EXPIRED Hyve "S-SKUD 2" - 1U, 8x LFF SATA6, 2x LGA 3647 Skylake-SP, 2x PCIe x16, 2x OCP PCIe x8 w/2x 25Gbps Connectx-4 installed - $149 + Shipping

Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.

Jaket

Active Member
Jan 4, 2017
230
118
43
Seattle, New York
purevoltage.com
For those that missed this deal, here's a similar setup for a standard Supermicro (or similar) 1U motherboard build I've had in my watchlist for quite some time (waiting for a price drop):


Doesn't quite qualify for a "Great Deal", not until they lower the prices more as the S-SKUD2 was $280 + tax once you figured in price + shipping, and you got 2x PCIe slots and mobo to boot.

Maybe someone can shoot off a few Offers for a discount on a couple, and report back?

EDIT: There's also this one I've had in my watchlist for a while, waiting on a discount...

Problem with both of these systems is they don't typically fit most racks. You need a extra deep rack to fit them if you plan on using rails. We have racks full of the Quanta ones which required some tweaks and what not. They are also a bit wider than the typical server as well the rails are basically just shelves which cause issues as universal shelves will cause the rack to bow outwards as well once you put around 18 systems into it.

Great systems otherwise.
 

eduncan911

The New James Dean
Jul 27, 2015
648
506
93
eduncan911.com
Update on power usage. So, sticking in a 2nd CPU and another 6x sticks of ram jumps the power usage by 30 watts!

CPU, Memory, Dual-SPF28Add-on CardsPower Usage @ Desktop Idle
1x 3106 CPU, 6x 2133Mhz, Dual SFP28 OCPNone53 W
2x 3106 CPU, 12x 2133Mhz, Dual SFP28 OCPNone84.7 W
2x 3106 CPU, 12x 2133Mhz, Dual SFP28 OCPIntel 905P, M.290.2 W
2x 3106 CPU, 12x 2133Mhz, Dual SFP28 OCPIntel 905P, M.2, SAS2 OCP97.0 W

All power settings below were taken 3 times via my original Kill-a-Watt and the average-ish taken.

Setup:

- BIOS set to defaults, BMC set to defaults.
- Fans set the Default (Quiet) mode
- Bronze 3106 1.7 Ghz (8C/8T)
- 4 GB 2133Mhz DIMMs
- Mellanox dual SFP28 installed, but no transceivers.
- Single PSU installed, 2nd one removed.
- Nothing else connected.

Measurements were taken with Ubuntu 22.04 LTS Desktop at idle with default power savings.

BMC (machine powered off): 9.3 W (idle), 16 W +peak first power on).
POST (no OS C-State power saving): 63 W (idle)
 

EasyRhino

Well-Known Member
Aug 6, 2019
483
344
63
all things considered that's good idle usage.
my old e5 2680 v4 system idles at 77W with only one cpu, four dimms, and no expansion drives or cards.
 
  • Like
Reactions: eduncan911

Markess

Well-Known Member
May 19, 2018
1,135
754
113
Northern California
all things considered that's good idle usage.
my old e5 2680 v4 system idles at 77W with only one cpu, four dimms, and no expansion drives or cards.
Not to crap the thread, but that seems high to me if you've got nothing attached/installed. Do you have a chassis with multiple PSUs or huge/fast fans, or something else that pulls a lot of power? One of my NAS started out with an E5-2680 v4 and it idled at a little over 30 watts on TrueNAS with 2xRDIMM, and 13 SSDs. Recently swapped the 2680 for a 2630v4 and idle is down to ~27 watts now.

@eduncan911 , nice write ups on your build and power measurements! Thanks for sharing.
 

eduncan911

The New James Dean
Jul 27, 2015
648
506
93
eduncan911.com
Not to crap the thread, but that seems high to me if you've got nothing attached/installed. Do you have a chassis with multiple PSUs or huge/fast fans, or something else that pulls a lot of power? One of my NAS started out with an E5-2680 v4 and it idled at a little over 30 watts on TrueNAS with 2xRDIMM, and 13 SSDs. Recently swapped the 2680 for a 2630v4 and idle is down to ~27 watts now.
How you measure wattage can be vastly different. The BMC reporting from the onboard PDU of your dual PSUs does not equal actual outlet power usage. Same with Windows software measuring the package usage. Also, UPS with displays that can show watts in real time is also a joke.

I measure at outlet with a Kill-a-Watt meter for the only true way to measure.

This method is always a higher number than ANYTHING else that registers a power reading. Everyone has to remember that: the actual power draw at the wall will be greater as it accounts for power supply in-efficiency at less-optium power draw (idle with no load).

You also have to remember there's a dual-port SFP28 PCIe OCP card (9W) and BMC (9W) using more power as well - that registers at the wall with 18W.

So, comparing Power Usage is a misnomer, unless you state the testing methods and location and all onboard components in use (so they can be disabled).

Therefore, I am stating my test method and readings for others to consume.

--

Edit: as a comparison, I just measured my E5 V3 bare bone server just now. The Supermicro PDU says the system is BIOS idle (no OS power savings) at 43 W. However, the Kill-a-Watt shows 61 W at the wall.
 
Last edited:

RolloZ170

Well-Known Member
Apr 24, 2016
4,956
1,448
113
I measure at outlet with a Kill-a-Watt meter for the only true way to measure
sure, if you want to know what to pay for electricity.
but if you want to know the power draw of the components its untrue.
the difference is also the AC voltage, 115/120 or 230V ?
range of efficiency of most PSU 70..96% depends on load range.
the component load can be from 70 to 90 watt at the same wattage of 100W at the wall on different PSUs.
a valid comparism is imho only possible if both systems use same PSU & AC voltage.
 

Markess

Well-Known Member
May 19, 2018
1,135
754
113
Northern California
Not to crap the thread, but that seems high to me if you've got nothing attached/installed. Do you have a chassis with multiple PSUs or huge/fast fans, or something else that pulls a lot of power? One of my NAS started out with an E5-2680 v4 and it idled at a little over 30 watts on TrueNAS with 2xRDIMM, and 13 SSDs. Recently swapped the 2680 for a 2630v4 and idle is down to ~27 watts now.

@eduncan911 , nice write ups on your build and power measurements! Thanks for sharing.

I guess I was pretty ambiguous with that post! :oops: I was actually trying to ask about @EasyRhino 's idle consumption, as both of my E5 v4 systems idle a decent amount lower (mine measured at the wall with a Kill-a-Watt).

I know next to nothing about LGA 3647, so wasn't meaning my post to be a comment on @eduncan911 's findings, other than to appreciate the approach and how thorough and organized they seemed to be.

Sorry for any confustion
 
  • Haha
Reactions: eduncan911

EasyRhino

Well-Known Member
Aug 6, 2019
483
344
63
Interesting. My motherboard or bios might be a power hog. It's a super micro x10dal-i so it doesn't even have a BMC.

Also your nas might be unusually efficient.

I remembering looking at my older e5 2637 v3 in hwinfo in windows. It didn't idle under 20 or 25w for just the CPU itself.
 
  • Like
Reactions: eduncan911

eduncan911

The New James Dean
Jul 27, 2015
648
506
93
eduncan911.com
Almost there (with 10x LFF HDDs).

Will pull the 2nd CPU and extra 6x DIMMs before its finished, and move to a pair of SSDs.

Overtime, I'll rotate in more SATA6-based HDDs - so I can remove the SAS2 OCP card for a very low-powered, disk-suspended backup server. Not bad overall.
 

Attachments

EasyRhino

Well-Known Member
Aug 6, 2019
483
344
63
One of my NAS started out with an E5-2680 v4 and it idled at a little over 30 watts on TrueNAS with 2xRDIMM, and 13 SSDs. Recently swapped the 2680 for a 2630v4 and idle is down to ~27 watts now.
So, to continue on this tangent, I have determined that your power usage is freakishly low.

I swapped motherboards on my system to a board that uses an i7 7700K. When nothing is plugged in it idles at under 20 watts. However, when I add a HBA, a 10G Nick, 4x 3.5 drives and 4x ssds, then the idle usage is closer to 80 watts in ESXi with the VMs booted but mostly idle.
 

eduncan911

The New James Dean
Jul 27, 2015
648
506
93
eduncan911.com
So, to continue on this tangent, I have determined that your power usage is freakishly low.
Supermicro PDUs (from Supermicro chassis) always gives "lower-than-actual" results, at anything less than 100W. the PDUs, all versions, all tend to be not accurate at all at low power usage. My sources for that statement are the various 7x Supermicro chassis's I've had, and have currently, and their many different PDUs - all against my Kill-a-Watt meter. The PDUs are always running lower - and at < 100 Watts it reads WAAAY lower.

And as I mentioned above, unless you are measuring at the wall with the same device that you are comparing too (e.g. Kill-a-Watt), then "comparing power usage" is just hot air.
 

Markess

Well-Known Member
May 19, 2018
1,135
754
113
Northern California
So, to continue on this tangent, I have determined that your power usage is freakishly low.
Freakish...that's me. Frankly, I was shocked it was that low. Helps that the Fujitsu motherboard doesn't have IPMI. Drives are SATA (Samsung PM843 & Micron 5100) & Intel Series 6 m.2 NVMe, so no high power draw stuff there either.

And as I mentioned above, unless you are measuring at the wall with the same device that you are comparing too (e.g. Kill-a-Watt), then "comparing power usage" is just hot air.
I use a cheap Kill-a-Watt EZ, and usually don't bother with PDU info other than as a historical comparison (i.e. "Gee, this machine is drawing half what it normally does...I wonder if something's wrong?").

I'd assume that if others post values they got using a Kill-a-Watt, and they are on the same mains voltage, then the measurements should be somewhat comparable??
 

RolloZ170

Well-Known Member
Apr 24, 2016
4,956
1,448
113
I'd assume that if others post values they got using a Kill-a-Watt, and they are on the same mains voltage, then the measurements should be somewhat comparable??
range of efficiency of most PSU 70..96% depends on current load. the required efficiency at 10% load and under must not be tested&granted to get 80 plus(except titanium). this goes down to 70% at some PSU circuit platforms.
the component load can be from 70 to 90 watt at the same wattage of 100W on the wall on different PSUs.
a valid comparism is imho only possible if both systems use same PSU & AC voltage.
Kill-a-Watt's have measure tolerances, use the same KaW-device to remove that issue.
 

EasyRhino

Well-Known Member
Aug 6, 2019
483
344
63
Freakish...that's me. Frankly, I was shocked it was that low. Helps that the Fujitsu motherboard doesn't have IPMI. Drives are SATA (Samsung PM843 & Micron 5100) & Intel Series 6 m.2 NVMe, so no high power draw stuff there either.
Still impressive/freakish because your whole system idle usage is around what just my 2680v4 idled at by itself.

Incidentally my numbers are from a tp link kasa smart plug. Don't know exactly how precise they are but the numbers check out with other known things like an air conditioner or my desktop computer.
 

Cruzader

Well-Known Member
Jan 1, 2021
525
528
93
range of efficiency of most PSU 70..96% depends on current load. the required efficiency at 10% load and under must not be tested&granted to get 80 plus(except titanium). this goes down to 70% at some PSU circuit platforms.
the component load can be from 70 to 90 watt at the same wattage of 100W on the wall on different PSUs.
a valid comparism is imho only possible if both systems use same PSU & AC voltage.
Kill-a-Watt's have measure tolerances, use the same KaW-device to remove that issue.
Efficiency and psu design definatly comes into play in the low wattages.

My old silent hosts was whiteboxes with single socket x10 supermicro and e5-2683 v4 when retired.
With just mobo/cpu/ram/120ssd for esxi i got results from 32w to 55w just by using various old psus i had sitting.
(measured on cable from wall with a fluke and same spec other than psu and over same timeframe after booted into esxi and idle)

And as for the reliability of hosts own reading i have a DL380 Gen9 with 12 spinners that say its using 19w :D
 

Cruzader

Well-Known Member
Jan 1, 2021
525
528
93
the PSU reads DC durrent (sec.side) and that must not be wrong.
As for my example it estimates it from spec and loads, add a few non-hp parts and it goes haywire with comedic results.
(thats the explanation ive gotten from HP previously)
At mid/high loads of just parts intended for the gen its fairly accurate tho.
 

nevergiveyouup

New Member
Dec 19, 2022
3
5
3
PiKVM until the BMC KVM is resolved.
Hello, We have solved the KVM/VMEDIA problem.
Please send it to anyone who have this problem. Thanks.
!!! NEVER UPDATE YOUR BMC !!!
1.SSH to bmc, with uname/pwd = sysadmin/superuser , If password is not correect, change it on web, user 0(the user before root, o matter what the user name displays , just change it and in ssh use sysadmin)
2. cd /conf/license
3. ls then you will see 2 files, count_days and status
4.Fill count_days with (quote not include) "1" ,you can use vi command to do this
5.Fill Status with 3 lines below:
KVM|30|2|ff:ff:ff:ff:ff:ff
MEDIA|30|2|ff:ff:ff:ff:ff:ff
RMEDIA|30|2|ff:ff:ff:ff:ff:ff
6. Now, exit ssh, login web console, you will see license is reset to 30 days. If it expires again, repeate all operates above.

Because of general superuser account exists, it is not recommand to expose your BMC on Internet.

FOR PEOPLE WHO FACING THIS PROBLEM ON SEARCH ENGINE:
hyve bmc ipmi license trial reset kvm vmedia
 
Last edited:

darkbane0520

New Member
Jun 12, 2023
5
2
3
So I have one of these, and it wasn't fun getting it running. I do not have an updated firmware for it, but I don't think it's needed. I had to reset the bios, and change the ACPI settings as it wasn't wanting to accept any other hardware adapters I tried to use (but had no issue with usb!).

Resetting the BMC via ipmitool is easy enough once you can boot to a live linux usb stick- and if you're really feeling adventurous you can get into the ami web bios with administrator/superuser (may have to mess with capitalization a bit), but it's accessible from the bmc screen once you're there.

I have had zero luck with ES processors, nor anything above 165TDP, it just doesn't handle it well.

The OCP v2 slots are easy enough to deal with once you get past that pin lock setup, and it does handle NVME without issue, but again only after you fix the ACPI headaches. The two onboard 25Gbe adapters work just fine, and I'm using 10gb twinax I picked up off of ebay with an arista 7124sx- totally doesn't care, but you will want to update the firmware on the mellanox card because it is old. The driver package off of nvidia handles it for you, but only under a windows os.

I gave in an installed win2019 just because I was lazy, but it will get wiped and become a member of a ceph cluster soon (tm)

I've attached pictures of the BMC interface, as well as the configuration with a pair of xeon 5120's, and a pic with 4 sas disks installed- if you haven't already figured it out, right angle pcie adapters are your friend with this box.

It's not terribly loud, but then I'm used to working in a datacenter, and I used to sleep in my lab without problem, YMMV
Did your BMC interface just look like that by default? I have a newer firmware version yet mine is still some old early 2000's hyve OEM bmc that i cant even do anything on. Also has an unusable java based KVM. looks nothing like the nice HTML5 GUI you have. I see user sl12 also has the same BMC as you and i just have no idea why mine isnt the nice looking one. I also have super limited ipmitool raw commands to the point where theres no reason to even use it. I attached a picture of my BMC
 

Attachments

  • Like
Reactions: Samir