dl185g5 cpu upgrade

[taggart_seattle]

Has anyone tried upgrading their dl185g5 to newer CPUs. First a disclaimer, I realize this won't be supported, but I am trying to determine if it _works_.

WARNING: to anyone with a dl185g5, beware of the problem with BIOS upgrades that can brick your system, details.

The dl185g5 uses Socket F and the marketing material claims it supports Opteron 2200 and 2300 series CPUs. We know that the later 45nm "Shanghai" versions of the 2300 work, HP supports those (at least the ful power and "HE" models).

The system uses the Serverworks HT-2100 and HT-1000 chipset combo.

I am wondering if anyone has tried any of these?

* "Shanghai" 2300 "EE" (energy efficient) models: 23KSEE, 2373EE, 2377EE

* "Istanbul" 2400 6-core modules (in particular the 2419EE)

I think the chipsets should work with the 2400 series, at least there are other HP models (dl386g6, dl165g6) that use that combo and support them.

Being able to put faster and more energy efficient CPUs in these would save them from the recycle bin.

Thanks,

Matt

[taggart_seattle]

After researching the chipset in these systems and learning the chipset at least supported the newer CPUs, I ordered an AMD Opteron 2419 hex-core CPU to test ($29 on ebay). First I tried swapping in the CPU with the existing firmware version which on boot led to a successful POST, listing 6 cores, and then an error from the BIOS saying CPU checksum was wrong. Then I upgraded the system firmware, stepping through each version (in order to avoid the bricking problem, ~15+ reboots). Once on the latest system/BMC BIOSes the system would POST and list 6 cores, but when the newer CPU was detected the system determined that this must be a dl165g6 system and flashed the BMC firmware to a different version. Then it rebooted, listed the latest firmware version of the dl165g6, and complained about 3 fans being missing (since the dl165g6 is 1U and has more smaller fans). Just for fun I took the cover off and plugged additional fans into the headers. This got it past the fan error, but then the system powers off.

The 2419 CPU uses far less power than the other CPUs supported by this board, so I don't think it's a matter of not being able to power things or hitting power/thermal limits. I suspect the problem is only in firmware and that it could be made to work with firmware fixes. But given the age of this product I know HP won't ever fix that (and have a financial disincentive to do so) and the closed-source nature of the BIOS is such that no one else could fix it short of some serious hacking.

In working on this and also due to the one bricked-by-firmware-upgrade system I have I realized something. Since these are HP's entry level systems and mostly outsourced design, they use pretty standard ATX components. So my next experiment is going to be swapping the mainboard for a much newer 3rd party board, but keeping the PSU (minus the management cable), using a newer pci-e HBA and connecting it to the existing backplane with a SFF-8087 to SFF-8484 cable, keeping the existing fans. I think the front panel connect should be the only tricky part.

[VelocityTracy]

can't help but asking, is this server in business use? or a play toy for you? also, are you saying that the firmware upgraded on its own to the DL165 G6 version?? My business is fixing errors and replacing parts I don't generally get to play around like this...interested to know what happens when you put the old procs back in as well.

[taggart_seattle]

I work for a non-profit with a pretty small equipment budget. A lot of the hardware we get is donated and also we have limited power available in our hosting location. So I am always trying to take old servers and put the most efficient performance/watt CPUs I can into them.

I forgot to mention, when putting the original CPU back in and booting the firmware decides it's a dl185g5 again, reflashes the BMC, reboots, and then it's back to printing the dl185 bios version on POST.

I probably won't spend too much time trying to make it work with a third party board, because the alternative is spending ~$675 to get a brand new 2U 12 x 3.5" case (and that backplane would have more bandwidth).

[VelocityTracy]

It seems I've been working with a lot of companies with a dwindling IT budget this year. Sad that IT people have to scour ebay to look for parts they need now. It's great that you try to maximize what you're working with! I'm happy to hear that it went back to the DL185 G5 when you put those old procs back in I was wondering about that. Good luck on your upgrades!

[Torsten.]

Not exactly sure about this, but I think I remember the DL165 G6 and DL185 G6 are using the same board.

But they have different jumper settings.

DL185 G6:
P26 2-3, default setting
P27 1-2 : default setting

DL165 G6:
P26 1-2, default setting
P27 2-3, default setting

[taggart_seattle]

I just checked on my my dl185g5 (NOTE: g5 not g6 like you mentioned, I don't know if yours was a typo).

P26: 1-2

P27: 2-3

[Torsten.]

It was a typo, sorry.

Just try to switch the jumpers and check what the system think it is ...

[Torsten.]

Check this link, it is a query about the DL185g5 system board, also used in ...

http://partsurfer.hp.com/Search.aspx?searchText=452339-001

[taggart_seattle]

Setting the P26/P27 jumpers didn't make any difference in the behavior, it still did the automatic BMC upgrade upon switching the CPU, still thought it needed more fans, and still powered off if I gave it more fans.

I also did some experiments with swapping out the system board for a standard ATX board. I ran into two problems:

1) determining the pinout of the front panel connector (457561-001) proved troublesome. I was able to locate the two pins associated with the Power LED, but that's about it. I suspect there may be more to the pins than just physically completing connections, because I couldn't seem to isolate a pair of pins that connected when the button was pushed. I was willing to live without the built-in front panel stuff (and just add my own) but...

2) if the additional management cable to the PSU isn't connected to anything, the system fails to power up. Some power is getting to the board, LEDs light, etc. but pressing the power button isn't working. I suspect the PSU won't supply power without some indication from the board. Fail-safe rather than fail-working.

Without those things the system really isn't usable as a general ATX case, or even a disk array to connect to another system. It's pretty much only usable as it originally shipped, old high-power cpu, expensive slow ram, etc.