Overclock your AMD Epyc

Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.

nero243

Active Member
Oct 28, 2018
108
78
28
MrCake tested the Modded bios Version 1.3, it's working (still not stable for ES).

I fixed the cosmetic issues on V 1.0C, the IFR edit is not needed for Supermicro boards. The download link is updated in the start post.
 
  • Like
Reactions: RageBone

nero243

Active Member
Oct 28, 2018
108
78
28
Hello!
I've updated the ZenStates app with Rome CPUID, however I haven't figured out the P-States (or manual overclock) yet.
There is no BKDG for family 17h and setting the well-known P-State MSRs doesn't change the CPU frequency. I have no access to NDA stuff.
If I set the P-State, then read it back it shows the newly set multiplier, but the actual frequency is still the old one. Same with the Linux script.
So I'm currently scratching my head over this.

It would help if you report that at least it detects the CPU correctly.
Performance Bias options "might" work, perhaps some of the others, but not P-State :(
I'm not even sure if the MSR are shared between Matisse and Rome.
Hopefully it won't "break" your system, but apparently I have no way to test.
EPYC1 shared the CPUID with desktop parts, but now Rome has a new CPUID.


Download zip from my github releases page: irusanov/ZenStates

Changelog:
  • Added Rome support (P-States won't work, but at least it should start without complaining about unsupported CPU)
  • Added support for 32-bit OS (tested on XP SP3 x86)
  • Reduced .NET Framework version to 4.0 (lowest I can go for the current codebase)

Update 21 Aug:
I have managed to change the frequency with commands to SMU. Unfortunately there's no public document describing all commands. I've updated the readme with what I've been able to find so far mostly by trial and error method. Should be able to provide at least manual overclock soon.

irusanov/ZenStates
https://developer.amd.com/wp-content/resources/55803_0.54-PUB.pdf

Here you go, FID, DID and VID are on page 151-152

Edit:
Compared it to the PPR from Naples, nothing apparently has changed.

So from what i've read the SMU commands/adresses have changed with Rome. Does this even matter, if you apply an OC to the P0 MSR, higher than the max turbo, the OC mode should automatically engage and the shadow pstates are discarded. Or do you have to enable the OC mode to set anything and does setting P-States even still work through the classical MSR? The PPR suggests yes, since the Bits are Read/Write.
 
Last edited:

Eliakim

New Member
Oct 7, 2019
5
0
1
i tried zenstate and only cpu1 is affected, the second cpu is not.
I have a H11DSI-NT with 7601 (2 cpu).

Anything i can do?
 

MrCake117

Member
Feb 28, 2019
32
24
8
25
Japan
i tried zenstate and only cpu1 is affected, the second cpu is not.
I have a H11DSI-NT with 7601 (2 cpu).

Anything i can do?

Hello

So about multiple CPUs configuration you just need to start AsusZenstates as usual,click on apply in order to apply your custom pstates settings to CPU0.

-Once is done go in task manager,find AsusZsSrv.exe (not AsusZenStates.exe) then click on define affinity.
-After you'll change Group 0 to group 1,check "use all core" and click Ok.
-Now you're able to modify the CPU1 pstates by clicking on apply in AsusZenstates app.

In order to revert to CPU0 just change group 0 and don't forget to check "use all core" in define affinity .
Each time you reboot the computer you'll need to apply this process.
If you get an "access denied" error when you want to change the cpu affinity just kill the process and restart it .

I also contacted supermicro in order to ask for a custom bios but they didn't approve my request unless you pay at least 500 motherboard/BIOS...

Imgur

I hope everything is clear for you

Thank you
 

nero243

Active Member
Oct 28, 2018
108
78
28
i tried zenstate and only cpu1 is affected, the second cpu is not.
I have a H11DSI-NT with 7601 (2 cpu).

Anything i can do?
I added the procedure to the start post.

Your running retail CPUs, right? Can you maybe try the custom core P-states from the modded v1.3 bios provided in the startpost?


Edit:

Added the first V2.0 Bios for the H11SSLi with the unlocked new CBS.
Consider this experimental, since idk which menu layout is the right one. There are like 2 full bios layouts in there all with their own variables and 2 CBS menus alone in the new layout. Also this weird thing isn't UBU compatible so i have to use MMtool and do a switcharound for BCP editing. I will write up a step by step DIY when i know it's working.

You'll get a few new settings on rome like XFR performance enhancer, where you can boost the max turbo up. Without being stuck in OCmode. Also DRAM OC goes now up to 4200MHz!

Naples and Rome just indicate which CBS menu is unlocked, both should run fine on both CPUs. Just the CBS Settings aren't going to work if you run the rome bios on naples.
 
Last edited:

charliehorse55

New Member
Oct 3, 2019
16
2
3
So, to summarize, do we have overclocking support on Rome now? With the zenstates application or only with a modded bios?
 

MrCake117

Member
Feb 28, 2019
32
24
8
25
Japan
So, to summarize, do we have overclocking support on Rome now? With the zenstates application or only with a modded bios?
It's untested, we don't know if Rome custom pstates or custom cpb settings works, and about zenstates, it still hasn't be updated for Rome if I'm right
 
Last edited:

nero243

Active Member
Oct 28, 2018
108
78
28
So, to summarize, do we have overclocking support on Rome now? With the zenstates application or only with a modded bios?
AMD Epyc Rome is in someway unlocked as far as i know. The Bios Mod P-States will most likely not work and Zenstates does not have Zen2 Support yet. I think i have everything to implement Zen2 support but infared/irusanov seems to have taken some time of from this project and the forums.


Also, the H11 v2.0 Bios versions are posting but the CBS setup is not showing up for naples. I probably need to edit the other AMITSE module...
 
Last edited:

Gordan

Member
Nov 18, 2019
39
8
8
I have an Asrock EPYCD8-2T with a 7402P and I can confirm that all BIOS settings relating to custom P states and TDP are completely ignored.

I got a beta BIOS out of Asrock to try to address some of the other bugs I ran into, but that BIOS not only doesn't fix the custom P states being ignored, it completely removes the custom P states section from the BIOS. It didn't remove the custom TDP options but they seem to do nothing, the exact same boost clocks are achieved under the exact same load regardless of whether I set the TDP to "Auto" (which presumably means 180W as per spec) or as high as 280W. And since the temperature never gets past 65C under full load, I can only assume the TDP settings are completely ignored.

Has anyone managed to overclock a Zen2 EPYC (7xx2) in any motherboard?
 

nero243

Active Member
Oct 28, 2018
108
78
28
Has anyone managed to overclock a Zen2 EPYC (7xx2) in any motherboard?
In the video i posted we have a guy claiming it's possbile and so is the word in the street.

To-Do List for a pleasant Rome OC experience:

Find and implement the right SMU addresses for Rome into ZenStates, if they are even different from matisse
Get PPT, TDC, EDC and BCLK/FSB working
manually test the "unlock" procedure
script/automate the unlock procedure

I'm pretty much useless for everything on this list.
 

Gordan

Member
Nov 18, 2019
39
8
8
I have made some progress with the EPYCD8 board.

Memory OC works fine. I settled for 3200 and tightened timings because 3200 is officially supported by the CPU and the board. The test BIOS v2.42 removes higher settings so I haven't tried going higher. BIOS v2.30 goes all the way up to 4200 IIRC.

P states are completely ignored.
Boost Fmax is completely ignored.

TDP limit is NOT ignored, and this lead to a finding that may sink the OC potential - as far as I can tell, this board is limited to 200W. Bumping the TDP form "auto" (180W) to 200W bumps the full load all-core boost from about 3000 to 3125, but going further makes no difference. The problem being that with P states ignored - preventing undervolting, and TDP limited to 200W, scope for OC is effectively non-existent because the TDP limit even prevents hitting the maximum boost clocks under full load on all cores.

Another observation - MCH counts toward the TDP. Going from 2933 memory speed to 3200 reduced boost clocks under full load. So there is a tradeoff to be chosen between memory bandwidth and boost clocks.
 

nero243

Active Member
Oct 28, 2018
108
78
28
I have made some progress with the EPYCD8 board.

Memory OC works fine. I settled for 3200 and tightened timings because 3200 is officially supported by the CPU and the board. The test BIOS v2.42 removes higher settings so I haven't tried going higher. BIOS v2.30 goes all the way up to 4200 IIRC.

P states are completely ignored.
Boost Fmax is completely ignored.

TDP limit is NOT ignored, and this lead to a finding that may sink the OC potential - as far as I can tell, this board is limited to 200W. Bumping the TDP form "auto" (180W) to 200W bumps the full load all-core boost from about 3000 to 3125, but going further makes no difference. The problem being that with P states ignored - preventing undervolting, and TDP limited to 200W, scope for OC is effectively non-existent because the TDP limit even prevents hitting the maximum boost clocks under full load on all cores.

Another observation - MCH counts toward the TDP. Going from 2933 memory speed to 3200 reduced boost clocks under full load. So there is a tradeoff to be chosen between memory bandwidth and boost clocks.
I think some of the CBS options won't work because these are leftovers meant for TR3. They're labled SSP (starship) and Rome is VH (valhalla). But as i said from the start bios P-states will not work.

You could just hook up a IR i2c dongle and bring your TDP back in place. Would also be interesting how that's handled by the SMU, since it looses control over anything the voltage controller does.

I might be able to mod your newer bios so you can use the higher memory OC options with the newer version, if there's any need for that.
 

Wasmachineman_NL

Wittgenstein the Supercomputer FTW!
Aug 7, 2019
1,872
617
113
I think some of the CBS options won't work because these are leftovers meant for TR3. They're labled SSP (starship) and Rome is VH (valhalla). But as i said from the start bios P-states will not work.

You could just hook up a IR i2c dongle and bring your TDP back in place. Would also be interesting how that's handled by the SMU, since it looses control over anything the voltage controller does.

I might be able to mod your newer bios so you can use the higher memory OC options with the newer version, if there's any need for that.
Ooooh, so that's what happened to Valhalla: it became Epyc!

@topic: Imagine EVGA (or some other madman brand) coming out with a SR-4 dual SP3r2 board with insane VRMs, holy shit. 128 cores at 4 GHz plus will make minced meat of every workload!
 

nero243

Active Member
Oct 28, 2018
108
78
28
Ooooh, so that's what happened to Valhalla: it became Epyc!

@topic: Imagine EVGA (or some other madman brand) coming out with a SR-4 dual SP3r2 board with insane VRMs, holy shit. 128 cores at 4 GHz plus will make minced meat of every workload!
Valhalla became Rome and Starship became Castle Peak. And the rest of the codenames probabaly won't even make sense to the insiders.

This is probably not going to happen as far as i'm concerned the whole Epyc OC stuff is just a whole lot of Bugs.

On Naples:
FID locked, but with a DID of 2 you're ready to OC
On Rome:
Everything locked but a new Bug is introduced to enable OC

Prettymuch like intel Haswell. Where all mobile Chips were unlocked on a early microcode, and the rest of the haswell lineup could apply the max 1c turbo to all core.

Or on skylake, where BCLK OC was patched later with a mc update.

AMD has a tight release schedule right now and probably had more important bugs to address.
 

Gordan

Member
Nov 18, 2019
39
8
8
I might be able to mod your newer bios so you can use the higher memory OC options with the newer version, if there's any need for that.
I'm not sure there is much point. Even going to 3200 from 2933 has impacted the TDP enough to reduce the boost slightly under full load. And there is only about an extra 10% I could get out of the timings, so 3600 would just turn into a massive time sink for questionable benefit.

Ooooh, so that's what happened to Valhalla: it became Epyc!

@topic: Imagine EVGA (or some other madman brand) coming out with a SR-4 dual SP3r2 board with insane VRMs, holy shit. 128 cores at 4 GHz plus will make minced meat of every workload!
You really need to stop fanboying EVGA. Having had two SR-2 motherboards, I can very confidently say they were the most garbage grade board I have ever had the misfortune of owning. There were so many firmware bugs and hardware design faults it wasn't even funny. It was unfit for purpose. Off the top of my head:

1) CPUs were rated for 192GB of RAM each. The board could only reliably handle 48GB. 96GB resulted in POST only succeeding about 25% of the time. Obviously a firmware initialisation timeout bug since once it managed to POST, it would work fine with 96GB.

2) A SATA-3 controller with two 6-Gbit ports connected via a single 5-Gbit PCIe lane.

3) Advertised as VT-d/VT-x capable, but features NF200 PCIe bridges on _every_ PCIe slot, which bypasses the IOMMU for DMA, which in turn completely breaks PCI pass through by not remapping the VM memory and trampling the PCI apertures. It took me weeks to get to the bottom of that bug and write a Xen HVM loader patch to work around it.

4) It's key feature of OC-ability never really worked properly due to vastly sub-standard clock generators used. The clock stability deteriorated to unusable levels at barely 10% above defaults and no mitigation worked to get it stable again.

The only reason I bought the 2nd was because I was initially convinced I had a duff board, but the 2nd one exhibited all of the exact same issues.

I eventually gave up, sold the SR-2s to fanboys in eBay for twice what I paid for them new, and for a fraction of the amount bought a Supermicro X8DTH-6i, and lived happily for the next 6 years - until a few weeks ago when I replaced it with EPYCD8/7402P.
 
Last edited:
  • Like
Reactions: Patriot

hoping

New Member
Dec 5, 2019
15
1
3
Why shouldn't they?
There is actually only one pro for overclocking: performance. It’s why gamers and PC enthusiasts overclock a CPU and a GPU (graphics processor unit). When you send more voltage to either the CPU or GPU, the graphics increase, response times within applications are reduced, and benchmarks can identify peak performance for complex software.

The biggest issue with overclocking is the reduction in a component’s lifespan. You can overclock a CPU, GPU, motherboard or RAM, but sending increased volts gradually damages these components. Damage is caused by heat generated from increased power. Additional heat doesn’t usually ruin a circuit immediately, so the damage is seen gradually over time.
 

Gordan

Member
Nov 18, 2019
39
8
8
The biggest issue with overclocking is the reduction in a component’s lifespan. You can overclock a CPU, GPU, motherboard or RAM, but sending increased volts gradually damages these components. Damage is caused by heat generated from increased power. Additional heat doesn’t usually ruin a circuit immediately, so the damage is seen gradually over time.
Having spent the last 25 years working with computers and electronics all day every day, I can honestly say that I have _never_ seen a single instance of ASIC failure due to OC-ing. Once it is dialled in and stable, an OC-ed system will go on for decades, until something happens that would kill it regardless of OC (dust clogging resulting in massive overheating, PSU failure, or something along those lines).

If you are worried about component aging, ASICs these days are pre-aged at the factory to make sure that what you get can consistently achieve what it can on day 1 for long enough that it will outlive it's usefulness.