Intel pentium 8505 support

Singman

Well-Known Member
Sep 13, 2019
47
1
48
58
Hi,

Are you sure the Pentium 8505 is supported by Proxmox ?
I have a few mini PC with this processor and they are all really unstable under Proxmox (latest version, intel firmware installed).
Under Linux, I tried many stress tests and they are rock solid.
Same for memory with memtest86+.
 
Hi,

Are you sure the Pentium 8505 is supported by Proxmox ?
I have a few mini PC with this processor and they are all really unstable under Proxmox (latest version, intel firmware installed).
Under Linux, I tried many stress tests and they are rock solid.
Same for memory with memtest86+.
I'm experiencing stability issues with an 8505 NAS board as well. All hardware stress tests pass, but Proxmox locks up after a random interval. It stops responding to pings, and when connected to a monitor the blinking cursor to login at the terminal isn't blinking. The system goes into a hard lock, and nothing useful is recorded in the logs.

Were you able to resolve your stability issues?
 
Not really but look like it's related to Proxmox custom kernel. The NAS run pretty well under Debian Bookworm. The next test will be to install the Debian, then transform it with Proxmox packages. Here is a dmesg for a Debian :
 

Attachments

I'm running 6.8.12-13 right now. I'm running a test after disabling turbo in the BIOS. As the lockup appears to be random, I haven't been able to find a trigger for rapid testing. If this locks up again, I'll look to testing with 6.14.x and will report back.
 
I've been up for over 24 hours now at base clock speeds. (Turbo disabled.) I noticed that when turbo was enabled, the CPU would occasionally sit at idle holding around ~80C. That temperature should only be seen when at the highest clock speeds, and only temporarily. With turbo disabled the system has been sitting ~33-35C. I'm starting to suspect that the CPU is held at PL2 and not released back to PL1. Possibly a fault with the governor/kernel? I haven't changed the governor's settings from performance to power saving, but I might for future testing.

When possible, hopefully later this evening, I'll pull the pre-installed CPU heatsink and copper block off the die. I'll report back with any potential mating issues, or thermal paste issues. (Missing paste, too little paste, improperly applied paste, etc.) The CPU fan was plugged into the system fan port when I received the board, so I won't be surprised if I find a "remove before installation" sticker on the heatsink's mating surface.
 
Current uptime with turbo enabled and manually setting PL1 and PL2 is at 1day 23h 18m.

I pulled the heatsink off of the copper block. This was no easy task as there was too much thermal compound and the two flat surfaces created a suction effect. I didn't pry very hard at the copper block as I didn't want the possibility of chipping the die. (Flashbacks to early Athlon chips.) I cleaned both the copper block and heatsink's mating surfaces with isopropyl and applied a fresh coat of MX-4. A mild improvement was seen with temperatures, but the heatsink quickly becomes heat-soaked and I feel it's at the limit of what it's capable of. I may look at replacing it with a low profile heatsink that's rated for 95W+ to keep fan noise to a minimum while still providing adequate cooling.

I noticed an oddity when re-enabling turbo. The BIOS has a section for PL1 and PL2 settings and one specifically related to boot clocks. Prior to disabling turbo, the PL1 and PL2 options were set to '0' which should pull default values. When turbo was re-enabled, there were "non-0 entries" for both PL1 and PL2. PL1 was set to 75W, and PL2 was higher. The system shouldn't allow more than 55W to be pulled, but if the CPU was boosting to 55W as a base clock speed it would explain the CPU sitting at high temperatures previously. I don't know why this only seemed to affect Proxmox, while all other stress testing passed without issue.

I'll update if I encounter another lock up.