This discussion has been locked.

You can no longer post new replies to this discussion. If you have a question you can start a new discussion

SG135 C2000 BUG

RobertoR over 5 years ago

Hello, I recently bought a used SG135 rev.1 which has AVR54 bug (I found out about this bug after I buy it).

For others vendors I found a possible fix, but not for Sophos.

There are two (as far as I know) fixes:

1. Replace a atom c2xxx with C0 revision

2. external pull-up resistor (18.2k to 10k), from LPC_CLKOUT0 or LPC_CLKOUT1, tied to 3.3V

Did someone tried to fix this bug? (I use home license, so I don't have any agreement with Sophos).

Best regards,

Robert

This thread was automatically locked due to age.

Parents

0 BAlfson over 5 years ago

Ahoj Robert and welcome to the UTM Community!

I haven't been inside an SG 135, are you telling us that the Atom processor cannot be replaced?

Cheers - Bob

Sophos UTM Community Moderator
Sophos Certified Architect - UTM
Sophos Certified Engineer - XG
Gold Solution Partner since 2005

MediaSoft, Inc. USA
Cancel
Vote Up 0 Vote Down

Cancel
0 RobertoR over 5 years ago in reply to BAlfson

Hello BAlfons, I didn't opened yet, but generally they are soldered to mainboard, so replacement is very, very hard.

Best regards,

Robert
Cancel
Vote Up 0 Vote Down

Cancel
0 Paul Sawatzki over 4 years ago in reply to RobertoR

Hi Robert,

yes, sure! Here they are (if you need more or more detailed ares let me know):

It would be great if there was a simple pullup to bring it back to life, but without schematics or layouts to hard for me.
Cancel
Vote Up 0 Vote Down

Cancel
0 RobertoR over 4 years ago in reply to Paul Sawatzki

I agree with you, without scheme wil be hard.
Cancel
Vote Up 0 Vote Down

Cancel
0 Paul Sawatzki over 4 years ago in reply to Paul Sawatzki

Hi Sophos Community,

is there any any hardware engineer from Sophos with access to schematics or layouts on board who could help? Any hint regarding the LPC_CLOCK_OUT pins would be highly appreciated.

THX,

Paul
Cancel
Vote Up 0 Vote Down

Cancel
0 RobertoR over 4 years ago in reply to Paul Sawatzki

Well I think you will not get an answer you hoped for. Intel has NDA with Sophos and so, they can't talk about it.

More about error:

https://www.eevblog.com/forum/microcontrollers/intel-atom-c2000-failures/
Cancel
Vote Up 0 Vote Down

Cancel
0 RobertoR over 4 years ago in reply to RobertoR

According to this document LPC_CLKOUT0 is GPIOS26 or AG51 pin. But for LPC_CLKOUT1 is note that Intel has not validated this mode.

So for now look on mainboard, maybe are there some marks about this pins.
Cancel
Vote Up 0 Vote Down

Cancel
0 RobertoR over 4 years ago in reply to RobertoR

Found this document and location of AG51 is visible in table 35-6 (Sheet 2 of 6). For ball out look at the table 35-3. Has someone try to remove C2000 processor from board?

The only thing we need is now line from mainboard (bottom of cpu)!
Cancel
Vote Up 0 Vote Down

Cancel
0 nd over 4 years ago in reply to RobertoR

It would be very nice to know which resistor have to be replaced exactelly....and I think this would be easier than replacing the whole cpu ;)
Cancel
Vote Up 0 Vote Down

Cancel
0 RobertoR over 4 years ago in reply to nd

You don't replace a resistor you just add a pull-up to right line/resistor (CLK line from CPU which is output doesn't works anymore).

I cannot tell you on which resistor you add an external pull up (my sg135 still works). But you can try your self.

What will I do when my die, I will short all resistor on the left top side (look on the picture) with resistor tied to 3.3V. When I find a right place it should power on / boot (I think you must press power on button).
Cancel
Vote Up 0 Vote Down

Cancel
0 Andreas Fischer2 over 4 years ago in reply to RobertoR

Having also an SG135 with defective Atom C2000, I stumbled upon the Nuvoton I/O Chip on the back of the mainboard, which should be connected to the LPC bus.

Maybe this is connected to the LPC_CLKOUT in some way, but i did not find anything confirming this. Could be a possible way for a DIY fix.

The datasheet mentions pin 15 as a Clock-Input for the LPC interface - just guessing...

https://media.digikey.com/pdf/Data%20Sheets/Nuvoton%20PDFs/NCT6776F,D.pdf
Cancel
Vote Up 0 Vote Down

Cancel
0 RobertoR over 4 years ago in reply to Andreas Fischer2

Well, more than I read more confused I am. Someone in this forum fixed LPC_CLKOUT0 with resistor 110 Ohm to 3.3V (actually my first post is not right regarding resistor, correct should be 8.2k to 10k, but this information I can't find anymore). So I'm not sure anymore. All fixes I saw are working with 100 or 110 Ohm resistor.

So be smart ...
Cancel
Vote Up 0 Vote Down

Cancel

Reply

0 RobertoR over 4 years ago in reply to Andreas Fischer2

Well, more than I read more confused I am. Someone in this forum fixed LPC_CLKOUT0 with resistor 110 Ohm to 3.3V (actually my first post is not right regarding resistor, correct should be 8.2k to 10k, but this information I can't find anymore). So I'm not sure anymore. All fixes I saw are working with 100 or 110 Ohm resistor.

So be smart ...
Cancel
Vote Up 0 Vote Down

Cancel

Children

0 Marcelo Clare over 4 years ago in reply to RobertoR

Hi Guys,

I've got a dead SG 135 that I would like to try and fix.

Does anyone know which pins to connect the resistor? Does anyone have a picture of the workaround?

Cheers
Cancel
Vote Up 0 Vote Down

Cancel
0 RobertoR over 4 years ago in reply to Marcelo Clare

For now everything is pure speculation. Nobody know were to put resistor.

Try an idea of Andreas. There could be LPC clock in but on the pictures is not visible if PIN15 is actually connected (need better picture of surroundings of NuvoTon NCT6776F chip).

Did you try to remove the battery for a few minutes?

For now, without scheme will be very hard to tell where to put resistor!
Cancel
Vote Up 0 Vote Down

Cancel
0 Marcelo Clare over 4 years ago in reply to RobertoR

Hi Robert,

Thanks for the information. Please see the picture of the Nuvoton chip. I have tried to remove the battery and start the unit again without success.
Cancel
Vote Up 0 Vote Down

Cancel
0 Marcelo Clare over 4 years ago in reply to Marcelo Clare
Cancel
Vote Up 0 Vote Down

Cancel
0 RobertoR over 4 years ago in reply to Marcelo Clare

Hello, on this new pictures is visible, that pin 15 is connected to open pin (in fact in two open pins) and one closed pin which I think goes to C910 (yellow dot). Need better picture of red square. From yellow point it goes up on open pin then right (need better picture) to closed pin (capacitor, resistor) and then left (open) and left (open) and finally to pin 15!

Red dot is 3.3V. Blue dot is pin15 (IOCLK) - This pin maybe wrong for fix, as is connected to 48Mhz clock???

EDIT: C910 is connected to X9 (it provides 48MHz clock to pin15!!!). Can someone measure this???

BTW:

This is pure speculation and without scheme is almost impossible to say to which pin you/we should add a resistor. I'm not responsible for any way!!
Cancel
Vote Up 0 Vote Down

Cancel
0 jamesharper over 4 years ago in reply to RobertoR

The synology fix is straighforward because the LPC clock in question is also used (multiplexed) as GPIO, and these GPIO pins are broken out to some sort of unpopulated connector that is easy to solder.

One of the unused or unpopulated jumper blocks on the the SG125 board may also be GPIO, but JP7 is the only possible candidate I can see with enough pins, and even then it would have to use the same GPIO pins on the C2000 that are mux'd over the LPC clock pins. It should be easy enough to test though.

Also, these are older pieces of hardware that have probably been running constantly for 4+ years, in all sorts of conditions. If you have a dead one then the C2000 bug is definitely a possibility, but there are plenty of other ways hardware can fail :) The one I am looking at right now started crashing every few days at first, then every few hours, now it won't POST. Lack of POST is definitely the defining symptom of C2000 fault (can't load BIOS from flash), but the crashing out with increasing frequency isn't, so I think my device may actually have a different fault.

James
Cancel
Vote Up 0 Vote Down

Cancel