MicroServer Gen10+ V2: unbearably noisy fans and extremely low write speeds on VROC RAID5

[WiFriendly]

Hello everyone.

I continue to experience problems with the Microserver Gen10 Plus V2. The problem with BSODs when using an NVME disk was solved by installing the HPE NS-204i-p board instead of Lenovo or QNAP, after which the server stopped causing problems with the NVME disk (BSODs, spontaneous reboots, error messages in the BIOS log, etc. disappeared).

But new problems appeared:

1. The fans spin in a 75-100% cycle, and this cannot be changed in any way. 1-2 minutes after startup, the server begins to roar like an airplane taking off; changing the power profiles and cooling profiles does not help. Waiting does not help—the server runs for weeks on a 100% fan cycle with no load at all.

Most likely, the reason is the installation of either the NS-204i-p expansion board or 4xWD Purple HDDs.

2. Extremely low write speed to HDDs, from 0.5 to 20 MB/s (see the Benhmark picture below), which is 100 (!) times lower than typical values. It is impossible to work with the server; writing small files freezes it, and file transfer over the network is interrupted (while the CPU, memory, and network load do not exceed 1-4%).

The HDDs are used as a RAID 5 array (built on VROC in BIOS). The array status in the Intel application is healthy (Normal), and array initialization is 100% complete (BTW, it took 8 days for the 4x12TB HDDs). Windows uses the SATA driver cp060020, applied via F5 on the OS installation stage. cp063622 and cp063620 are also installed just in case.

Since reading from disks works as it should and is quite fast, it is logical to assume that the problem is in the write cache, or rather, in its absence. But it is impossible to enable cache in write mode for this device, Intel Disk Drives -> RAID Volume.

The checkbox does not allow itself to be checked, citing the fact that this mode is not supported.

An attempt to enable the checkbox via the registry or the Microsoft Dskcache.exe utility does not give anything (the mode cannot be enabled). The registry entries are returned to the default values, and the utility only enables Power Protect:

dskcache.exe +p +w d:

New settings:
Write Cache is disabled.
Power Protect is enabled.

I am sharing a report on the RAID array from the Intel Virtual RAID utility.

General data:

1. Windows Server 2022 OS, fresh installation.
2. The server is not loaded (zero third-party applications, zero users, and CPU and network load do not exceed 1-4%)
3. Room ambient temperature is +17 degrees Celsius (63 F).
4. All latest BIOS, latest iLO, and latest service pack P76090_001_gen10spp-2024.11.00.00-SPP2024110000.2024_1113.16 are installed.
5. And here you will find the report of the program HWinfo in the part that may be important for the questions asked. Excessive information has been discarded. 

And the question is: does anyone have any ideas on the essence of any of the above problems?

[support_s]

System recommended content:

1. Gen10 Service Pack for ProLiant 2021.10.1 - Component Release Notes

2. Component Release Notes for Gen10 Service Pack for ProLiant, v2023.03.00.00

Please click on "Thumbs Up/Kudo" icon to give a "Kudo".

Thank you for being a HPE valuable community member.

[WiFriendly]

Here you can see a benchmark of an old home PC, which is over 10 years old, against a Microserver Gen10+ V2 from 2024. As you can see from the pictures, writing to the disk array is almost inoperable.

10-year-old home PC:

Brand new MS Gen10+ V2:

[WiFriendly]

And here are some more pictures from iLo:

[WiFriendly]

More observations.

The Enhanced CPU Cooling profile looks like the most acceptable. The system starts with a fan cycle of about 20%, then gradually increases at a rate of approximately 1 percent per minute, and only then begins to jump sharply. Then, when the speed reaches about 80–90%, the increase slows down. It takes a couple of hours to reach 99%. And the whole day or something to get 100%.

The trick with fan control via the iLO API does not work after about 50%. of fan speed If the fan spins 20-40%, then sending the FanPercentAdjust=50 helps, and the cooler literally goes silent. But if there was a jump of 50% or something like this, then sending this parameter does not change anything at all, not even by one percent.

Even in this case, we have the roar of the aircraft turbines, just not 5 minutes after pressing the power button, but 40-60 minutes later. And using the server becomes impossible even in the office—it can be heard through a closed interior door.

[Rama2]

The HPE NS-204i-p Board is not recommended for this server based on its configuration, which may be causing the loud fan noise issue.

For more details, please refer to Page 5 of the QuickSpecs document:
HPE ProLiant MicroServer Gen10 Plus v2 QuickSpecs

Kindly raise a support ticket with HPE for additional details 

[WiFriendly]

Dear Sir,

I appreciate your attention to this matter, but could you clarify why only the optional expansion card from HPE behaves in this manner? Meanwhile, similar cards from QNAP, Lenovo, and Dell operate silently. It appears that this design choice intentionally prevents your server from functioning optimally with an HPE expansion card installed.

This gives the impression that HPE’s marketing strategy deliberately creates inconvenience for customers, potentially pushing them towards higher-end equipment. Such an approach seems to constitute a deliberate act of impairing the product’s features and usability.

Could you please explain this reasoning?

Also, could you clarify which ticket you are referring to? There has been no response, no resolution to the issues raised. The information provided so far feels dismissive and gives the impression of support in name only, rather than actual assistance.

Would you consider updating the firmware to address the issue with your reactive fan? Perhaps implementing a tangible solution to resolve the situation, rather than continuously referencing quick specs, which has been your approach for the past two months.

What exactly is preventing you from taking meaningful action, other than stringing the customer along?

That said, I understand it’s unlikely that you intend to take any real steps. It appears far more cost-effective to simulate support by merely pasting links to documents on the HPE website.

I have no illusions. Most likely, you will never take any action to help.

[Rama2]

We appreciate your feedback regarding the HPE card. To clarify, if the server is being used with third-party or unsupported cards, compatibility and performance cannot be guaranteed, as testing and optimization are conducted exclusively for HPE-certified components listed in the QuickSpecs.

We also acknowledge your comments about the ticket resolution process and regret if our previous responses did not meet your expectations. Our goal is always to provide helpful and actionable support.

Regarding the HPE NS-204i-p Board, it is not recommended for this server configuration. Since it has not been tested or validated by the HPE engineering team, we are unable to provide firmware updates or alternative solutions for this component.