• Welcome to Andy's Workshop Forums. Please login or sign up.
 

Hacking an HP Z800 motherboard into a standard PC case

Started by Andy Brown, November 01, 2014, 04:53:46 am

Previous topic - Next topic

Andy Brown

Quote from: badboy2 on February 23, 2015, 09:05:29 pm
Hi all, me again >.<. I try to put the cpu heatsink on, use 1 cpu e5520 to make sure old bios could handle this. I try many time to turn my system on but computer power supply shuts down after a few seconds and I have no idea why's that. Anyone could give me a suggestion for this trouble?


Unfortunately those symptoms are associated with a wide range of issues. If your power supply is faulty or incorrectly fitted then this will happen. If your CPU is faulty then this will happen.

Are you certain that your PSU is wired correctly and that all connectors are fitted (board power, CPU power, memory power)?

Have you got the correct memory type installed in a supported configuration?

Is this the same CPU that was previously running without a heatsink? As you've discovered it has automatic thermal shutdown when the core temperature approaches damaging levels but this isn't a feature you'd want to test. It is possible that the CPU has been damaged.
It's worse than that, it's physics Jim!

digitaltrousers

Quote from: Andy Brown on February 21, 2015, 08:17:19 am
I've had some more time to play around with the BIOS SPI header on an 002 motherboard. I've verified that when VSS and VDD are supplied with 0/3.3V then WP and HOLD are both high at the legs of the BIOS IC. This is good. It means that the SPI flash device is write enabled and should accept programming commands.

However, I've tried sending it the JEDEC RDID command and it's not responding, SO stubbornly stays high. I know my CE/SCK/SI protocol is correct because I've verified it with a logic analyser at the master end and I've communicated like this with dozens of SPI ICs before now. Not sure what's going on, maybe I'll probe it at the legs of the BIOS IC to see if the signals are not getting there intact.


Have you tried both logic low and high on pin 1 of the SPI header (one with the square pad)? I'm not sure it's CE as in your graphic (at least on my board anyway).

Andy Brown

Quote from: digitaltrousers on February 25, 2015, 03:03:30 amHave you tried both logic low and high on pin 1 of the SPI header (one with the square pad)? I'm not sure it's CE as in your graphic (at least on my board anyway).


Only active low, which follows the datasheet protocol. I've never yet seen an active high SPI IC, what did you have in mind? I also tried both the supported SPI modes with no change in response. Clearly something's amiss though so I'll be poking around a bit further with the logic analyser to get a closer look.

Pin 1 on the SPI header is a dead-short (<1 ohm resistance) to pin 1 on the BIOS IC on my board. Isn't it the same on yours?
It's worse than that, it's physics Jim!

digitaltrousers

February 25, 2015, 06:57:17 pm #63 Last Edit: February 25, 2015, 07:39:05 pm by digitaltrousers
Quote from: Andy Brown on February 25, 2015, 01:29:14 pm
Pin 1 on the SPI header is a dead-short (<1 ohm resistance) to pin 1 on the BIOS IC on my board. Isn't it the same on yours?

No, mine's different. I don't have the info in front of me right now, but the trace goes to one or two tiny (unpopulated) resistors nearby, the other side of which seems to be tied up with a FET which itself seems involved with #WP. I was actually getting an unusual reading from my multimeter re the FET (beeping for continuity but still showing open circuit on the display - which I'm not sure how to interpret), but definitely not a direct short to the CE pin on the SPI flash. Have you soldered pins to the pads? Could you have shorted pin 1 to the neighboring pin? It does seem strange to me that there would be two pins on the header for exactly the same thing.

Obviously though, it's a curiosity whether the -003 bootblock is only compatible with the more recent chipset stepping, i.e on the -003 version of the board. See http://h30499.www3.hp.com/t5/Workstations-z-series-xw-series/Z800-2nd-gen/m-p/2367973/highlight/true#M12100

Andy Brown

Quote from: digitaltrousers on February 25, 2015, 06:57:17 pm
I was actually getting an unusual reading from my multimeter re the FET (beeping for continuity but still showing open circuit on the display - which I'm not sure how to interpret),


If it's a logic level FET then you may have triggered the gate with the multimeter test current. They can be very sensitive.

Quote from: digitaltrousers
Have you soldered pins to the pads? Could you have shorted pin 1 to the neighboring pin? It does seem strange to me that there would be two pins on the header for exactly the same thing.

I've soldered some short solid core wires to each of the pads to allow easier hacking but all continuity tests were done before I did that so I am sure of the readings.

Quote from: digitaltrousers
Obviously though, it's a curiosity whether the -003 bootblock is only compatible with the more recent chipset stepping, i.e on the -003 version of the board. See http://h30499.www3.hp.com/t5/Workstations-z-series-xw-series/Z800-2nd-gen/m-p/2367973/highlight/true#M12100


Yes that is a concern. Before starting this I read the differences between the steppings in Intel's docs and there was nothing in there that I thought would affect a BIOS's ability to boot. Of course the bootblock could have code that looks at the stepping and flat-out refuses to proceed if it sees the wrong version.
It's worse than that, it's physics Jim!

digitaltrousers

Andy, I had a play with the SPI chip today. What programmer are you using? My Bus Pirate doesn't seem strong enough to drive the decoupling capacitors / other circuitry connected to these lines. I am considering trying to communicate with the chip with the board powered up.
If you indeed have a programmer capable of powering the chip in situ, have you tried holding pin 1 of the SPI header high? I'm still not convinced it's a second CE pin. Interestingly, when I put 3.3V on it (and this pin only), I get just under a volt at the Vdd terminal on the SPI flash chip. I'm guessing the voltage would be higher if the Bus Pirate had more current capability. If you were treating it as a CE pin, and it instead had some other function, it might explain why you weren't able to get a response from the chip.
You wanna try with your better programmer?

digitaltrousers

Also, did you ever try your 002 board with dual processors?

Andy Brown

QuoteWhat programmer are you using

I was using an Arduino clone running at 3.3V. 3.3V on that board (a Seeeduino) is provided by an LD1117 clone that should be up to the job. I haven't tried again since my first attempt but will set aside a few hours hopefully soon to have another go. This time I'll take some readings off the chip legs instead of at the Arduino pins.

QuoteAlso, did you ever try your 002 board with dual processors

No. The chance to get an 003 board came up at a price I couldn't pass on so I snapped it up and then bought and added the second CPU a few weeks later.
It's worse than that, it's physics Jim!

x2xx

quick info:

i was searching for max power output thought 6pin GPU connector and i find this pdf
http://www.nvidia.com/content/quadro/maximus/di-06471-001_v02.pdf
where they use  6-pin to 8-pin adaptors for tesla and quadro fx5xxx in Z800

because i want to try SLI GTX980 and dont want case full of molex conversion cables

Andy Brown

Quote from: x2xx on March 11, 2015, 10:26:07 am
quick info:

i was searching for max power output thought 6pin GPU connector and i find this pdf
http://www.nvidia.com/content/quadro/maximus/di-06471-001_v02.pdf
where they use  6-pin to 8-pin adaptors for tesla and quadro fx5xxx in Z800

because i want to try SLI GTX980 and dont want case full of molex conversion cables



Good that they refer directly to the Z800 in that PDF, that is helpful. FYI: The limiting values for power that can be drawn down by a graphics card are:

75W from the PCIe slot
75W from a 6-pin GPU power cable
150W from an 8-pin GPU power cable

If you do use Molex conversion cables then look for nice fat wires and a good quality connector because 75W is around 6A from the 12V line which comes close Molex's own rating of 6.25A for the connector. Some of the flimsy clones may not be able to handle as much current.
It's worse than that, it's physics Jim!

JonMS

Quote from: x2xx on March 11, 2015, 10:26:07 am
quick info:

i was searching for max power output thought 6pin GPU connector and i find this pdf
http://www.nvidia.com/content/quadro/maximus/di-06471-001_v02.pdf
where they use  6-pin to 8-pin adaptors for tesla and quadro fx5xxx in Z800

because i want to try SLI GTX980 and dont want case full of molex conversion cables


Z800 board is not GeForce SLI certified so you will need to use an SLI hack to make it work. I did this back in the day on an Asus P5Q-E board. As far as powering them if you're using the standard Z800 case you can run dual 5.25" drive-bay PSUs (1 for each GPU) I'm doing this for my dual AMD R9 280X set up as Crossfire requires no certification. Here is a picture of my set up:

digitaltrousers

March 16, 2015, 03:34:18 pm #71 Last Edit: March 16, 2015, 03:45:05 pm by digitaltrousers
Hi All,
I initially tried an -002 board with an X5672 processor - it didn't work, as expected. The processor fan just ran flat out, no video or other response. I then got hold of an E5504 in order to do the BIOS update and see if that would allow it to boot with the X5672 (or two hopefully). This datasheet lists it as an available processor, so shouldn't be a problem, I thought: http://h10010.www1.hp.com/wwpc/pscmisc/vac/us/product_pdfs/z800_datasheet_mar09_NoSkyRoom.pdf
However when I installed that one, same thing as with the X5672. I then noticed that in the more recent datasheet here: http://www.bbt.si/modules/uploader/uploads/s_product/file/hp-z800-1_copy3.pdf that processor is NOT listed as an option. I don't know which motherboard revisions relate to which datasheet (with there being three motherboard revisions). Maybe I've only ever put unsupported CPUs on my motherboard...
I thought it would be worth asking though, has anyone else had the same symptoms, and/or could shed any light on the problem?
PS, all the voltages at the motherboard test fine.
Thanks!

Andy Brown

Quote from: digitaltrousers on March 16, 2015, 03:34:18 pm
I thought it would be worth asking though, has anyone else had the same symptoms, and/or could shed any light on the problem?

Silly question: what's your memory configuration (type and slots used)?

Also, have you got a case speaker you can rig up to hear the diagnostic beeps? (the meaning is in the service manual under "Diagnostic LED and audible (beep) codes")
It's worse than that, it's physics Jim!

BenGman

March 16, 2015, 05:45:44 pm #73 Last Edit: March 16, 2015, 07:37:53 pm by BenGman
Good Day,
Thanks for your worklog. I have two issues:
1. How do you remove the motherboard standoffs that come with a new board and
3. How do you remove the CPU fan  brackets from the motherboard? The CM 212 Evo fan mount standoffs cannot fit.

This link has photos of my build log so far, including pics of the motherboard standoffs:
https://www.dropbox.com/sh/ndo0kl2q4txi1v8/AAB4RlnKAWsvTB-CSYkwur9Da?dl=0.

Thanks in advance.

digitaltrousers

Quote from: Andy Brown on March 16, 2015, 03:45:13 pm
Silly question: what's your memory configuration (type and slots used)?

Also, have you got a case speaker you can rig up to hear the diagnostic beeps? (the meaning is in the service manual under "Diagnostic LED and audible (beep) codes")


Not a silly question - I was actually wondering if my 1333MHz memory would clock down to 800MHz for the E5504 - I had assumed it would. I've got 3x4Gb sticks in the black slots for CPU0. It's genuine HP memory (registered I believe) with the part number 500203-061 (http://partsurfer.hp.com/Search.aspx?searchText=500203-061). I've also tried with just one stick installed. I have some "normal PC" memory that runs at 800MHz but I don't think one can install such memory in this one?
I have rigged up a little speaker and the HDD LED which is also involved in displaying the codes. I don't get any codes when I try to boot up, but if I turn the power supply off at its switch, it beeps/flashes four times, which is a power supply overload error. I had been chasing this, but then I thought it might be because of the hacked wiring (Chinese adaptor cable) or other differences from standard. I.e, standby lines staying active a little longer when the switch is turned off or something. It's a 1050W power supply and all voltages are normal at the connectors during "running".