Overclock your AMD Epyc

[neobenedict]

825W from the wall, 18000 19000 score breached!

Still headroom on those VRMs. This is pretty insane though.

47k in r23 but it set off some sort of continous beep alarm, maybe overcurrent or overtemperature. I'll dial it back a bit...

 

[Subatomic]

825W from the wall, 18000 19000 score breached!

Still headroom on those VRMs. This is pretty insane though.

47k in r23 but it set off some sort of continous beep alarm, maybe overcurrent or overtemperature. I'll dial it back a bit...

Sounds like a VRM temp spike above the limit. I think this should be logged by IPMI. If the cpus downclocked to something like 1 GHz and stay there than that is likely what happened.

 

[neobenedict]

Anyone know if memory can be overclocked (eg. from 2133mhz to something higher) at all on a H11dsi?

 

[alex_stief]

You can try to increase the frequency. But i never had any luck with that

 

[lixinran0809]

With the USB005 device, I was able to disable OCP and achieve benchmark stable at all core 3.75Ghz@1.4V with a 2s1905 on an H11SSL-C. Here are some screenshots I got. Based on the PowIRCenter readings, the VRM was moving more than 340A into the CPU at max (default OCP protection triggers at 240A). The maximum package power on hwinfo64 was about 550W. And the VRM's temperature was rising quite rapidly at the end of each Cinebench run, easily reaching around 95C. It is worth noting that the VRM is under a custom water block too, it will probably not be able to finish the run with the original heatsink.

I was able to run R15 with all core at 3.8Ghz (the top result). It is less demanding than the newer cinebenches. But anyway it was too unstable even just for benchmarks.


Based on the R23 result we are seeing a 10% increase over 2990WX! This makes me pretty happy but unless I can figure out a way to cool the VRM even better I guess the daily maximum overclock will still be around 3.4Ghz.

 

[esses]

So out of interest -

Is it possible to upgrade the EEPROM chip in a rev 1.x board without completely destroying everything?

Additionally, I use my board for some awful workstation/VM hybrid combination and I'm hunting for exposing further H11DSI BIOS settings such as BAR resize, does anyone know if there's been modding activity on that end?

I did try Supermicro support on both counts, but they're pretty useless.

 

[neobenedict]

So out of interest -

Is it possible to upgrade the EEPROM chip in a rev 1.x board without completely destroying everything?

Additionally, I use my board for some awful workstation/VM hybrid combination and I'm hunting for exposing further H11DSI BIOS settings such as BAR resize, does anyone know if there's been modding activity on that end?

I did try Supermicro support on both counts, but they're pretty useless.

I believe the only difference is the eeprom, but there may also be changes to the power delivery to support the higher TDP CPUs. You would still need to solder it on yourself though. Someone else may be able to comment better.

 

[Culbrelai]

So how high did you guys get stable? 3.8ghz?

I notice that there are no 16 or 8 core Epycs mentioned, any reason why? You think they'd be better overclockers?

 

[neobenedict]

I'm running at 3.1ghz for dual socket 7551 24/7. Anything higher drops performance or hits a VRM temperature limit. Plus the power usage is just absurd.

 

[JamesHawk]

Hello guys,

I have a h12 ssl-i mb and I follow these steps to modify a bios. But I can't flash it with IPMI.

It says "BIOS can not pass secure validation."

How can I flash it?

Thanks.

 

[Zhang]

Hello guys,

I have a h12 ssl-i mb and I follow these steps to modify a bios. But I can't flash it with IPMI.

It says "BIOS can not pass secure validation."

How can I flash it?

Thanks.

Supermicro probably added BIOS verification for newer H12 boards? As far as I can recall, on H11 boards I could flash any modified BIOS without security verification.

 

[Moopingkrup]

System epyc 7571x2

 

[BabbleBones]

Hello I'd like to try replacing the BIOS on some cheaply available v1 boards however I need somebody to tell me the model number written on top of the rev 2 EEPROM chip, please. I can order a few, flash it, and swap them out, let you know if it works.

Drop the model number of your rev 2 EEPROM 256Mbit flash chip here if you would, kindly.

 

[systemviper]

@Moopingkrup, Love the pics, Great Setup...

 

[Alex0220]

AMD EPYC 7401P @ 3.7 GHz 1.3V

 

[kbhinkle]

Hello,
I am looking for 64 Core Rome ES CPU's. As many as 8-10. 'JUG5' ES p/n's are fine. If anyone has any insight (other than Ebay) a connection would be much appreciated. Thanks in advance.

 

[hiuyu]

I've got a question. I did overclock the processer by editing the p-state, but the frequency still remain same in the virtual machine. Doesn't successful change the performance for vm or there is no change?

 

[AXm77]

Did anyone try to run pair of 2S1905A4VIHF4's on H11DSi rev 2.0 ?

 

[Zhang]

Did anyone try to run pair of 2S1905A4VIHF4's on H11DSi rev 2.0 ?

I can confirm a pair of 2S1905A4VIHF4 works flawlessly on H11DSi rev 2.0. They can be easily overclocked to all core 3.0 to 3.2 GHz, or maybe even higher if you have good VRM cooling.

 

[AXm77]

I can confirm a pair of 2S1905A4VIHF4 works flawlessly on H11DSi rev 2.0. They can be easily overclocked to all core 3.0 to 3.2 GHz, or maybe even higher if you have good VRM cooling.

Thanks. That is good news