HPE EC200A Xeon D-1518 'hybrid server' on ebay ~$150

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

Samir

Post Liker and Deal Hunter Extraordinaire!
Jul 21, 2017
3,215
1,407
113
49
HSV and SFO
I did this with an EC200a from Japan. Confirmed I can change the fan control settings, but it's unfortunate that the settings are not persistently saved.
But the Japan version had a different processor, didn't it?
 

DarthCirrus

Member
Dec 12, 2022
51
51
18
It’s amazing to see that 2 years ago these Xeon D-1518’s were available for $150 when today they are >$700 on eBay; and that’s just for one in a SM motherboard, not even a whole system..
 
  • Like
Reactions: Senses3 and Samir

zac1

Well-Known Member
Oct 1, 2022
432
357
63
It’s amazing to see that 2 years ago these Xeon D-1518’s were available for $150 when today they are >$700 on eBay; and that’s just for one in a SM motherboard, not even a whole system..
That comparison seems a bit apples and oranges... Maybe even bananas.
 
  • Haha
  • Like
Reactions: Aluminat and Samir

DarthCirrus

Member
Dec 12, 2022
51
51
18
That comparison seems a bit apples and oranges... Maybe even bananas.
You’re right - I don’t have an apples/apples comparison because there aren’t any of these HP systems available on eBay. I suppose without pcie expansion these are somewhat niche - but maybe I don’t understand how much of a discount these were sold at due to their limitations.

I just wouldn’t expect that the price on any Xeon D system would have that cheap 2 years ago if and to still have those chips be in demand and sold at a premium.

I think I’d certainly be happy paying $150 for one of these today given the prices for other low power routing solutions - even without the option to add a NIC…

So, poorly formed comment; I should have just said “wow can’t believe 2 years ago you could buy a 1518 for $150.”
 
  • Like
Reactions: Marsh and Samir
May 4, 2015
37
16
8
40
I ran mine until the clock drift caused issues since it was doing firewall duty. Battery replacement didn't solve. So I chucked into recycling. wasn't worth the time to troubleshoot it
 
  • Wow
Reactions: Samir

Samir

Post Liker and Deal Hunter Extraordinaire!
Jul 21, 2017
3,215
1,407
113
49
HSV and SFO
I ran mine until the clock drift caused issues since it was doing firewall duty. Battery replacement didn't solve. So I chucked into recycling. wasn't worth the time to troubleshoot it
Damn, still would have been a great 2-bay nas unit. :(
 

johndball

New Member
Nov 27, 2021
4
9
3
Yeah, I tried that last week before posting here. (via iSCSi on my FreeNAS box) I also got it installed, but I couldn't get it to boot via iSCSI. Are we best friends?

I picked up the $3.50 USB 3.1 32GB Microcenter thumb drives. Same thing.
I also figured this may just be a BIOS incompatibility. So I downgraded, and downgraded, and downgraded:
ILO 2.75 > 2.70
BIOS v2.66 > v2.60
BIOS v2.60 > v2.56
ILO 2.70 > 2.56
Turned off USB 3 support, thought I did that last week, booted once, did not boot subsequently.

One thing I noticed is that after a failed boot (Fatal Error 8), it will reboot after a minute or two with no intervention. On the subsequent reboot, the BIOS won't see my USB drive, rather it seems the Phison xxxx boot, which is not what is normally seen. I think this is a BIOS issue, not a ILO issue, not a VMware issue. I can't believe there aren't more people talking about this.

While I was at Microcenter, I also picked up 2x cheap MicroSD readers ($.99). Installing onto a 32GB Samsung PRO Endurance MicroSD. Maybe that will work.

Edit:

Narrator: It still didn't work. USB is garbage on this. I saw something at some point mention swapping the CMOS battery, I'll try anything at this point.
I know this post was back in 2020, but I just installed ESXi 7.0.3i February 2023 on the EC200a. Immediately ran into the USB boot issue using Sandisk drives. I have two EC200a systems so I was able to get creative with mix-n-match settings. Going into the BIOS, the settings that solved my perpetual boot issues after install were:
1) set the BIOS to reboot and try boot indefinitely if boot failure
2) remove unnecessary boot options (PXE boot for example) <-- mainly just to make USB boot a faster priority
3) disabled USB 3.0 support - it was set to auto by default
4) used the front USB 2.0 port - I tried every USB 3.0 even with support disabled and got the red error on ESXi boot - disabling USB 3.0 support and moving to a USB 2.0 port solved it

Happy to see my EC200a still chugging along with 2023's 7.0.x image :)
 

tammyTUX

New Member
Oct 25, 2023
2
4
3
Germany
Hello folks,
after some years I suspended my ES200a from serving as a labserver, but I want to give it a second life for using with OPNsense.
Therefore I wanted to expand it with the communication board, but this seems to be hardly to find, neither on Ebay nor somewhere else.
Maybe someone here has a communication board unused in the drawer and want to sell it to me?
 
  • Like
Reactions: Samir and Helzy

Helzy

Active Member
Oct 19, 2017
189
59
28
56
Hello folks,
after some years I suspended my ES200a from serving as a labserver, but I want to give it a second life for using with OPNsense.
Therefore I wanted to expand it with the communication board, but this seems to be hardly to find, neither on Ebay nor somewhere else.
Maybe someone here has a communication board unused in the drawer and want to sell it to me?
I"m in the same boat. Line me up as #2 behind @tammyTUX for a communication board if anyone has one or knows of a source thats not thru the roof expensive.
 
  • Like
Reactions: nophtosh and Samir

logan893

Member
Aug 12, 2016
67
12
8
44
Has anyone made further attempts at proper fan replacement mods?

I want to attempt a fan swap on my TM200 (Japanese unit). As is popular for various fan mods, a Noctua NF-A4x20 would be suitable, and I think even at default fan curves or worst case full power, maybe with a touch of manual speed control should do very well.

Thread history

I have seen only a few attempts at getting a fan swap to work. @duckduck identified the pinout a few years ago ( https://forums.servethehome.com/ind...8-hybrid-server-on-ebay-150.29619/post-281046 ) but didn't seem to have much luck otherwise with the fan error detection. They also report no luck with the PWM signal and a Noctua fan ( https://forums.servethehome.com/ind...8-hybrid-server-on-ebay-150.29619/post-297465 ).

@chereszabor had some more details ( https://forums.servethehome.com/ind...8-hybrid-server-on-ebay-150.29619/post-297460 ) but was also unsuccessful with bypassing the fan error detection.

Another attempt I saw in this thread was @Rubenel using the stock fan without blades mounted outside the case. ( https://forums.servethehome.com/ind...8-hybrid-server-on-ebay-150.29619/post-297898 )

Initial testing and promising results

When I open up my unit completely and remove the fan, I of course get the "fan failure" error when trying to boot.
Out of curiosity, and perhaps stupidity, I have concluded that if I short to Ground the fan header pin that sits next to Ground, I bypass the fan error and can boot without a fan. Current draw is roughly 0.67 mA when I short through my multimeter current measurement mode, so it seems safe enough. If I stop shorting the pin after boot then the system will detect a fan failure which after 60 seconds unless cleared triggers a shutdown.

I don't like the idea of a permanent short of PWM Signal to Ground, even though the current draw is low. I found that with a 2.2k pull-down resistor I get the same effect of bypassing the fan error detection with reduced current draw, at only 0.47 mA. With a 4.7k pull-down resistor the fan error is still triggered.

With server power disconnected, the PWM Signal pin measures at 4.85k Ohm resistance to ground, which is gone and measures infinite (above 200M on my multimeter) when power is connected to the server. Not sure if this is significant.

With either of the pull-down resistor or short to ground I get a positive current reading with negative lead to ground and positive lead to the PWM Signal pin; current flows from PWM pin to ground. When I connect a regular PC fan (I only have one 4-pin PWM fan readily available, a Scythe Kaze Flex Slim PWM 92mm) then I get a negative current measurement; current flows to PWM pin from fan. I also tried to pull down the PWM pin to ground while connecting the Scythe fan PWM - it was insufficient pull-down with 2.2k resistor so fan error is triggered, and there is no PWM control for the fan. Lowering the pull-down resistor to 1k again bypasses the fan error, but still no PWM control.

Voltage I can measure from the PWM pin when it's floating/disconnected is just below 3.3V, and at least my multimeter can't detect any change in voltage when left floating regardless of whether I set PWM to 15% or 100%. Not sure if this is relevant, and I don't have better equipment to check the signal nor do I know what to expect when it's otherwise disconnected.

Temperatures with and without stock fan

With the case open and booted into an idle proxmox (no VMs running), with the fan removed and disconnected and shorting the PWM pin to Ground on the motherboard, the system runs fine. At 23 C ambient running (idling) for 60+ minutes without active cooling, the CPU temperature has slowly climbed up to being stable around 80-81 C.

For reference, my system (Xeon D-1518, 2x32GB RAM and a pair of Samsung SM863 SSDs) with the fan mounted and blowing at stock fan curve, the CPU temperature at idle is roughly 51 C and fan speed stays low (roughly 15%). I have not done any repasting of thermal paste yet. Full fan speed (PWM min set to same as stock PWM max, which is 204) brings CPU idle temps down to roughly 38 C.

To check minimum viable cooling required I've performed a max CPU load test with "stress" in Linux ("stress --cpu 8 --io 8 --vm 8 --vm-bytes 6G") and the CPU keeps below 100 C (at 23 C ambient) with roughly 15% fan speed. Bodes well for a fan swap.

Interesting observations

With the fan connected I can force the fan to stop using the modified iLO where I have access to the "fan" command. I set max PWM to 1, fan stops, and system keeps chugging. So, the system has a kind of "zero RPM" mode support.

With the max fan PWM set to something low (e.g. "fan p 0 max 10", where 10 is the minimum for the fan to start in my testing), sufficient for the fan to spin at its lowest, and stopping the fan with my finger, the system detects fan the fan stopped and tries to re-start it after a few seconds. If I keep holding the fan blades to prevent it from starting then the system issues a shutdown after about 60 seconds.

Setting the max fan PWN to a value sufficiently low for the fan to stop on its own (1-7 in my tests) then the system doesn't care that the fan is stopped nor does it trigger any shutdown nor fan failure detection.

Conclusions

Fan speed via RPM / tach sense doesn't have any impact on the fan error detection.

The fan error detection system can be tricked with PWN signal pin pulled down (~2.2k or less) or even short to ground (might be risky).

Fan stopped due to low max fan PWM set from iLO is OK and will not cause a fan error or shutdown (but these settings are not persistent).

I still have no idea how to get the PWM signal to work with a traditional PC fan, but at least it's possible to completely remove and replace the OEM Delta fan.
 
  • Love
Reactions: Samir

logan893

Member
Aug 12, 2016
67
12
8
44
Couldn't shake the feeling something was wrong, so I did some more experimentation.

The PWM signal comes from the 3rd pin counting from ground, the pin next to the 12V power pin, corresponding to the blue cable to the fan. It is not the 2nd pin, next to ground, going to yellow fan cable as previously mentioned by others. My multimeter can detect a voltage between 3 V (PWM set to 1, close to 1%) and 0 V (PWM set to 255, 100%) depending on the PWM duty cycle I force via iLO. The max fan speed the system allows by default, PWM value 204 (max according to the default settings), corresponds to roughly 0.64 V as measured by my multimeter.

So, the PWM is inverted and this seems to also mean current flows from the PWM signal output into the fan, not sourced from the fan as is typical for normal PC PWM fans.

Since my Scythe PWM fan has a built-in pull-up resistor for PWM signal to around 4.85 V and expects a non-inverted PWM signal where 0V is 0%, I need to invert (un-invert?) the signal from the TM200 for it to be useful. I built a small circuit on a breadboard with an NPN BJT (2N3904), a 1k resistor between base and PWM signal, emitter to ground and collector to fan PWM input.

This works, and seems to give me the top and bottom fan speed range (can't easily confirm fully without checking the fan RPM). However, it is also a design that Noctua recommends to avoid, though I don't know if that really applies in this case. The Scythe fan has a maximum current to ground of roughly 0.277 mA, which is very low and perhaps as Noctua warns maybe it could cause a slow turn-on slope. It would be best to watch what happens using an oscilloscope, though I don't have one at hand. Looking at a generic datasheet for the 2N3904 and extrapolating turn-on time, if collector current is in this 1/4 mA range, the turn-on time is less than 0.4 microseconds, which is 1% of a 25kHz PWM duty cycle. Plenty fast enough, if that is the case.

I also don't know for sure what the PWM frequency is out of the system. Hopefully it's the typical 25kHz.


As for fan failure detection, since the 3rd pin is PWM and by pulling 2nd pin low there is no fan fault triggered, it may simply be that there is no RPM / tach sensor or reporting from the fan. Just a simple "Rotation Detection O/P (R00)".


Next I'll source some Molex PicoBlade cables so I don't have to butcher the original fan or modify the board, and consider whether I need to make a more elaborate circuit to harness the PWM output from the board, or go with some manual control.

Looks promising for a fan replacement.
 
Last edited:

logan893

Member
Aug 12, 2016
67
12
8
44
I went a little deeper with the tach sense.

Most interesting is the stock Delta fan's behavior.

The amount of current flowing from the board's tach sense pin into the fan tach wire when the fan is in a "fan OK" state is... *drum roll* right at the 0.67mA level! It's a simple short to ground!

When the fan is in a "fan failure" mode, the tach wire is left floating and gets pulled high by the sense side. The voltage I measure is just below 3.3V so they seem to be using a ~4.7k Ohm pull-up resistor to end up with this current sink.

Summary of the fan behavior I observed

The fan may have some logic built-in to allow for the behavior observed.

The fan supports a zero-RPM mode. When the fan is off due to a very low PWM duty cycle (or steady voltage applied to PWM input at 3V or more), then it is not triggering the fan failure. Tach wire remains an open collector, essentially shorted to ground.

When I disconnect the PWM input to let the fan run freely, the current sink drops slightly, down into the 0.5mA range when at full speed. This may indicate some form of RPM measurement is still possible, but I can't confirm without an oscilloscope.

When I keep the fan stopped with my finger while it should be spinning, it will try to start every few seconds.

When the fan fails to start while it should be spinning (i.e. not a zero-RPM level PWM signal), the fan will detect that its blades are blocked and the tach line will be left floating, triggering the "fan failure" mode. Even after letting the fan start to spin freely again, the tach line will be left floating for a little while, seemingly until the fan reaches its desired speed (e.g. it has to ramp up to its max fan speed if PWM is set to 100%).

Regular PC fan

Hooking up my Scythe fan's tach wire directly to the sense pin, the HPE server will fluctuate between "fan OK" and "fan fail" every few seconds. So, a direct connection doesn't fully work, and might not even get past the boot check (I didn't try this yet).

I suspect this is because the tach sense wire of a regular fan will have a 50% duty cycle. Not only does it only sink roughly 50% of the current compared with the Delta fan, but it would also leave the pin floating or pulled high half of the time. I tried adding a counter pull-down resistor to ground to partially sink the current from the board's tach sense pin, yet this was insufficient to prevent the continuous toggling between "fan OK" and "fan failure" state. Also, any sufficiently low resistance value would possibly cause the "fan OK" state on its own if the fan's tach is floating, and so is not useful.

If "fan failure" detection is wanted, not simply bypassed, then some kind of translation circuit has to be built. I won't attempt this just yet, not even sure if I have all the necessary components to make one.
 
Last edited:
  • Like
Reactions: Samir

Samir

Post Liker and Deal Hunter Extraordinaire!
Jul 21, 2017
3,215
1,407
113
49
HSV and SFO
Found one that ships to NA

Found 4x down under, so shipping is expensive to the US:

Another 4x with drives down under:
 

Helzy

Active Member
Oct 19, 2017
189
59
28
56
I have one sitting about 3 feet from me running 24/7. It has a hum, but nothing terrible. If you had one set up near a TV for instance I doubt very much you would hear it.