VRM modify ICC_MAX to run high TDC OEM cpu

[RolloZ170]

change ICC_MAX of VRM controller to match thermal design current(TDC) of the processor for LGA4677 platform
Could you write a program for 4677 platform to change TDC when you free?

we never stop learning, update to this:

on LGA4677 platform VR we have Iccmax and Iccmax HC.
used is the sum of the Iccmax plus HC value.
i lowered the Iccmax value only, understandable a 500A CPU not stopped working.
some can try to make a low TDC platform compatible with normal oder high TDC processors.
but you need vendor tool and software, at least for the MPS2973 all is available.
Gigabyte and Supermicro use MP2973 (MPS, preprogrammed 2971)
ASRrock Rack use RAA229126 (renesas)

 

[whitelilis]

no. check what VRM controller is used (near area around the power stages / chokes/coils(grey blocks) )

please provide a picture of VRM section. there are two versions, one with 5+1(ok) and one 4+1 (too weak) power stages.

I have the same mother board.
The follow is the overview. Which place to find the VRM? The board is so big, can you give me some scope ? I will provide more pictures

I620-g30服务器主板英特尔c621芯片组双lga 3647 Ddr4 60p24-us Rev V1.1 V1.2 - Buy Motherboard c621 Chipset lga 3647 Product on Alibaba.com

I620-g30服务器主板英特尔c621芯片组双lga 3647 Ddr4 60p24-us Rev V1.1 V1.2 - Buy Motherboard c621 Chipset lga 3647 Product on Alibaba.com

chinese.alibaba.com

 

[RolloZ170]

The follow is the overview. Which place to find the VRM? The board is so big, can you give me some scope ? I will provide more pictures

 

[gug296]

I'm working on the TDC unlock mod for an Inspur NF5280M5. While i see someone has PXE1610C controlleron their one , my board seems to have a different VRM chip.
Based on the markings (see attached photo), it appears to be a Texas Instruments chip with the code SN1701021. It looks like a custom OEM part number.
Has anyone encountered this specific TI controller before? I'm unsure which tool or register map applies here (e.g., is it compatible with TPS53679?).
Thanks in advance for your help!

 

[RolloZ170]

Has anyone encountered this specific TI controller before? I'm unsure which tool or register map applies here (e.g., is it compatible with TPS53679?).
Thanks in advance for your help!

SN1701021 = preprogrammed TPS53678 (my CLI app supports both 79 and 78)

 

[gug296]

Thanks for confirming it's a TPS53678! That explains a lot.
Since I already have direct access to the bus via Linux, is it possible to just use i2c-tools to unlock it manually?
I've located the VRM addresses at 0x62 and 0x64. Could you please confirm if the following raw sequence is correct for this specific chip?
Write 0x00 to 0x10 (Unlock protection)
Write new OCP limit to 0x46
Write new TDC limit to 0x4A
Send command to 0x15 (Store to NVM)
Also, a quick question about your tool: If I do use an MCP2221A, would I need to modify any config files? I'm asking because my chip seems to report a custom Model ID (or sometimes just empty 00s due to OEM settings), and I'm not sure if the tool will auto-detect it.
Thanks again!

 

[gug296]

i modified like this using i2ctool:
user@livedvd:~/Desktop$ sudo i2cget -y 0 0x58 0xad w
0x7801
user@livedvd:~/Desktop$ sudo i2cget -y 0 0x60 0xad w
0x7801
user@livedvd:~/Desktop$ sudo i2cset -y 0 0x58 0x00 0x00
user@livedvd:~/Desktop$ sudo i2cset -y 0 0x60 0x00 0x00
user@livedvd:~/Desktop$ sudo i2cget -y 0 0x58 0xda w
0xbfe8
user@livedvd:~/Desktop$ sudo i2cget -y 0 0x60 0xda w
0xbfe8
user@livedvd:~/Desktop$ sudo i2cset -y 0 0x58 0xda 0x00ff w
user@livedvd:~/Desktop$ sudo i2cset -y 0 0x60 0xda 0x00ff w
user@livedvd:~/Desktop$ sudo i2cget -y 0 0x58 0xda w
0x00ff
user@livedvd:~/Desktop$ sudo i2cget -y 0 0x60 0xda w
0x00ff
user@livedvd:~/Desktop$ sudo i2cset -y 0 0x58 0x11 c
user@livedvd:~/Desktop$ sudo i2cset -y 0 0x60 0x11 c

 

[RolloZ170]

Also, a quick question about your tool: If I do use an MCP2221A, would I need to modify any config files?

no, i detect exact model before programming.
MCP2221a_iccmax_FF.exe -TPS53678 62 64

 

[RolloZ170]

Since I already have direct access to the bus via Linux, is it possible to just use i2c-tools to unlock it manually?

i do not recommend that. you get not owner of the bus, other busnaster(PCH,BMC) can write between your read write cmd's.

 

[whitelilis]

View attachment 47531

whitelilis.github.io/sogon_motherboard.jpg at master · whitelilis/whitelilis.github.io

Contribute to whitelilis/whitelilis.github.io development by creating an account on GitHub.

github.com

And the text on the chip:

 

[RolloZ170]

And the text on the chip:

PXE1610C is oK. you need to trace to a I2C header or solder.
( note: a restistor can be in serial with sda and or scl line )

 

[whitelilis]

PXE1610C is oK. you need to trace to a I2C header or solder.
( note: a restistor can be in serial with sda and or scl line )

Thank you very much, I will have a try some days later.

 

[whitelilis]

PXE1610C is oK. you need to trace to a I2C header or solder.
( note: a restistor can be in serial with sda and or scl line )

By the way, what value resistor should I use? Thanks again.
And there are two pxe1610 chip each above a cpu, do they all need to be modified?

 

[RolloZ170]

By the way, what value resistor should I use? Thanks again.

there should be a restistor in line from VRM IC to I2C header. if so don't connect direct to VRM IC.
resistor is mostly 20...30 ohm and located close to the VRM IC.
edit:
i found I2C VR header on A620-G30 Server Motherboard (EPYC)
I620-G30 should have similar.

 

[gug296]

i do not recommend that. you get not owner of the bus, other busnaster(PCH,BMC) can write between your read write cmd's.

thank you a lot for replying me.anyway，i tried using i2ctool first and lucky it works on my board and successfully booted 8259cl

 

[whitelilis]

there should be a restistor in line from VRM IC to I2C header. if so don't connect direct to VRM IC.
resistor is mostly 20...30 ohm and located close to the VRM IC.
edit:
i found I2C VR header on A620-G30 Server Motherboard (EPYC)
I620-G30 should have similar.

So can you please show the picture on A620-G30 ? I can try to find the header.

 

[daunce]

Thanks @RolloZ170. Worked a treat on Dell R740 with Adafruit MCP2221A & 4pin JST. I now have Intel Xeon 8259CL working.

 

[xf-vc]

Bios 3.9 had a extra spector fix 3.4 does not seem to have.

Getting closer using bios 3.4 can get nicer single core speed and still somewhat good multithread scores. Not as hight as the Asrock board but getting closer.

View attachment 34072
View attachment 34073
View attachment 34074
View attachment 34075

BIOS CPU setup Hardware P‑States = native OK！

 

[Andreiva]

For the HPE DL380/ML350 (I imagine they are close enough) G10 does the server need to be on for detection to work? With it just plugged in I get nothing and when I tried to start it up something showed up on address 62, but nothing on 64 (only 1 cpu installed), if I want to mod both do I need dual CPUs? I have 2 8259CL but I don't think they will boot until the mod is done...

 

[Porlam]

For the HPE DL380/ML350 (I imagine they are close enough) G10 does the server need to be on for detection to work? With it just plugged in I get nothing and when I tried to start it up something showed up on address 62, but nothing on 64 (only 1 cpu installed), if I want to mod both do I need dual CPUs? I have 2 8259CL but I don't think they will boot until the mod is done...

You need to install CPU 1 and 2 with normal CPU to modify both VRM.