDiskMark, the response time got all the way up to 250.000+ ms in task manager at one point. The results are very inconsistent.
Here's an example: https://paste.pics/R9RY5
Weirdly enough, the last time I tried to see what's the problem, the crystal disk results were fine with the exception of the RND4KQ1T1 being rather low. Now the read and write speeds fluctuate between the advertised 5GB/s and 90Mb/s or so.
The temperature got up to 70°C during the test, there is a heatsink on the SSD. When downloading through steam, the temperature goes up to around 61°C. When idle its around 50°C

My other specs:
Win 10
7900x3D
Asus ROG Strix B650E-E
64GB DDR5 3000MHz CL30

How can I go about finding the culprit? Thanks!

Reseat the ssd in the slot a couple of times.....test.
Swap slots.....test.

 

[MWink64]

How full is it? How much are you writing to it at one time? Do you know if TRIM is enabled and functioning? It's quite possible the pSLC cache is filling up. Since it's a DRAM-less QLC drive, when you exceed what the cache can handle, performance will absolutely tank. See this review. If it's getting full, you could try freeing up some space, trimming the drive, then letting it sit idle for a few hours.

 

[USAFRet]

For example, when downloading a game over steam, it cannot sustain the full 600mbit download speed with the active time in the task manager going to 100% periodically.

The performance limit here is what Steam provides at any particular moment, not the drive speed.

 

[MWink64]

The performance limit here is what Steam provides at any particular moment, not the drive speed.

If the drive is maintaining a 100% activity level, that suggests it's the bottleneck.

 

[TehPenguin]

The performance limit here is what Steam provides at any particular moment, not the drive speed.

It only slows down when the active time goes to 100%

Reseat the ssd in the slot a couple of times.....test.
Swap slots.....test.

Huh, didn't think this could be a mechanical issue. Will do this on the weekend!

How full is it? How much are you writing to it at one time? Do you know if TRIM is enabled and functioning? It's quite possible the pSLC cache is filling up. Since it's a DRAM-less QLC drive, when you exceed what the cache can handle, performance will absolutely tank. See this review. If it's getting full, you could try freeing up some space, trimming the drive, then letting it sit idle for a few hours.

It's 75% full, but it did this even when 50% full.