Users of GigaByte MZ73-LM2 & MZ33-AR1 - comprehensive list of all issues and solutions

[fatherboard]

I'm re-doing my home work again for the third time about GigaByte MZ73-LM2, after my earlier verdict "stay away from it".
I saw some posts that gave some solutions, but before buying a MZ73-LM2, and in absence of similar Motherboards that would fit in an E-ATX case, I have to re-assess.

Together we could avoid falling again and again in the same issues. E.g. there are tons of "waiting for chipset to initialize" posts that keep re-appearing over and over again.

The idea is to list any and all issues and their respective solutions, with the reference. I will update/correct this very first post with a list from your comments, so there is no need to keep scrolling and collecting them.

-------------------------------------

"waiting for chipset to initialize"
→ BIOS update to latest R19_F40
→ Two CPUs, not just one
→ Better fill the 24 RAM slots

Ref: post


---------------------------------------
MZ33-AR1 design flaw
" The distance between the end of pcie slot and memory slot/cpu socket is too tiny. I installed 12 memory and the everything I used on PCIE slots need to be low profile type to prevent the memory getting into the way. Honestly I think this mobo has some design issues."
→ I have to use PCIE raiser to install the GPU otherwise it won’t even fit unless I remove most of the memory. REF

----------------------------------------

@Squuiid Have you used a dual 9965 192 cores on the MZ73-LM2?
Has anyone else used a dual 9965 with MZ73-LM2?
@terryww You had some PSU issues, was it the cause of the problem? and what was the problem?

 

[terryww]

Thanks for a new thread. I started a thread about LM2 and Epyc 9755 ES. The current status is as follows:
- @RolloZ170 confirmed they could boot with 2 CPUs only with the latest BIOS and that for 2 socket motherboards, it's required to have 2 CPUs to boot
- the seller claims they can boot with DDR5-6400. I tested with 1 CPU with DDR5-6400 and DDR5-5600, didn't manage to get past "waiting for chipset to initialize".
- for 2 CPU setup I tested with a LC Power 1650W (single rail 12V@130A) and Cannon ATX3.1 2500W (single rail 12V@200A). Both shut down after a second (and it looks like the motherboard's over current protection LED starts to blink). Last thing to test is the hypothesis of high transient currents which require a proper server PSU (to not trigger the PSU's or motherboard's over current protection circuits).
- using latest BMC and BIOS versions, the BMC doesn't update the BIOS version after flash. One member suggested the version number gets updated on system POST (which the system didn't manage to do since stuck at that "waiting for chipset to init" screen)
- any changes to BIOS are not saved (and I checked the jumper on the motherboard is not set to BIOS recovery)

All of the 24 RAM slots need to be populated to boot? Can somebody confirm that DDR5-6400 or DDR5-5600 modules work?

 

[fatherboard]

- the seller claims they can boot with DDR5-6400. I tested with 1 CPU with DDR5-6400 and DDR5-5600, didn't manage to get past "waiting for chipset to initialize".

All of the 24 RAM slots need to be populated to boot? Can somebody confirm that DDR5-6400 or DDR5-5600 modules work?

Good Lord, there is a lot more going on than I thought. Let's address one by one, elimination, assuming the earlier problem is addressed.
So first let's assume a Dual CPU board MZ73-LM2 has 2 CPUs.
Does RAM DDR5-6400 and DDR5-5600 not work for you because of 1 CPU? or did DDR5-6400 and DDR5-5600 also not work with 2 CPUs?

In a few weeks I will test the DDR5-6400 or DDR5-5600 with 2x 9965, that will be one hell of an expensive test, as no other E-ATX/SSI-EEB board takes 2x 9965.

- for 2 CPU setup I tested with a LC Power 1650W (single rail 12V@130A) and Cannon ATX3.1 2500W (single rail 12V@200A). Both shut down after a second (and it looks like the motherboard's over current protection LED starts to blink). Last thing to test is the hypothesis of high transient currents which require a proper server PSU (to not trigger the PSU's or motherboard's over current protection circuits).

A motherboard designed for E-ATX doesn't take ATX PSUs and requires Server PSUs? I thought ATX PSUs had protection:

Let's put it like this: tell me what PSUs you want me to test with, and I will do it if I can find it. Initially the plan is 2x 2500 Single rail in a few weeks (high wattage because of 2x 9965 and blackwell 6000 GPUs could join the party).

- using latest BMC and BIOS versions, the BMC doesn't update the BIOS version after flash. One member suggested the version number gets updated on system POST (which the system didn't manage to do since stuck at that "waiting for chipset to init" screen)
- any changes to BIOS are not saved (and I checked the jumper on the motherboard is not set to BIOS recovery)

is this BIOS thing solved? were you able to successfully update to R19_F40 and get past the issue?

Are you using MZ73-LM2 Rev 3?

By the way is your MZ73-LM2 now working?

 

[terryww]

or did DDR5-6400 and DDR5-5600 also not work with 2 CPUs?

Didn't manage to power up 2 CPUs, as said, tried 2 PSUs already, last idea is to use a server PSU (the high transient load triggering over current protection circuits problem should be better handled in server PSUs according to ChatGPT)

Let's put it like this: tell me what PSUs you want me to test with

Since yours are also 500W CPUs, let me know about ANY PSU you manage to power up the system. As said, I tried Cannon 2500W, 208A@12V single rail, spins up the fans for 2s, then shuts down. So my next step is trying out a server PSU.

and blackwell 6000 GPUs could join the party

I wonder how you plan to use multiple GPUs when the PCIe slots are so tight together.

is this BIOS thing solved? were you able to successfully update to R19_F40 and get past the issue?

No, that's the current status.

Are you using MZ73-LM2 Rev 3?

I think it is, but I'll have to check (will be back in the lab next week)

By the way is your MZ73-LM2 now working?

It is working in so far as I see the "initialization" screen when using 1 CPU. Haven't managed to power up with 2 CPUs.

If your use case is GPUs, I've been looking at TURIN2D24G-2L+/500W, but with ES CPUs somebody mentioned it would be a project on another level, also one apparently needs some adapters for the power cables.

 

[fairydreaming]

Regarding the power supply I found that Puget Systems sells configurable dual Epyc workstations based on MZ73-LM2 motherboard (they call this workstation E210-XL). They use Super Flower Leadex Titanium 1600W PSU, so I don't think a server PSU is absolutely necessary. In their workstation configuration form it's possible to select 500W Epyc 9755 and Epyc 9965, so I suppose they verified the correct operation of such configuration.

 

[terryww]

workstation E210-XL

Oh, that's interesting info, thanks. So it works with an ATX PSU that offers 133A@12V. Hm... ok, then not sure if it's a bit special because it handles transient currents better and it's over current protection circuit isn't triggered OR there's sth wrong with my motherboard.
Would be interesting to know what other ATX PSUs people use.

 

[kromberg]

I am thinking about using this motherboard with a set of 9B14/9654 CPUs. Could of questions I have:

1) Does the board support the CPUs? From my initial finds I think it should.
2) What is a recommended PSU for it? First time I have seen a MB needing additional 6-pin 12v connectors. Are those just 6-pin PCIe power connectors?

 

[mrpka]

R19_F40 here with 2x 9755 ES and 24x 6400 RAM. System works fine with Asrock PG1600 PSU and RTX 6000 Blackwell. Using 2x Silverstone 360 AIO coolers, just like the Puget configuration. It was important to tighten both the socket screws and cooler plate to specification. I did it too loose the first time and the system was not booting.

 

[Venturi]

So, I've had success on these boards, but went through a significant learning curve.
Do a lot of ML / NLP / LLM / and other projects mostly related to imaging and also to isotopes.

Mt most recent home PC is based on MZ73-LM2 Rev 3.0, Along the way I've learned all sorts of stuff to optimize the bios, work arounds, and how to overcome some issues.

Glad to discuss and help you with some of the challenges, bios, settings, and configuration

https://forums.servethehome.com/index.php?threads/mz73-lm2-rev-3-dual-epyc-9684x-dual-5090-rtx-relatively-small-footprint.52052/

 

[fatherboard]

Along the way I've learned all sorts of stuff to optimize the bios, work arounds, and how to overcome some issues.

Glad to discuss and help you with some of the challenges, bios, settings, and configuration

I've see your post. Great that it works. Your post is mainly about bringing things together in a nice way. Whereas the challenge is to make this motherboard work in the first place.
It would be great if you could write the motherboard problems you've encountered and how you solved them (motherboard, bios, settings, configuration...)

 

[Venturi]

I included the assessment on the bios and settings that solved many of my issues, please check the parent thread

I included bios screen pics as well as the reasoning on some of the bios settings choices

J

https://forums.servethehome.com/index.php?threads/mz73-lm2-rev-3-dual-epyc-9684x-dual-5090-rtx-relatively-small-footprint.52052/

 

[Venturi]

In first startup, on the MZ73-LM2, on the bios other than R04-F32, I had to only populate Socket 0 and only 2 sticks of ram. Then after getting past the 4 cyclesof chipset initializing message (about 11 minutes, and a successful bios screen post, I would install the second cpu and all the ram.

after figuring out that bios R04-F32 let the sockets work fully populated, on the second board, same model, I just used the bmc bios flash, force reboot, and all posted well on the first try.

mrpka mentioned, be sure to seat the coolers all the way down

I had also a a performance hit on the MZ73-LM0 Rev 3 when using any bios after R04-F32, reverting to the R04-F32 restored performance and stability

Same with the MZ73-LM2 Rev 3, there was a performance and stability hit after R12-F37, reverting to the R12-F37 restored performance and stability

 

[terryww]

The latest BIOS update for LM2 is R20_F41. Do you also see a performance hit with it? What kind of perf hit do you see - 10% or more? Can you share the workload?

 

[terryww]

R19_F40 here with 2x 9755 ES and 24x 6400 RAM. System works fine with Asrock PG1600 PSU and RTX 6000 Blackwell. Using 2x Silverstone 360 AIO coolers, just like the Puget configuration. It was important to tighten both the socket screws and cooler plate to specification. I did it too loose the first time and the system was not booting.

Just to clarify, by not booting you mean the fans not even spinning up or spinning up for a sec and shutting down? Also, have you perhaps tried with any other PSU?

 

[Venturi]

The latest BIOS update for LM2 is R20_F41. Do you also see a performance hit with it? What kind of perf hit do you see - 10% or more? Can you share the workload?

Yes, I'm sure a lot of it is from the security mitigations, BUT, shamefully so, security should NEVER compromise performance, ~unless Intel/AMD are writing checks back to the purchasers for performance deprecation

I have all the bioses for these boards and the R04-F32 is fastest, the latest bios is the slowest ~in my ecosystem. milage may vary for others.

I notice the drop in read/write speeds as well as mGPU performance, when all things remain equal other than bios versions

Reverting to R04-F32 restores performance.

However, I also recognize that I don't have the latest 9005 CPUs, so newer bioses may be necessary for those CPUs. I'm happy with my 9684X's for now.

 

[fatherboard]

@mrpka on a side note, I see you're using the AIOs, how are they performing with your 500W CPUs (9755 ES)? I'm asking because I contacted Silverstone and they said it could possibly handle 500W CPUs if I don't do anything crazy with it. I would truly appreciate if you could let me know if the AIOs have kept your CPUs cool under heavy load.