How to fix Asrock c2750d4i with C2000 bug

[Ericloewe]

Full disclosure: This whole C2000 LPC saga has me increasingly confused the deeper I look into it.

If any of these fixes are to work, what we need is to supply a high rail to be pulled down by the CPU when it wants to output logic low (or it wants to detect the boot mode on power-up) - that's not 3.3V, it'll depend on the IC, but Intel specifies 2 V for the LPC bus. So, as long as logic high is consistently decently above 2 V, all's good.

That said, I'm not following the benefit of the divider versus a straight pull-up.

 

[cj0]

Thanks to @Ericloewe and @nogates. With an external power supply, testing leads, clips and the 10 second delay I was able to boot my no longer booting Asrock c2750d4i board (which previously showed near 97ºC CPU temperature in IPMI).

The Picotest power supply says that PCICLK consumes only 6 mA (@2.00 volts), see pictures.

A Linux reboot command restarts fine without having to do the 10 second delayed PCICLK power supply hack.

PS How hard can it be for Asrock to supply a TPMI header add-on, as a form of customer service, instead of denying warranty exchange requests. I am even willing to pay for the thing.

 

[cj0]

Though I can't reproduce @nogates statement: "Of course, if I disconnect the PCICLK power, the computer freezes completely".

As long as the UEFI/BIOS boot has been handed over processing to the operating system, then I can remove the +2V signal from PCICLK without freezing the computer.

During BIOS boot, the boot process pauses until +2V signal returns. One time boot seemed to stall until I increased 1.99 volts to 2.09 volts.

 

[Ericloewe]

During BIOS boot, the boot process pauses until +2V signal returns.

That seems to be coherent with the observations over on the other thread, where the CPU seems to get around to booting, but then the boot process doesn't really go anywhere.

 

[cj0]

My previous comment in #43 about the voltage level is not true. I can even boot my C2750D4I at 1.00 volts into the OS. With the external PSU turned off, the C2750D4I will no longer boot.

I wish myself the pictures from @RZSN where the 5+2 C2000 bug fix resistors are soldered onto the C2750D4I.

 

[cj0]

@

I wish myself the pictures from @RZSN where the 5+2 C2000 bug fix resistors are soldered onto the C2750D4I.

Because servethehome only shows hires pictures to members, the helpfull C2750D4I rev. 1.02 repair pictures are duplicated here.

Under the CPU:

Under the TPM-S header repair by Asrock ("MFG Board Level Repair") according to @bfarnam:
Note: marking "12A" = resistor 130 ohms / 1% tolerance.

Under the TPM-S header repair done by RZSN:

 

[cj0]

C2750D4I TPM-S voltage divider on a header

The 2 mm spaced 9 pin headers arrived today.

A first attempt with 470 and 470 Ohm doesn't work.
Also the clipped off terminals don't look good.

The resistor leads were routed inside the terminal header. Thus blocking the motherboard pins that should fit in that opening:

In a second attempt the resistor leads were clipped off shorter.
The unused terminals were pulled out of the header (instead of clipped).
Bending and soldering done neater.
Also the resistor values were chosen as in the opening post: 100 ohms on the left side, and 470 ohms on the right side.
The header sits nicely on the mainboard.
And last but not least, the C2750D4I mainboard boots again.

 

[Kortex]

I just used the 100 + 470 ohm method and it worked. I almost can't believe it. Do I need to upgrade my firmware or disabled anything in bios for longevity's sake?

 

[Ericloewe]

You must update the BMC's firmware to avoid an unrelated bug that destroys the BMC EEPROM.

 

[Al Fuller]

You must update the BMC's firmware to avoid an unrelated bug that destroys the BMC EEPROM.

Is updating the BMC firmware something that can be done from the IPMI web interface? Or is there a different path?

 

[Redcoat]

Is updating the BMC firmware something that can be done from the IPMI web interface?

Yes.

 

[XaNaX]

Another success story. Mainboard just died when running fine last week. I am remote right now and have no access to the board, bit asked a friend to make this divider connect and boot. Tested without it no vga output. With the divider works fine, boots, reboots works like a charm at least for now.

 

[Ericloewe]

You'll probably want to keep those wires shorter, for better signal integrity.

 

[Prong0035]

I did what the first post mentioned, using the same values, placed the same way, with short legs on the resistors and got my server back to working condition.

 

[rayrayrayraydog]

Hi, I have come across a very cheap C2750D4I, revision 1.03, that was supposedly working. I cannot get it to boot up however, and I've eliminated all variables like power supply and RAM. The IPMI lights come on and I can access it, but the system does not POST or display anything over VGA. I have cleared the CMOS and tried with one stick of RAM. I have also tried the 100/470 resistor trick in the OP with no luck.

Here is what I see in the IPMI after I attempt to power on the board. I don't have an oscilloscope to get any readings from it. Any ideas?

 

[Ericloewe]

Hmm, the screenshot does paint the picture of a dodgy power rail or two...

The IPMI lights come on and I can access it, but the system does not POST or display anything over VGA

Do the fans spin up?

I don't have an oscilloscope to get any readings from it. Any ideas?

An oscilloscope wouldn't be the starting point, but a multimeter would. The board will surely have test pads for those voltage rails somewhere (finding them is likely to not be super fun, but they might be labeled). Cross-checking the BMC's measurements should narrow things down a bit.

If it is a power delivery issue, a repair will probably be too complex for most people to be comfortable with it, but any decent repair shop might be able to take it on, assuming you can narrow it down and acquire the necessary parts.

 

[rayrayrayraydog]

Hmm, the screenshot does paint the picture of a dodgy power rail or two...

Do the fans spin up?

An oscilloscope wouldn't be the starting point, but a multimeter would. The board will surely have test pads for those voltage rails somewhere (finding them is likely to not be super fun, but they might be labeled). Cross-checking the BMC's measurements should narrow things down a bit.

If it is a power delivery issue, a repair will probably be too complex for most people to be comfortable with it, but any decent repair shop might be able to take it on, assuming you can narrow it down and acquire the necessary parts.

Yes, fans connected to motherboard headers do spin up. I will bust out the multimeter and start searching. Thanks.

 

[rayrayrayraydog]

Yes, fans connected to motherboard headers do spin up. I will bust out the multimeter and start searching. Thanks.

I have powered on the board and let it sit for bit and read the following with the meter.
PCICLK - 0.435V
3v - 3.28V

EDIT:

If I connect the 100/470 resistors from the OP and press the power button, the board will sometimes keep power cycling itself on and off every few seconds. It might do this 5 - 10 times before it stops and stays on, but it never POSTs. If I connect the power supply to the board with the resistors in place, and wait 30 seconds or so before powering on, it won't power cycle over and over but won't POST either.

 

[Ericloewe]

PCICLK - 0.435V

If that's the LPC clock (should be, the board doesn't have any conventional PCI devices), then it definitely screams LPC degradation. Should be reading 2.3 V or so. So it does sound like you have two separate issues ruining your day.

 

[rayrayrayraydog]

I was reading from the TPM pins PCICLK and GND for the 0.435v. I also tried the version of the fix with dual 470 ohm resistors and had the same result. Probably about time to recycle this one. :)