Arch Linux

Clifforus

Member

Registered: 2023-05-12

Posts: 1

Why might I suddenly need nvme_core.default_ps_max_latency_us=5500 ?

Hello all!

I've had the same M.2 NVMe SSD, a 1TB addlink S70, for about 3 years now, and have not had to use this kernel parameter:

nvme_core.default_ps_max_latency_us=5500

My system has been stable. However, over the past several weeks I've had system freezes that have gotten more and more frequent. I eventually managed to pin it down to the NVMe SSD (which is my root drive) failing to write and timing out.

After applying this kernel parameter, my system appears to be stable again.

As I was troubleshooting I tried a number of other things. Booting into kernels from before the freezing started, and LTS kernels, did not help. Strangely, disabling AMD Cool 'n' Quiet and Core Performance Boost in my BIOS did.

To be clear, I've done a lot of reading up on this kernel parameter and I think I more or less understand why it can help with NVMe drives causing system freezes and I am not looking for an explanation for what it does.

What I am curious about though, is why my system has been stable for years without this parameter, and now seemingly out of the blue, I need to use it? Was there a change in a package that might explain it? Is this drive failing and the kernel parameter is just working around it?

I'm really just trying to understand what's happened here for my own knowledge and curiosity. I'm replacing the SSD anyway so it's not really a huge deal, I would just like to get to the bottom of it. Thanks in advance for your replies!

joshtau

Member

Registered: 2014-01-28

Posts: 10

Re: Why might I suddenly need nvme_core.default_ps_max_latency_us=5500 ?

I was also having issues, but it was more around slow booting than freezes that was causing me grief.

Turning off that will disable ASPM, which might not be a great solution depending on your usecase.

Background: My motherboard is an x570, where the first NVMe goes to the CPU, and the second/third go to the chipset. Running the second/third were causing a bit of heat, so I am now using ane expansion card to run it off the CPU instead. Got a nice little speed bump too.

The problem was that I had LCLK DPM disabled in the BIOS, as I was having issues with my 5600XT graphics card and that solved it. However it causes issues with expansion cards (mine is a EZDIY-FAB Quad M.2 PCIe 4.0/3.0 X16 Expansion Card, 4 slots).

So yeah, double check you haven't disabled (or enabled, I guess) any power saving measures in your BIOS.