MZ73-LM2 Rev 3, Dual Epyc 9684X, Dual 5090 RTX, relatively small footprint

[Venturi]

Walmart, 97 to 99%just dry it afterwards,
Wipe wet and dry with a microfiber, don’t get the microfiber snagging on on the central circuitry


on board revisions: the MZ73-LMO accepts turins up the 450w variety, rev 2 and 3

I have two mz73-in a box rev 2 and 3 that were tested with turins, then I decided to stay with my 9684X so installed a MZ73-LM0, and returned the Turins

complicated, but was getting better performance from the 9684x (that deserves its own thread) so stuck with them


Through several revisions of boards and CPUs I have not encountered the issues you are having except for

1 on MZ73-LM0 the early bios would not allow 9684x to boot until the updated bios was installed by BMC, and to do that I had to start with a single lessor processor, update, then install both 9684x



on MZ73- LMO first board had bent pins and did a return and got a new one

a second MZ73 LM0 was rev 2, and then a rev 3 later , both worked great

on

mZ73- LM2, boards looked perfect but, if I seated the CPUs screwed down all the way there were instabilities, solved by backing of the screws about 1 turn - found tha out by AMD support on issues they encountered on differences in heatsink manufacturers, screws and springs

changed heatsink and the issue went away and screwed all the way down no issues, but backed of a turn a day later from all the voices in my head

I have also seen the opposite, if you don’t screw down enough to apply the correct pressure.

So it can be a bumpy road till you figure out what’s doing it,

 

[terryww]

For 9xx5 Epycs, the prescribed torque for screws is 1.5Nm.

 

[DanRR]

For 9xx5 Epycs, the prescribed torque for screws is 1.5Nm.

Using torque as measure of optimal pressure on processor is utterly dumb and retarded. Pity all use this, including AMD itself. It has nothing with the actual pressure specifically when with time rust, heat and dust corrodes the surfaces. Necessary to develop new methods to reliably define the pressure.

My screws were so rusted after two years of use so that i used silicon lubricant to make them just to turn

 

[RolloZ170]

Using torque as measure of optimal pressure on processor is utterly dumb and retarded. Pity all use this, including AMD itself. It has nothing with the actual pressure specifically when with time rust, heat and dust corrodes the surfaces. Necessary to develop new methods to reliably define the pressure.

My screws were so rusted after two years of use so that i used silicon lubricant to make them just to turn

1.5Nm is more than required, but less than making a damage to the threads.
the pressure is defined by the socket pins. once the processor pcb gently touches the socket frame plastik, more torque will give nothing,
just bent socket retention mechanisms. required is only balanced turn scrosswise, not one first to the end, then next...wrong !

 

[Venturi]

Hence,
Seat the screws all the way down, only till they stop, go no further, and then back them off a turn

The pressure from the heatsink springs on the screws, pushing the heatsink down, more when they are compressed, is the variable

INMHO

 

[RolloZ170]

in red marked the plastik frame where the CPU lays down right before 1.5Nm. more pressure after that is not effective.

 

[DanRR]

Walmart, 97 to 99%just dry it afterwards,
Wipe wet and dry with a microfiber, don’t get the microfiber snagging on on the central circuitry


on board revisions: the MZ73-LMO accepts turins up the 450w variety, rev 2 and 3

I decided not to do that after reading that it is not safe for plastics. No one knows what components AMD plastics have.

Please confirm that you have MZ73-LMO Rev2 and you tried Turins on it. I spoke to Gigabyte and they told me in plain Englsih what specifically is turned off in it to not fully accept them. Just feeling the 9755 power is a bit larger for it even if I upgrade all heatsinks increasing them several times and put also blowers

Can you give a link for your DUI case? I have similar one but your look larger what I need after adding monstrous power supply

 

[DanRR]

in red marked the plastic frame where the CPU lays down right before 1.5Nm. more pressure after that is not effective.
View attachment 46022

I think that just using the same amount of screw rotations one after another in cross order just by one rotation at a time (I used from 6 to 9 and all worked. I also have torque screw driver and initially also used it to find two years later that I can not do a single rotation with it so much screws got dirty with the dust) would be the best way to have constant areal pressure on chip.

This comes from the physics. The SP5 socket has rectangular shape , and same shape has the EPYC both Zen5 and Zen5c internally. Hence the pressure of two central screws has to be larger than the four peripheral ones. And this is exactly how the mount was designed:you initially tight two central screws until the mount lowers and other 4 start to touch their threads on the motherboard. At that moment the springs on these two central screws already have pressure while others have it zero. After that you start tightening all of them one by one.

The torque 1.5Nm or whatever has zero relationship to the correct homogeneous pressure distribution and probably is even dangerous. Because even in ideal conditions at the end all screws will have the same torque readings (same local forces on chip) but that is wrong to get homogeneous pressure distribution over entire chip area

 

[DanRR]

which file again ?
(btw: asrock rack has only the BIOS bin in the package, nothing else,
at supermicro you can't flash BIOS without working BMC....)

Hopefully I did not lost it with all my rebuilds. Will post when find. That was the major run file in subdirectory which starts the whole process

 

[Venturi]

I decided not to do that after reading that it is not safe for plastics. No one knows what components AMD plastics have.

Please confirm that you have MZ73-LMO Rev2 and you tried Turins on it. I spoke to Gigabyte and they told me in plain Englsih what specifically is turned off in it to not fully accept them. Just feeling the 9755 power is a bit larger for it even if I upgrade all heatsinks increasing them several times and put also blowers

Can you give a link for your DUI case? I have similar one but your look larger what I need after adding monstrous power supply

Plastics?

you mean wiping the contacts clean on the cpu?
As in the substrate?

I’m not suggesting immersion, just wiping them off


most Tim Suppliers will also include alcohol swabs

 

[DanRR]

Plastics?

you mean wiping the contacts clean on the cpu?
As in the substrate?

I’m not suggesting immersion, just wiping them off


most Tim Suppliers will also include alcohol swabs

Substrate is of course silicon but the material surrounding contacts made of plastic like in motherboards. This could cause discoloration etc and the chance to return processors still exists. Cotton swabs I agree are definitely bad idea, I saw under microscope a lot of fibers but the fun is that my second processor finally started to work after using swabs and then synthetic cloth

I could not find in USA anyone who sells anything close to the pure medical or technical alcohol because it is forbidden to sell to the public in high concentrations . Who is Tim Suppliers ?

 

[twin_savage]

I could not find in USA anyone who sells anything close to the pure medical or technical alcohol because it is forbidden to sell to the public in high concentrations .

If you live in a free state without CARB you can buy denatured alcohol at many home improvement or more industrial stores. If you live in California you can buy what they call "culinary solvent" which is pure ethanol, however I've never actually seen it in a store so you'd probably have to buy it online.

 

[RolloZ170]

Substrate is of course silicon but

???
For decades, the so-called organic substrate, which consists of a woven structure through which signal and power lines are routed, has been used. The organic substrate replaced the inorganic ceramic package 30 years ago.

 

[Venturi]

Substrate is of course silicon but the material surrounding contacts made of plastic like in motherboards. This could cause discoloration etc and the chance to return processors still exists. Cotton swabs I agree are definitely bad idea, I saw under microscope a lot of fibers but the fun is that my second processor finally started to work after using swabs and then synthetic cloth

I could not find in USA anyone who sells anything close to the pure medical or technical alcohol because it is forbidden to sell to the public in high concentrations . Who is Tim Suppliers ?

TIM, thermal interface material, example, noctua includes three alcohol pads with every tube of TIM.


At this point I'm just suggesting clean the contacts with something that leaves no residue

It was not meant to be this complicated, I was just trying to help

 

[RolloZ170]

I saw under microscope a lot of fibers but the fun is that my second processor finally started to work after using swabs and then synthetic cloth

synthetic cloth: where the micro/nano parts are from.

 

[DanRR]

It was not meant to be this complicated, I was just trying to help

Yea, but here is California. California prohibits the sale of 190-proof alco in retail stores. And many many many many many many many many other things. The trick to order something online also rarely gets through. But I will try. Thanks twin_savage for the hint.

 

[DanRR]

???
For decades, the so-called organic substrate, which consists of a woven structure through which signal and power lines are routed, has been used. The organic substrate replaced the inorganic ceramic package 30 years ago.

Was silicon, then organic, soon again silicon in the form of glass, silicon oxide. But ok, the correction is taken. The point was not to damage in any way the contacts area which is some kind of organic material

 

[DanRR]

synthetic cloth: where the micro/nano parts are from.

Look yourself, the amount of dirt was shocking. Thermopaste, dust, everything. And the fun is it is invisible. I found it just by chance. On first glance all looks almost perfect like this, just use some blower and you're done :


But after many unsuccessful attempts to boot, pulling processor and blowing again, at one specific angle in one specific environment see what was revealed:


When cleaning with the swabs I removed tough sticky thermopaste and dirt but added more in the form of cotton fibers. Then synthetic cloth (for cleaning glasses) removed the fibers and processor finally booted

 

[DanRR]

Venturi,
Can you stress test your Genoas with some app which runs all its cores and measure temperature of processors? The temperature data with the graph showing the time history is called Psensor where sensor name with highest temperature is Tctl, but could be other apps, depending on availability for your OS.

 

[Venturi]

Venturi,
Can you stress test your Genoas with some app which runs all its cores and measure temperature of processors? The temperature data with the graph showing the time history is called Psensor where sensor name with highest temperature is Tctl, but could be other apps, depending on availability for your OS.

never really been a fan off burners and artificial stress tests, for what purpose would this be?