Supermicro X10 incompatible with Adaptec 7805Q

[ptr727]

Glad to see the site up again

While waiting the site to come up I posted here, not much help, hope somebody here can be of more help:
Supermicro X10SLM+-F and Adaptec 7805Q fail - [H]ard|Forum

The problem is as follows:
Using Supermicro X10SLM+-F and Adaptec 7805Q.
On cold boot, the Adaptec controller BIOS waits for the controller to come online.
This fails, and the Adaptec controller reports “"Controller monitor disabled or failed to start", "Controller not started", "Press any key to continue".

Using exactly the same board, CPU, and memory the Adaptec 81605ZQ works fine, but I need to get the 7805Q working.

I upgraded the X10SLM+-F BIOS firmware and IPMI and to latest, and it made no difference.
I moved the 7805Q to a differnet machine and updated the firmware, no difference.
I tried the PCIe v3 and the PCIe v2 slots, same problem.
I disabled ASPM, PCI logging, enable PCI compliance, no difference.
I cleared the CMOS and used hte PCIe v2 slot, no difference.

I also tested the following combinations:
Adaptec 7805Q + Supermicro X9SCM-F works fine.
Adaptec 81605ZQ + Supermicro X10SLM+-F works fine.
Adaptec 7805Q + Supermicro X10SLM+-F fail.
Adaptec 7805Q + Supermicro X10SLH-F fail.

In a working system the 7805Q boots and the red lights go into "knight rider" mode, on the X10 systems all the red LED's are on permanently.

I opened support cases with Adaptec and Supermicro.
Supermicro had me try a couple settings and CMOS reset, then requested I open an engineering case, no ETA on when it will be heard.
Adaptec said the Supermicro X10SLM-F is on the hardware compatibility list for the 7805Q using BIOS v1.1a, they cannot confirm operation with other boards or BIOS versions.

Any ideas?

P.

 

[MiniKnight]

Is using the X9SCM-F not a possibility? You are going to spend more time and money troubleshooting than just swapping. Not like V3 is that great of an upgrade over V2.

 

[ptr727]

Except I already have two new servers built and racked, so yes new motherboard, new processor, new memory is an option, an expensive option (The X9 plus 7805Q is from an already deployed system).

Btw, SM support just let me know that they tested a 7805Q in a X10SLM+-F board ... and it works.
They then told me my memory is not supported, I used Kingston KVR16LE11/8EF modules.

There are a couple threads around on X10 boards and Kingston memory, when I ordered this equipment, some time ago, Kingston had this memory listed as compatible, but later pulled it, currently Kingston does not list any memory as compatible with any X10 boards. From what I can gather Kingston changed chip suppliers, and no new modules are certified for X10 use.

The certified compatible 4GB Hynix modules are $110! a pop, just ordered 4 to try out.

P.

 

[abulafia]

Have an X10SL7-F and an X10SAE-V. I've used a bunch of (different) Crucial ECC UDIMMs off Amazon in the past, including 1866mhz and DDR3L. Some can be had for $150 for a pair of 8GB sticks. Let me know if you want part numbers.

 

[ptr727]

I got my $600 for 16GB SM certified Hynix memory, compared to the $400 for original 32GB Kingston, same problem.

Thanks to a user at HardForum, who was having similar problem, and working with NL SM support, they gave him a pre-release BIOS that solved his problem.
I told US SM support, and they gave me the same BIOS, tested it, and it works fine.

Glad and mad at the same time.

P.

 

[abulafia]

That's nuts. I pay no more than $370, and usually closer to $320 for 4x8GB. Also, I hate Kingston for some SSD nonsense, and figure that if a company sends out one SSD for review, swaps the chips when actually at market, eh f'. em. Hynix I think of as somewhat overpriced and generally as a factory server OEM part/option (and very overpriced). Possibly why they are on HCLs, I guess.

I've stuck to Samsung (mostly for RDIMMs, specifically 16GB) and Crucial (for 2x8GB UDIMMs for quick Amazon Prime or local B&M purchase options) and have had no problems. Boards include three Supermicro X9 2011s, and in UDIMM land, an X9 (216) 1155, X10s as above, ASRock 1150s (224/226) and the C2750 Avoton board. All Xeon, all ECC.

For the Crucial UDIMMs, I think I've used three different parts (1.5v 1600, 1.35v 1600, 1.5v 1866) based on what the board supports (i.e. best spec) or going DDR3L when there's little price difference and plain 1600 DDR3 when there is. I probably have over a dozen sticks purchased since March. I wouldn't hesitate to give Samsung UDIMMs a shot either; if Amazon carried them I would probably have more of a mix between them and Crucial in UDIMMs, especially since I went Haswell only for power consumption/electrical issues, not by preference, and so had been buying Samsung up to that point. Haswell's practical 32GB ceiling hurts me badly (my Photoshop can idle at 17GB, and that's with "bad" non layering habits formed by having to work around a limited RAM canvas). Really like the Samsung sticks - I have eight of the 16GB sticks over two machines. Solid.

 

[abulafia]

Incidentally, while I do understand the concept for HCLs (especially if acting as an integrator/reseller or CYA for IT department purchases) but I can't be the only one that thinks it's utterly ridiculous to buy name brand ECC server/workstation grade RAM, by definition in JEDEC spec, matched with same grade CPUs and board and have to worry about compatibility? Doesn't quite scream "reliability" or "stability," does it?

 

[ptr727]

Just when I thought I was making progress with SM, I get this when I push them to give me an official release date:

There won’t be another BIOS release any time soon.
At this time, I’m not sure if it’s a BIOS issue or card issue.

The BIOS could be merely a work-around to fix card issue.

I cannot duplicate issue so resource allocation is not a priority for this case.

Anyways, I will check with BIOS engineer to see if this fix is specific
to the card or whether it would be included in next release.

The next release may not happen until another 3-6 months.