CPU clock drop when using 16*64 ram. I really dont understand why.
- Thread starter sharonyue
- Start date
Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.
Looks like you tried a lot of stuff, but I am missing a few things here:
1) it seems like you are running a memory only stress test, and judging by CPU frequency. Nobody knows what this test does exactly, so reduced CPU frequency might actually be normal
2) Check the sensor values. In particular, temperatures for CPU, CPU VRM, memory and memory VRM.
3) last not least: RDIMM can run hot. And modules with higher capacity tend to run even hotter. If this is in a workstation case without any special cooling for the RAM, 16x64GB is very likely to overheat during a memory stress-test. I have experienced this first-hand with a nearly identical system. 81°C is the threshold for DDR4, any more than that, and something will throttle. Again, looking at the sensor values would help.
You can either install this software from Supermicro: SuperDoctor® 5 (SD5) | Supermicro
Or connect another system to the IPMI port.
1) it seems like you are running a memory only stress test, and judging by CPU frequency. Nobody knows what this test does exactly, so reduced CPU frequency might actually be normal
2) Check the sensor values. In particular, temperatures for CPU, CPU VRM, memory and memory VRM.
3) last not least: RDIMM can run hot. And modules with higher capacity tend to run even hotter. If this is in a workstation case without any special cooling for the RAM, 16x64GB is very likely to overheat during a memory stress-test. I have experienced this first-hand with a nearly identical system. 81°C is the threshold for DDR4, any more than that, and something will throttle. Again, looking at the sensor values would help.
You can either install this software from Supermicro: SuperDoctor® 5 (SD5) | Supermicro
Or connect another system to the IPMI port.
Last edited:
I understand it is not happening when running Linux? If yes it might be the benchmark, check method or some energy related Windows setting.
As long as you don't perceive any obvious slowdowns in your real day to day tasks I would advise you to stop worrying.
As long as you don't perceive any obvious slowdowns in your real day to day tasks I would advise you to stop worrying.
It's package power, which definately includes MC.
My guesses are two possibilities: CPU temperature, or VRM temperature.
Hi guys, thanks for the reply.
Yes, I tried lots of methods. We are vendor. We use aida64 to stress our workstation before we send them to our customer. We found just ranning RAM test would cost less power, typically we have dozons of workstations running a test in the mean time, so we just run RAM testing to save power. The CPU clock should not drop either I run CPU test or RAM test. We follow the exact stessing rule before we send our machines to customers.
Since this does not happen quite often, for a new workstations, maybe 1 out of 100 or 150. So right now I have no idea whats the problem.
Typically when we have a workstation that cannot pass the stress test, we would put it away until we have time to deal with them. I will keep this updated until I have any more updates. Thank you all guys. I really appreciatie your replies.
Yes, I tried lots of methods. We are vendor. We use aida64 to stress our workstation before we send them to our customer. We found just ranning RAM test would cost less power, typically we have dozons of workstations running a test in the mean time, so we just run RAM testing to save power. The CPU clock should not drop either I run CPU test or RAM test. We follow the exact stessing rule before we send our machines to customers.
Since this does not happen quite often, for a new workstations, maybe 1 out of 100 or 150. So right now I have no idea whats the problem.
Typically when we have a workstation that cannot pass the stress test, we would put it away until we have time to deal with them. I will keep this updated until I have any more updates. Thank you all guys. I really appreciatie your replies.