VRM modify ICC_MAX to run high TDC OEM cpu
- Thread starter RolloZ170
- Start date
Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.
Hello!
I used a non-standard method, fully adapted the latest BIOS version for the OEM CPU and some things else.
For the experiment I have Supermicro X11SPi-TF, Xeon Gold 6268CL, Xeon Platinum 8272CL, Xeon Platinum 8124M (240W TDP, work with VRM mod firmware + BIOS), Xeon Gold 6150., I did not use the outdated BIOS, at the moment when I started working I had BIOS 4.3, TDC 228A. The TDC 228A to 255A unlocking was done with the help of additional .efi with added BIOS menu item and the parameters by default TDC 255A. The Resizable BAR functionality was also added, if did not pay attention to the answers of Supermicro's "technical specialists" - "here is not supported at all". Adding RAM timing settings added an additional 15% of the performance with minimal adjustment. I also created a TPM2.0 SPI type 10 pin, specifically for this board, I used SLB9672 FW:16.13, because Supermicro didn't this, or I found in sale outdated TPM World versions for a more 100€ price or 30€ for crooked Chinese fakes with Chinese firmware. In this way I got SHA-1, SHA-256, SHA-384, TXT Support. This BIOS has transformed a so bad, so slow routine PC into a beautiful workstation in a wonderful way!
I used a non-standard method, fully adapted the latest BIOS version for the OEM CPU and some things else.
For the experiment I have Supermicro X11SPi-TF, Xeon Gold 6268CL, Xeon Platinum 8272CL, Xeon Platinum 8124M (240W TDP, work with VRM mod firmware + BIOS), Xeon Gold 6150., I did not use the outdated BIOS, at the moment when I started working I had BIOS 4.3, TDC 228A. The TDC 228A to 255A unlocking was done with the help of additional .efi with added BIOS menu item and the parameters by default TDC 255A. The Resizable BAR functionality was also added, if did not pay attention to the answers of Supermicro's "technical specialists" - "here is not supported at all". Adding RAM timing settings added an additional 15% of the performance with minimal adjustment. I also created a TPM2.0 SPI type 10 pin, specifically for this board, I used SLB9672 FW:16.13, because Supermicro didn't this, or I found in sale outdated TPM World versions for a more 100€ price or 30€ for crooked Chinese fakes with Chinese firmware. In this way I got SHA-1, SHA-256, SHA-384, TXT Support. This BIOS has transformed a so bad, so slow routine PC into a beautiful workstation in a wonderful way!
Why are you asking and asserting so much, I gave screenshots, and besides, the guy has already tested his own X11SPi-TF for a video where he clearly makes it clear that modification is necessary.
Last edited:
This guy used what I gave him for his 6268CL and I did not modify his SPS and it requires TDC-255A, if the TDC-228A limit is left the 6268CL will not work. You can remove the TDC limit check, then the OEM CPU will work, but their performance will remain poor. I currently have several OEM CPUs coming that will also require TDC-255A.
I was recently scammed by eBay, the seller sent a damaged Platinum 8259CL (210W) instead of the one advertised as new, and the test failed. During the disassembly of these CPUs, the Chinese use screwdrivers to remove the heatsink and they damaged the CPU PCB, this CPU had to be returned.
But I can assure you that the ability to switch TDC in the BIOS is a very useful option, which allows you to use almost all LGA-3647 compatible CPU. Because disabling TDC checking does not solve the problem.
understand...
well i have seen weird things with 6268CL and co.
the newer the BIOS is, the less the performance was. maybe SM applied a lower max. current.
Last edited:
Maybe not a bios, just new microcode from Intel. I upgrade the microcode for p-8124 3stepping and can't see the trouble with performance between 2018 and 2023 version on my x11dpi-n. But my current on vrm(mp2955a) for PL1 250A(50a per phase), last bios 4.3. For less current indication with tdp limit, maybe need shunt mod)
on vrm max PL2(TDC)=PL1+20%
on cpu TDP=PL1 I'm not remember, maybe pl2=pl1+25% in stock bios. You can check current limit override setting , pl2 can be 0~25%
165+25%=205w and can be lower, RAPL setting i never touch - only troubles.
if cut off 4 cores from 6268cl it is work fine with 3.4ghz without c-states(c0\c1).If not cut off the cores with no limit c-states, situation will be better, but not for long workload.
8275CL so hot, but stable. 8259cl cold, but stable turbo boost. 6248R best choice.
If enable c-state c1e, limit for temperature will be higher by 4-6 degrees Celsius
Attachments
Last edited:
I don't think I ever saw this specifically mentioned, but will ES CPUs work with these mods?
I have a Dell T7920 tower, and I'm torn between grabbing a pair of 8260 ES, 8259L ES, or 8259CL (non ES) chips off eBay. There's also a 8260M QS chip, but they're nearly double the price (and only 165W).
Also do any of the "known good" chips support DDR4-2933 memory speeds? If they're all stuck at DDR4-2666, I guess packing the DIMM slots with Optane memory might not be a bad idea.
Thanks!
I have a Dell T7920 tower, and I'm torn between grabbing a pair of 8260 ES, 8259L ES, or 8259CL (non ES) chips off eBay. There's also a 8260M QS chip, but they're nearly double the price (and only 165W).
Also do any of the "known good" chips support DDR4-2933 memory speeds? If they're all stuck at DDR4-2666, I guess packing the DIMM slots with Optane memory might not be a bad idea.
Thanks!
I have a 7920T on BIOS 1.7.1 and I am planning to upgrade and update it. Would the 8259CL be the best chip (with this mod) or does the new-ish microcode make this chip no longer worth the effort (0x5003003)?
What other CL/ES/QS options does this mod open up that are compatible with Dell's latest bios? I'm using 2666 ram and plan to add optane.
What other CL/ES/QS options does this mod open up that are compatible with Dell's latest bios? I'm using 2666 ram and plan to add optane.
NEC R120H-1E (HPE DL160 G10*) modding Works. Thanks for everyone, special Thanks to RolloZ170 !
My work photos:
Please refer I2C pad. (Red wire = SCL, black wire = SDA)
* oem name revised
My work photos:
Please refer I2C pad. (Red wire = SCL, black wire = SDA)
* oem name revised
Last edited:
you can only mod the 2nd CPU VRM with a CPU populated there and in first socket, the controller are disabled without CPU populated.
Last edited:
Hello, @RolloZ170 i did vrm mod on Lenovo SR630, CiscoHX 220c M5 an Inspur SA5112 M5, if you can add in the list i i will send pictures and connectors details.