Running a Xeon E5-2687W in a "135W" rated motherboard

[danwood82]

I keep coming across this issue - the only motherboards that support the Xeon E5-2687W are officially designated "workstation" motherboards. All the rest are arbitrarily limited to the more expensive and slightly slower 135W E5-2690.

I've found the ideal Supermicro board for my needs, the X9DRH-7TF, with an LSI 2208 controller and dual onboard 10GbE ports, but it's a "server" board, so it only supports the 2690.
They also do the X9DAX-7TF, which is exactly what I'm after, but annoyingly only comes in an EE-ATX factor, while the enclosure I want to use only supports E-ATX.

Supermicro won't tell me, as it's outside the bounds of official support, so I wonder if anyone here might know from experience. What would happen if I tried to run 2x 2687W chips in a board which officially supports only 135W CPUs?
Would it not be able to supply enough power? Would the firmware not recognize the chips? Would I void warranty or something?
It seems like a really odd arbitrary limitation to slap on one single CPU out of an entire range.

 

[renderfarmer]

Funny you should bring this up. I got a pair of E5-2687W CPUs off of ebay yesterday and I was wondering if any of my supermicro servers would take them to test while I waited for the Dell workstation I ordered.

I decided against it for fear that the power control and supply systems on teh board (all of those MOSFETs and Capacitors around the CPUs) might get damaged.

 

[awedio]

Funny you should bring this up. I got a pair of E5-2687W CPUs off of ebay yesterday and I was wondering if any of my supermicro servers would take them to test while I waited for the Dell workstation I ordered.

I decided against it for fear that the power control and supply systems on teh board (all of those MOSFETs and Capacitors around the CPUs) might get damaged.

render, were they ES's?

 

[renderfarmer]

render, were they ES's?

Not this time. I looked at ES-2687Ws but they start off hobbled at 3GHz instead of the retail 3.1GHz. The ones I bought are OEM pulls.

 

[danwood82]

I almost risked trying out some ES chips a while back, but I read way too many stories of stability issues with the 3.0GHz 2687W ESs. Sounds like they're very much discounted by all release-version motherboard firmwares, and you're well and truly on your own if anything doesn't work right.

Supermicro actually did get back to me in the end... although I'm still trying to pin it down. At first they told me the 2687W was designed for workstation boards, and recommended some... later they've vaguely suggested that the X9DRH-7TF "might have the thermal issue with Intel 2687W".

If I can actually get them to confirm for certain that the only limitation is thermal, and definitely not electronic or firmware, then I'll order the board. I'm pretty sure my planned build can handle the heat.

 

[nitrobass24]

FWIW I used Dual W5580s in a plain Supermicro Server Board without any issue.

Sure the max TDP is higher, but the voltages are the same.

 

[renderfarmer]

Supermicro actually did get back to me in the end... although I'm still trying to pin it down. At first they told me the 2687W was designed for workstation boards, and recommended some... later they've vaguely suggested that the X9DRH-7TF "might have the thermal issue with Intel 2687W".

Ask them if the CPU would have a thermal issue or the motherboard. Because if it's eeh CPU then what motherboard you have them plugged into has nothing to do with it. You could have 9lbs vapor chamber copper heatsinks on this thing for all they know. I'm inclined to believe that if they have any hesitation it's because they may end up having to RMA your mobo and that costs them money.

 

[renderfarmer]

FWIW I used Dual W5580s in a plain Supermicro Server Board without any issue.

Sure the max TDP is higher, but the voltages are the same.

Any commercially available product will have a healthy safety margin built in. The difference between 150W and 135W is 11%. Since teh Voltage remains teh same then that's all amperage. Maybe there are some EE's here that can shed some light on how much of an impact 11% more amps would be on a mobo?

 

[TheBay]

All depends on what VRMs are used on the board, might be fine for a bit but put that under 100% load and the VRM's may not have 11% tolerance, not to mention other components, then there is the thermal factor but that's the least of your worries. Plus server boards seldom use cooling on any of the power mosfets etc.

 

[renderfarmer]

All depends on what VRMs are used on the board, might be fine for a bit but put that under 100% load and the VRM's may not have 11% tolerance, not to mention other components, then there is the thermal factor but that's the least of your worries. Plus server boards seldom use cooling on any of the power mosfets etc.

That's what I figured. Thanks for the info.

 

[danwood82]

Well, I actually did get a good, direct response from Supermicro in the end:
"BIOS will recognize the 150W 2687W CPU. Problem is VRMs are only rated to run with CPUs up to 135W. It’ll probably boot initially and then under high load you will see issues."

I looked at the board layout photos, and it does indeed appear that 150W-rated boards have 7 VRM/Diode pairs while the 135W-rated server boards have only 6. So... it looks like the workstation boards probably have ~14% extra headroom.

So yup, TheBay, you were bang on the money. Seems it is best to keep the 2687W clear of the wimpy server boards!

(disclaimer, I know they're anything but wimpy )

 

[FlashG]

rated for 150W

This supermicro board is rated for the 150W 2687w v2:

Supermicro | Products | Motherboards | Xeon® Boards | X9DAX-7TF

Enjoy!!

FlashG