Xeon Platinum P-8124 & ASROCK EPC621D8A

Aug 28, 2019
99
43
18
Also a tip for anyone, if you're using 64GB RDimms like I am, try to get ones running at 2666 for these chips. 2933 or higher seems to have issues when running and will spit out ecc errors even if the ram is 100% working. I stuffed it with 8x samsung 64GB 2933 dimms and it kept reporting all dimms as bad.
 
  • Like
Reactions: iamnypz
Aug 28, 2019
99
43
18
Final tip, if your system crashes often for no reason follow these steps.

Enter Bios
Go to CPU Configuration
Set the package c state support from auto to either C0/1 or C2
Save and restart.


The auto configuration or any other option will cause your system to crash on startup or during operation. C0/1 or C2 allow the cpu to run normally.
 
  • Like
Reactions: iamnypz

RolloZ170

Well-Known Member
Apr 24, 2016
1,933
501
113
55
I did also find of the 7x p-8124 chips I have with me atm that 2x of them instantly hit 107c under 100% load so looks like i'll be delidding some chips. Any recommendations or tips for the lga 3647 platform is appreciated.
img_20220830_162356cc86c3q.jpgP-8124(some cores 107C) delidded with WS 3647(der8auer)
just clean it: i have removed the black high-temp.silicone only on the IHS and used MX-5
 
Last edited:
Aug 28, 2019
99
43
18
So I got it delidded with a razor, it was bad thermal paste, sanded the top of the chip so it's smoother, applied liquid metal and plopped it in, max of 77c full load. Also I did make a work around for my proxmox issue. I just installed proxmox on debian, then added 2x sata ssds, installed proxmox hard onto the drives in raidz1 and then put them in the system and it worked. I just ended up cloning the vm across more drives for simplicity.
 
Aug 28, 2019
99
43
18
sanded ? one thing what never be done thought. foreign atoms could migrate into the silicon and destroy the chip over time.
Sorry, meant sanding the IHS. I did actually ask a friend at Intel who manages one of the FABs if he can "sand" the die itself which can be done. They do that for testing so I might be able to get a bit of extra headway next week when I drop the chip off.
 
  • Like
Reactions: RolloZ170

jasonsansone

New Member
Sep 15, 2020
29
10
3
With very rare and limited exception, Intel CPU’s topped out at 165W TDP prior to SKX. The jump up to 240W is significant. I suspect the long term failures of thermal paste and some cores in P-8124 are related to engineering larger power draw CPU. Also possible cooling for the hotter chip at Amazon or wherever was inadequate. I haven’t had these problems with the P-8136 at 165W.
 

RolloZ170

Well-Known Member
Apr 24, 2016
1,933
501
113
55
The jump up to 240W is significant.
only custom OEMs
I suspect the long term failures of thermal paste and some cores in P-8124 are related to engineering larger power draw CPU.
goes in hand with the high core frequency, the frq/draw courve in not linear.
Also possible cooling for the hotter chip at Amazon or wherever was inadequate.
time is money, they don't care.don't wait for failure, just replace after X month and go...
I haven’t had these problems with the P-8136 at 165W.
lower frequencies.
have a Microsoft(Azure) Olympus dual 3647 motherboard, it has two BIOSs - one includes B1 microcode.