Gigabyte MD71-HB0/Datto S4E18 CPU Throttle

[STANirvanaIND]

Hey Everyone,

Having an issue with a reflashed Gigabyte MD71-HB0 motherboard. The short and sweet is whenever a PSU is unplugged (I have 2 1280W PWS-1K28P-SQ) the CPU(s) drastically downclocks, to the point of being unusable. This occurs regardless of the load on the CPU (currently testing it with just the board and an HBA plugged in), and the current draw is more than in line for 1 PSU (less than 400W). Anybody have any idea why this would be? I've already reached out to Gigabyte support and their best guess is the PSU's aren't supplying enough power. Currently the PSUs are drawing ~380W and ~12W (active/passive).

Any insights are greatly appreciated! I've dealt with pre-built servers for years but this is my first "hage-page" (with no support!).

 

[RolloZ170]

if you have a PDB and SMBus connection to the motherboard:
this can be normal behavior (PROCHot) the BMC thinks one PSU is dead and forces the server to impede data loss.
i think you need to tell the BMC one PSU is enough to keep normal operation.
just a thought.

 

[STANirvanaIND]

if you have a PDB and SMBus connection to the motherboard:
this can be normal behavior (PROCHot) the BMC thinks one PSU is dead and forces the server to impede data loss.
i think you need to tell the BMC one PSU is enough to keep normal operation.
just a thought.

Haha now if I could just figure out how to tell the BMC that, we'd be golden! Unfortunately I can't grab any PSU stats in the BMC (not that it appears to have an option to configure something like this anyway), but can from Windows. I'm assuming this throttling is coming from the BIOS/BMC since it happens as soon as the system starts, before any OS has loaded.

I'm thinking of just removing the PDB cable since I can only read stats in the booted OS anyway and have a UPS to grab more detailed power info (and anything more detailed than that I don't have a use for).

Of note is that Datto calls this standard behavior and literally has an article about CPUs in the SIRIS 4 throttling if only one PSU is present....of course the fix is to get a working 2nd PSU or reach out to Datto support. (The Datto base version comes with 2x 800W PSUs). I've just never seen something like this on any systems I've used in the past (Dell & HP) and these had more power hungry CPUs with weaker PSUs.

 

[nabsltd]

The short and sweet is whenever a PSU is unplugged (I have 2 1280W PWS-1K28P-SQ) the CPU(s) drastically downclocks, to the point of being unusable.

I'm not familiar with the Gigabyte board, but I know that Supermicro boards with BMC connected to a dual power supply behave very differently depending on whether you remove the plug from the power supply (BMC complains) or if you remove the power supply completely (BMC is happy as long as the board is getting enough power).

So, what do you mean by "unplugged"?

 

[allish]

I'm not familiar with the Gigabyte board, but I know that Supermicro boards with BMC connected to a dual power supply behave very differently depending on whether you remove the plug from the power supply (BMC complains) or if you remove the power supply completely (BMC is happy as long as the board is getting enough power).

So, what do you mean by "unplugged"?
Regarding online games: playing online can also cause different reactions depending on the specific platform or game. For example, aviator-jeux can save your progress even if there are temporary problems with your internet connection, while others may require you to restart the game completely. It is important to know which platform or game you are using so that you can understand how it reacts to such situations.

If you check the BIOS/BMC settings for parameters related to processor power management. Maybe you can disable power management there.

 

[RolloZ170]

If you check the BIOS/BMC settings for parameters related to processor power management. Maybe you can disable power management there.

you can not overrule PROCHOT.

 

[STANirvanaIND]

I'm not familiar with the Gigabyte board, but I know that Supermicro boards with BMC connected to a dual power supply behave very differently depending on whether you remove the plug from the power supply (BMC complains) or if you remove the power supply completely (BMC is happy as long as the board is getting enough power).

So, what do you mean by "unplugged"?

Interesting, this behavior was observed removing the plug, I haven't tried removing the one of the PSUs themselves.

If you check the BIOS/BMC settings for parameters related to processor power management. Maybe you can disable power management there.

Unfortunately doesn't look to have had any effect (at least with the settings I used, anything related to C & P states)

you can not overrule PROCHOT.

So PROCHot features can kick off even if the CPU is well within safe temps (reporting ~20C)?

 

[RolloZ170]

So PROCHot features can kick off even if the CPU is well within safe temps (reporting ~20C)?

sure. its a pin to the CPU socket. sometimes the CPU can not know what happens outside.
the BMC has not many tools to stop the CPU from operation.