PC Processors

[reeflex]

Hi there,

Are there any adjustments needed to run 1TB of RDIMMS with AMD Ryzen Threadripper PRO 3995WX? Any adjustment to VSOC values, etc.?

Motherboard: Supermicro M12SWA-TF

Chassis: 747BTS-R2K20BP-OTO-11

RAM: Micron MTA144ASQ16G72PSZ-3S2E3

Front fans: 3x FAN-0138L4 (7500 RPM); 1x FAN-0114L4 (5000 RPM)

Rear fans: 2x FAN-0082L4 Rev.B; Ordered 1x MCP-320-00046-0N-KIT so there will be 3 rear exhaust fans

GPU: Zotec RTX 3070

I just installed 8x 128gb ram and my system restarts after a couple minutes of TestMem PCBDestoyer config

Not sure what else might be happening, but VRMABCD and VRMEFGH are overheating:

Thanks for any help

[misterj]

Thanks, reeflex. Go back to your original memory and see how your system runs Cinebench R24 Single and Multicore. Please tell SM that the 3995WX is NOT unlocked for overclocking. I will check with AMD and make sure RM should install and run on your 3995WX. John.

[reeflex]

Hey John,

I'm back on the original memory and ran cinebench R24:

where would be a good place to mention to SM this piece: 'tell SM that the 3995WX is NOT unlocked for overclocking'? This is to do with getting RM installed onto the system?

Thanks

[misterj]

reeflex, yes.

I would like to see the RM screenshot from Cinebench, and if not your temperature screenshot. John.

[reeflex]

Ok let me check. 

Here are the temps when running cinebench:

[misterj]

Thanks, looks good reeflex, John.

[reeflex]

Hey John, I'm back 
I ended up buying the QVL ram MEM-DR412MH-ER32-MB 128GB DDR4 3200 RDIMM RAM Upgrade for Supermicro Servers | eBay

But, the CPU-Z info is really weird with these, not sure what to make of this. 2 of them are undefined manufacturers, should I get these replaced?:

Anywho, the VRMEFGH is still getting up to 95C during the stress test, but now the system isn't crashing at least. So there was some other compatibility issue with the last memory.

I stopped the test at 2 hours since I'm concerned 95C for a long period of time will degrade the system?

This is with the the heatsinks and a 40mm fan blowing onto the VRMEFGH.

Maybe I need to try a 60mm fan or a different brand that is stronger? Or pull the heatsinks off and replace the tape with thermal adhesive?

I'd like to rig something like this in there to get a bigger fan blowing onto them, but doesn't seem compatible.

https://www.silverstonetek.com/en/product/info/fans/FDP01/

They also have VRM ducts, but not sure this fits:

https://www.ebay.com/itm/303995970762

Anyways, this is my latest update haha. Please let me know if you have any thoughts. I'll see what else I can do to cool down these VRMs, unless 95C for days or weeks on end isn't an issue.

Thanks

[misterj]

reeflex, it is a common saying "if it works don't fix it." It looks like to me that the eBay seller slipped in a non Samsung stick. If it is stable, leave it be. How long do you have for warranty or refund? Sure wish you would look at Amazon or other sellers that at least I have more confidence. I suspect that VRM module can operate at higher temperatures than processors maybe even as high as 120C. You need to ask your system vendor the maximum for all the temperatures their applications  shows. How much memory are you running now? Please post a screenshot of RM running Cinebench R24 multicore. If you cannot get RM running (what did vendor say) post image of the utility you are using. John.

[reeflex]

Thanks John.

Looks like I have 14 days to exchange or return. I'm tempted to contact them to exchange the two 'Undefined' sticks I have. I'm paranoid now that the system would run better with 2 more Samsung sticks. 

Yea it's tough, the compatible nemix memory is 50% more on amazon: https://www.amazon.com/Supermicro-Compatible-MEM-DR412MH-ER32-NEMIX-RAM/dp/B09G8CCH66

I'm running 8x 128gb memory again. 

On another note, SuperMicro support initially said this is VRMEFGH:

Then this:

Now finally this after 'confirming with [their] engineer team'. 

 

So I can start with getting a heatsink and pointing the fan in the right location for VRMEFGH.

They just say the system prevents installation of RM. I'll try hwinfo and running cinebench when I can.

[elstaci]

Went to Super Micro Support for your Motherboard QVL List and it just shows one RAM Part Number using 128GB RAM set: https://www.supermicro.com/en/support/resources/memory?sz=128&mspd=3.20000005&mtyp=139&id=632be86ff1...

The rest of the other type of RAM the most it showed was one Part Number for a total of 32GB. Extremely limited QVL List by Super Micro. Showing only one RAM Part Number per type of RAM the Motherboard supports.

Just to see if for some reason your 1 TB RAM set isn't fully compatible try installing first 1 RAM stick and see if the issue occurs again. If it works fine than install a second RAM Sticks and test again and so forth.

Some Motherboards may not be fully compatible with certain size RAM sticks and the amount of DIMM Slots being populated. Generally you would first install 1 RAM Stick to see if the problem occurs if it is a incompatibility issue then add more RAM sticks until the problem starts again.

From this Processor Website CPU MONKEY the Threadripper Pro 3995WX does support 2 TB of RAM Memory: https://www.cpu-monkey.com/en/cpu-amd_ryzen_threadripper_pro_3995wx

I imagine you contacted Super Micro Support to make sure your RAM Set was compatible correct?

If you haven't contacted Super Micro Support I would ask them if your RAM, by your PC behavior, is due to not being fully compatible with all Dimm Slots being populated with 128GB Ram Sticks.

Downloaded your Motherboard's Manual and it doesn't mention much about RAM:

Use the above DIMM Slot chart to install less than 8 RAM Sticks on your Motherboard.

Does your Motherboard CPU LIST show the ThreadRipper Pro 3995WX as being compatible?

Also try to upgrade your Motherboard BIOS to the latest version which can help with incompatible hardware to become compatible.

NOTE: You didn't mention which OS you are using but here is Super Micro OS Compatibility Chart for your Motherboard: https://www.supermicro.com/Aplus/support/resources/OS/OS_Comp_WRX80.cfm

[elstaci]

NOTE: Check in your Motherboard BIOS to make sure it is showing the full 1 TB of RAM that you have installed.  That is one way to see if the RAM set is compatible in all 8 DIMM Slots.

EDIT: Reread your OP again. One 4 Dimm slots (ABCD) are much cooler than the second 4 Dimm slots (EFGH) on your motherboard.

Since your PC is booting up that at least means your RAM is compatible to be able to Boot up your PC.

Try using only the first 4 Dimm Slots (ABCD) and see if the problem occurs if it doesn't and it does with 8 Dimm Slots populated (ABCDEFGH) and the PC shuts down due to overheating, I would Contact Super Micro Support to see if your Motherboard needs to be checked or it is a compatibility issue using 8 Dimm slots with your specific RAM Part number.

Your PSU 3.3/5.0/12VDC outputs are excellent so it doesn't look like a power issue due to the PSU.

[reeflex]

Hey elstaci,

Thanks for the research. This RAM isn’t on the QVL unfortunately, but it does seem to work and register all 1TB:

I can check with SuperMicro support tomorrow to see if they have any insights. I’m not optimistic here since they will probably say it isn’t on the QVL and leave it at that. I wouldn’t blame them.

I did buy the system off newegg so the CPU and Mobo should be compatible. Everything worked when using 8x32gb RAM. It is this system: https://www.newegg.com/supermicro-superchassis-747bts-r2k20bp-oto-11-tower-rack-mountable/p/N82E1685... 

Sorry, I’m using Windows 11 Pro: 

I can try updating the BIOS. I’m concerned about this warning here on this page:

https://www.supermicro.com/en/support/resources/downloadcenter/firmware/MBD-M12SWA-TF/BIOS 

Should I still try updating the BIOS?

I did try 4 sticks in D, C, G, H according to the motherboard manual and it failed the memtest still. Should I try the 4 in ABCD?

Please let me know if I missed anything or any other info I can provide to help. 

Thank you.

[misterj]

reeflex, I'm still doing research but have some questions. Do you really have two 2200 Watt power supplies? Looking at the MB manual, there are no big power FETs or inductors around the processor. How is the core voltage and memory voltages created? Pictures of the memory sticks do not have any heat sinks. This surprises me. Do they really not have heat sinks? Does not look like the CPU cooler can do the job but I need to see the RM screenshots. If you contact the system builder, please also ask them if they block RM from installing and what you can not do that would void the warranty, e.g., can you remove their software? John.

[reeflex]

Thanks for the research John.

Yea it does have two of these 2200W power supplies:

https://store.supermicro.com/us_en/2200w-1u-pws-2k20a-1r.html

Not sure how to find this answer, what do I need to do to determine this?

---

How is the core voltage and memory voltages created?

---

The memory doesn't have heatsinks. Should I get some?

I haven't reached out to SM support yet but I'll fire off an email to them. Need to see what they say about properly running 8x 128gb memory on this and what I can do to install RM on it. I'll post back what I hear from them.