z820 motherboard variants for replacement

[TJMacDaddio]

Hi. I have a z820 that I believe needs its motherboard replaced.

Its current board is an SP: 708610-001 / AS: 618266-003 with the J63 v03.94 BIOS with a Boot Block Date of 03/06/2013. It is running dual E5-2680 2.70GHz CPUs.

A board I've found to possibly purchase is an SP: 708464-001 / AS: 618266-001. I do not know the BIOS version or Boot Block Date.

What additional information do I need to confirm that my current CPUs are compatible with the replacement mobo?

Many thanks!

[DGroves · Accepted solution]

your current board is a 1.02 board revision, that IS capable of running the v2 ivy bridge xeons

there are only two versions of this board, 12/28/11 for the the v1 which can only  run the original Sandy Bridge Xeon processors

and 3/6/13 for the later v2 board which has a updated c602 chipset revision

that allows reliable usage with the E5-26xx v1 Sandy Bridge and the newer E5-26xx Ivy Bridge "v2" processors

HP and intel had hoped that the orig  c602 chipset would be compatible with both the v1/v2 E5-26xx xeons, but found this not to be the case and had to do a revision to one of the c602 chips in order to resolve v2 E5-26xx cpu compatibility

Board Part number 618266-004 definitely supports v2 processors

 motherboards with "v1.03" revision work with v2 processors.

v1.02 board revision motherboards SOMETIMES support v2 processors,

depending on the "Rev number" printed on the sticker

"OR" is a v2 capable xeon board revision

[TJMacDaddio]

Thank you for your reply. There's wealth of information there that goes way above my pay grade, so to speak.

It sounds like the mobo I've found for purchase is an early variant - possibly the first variant - and chances are low to none that it would be capable of running my existing processors. Do I have that right?

I'm curious how to establish the version of my CPUs. I've reviewed the BIOS System Information, Windows System Properties and the HP Performance Advisor. None offer what I recognize as version information on the CPUs beyond "E5-2680".

Thanks for your time and expertise.

[TJMacDaddio]

I'm having trouble finding info "for dummies" on identifying the version of a Xeon processor - but I suspect that if I'm seeing just "E5-2680" with no "v2/v3/v4", then my CPU probably <isn't> a v2/v3/v4?

I found a list of Xeon processors that seems to confirm that I have the "OG", by virtue of its number of cores (8) and frequency (2.7GHz). Am I connecting those dots right?

In that case, I should be able to run the CPUs I have in the SP: 708464-001 / AS: 618266-001 motherboard - because that board was limited to running Sandy Bridge-based CPUs, and it appears I have Sandy Bridge-based CPUs. True?

Thanks for the help!

[TJMacDaddio]

I haven't found a "for dummies" guide on how to identify the version of an Intel Xeon processor, BUT... I suspect that if I'm not seeing a "v2/v3/v4" specified, I don't <have> a "v2/v3/v4"?

Also, by virtue of the number of cores (8) and the clock speed (2.7GHz), it seems like I have the original Sandy Bridge-based CPU. Do I have that right?

In that case, the SP: 708464-001 / AS: 618266-001 mobo should work with my CPUs. True?

It just so happened that my system came with the slightly newer mobo that <could have> run a E5-2680v2 - but I don't have the v2s, so the earlier version mobo would work for me. Yes?

I know to the very experienced, I must be repeating myself - but I'm just trying to make sure I understand.

Thanks for the patient help.

[TJMacDaddio]

And apologies for the several repetitive posts above. Last night, they were not showing up and I even got the message at one point that they were being removed and marked as spam. Today, I see them and I'm unable to delete the duplicate.

Anyway, I'll appreciate the confirmation of my thoughts or an explanation of why my conclusions are wrong. I'd like to purchase the mobo and make my attempt to repair this machine soon.

Thanks for the help.

[DGroves]

look carefully at your cpu(s) and you will note another alphanumeric 5 or 6 digit code that's the cpu's "SPEC" which is used to id each revision that intel makes to their cpu's "SR1A6" is one of them for the v2 e5-2680

revisions are done to fix or add things to the cpu and a cpu may or may not  get these updates but most cpu's will get at least one revision during their production lifetime

print out the v1/v2 cpu specs and you will then see the differences between the revisions

core count and base freq/turbo feq

(v1 Orig release) 2680 specs "Sandy Bridge" there were six diffrent SPEC numbers for this cpu

https://www.intel.com/content/www/us/en/products/sku/64583/intel-xeon-processor-e52680-20m-cache-2-7...

Lists "SPEC"

https://www.cpu-world.com/CPUs/Xeon/Intel-Xeon%20E5-2680.html

(v2) 2680 specs "Ivy Bridge"

https://ark.intel.com/content/www/us/en/ark/products/75277/intel-xeon-processor-e52680-v2-25m-cache-...

Lists "SPEC"

https://www.cpu-world.com/CPUs/Xeon/Intel-Xeon%20E5-2680%20v2.html

[TJMacDaddio]

Thanks very much for your thorough and informative reply!

I haven't yet removed the heatsink to take a look at the chip. I need to get some thermal paste first so I'm ready to put Humpty Dumpty back together.

However, using HWiNFO64, I've confirmed my E5-2680 is "Sandy Bridge" - which I <think> makes me confident enough to order the SP: 708464-001 / AS: 618266-001 mobo. It's a much better price than other Z820 system boards, and now I think I understand why: I won't be able to upgrade CPUs without changing the mobo, too. That's a limitation I'm willing to accept in an effort to get the system back up and running reliably.

Perhaps you could offer your opinion on whether replacing the motherboard makes sense?

A UPS self-test failed spectacularly, abruptly pulling the power rug out from under my system. Following that event, these symptoms occur in various combinations.

Despite replacing the CMOS battery, the time drifts - and not by just a minute or two; it can be off by hours in a very short amount of time. I've reset it in the BIOS and/or used Windows to sync the time, but it can and does still drift.

Cold boots usually complete without complaint.

Once the system has been running for a time (length varies), fans ramp up - though there is no load taxing the system.

I've tried various utilities - HWMonitor, HWiNFO64, SpeedFan - to read system temp and fan state, but none detect the system fans. The only utility that knows that the fans exist and what they're doing is HP Performance Advisor, so I've chalked up the other utilities' inability to read the fans' states to proprietary hardware and sensors?

Refreshing sensors in HP PA returns the fans to a lower speed, but results in various sensors disappearing, or showing 0 RPM on various fans, or showing 0 or 255 degree Celsius CPU temps. The system becomes unstable while HP PA tries to refresh, then seems to operate smoothly again. Opening the chassis, I can verify that fans are still running and the system temp is within a range I'd call "normal". Full disclosure, the one 60mm memory fan is <not> running well and I plan to replace. All other fans (chassis, squirrel-cage memory, CPU) seem to be running fine.

Rebooting the system in this state often hangs. Windows 10 will shut down, the screen will go black, but the power button remains blue and system fans continue to run. Long-pressing the power button sometimes works, but I have occasionally pulled the power cord.

Once the system has been forced off, it's finicky about restarting with a simple push of the power button. Once I get it to respond to the power button, if it's still warm, there's usually a combination of POST errors about the thermal sensor not being programmed or chassis, memory or CPU fans not being detected. If the system has cooled, it will usually proceed without complaining about any fans or thermal sensors, but...

On reboot, the RAID volume may be marked as offline and I'll have to Ctrl+C into the setup to mark it active.

Occasionally, the system will boot into Windows with some USB ports missing-in-action - for instance, my WiFi dongle not being recognized (a front USB port) or the connected UPS not being detected (a back port). Otherwise, I have to admit that Windows seems to run normally even while the fans are running on high for no apparent reason.

So... that's why I'm thinking replacing the system board is in order. Do you think that's jumping the gun?

I am questioning the fan assembly connection and the fans themselves, but don't know how to test the connection or the fans beyond verifying visually that they are running while the system is on. The other part that I wonder about is the power button/speaker/thermal sensor assembly. But even if the sensor was misreading the ambient temp, I wouldn't think it would cause the fans to disappear from HP PA or go missing during POST.

I'm certain that I don't have your degree of expertise or (I suspect) equipment / spare parts, but are there other diagnostic steps you would recommend to a less experienced but capable amateur?

I'm running on a backup XW6600, so I have some time to tinker. Still, I don't want to miss the chance to buy an affordable mobo, if that's what it will take to get back up and running.

Thanks very, VERY much for your time and help.

[SDH]

A good little free utility is CPUz from CPUID.com. At the top of the front page of its results is the actual official name of the processor it sees. You'll note that a E5-2680 detected by the utility will show no v1 after that, but a E5-2680 v2 will show the v2, and on with the v3 and v4 processors if you're running a ZX40. You could have a v2 Z820 motherboard that can run 1 or 2 E5-2680 processors but if CPUz does not show "v2" in its name it is a v1. To my knowledge there is no way to probe with software for the sSPEC code of a processor. You can look on the top of the stainless heat spreader of the processor for that laser etching. DGroves mentioned 6 sSPEC codes for one of these... I only see 2 at the max but there are "qualification" samples with a Q in the code... you never want one of those test samples.

Think of Sandy Bridge as rough and gritty while Ivy Bridge is soft and more sophisticated. Earlier vs later, worse vs better, slower vs faster. All the version 1 processors are known by their alphanumeric by itself, without a "v1" to follow that. However, for clarity many of us give them a v1. Then came the v2, v3, v4 officially added on to the names. The ZX20 workstations can never run a v3 or a v4. Some of the ZX20 workstations can run either a v1 or a v2 processor. The earlier v1 motherboards can only run a v1 processor. The ZX40 workstations can run v3 or v4 processors and never a v1 or a v2... no need to know the boot block dates with them. Confusing enough? Here's more, which I put together for our IT guys and gals because I got tired of explaining it over and over... a cut and paste:

For the Z820 motherboard the v2 versions that can run either the Sandy Bridge and Ivy Bridge (v1 and v2 processors, respectively) and on all v2 motherboards there is a HP bar code label that has a right-hand "assembly part number" of 618266-003 or -004. There also is a "Rev" alphanumeric just below that, for example "Rev 0L" or higher (0M, 0N, 0O and on up to 0W) for all v2 motherboards. The left-hand barcode "spares part number" has nothing to do with the v1 vs v2 motherboard issue. That simply tells you what the motherboard is "branded" for. That has to do with codes burned into the firmware that relate to money paid or not paid to Microsoft by HP for licensing of certain Windows OS versions, and maybe also to Intel for use of certain Intel software (RSTe, for example).  If not branded for Windows the motherboard was usually for Linux, or BYOL.

Boot Block Date (which you can only see if you boot into BIOS): In a ZX20 if it ends with 2011 it is a v1 motherboard and can only run specific Sandy Bridge processors. "2013" instead means it is a v2 motherboard and can run either specific Sandy Bridge or specific Ivy Bridge processors.

I've seen v2 motherboards with "PCB REV: 1.02" and also 1.03... and some v1 and some v2 motherboards with 1.02. So, that does not help much. 1.03 helps more but I don't use the printed REV number at all. I just use the right-hand part number ending in -003 or -004 to ensure it is a Z820 v2.

"E5-2680" equals "E5-2680 v1" for practical purposes. This has two sSPEC codes, and both are "Sandy Bridge".
SR0GY (the initial C1 stepping, max 1600 FSB)
SR0KH C2 (the later improved stepping, also max 1600 FSB)

"E5-2680 v2" 1 sSPEC code, a later better potentially faster "Ivy Bridge" processor:
SR1A6 (only 1 release, 1866 max FSB)

"E5-2680 v3" 1 sSPEC code: Note that v3 and v4 run in ZX40, not ZX20
SR1XP (only 1 release, 2133 max FSB)

"E5-2680 v4" 1 sSPEC code:
SR2N7 (only 1 release, 2400 max FSB)

[SDH]

"I was initially discouraged by -002 and-003 mobo prices, but a more careful search has turned up some that are only a little more money than the -001 that caught my eye for less than $100. The -001 is new, while the others are pre-owned, but it definitely deserves consideration for the reasons outlined."

No, don't buy that. The right-hand part numbers 618266-001 and -002 are both v1 Z820 motherboards. You only want a v2 motherboard and two fast v2 processors for sure. If you need to replace two E52680 v1 processors it will cost you about 25.00 to get two E5-2680 v2 processors, free shipping recycled via eBay. Look up SR1A6 and also E5-2680 v2 there. Look for price and shipping lowest, and "Top Rated Plus" US seller. I don't buy recycled processors from China unless there is a big price difference unless the seller is highly regarded and I'm not in a rush.

I've posted on recycled server memory here that works at the top speed the v2 ZX20 motherboards can ever run at, for about 10.00/8GB stick. The faster ZX40 workstations take a different type of memory.

A slow processor will slow down fast memory to its same slow speed, and slow memory will do the same to a fast processor.

Regarding your current motherboard: It sounds like a v2 but it can be confusing. Look at that bar code label on your motherboard.

Regarding the UPS event triggering oddness: You might have a partially corrupted BIOS (that does happen). It takes more than just pushing the motherboard's yellow button or even pulling the battery and putting it back in to fix that. You need to unplug from wall, fully drain all motherboard capacitors, literally unplug everything, and do a deep level reset of CMOS with the battery out. Some PCIe cards actually have batteries on them. Read up on how to do a deep level CMOS reset. I've even seen a plugged-in monitor hooked to an otherwise unplugged HP workstation light up a motherboard LED... even video cables can carry power. This deep reset process should also reset your BIOS settings to factory defaults and your time/date to something incorrect, which is good. That means you did it.

Regarding changing processors, Noctua original thermal paste is the only one I use.... nonconductive and spreads very thin and easily with a latex-gloved finger. I tried the second version and went back to the first.