Finally: Overclocking EPYC Rome ES

chraac

New Member
May 5, 2019
21
0
1
Hi guys,
I recently purchase a 7452 QS 57-04 on EPYCD8, 8x16GB 2400.
And I found that some temperature in HWInfo64 was ridiculous high:
1608389958961.png
As you can see that "CPU Die (average)" was around 72°C and when I run the Prime95 torture test it'll quick raise up to 100°C (TjMax), but the CCD temp seems pretty low (30-40°C).
Can someone tell me what this temp actually is and it's ok to ignore it?
Thanks
 

dragonian

New Member
Jan 3, 2020
19
10
3
I recently put together a system with a ZS1711E3VIVG5 (32c) on a H11SSL-i, and I'm pretty impressed.
I was originally looking a retail 16c Rome (or waiting for Milan) for whole house virt server, and found this processor for 1/2 the price, with 2x more cores.

Some little benchmarking:
FreqCinebench R20 Killawatt at wall Hwinfo VIDComments
default1700 MHz7392129W0.844(seems all cores at 1.6)
"max multi core":3450 MHz14998380W1.24
"best of both"3300 MHz7311213W1.18 (most cores running at 2.7-2.8)
"best of both - locked"3300 MHz14657337W1.18 (all cores at 3.3)
"best of both - locked"3200 MHz14354305W1.11(all cores at 3.2)

I think I'm inclined to just run best of both with the lock setting @ 3.3 or 3.2. Any real reason not to? This way I don't have to futz with the EDC bug.
Temps seem fine at load CPU: 74 VRM: 58
Idle at the wall seems to be between 80-90W mostly independent of setting/freq.

Next to try this in linux with Proxmox.
 
  • Like
Reactions: lixinran0809

mirrormax

Member
Apr 10, 2020
77
30
18
I don't trust any voltage by hwinfo/aida64 on these epyc cpus, they are all over the place from my comparison to direct monitoring from the vrm debug port, ipmi sensors seems more accurate but dont always show everything.

That said i've also seen 1.3v+ when switching between presets, i think there's a bug when going from for example high single to high multi where it ends up at 1.344v. restarting the tool and applying high multi again sets it back to 1.05.

Also remember that VID is not the same as actual voltage to the cpu(vdd)
VID is the chips expected voltage at certain freq.
for example high multicore on 64c sets 1.05v and 3800, the reported VID jumps to 1.35v, the actual voltage under load is 0.953v.
though i'm guessing the VID and VDD will be closer on the 32cores as the presets don't use the EDC bug. still wouldn't expect 1.3+ unless you are going for 4ghz or something.
 
  • Like
Reactions: chraac

Wilhelm

New Member
Aug 1, 2019
4
0
1
I am running 2 x ZS1406E2VJUG5 on the H11DSI, and the motherboard is reporting a P1_SOCRUN of 1.174V (upper critical). Is anyone able to share how they solved this? I have placed a few 40mm fans on the motherboard heatsinks but so far it has made no difference.
 

chraac

New Member
May 5, 2019
21
0
1
I don't trust any voltage by hwinfo/aida64 on these epyc cpus, they are all over the place from my comparison to direct monitoring from the vrm debug port, ipmi sensors seems more accurate but dont always show everything.

That said i've also seen 1.3v+ when switching between presets, i think there's a bug when going from for example high single to high multi where it ends up at 1.344v. restarting the tool and applying high multi again sets it back to 1.05.

Also remember that VID is not the same as actual voltage to the cpu(vdd)
VID is the chips expected voltage at certain freq.
for example high multicore on 64c sets 1.05v and 3800, the reported VID jumps to 1.35v, the actual voltage under load is 0.953v.
though i'm guessing the VID and VDD will be closer on the 32cores as the presets don't use the EDC bug. still wouldn't expect 1.3+ unless you are going for 4ghz or something.
I have and 32c 57-04 qs, and the VID is 1.26v+ in 3.45G. When I manually set the voltage to 1.05v ,the cinebench r20 score is lower than 1.26v+, even lower than 3.0G.
 

bayleyw

Member
Jan 8, 2014
86
21
8
I have and 32c 57-04 qs, and the VID is 1.26v+ in 3.45G. When I manually set the voltage to 1.05v ,the cinebench r20 score is lower than 1.26v+, even lower than 3.0G.
That's the SMU at work there - based on the core voltage, the SMU will clock stretch to avoid crashes. It's why, for example, you can't undervolt the desktop Ryzen parts much.
 

chraac

New Member
May 5, 2019
21
0
1
That's the SMU at work there - based on the core voltage, the SMU will clock stretch to avoid crashes. It's why, for example, you can't undervolt the desktop Ryzen parts much.
Maybe there are some internally freq throttling, I see the freq is constantly 3.45G@1.05v
 

dassiq

Member
Jul 10, 2017
32
9
8
46
Hi

I have the same setup, do you happen to find the Rome bios ?

All we need is the Bios v2.0 for the h11DSU board Rev 02. We can extract the last half and use it on the Rev 01 board.

Regards

Hey everyone, first post here. I tried going through the thread but didn't find any specific answers.

Could someone point me to a version/stepping reference for epyc processors? Specifically, QS (stepping 0?) Epyc on ebay sometimes claim a lower frequency than retail, sometimes not. Do they turbo up to acceptable frequencies without tinkering? 64 cores for $2000 sound appealing but I'd rather not rely on software tools to set the frequency.

I have a SM server with an H11DSU-IN rev1, currently running a pair of 7371s. Are there modded bioses with Rome support going around for this board? I've only seen the ones from OP for standard ATX systems.

Thanks!
 

chraac

New Member
May 5, 2019
21
0
1
I don't trust any voltage by hwinfo/aida64 on these epyc cpus, they are all over the place from my comparison to direct monitoring from the vrm debug port, ipmi sensors seems more accurate but dont always show everything.

That said i've also seen 1.3v+ when switching between presets, i think there's a bug when going from for example high single to high multi where it ends up at 1.344v. restarting the tool and applying high multi again sets it back to 1.05.

Also remember that VID is not the same as actual voltage to the cpu(vdd)
VID is the chips expected voltage at certain freq.
for example high multicore on 64c sets 1.05v and 3800, the reported VID jumps to 1.35v, the actual voltage under load is 0.953v.
though i'm guessing the VID and VDD will be closer on the 32cores as the presets don't use the EDC bug. still wouldn't expect 1.3+ unless you are going for 4ghz or something.
Hi guys, I'm finally running my 7452qs at 3.2ghz 1.1v, just want to know that if this voltage is ok for daily usage and whether the chip will degrade
 

dassiq

Member
Jul 10, 2017
32
9
8
46
@ExecutableFix , I am trying to mod the h11DSU bios R2.0 for Rev01 board. Your posts says to just take the last half of the ROME bios, what does that exactly mean ? does it have to be exactly half ? or should I just cut the bios from the address where there is some data starting after the half point (hex address) . But this will result in a slightly smaller file then exactly 16mb.

"
But the CPU doesn't post on my -insert brand- motherboard!!!
That's what these things do. Engineering sample CPUs just aren't supported by every board out of the box. There is a simple trick to make it work on any board, even on Rev. 1 boards. The only thing you need to do is copy the second half of the bios (the Rome part) and duplicate it on Rev. 2 boards or just upload that on Rev. 1 boards. Make sure the bios has an Agesa < 1.0.0.3 because otherwise it will not work. The engineering samples do work out of the box on the Asus KRPA-U16. I’ve included some pre-‘modded’ bioses below. "
 

madbrad

New Member
Dec 10, 2020
15
0
1
Guys, I need some urgent help, please.
Finally all my components arrived and I mounted a brand new H1SSL-I-O and I can't power it on. No led, no fans spinning, nothing. I haven't put the CPU in in the socket yet. I tried powering it on with and without the cpu 8-pin. I have the power button connected to pins 1 and 2 of the JFP1. It is in the chassis of my old PC so the PSU works, I had just previously powered my old mono with it. What can I try? Do I need to power the 4-pin as well? Some jumper stuff? Some clear CMOS magic? Do I need to install the ES CPU and cooler? I want to flash the custom BIOS via IPMI.
 

madbrad

New Member
Dec 10, 2020
15
0
1
Guys, I need some urgent help, please.
Finally all my components arrived and I mounted a brand new H1SSL-I-O and I can't power it on. No led, no fans spinning, nothing. I haven't put the CPU in in the socket yet. I tried powering it on with and without the cpu 8-pin. I have the power button connected to pins 1 and 2 of the JFP1. It is in the chassis of my old PC so the PSU works, I had just previously powered my old mono with it. What can I try? Do I need to power the 4-pin as well? Some jumper stuff? Some clear CMOS magic? Do I need to install the ES CPU and cooler? I want to flash the custom BIOS via IPMI.
Update: It powered on after I installed the CPU but died after I flashed BIOS via IPMI.
I installed the CPU (the 32 core ES mentioned in the OP), pressed the power button, nothing. Then I left for a minute and when I came back the board was powered on. I flashed the Rome rev 2 bios from @ExecutableFix and after it the the board is rev 2.0 (version is printed on the actual board). About 2 minutes later it shut itself down and won't start anymore.
 

madbrad

New Member
Dec 10, 2020
15
0
1
Update: It powered on after I installed the CPU but died after I flashed BIOS via IPMI.
I installed the CPU (the 32 core ES mentioned in the OP), pressed the power button, nothing. Then I left for a minute and when I came back the board was powered on. I flashed the Rome rev 2 bios from @ExecutableFix and after it the the board is rev 2.0 (version is printed on the actual board). About 2 minutes later it shut itself down and won't start anymore.
Update2: I cleared CMOS via shorting the pads on the board and now I only get the LEDM1 flashing. I can connect via IPMI but I can't power on the system. Any hints appreciated. These are currently the values shown in the IPMI.
Firmware Revision: 01.52.00​
Firmware Build Time: 11/18/2019​
BIOS Version: 2.1
BIOS Build Time: 02/21/2020
Redfish Version: 1.0.1
CPLD Version: 02.b4.02
 

gsrcrxsi

Member
Dec 12, 2018
59
5
8
Update2: I cleared CMOS via shorting the pads on the board and now I only get the LEDM1 flashing. I can connect via IPMI but I can't power on the system. Any hints appreciated. These are currently the values shown in the IPMI.
Firmware Revision: 01.52.00​
Firmware Build Time: 11/18/2019​
BIOS Version: 2.1
BIOS Build Time: 02/21/2020
Redfish Version: 1.0.1
CPLD Version: 02.b4.02
reseat the CPU, and make sure it's not backwards in the socket. sometimes these ebay sellers are careless and don't put the CPU the right way in the carrier tray. it's happened to me and a friend of mine.

the system isnt supposed to turn on if the CPU isnt installed (your first post) so that's normal.
 

madbrad

New Member
Dec 10, 2020
15
0
1
reseat the CPU, and make sure it's not backwards in the socket. sometimes these ebay sellers are careless and don't put the CPU the right way in the carrier tray. it's happened to me and a friend of mine.

the system isnt supposed to turn on if the CPU isnt installed (your first post) so that's normal.
The CPU came without a carrier tray, I placed it in manually (whilst shitting my pants), and it gets correctly detected via IPMI as Processor: AMD Eng Sample: 2S1705E3VIVG5_20/17_N .
Turns out the problem is the power button does not work. I am so relieved and at the same time I can't power on the damned thing. I can get to BIOS if I power it via IPMI but I can't even power it on by shorting the pins 1 and 2 of the JF1. Is it possible that it's defective?
 

madbrad

New Member
Dec 10, 2020
15
0
1
Update2: I cleared CMOS via shorting the pads on the board and now I only get the LEDM1 flashing. I can connect via IPMI but I can't power on the system. Any hints appreciated. These are currently the values shown in the IPMI.
Firmware Revision: 01.52.00​
Firmware Build Time: 11/18/2019​
BIOS Version: 2.1
BIOS Build Time: 02/21/2020
Redfish Version: 1.0.1
CPLD Version: 02.b4.02
Update3: turns out I am an idiot and the issue is the power button / JF1 connector. I have a Thermaltake Core P6 case and the power button worked fine on the old moboard. It won't even work if I try to short it with a screwdriver. I am now turning it on via IPMI but obviously this is not a solution. Any advice here?
 

gsrcrxsi

Member
Dec 12, 2018
59
5
8
Are you using the right pins? On that board pins 1 and 2 for the power button are two pins closest to the USB3 front panel connector.
 

mirrormax

Member
Apr 10, 2020
77
30
18
You have 2.1 bios pretty sure someone mentioned in here that dos t work with some of the ES CPUs(zs ones AFAIK) . Try 2.0 maybe
 

madbrad

New Member
Dec 10, 2020
15
0
1
Update3: turns out I am an idiot and the issue is the power button / JF1 connector. I have a Thermaltake Core P6 case and the power button worked fine on the old moboard. It won't even work if I try to short it with a screwdriver. I am now turning it on via IPMI but obviously this is not a solution. Any advice here?
Update4: Thank you all for replies but everything is dead again, I only hear a buzzing noise from the PSU/mobo (which I had heard before).
I had to leave for a few hours but returned. I connected the power button to the correct JF1 pins and I could power it on and was happy. However, after I powered off the PSU so I could reseat the external gpu (I had to turn the case over and rearrange some PSU cable management for this), it's all dead again. I don't even get the LEDM1 flashing. I only hear a gruesome buzzing noise as if there's something powering some coils. It seems to be coming from both the PSU and mono. I had previously heard this before several times when the LEDM1 was not being powered. I assume it's either short circuiting somehow or it may have something to do with the CPU being put in without a tray (maybe it shifted when I turned the case over?). I applied liquid metal cooling compound on the CPU and would hate to have to remove it, so I'd rather hope it's something else.
This is now true despair.
 

gsrcrxsi

Member
Dec 12, 2018
59
5
8
sounds like a PSU issue honestly. try a different PSU.

were you using the wrong pins for the power button before?