BL465C G1 - Quad-core upgrade fails

[Robert Toth]

Hi everyone,

After reading everything that was available, I decided to upgrade one of our BL465C-G1 blades from two dual-core 2214-EL processors to the latest Quad-core 2354 versions. This blade is our SQL server (Win2K3 x64, SQL 2005 STD x64) and we could use the extra horsepower going forth with our Sharepoint 2K7 install.

In a word: It DOESN'T work.

I did everything in the right sequence (ROM upgrade first to the 2008.11.09 for the C13, ILO upgrade to the 1.70, and finally the 2.32 upgrade for the OA. I even went so far as to upgrade the 2.72a5 on the QL 1150 mez cards.

When I first plunked in the Quad core (just one to start), the blade red-lighted. OK, no problem. Do the config reset (SW1-6) process back with a dual-core. Came up blank on the dual-core right after. As it should.

Plunked the 2354 back in, red-lighted again.

OK, back on the dual-core, do the first stage config clear again, and then yanked the battery out of it for an hour. Clear the RTC/CMOS the old-fashioned way.

Plugged the Quad back in, put back the battery, made sure the reset SW1-6 was off... Success!!! It comes up and is recognized properly.

Only have half of my RAM tho (the other bank needs the second processor to enable it), but we're getting somewhere.

Plunked in the second quad - red-lighted right away. ARGGGGGG...

OK, back off the second quad and it powers up. So now I swap the quads - just make sure it's not defective. Powers up fine.

Take the first Quad now, plug it into Socket 2 - red-light... At this point, I'm thinking a big-heavy truck, one blade, and a small meeting together is looking pretty good.

Anyways, I yank both of the quads - replace them with both of the old Dual-cores. Just making sure we didn't do something stupid on the second socket. Comes up fine. And with all of the memory now showing ok. Technically, everything works on the blade. so far so good.

Put back both quads. Red-light.

Yank the second quad - fires up just fine but again, I'm down to 1/2 my RAM.

So I put back the duals, light up the blade and make sure everything is back as it should with configurations, IP's, WWM to the EVA - everything is working fine and I step away from the chassis before I do something I might regret later.

And now, kind folks, I turn to you for insight / advice / help. I know both my 2354's are fine. I know the board/chassis/OA is at the correct ROM levels. I know the system board is working fine as it works perfectly with the Duals. And I know it's something to do with the BIOS ROM and dual Quad-cores on the blade. Because I repeated the whole process on a second BL465C-G1 blade with exactly the same outcome.

So, has anyone else tried upgrading an existing Dual-core 2214 blade to dual Quad 2354's before? Did I miss anything stupid? I have searched all over and nothing I have found is of any help. Everything says it should work fine.

Sorry for the long post - I'm just hoping someone out there can help shed some light as to why this isn't working. This is turning into a very expensive mistake which I am hoping will not be. Thanks!

[Ken Henault]

I know how to fix it. Please email ken1(at)hp.com

[Robert Toth]

Ken,

Thanks again for pointing out the system board issue. I just love undocumented issues (at least publicly undocumented ones).

Having checked the warranties as you recommended, we'll be calling HP support to initiate a ticket with them for the replacement(s) we need.

I'll update back with news once that is done. But I'm pretty confident what you specified is the issue, and this will fix the problem.

Thanks!

[davidgrahamjr]

I just got off the phone with HP and they stated flat-out that upgrading from an Opteron 2218 to a 2354 is not supported.

[Robert Toth]

Actually, the upgrade works quite well, once the source of the problem is fixed. And many apologies for not having gotten back to this earlier to indicate that we're all happy campers over here now.

In a word, revisions... You simply have to have a recent enough revision of mainboard to run the dual quad-cores. That was my problem, I had a pre-spec board that couldn't handle more than 4 cores (two duals, or 1 quad).

Many thanks to Ken on getting me on the right track to getting this resolved. My SQL server blade is now happily humming along using multipath access to the EVA-4000. And my users thank you for the extra-quick time it takes to blow up their queries now.