Z620 Workstation - Static noise

[tomstephens89]

I should probably say, that I am an avid Music enthusiast / audiophile and have owned every high end sound card from creative and ASUS in the past 10 years. Both internal and external, external is ALWAYS better anyway. I'll never reccomend an internal sound card anymore but thats a seperate discussion.

I use a Cambridge Audio 851D connected via USB as my 'sound card' nowadays. Ground on the data line is lifted, plus filtered with an Audioquest Jitterbug.

This whine noise from the Z620 is not caused by electrical interference. It is most certainly coil whine as the choke coils on the board are resonating at a high frequency. This will explain why the noise changes/starts/stops as power consumption/demand changes when the system is idle or used.

I will post back after I have dismantled the machine and investigated. My goal is to locate the coils and cover them in hot glue.

[JderulWoodsend]

Hi Tom ! I am waiting for your results more intense than I wait for Santa Claus.

I wish you could confirm the hot-glue technique fixed the coil chokes noise.

I brought my Z620 to any certified/uncertified HP service and I am 80-90% sure that the glue in the chokes is a patented problem of these workstations.

[MagikSeven]

I found a solution!!!!

Ok so bare with me. I also have a HP z620 with xeon e5 2665. I too was hearing that god awful noise. Almost like a dot matrix printer. When i researched it a bit I found it was'nt limited to HP. Alot of people with X99 boards and even nahelem had the problem too and a few people thought it was related to how windows control the c2, c3 states. 

Things about the problem that I found were:

1. Problem was occuring when within windows when CPU wasn't doin much. i.e surfing the net or something not too demanding.

2. Problem was not occuring when processor was being heavily stressed. ie a heavy workload where all cores were being put to use. 

3. The problem didn't seem to occur in bios and the bios isn't really using any itensive CPU power. 

Now that would make me think that it was in fact the way Windows was controlling the power states on the cores when some cores were not being used. I do still think it is a coil whine related noise. I think it might be the way the hardware on a deeper level passes power to the cores and then windows might attempt to disable the cores and put them into idle and the excess energy is being dissipated causing resonance on the coils. But I'm no expert so who knows. 

But i did find a fix and this is it:

1. Open regedit

2. Locate HKEY_LOCAL_MACHINE SYSTEM\ CurrentControlSet\Control\ Power \PowerSettings \54533251-82be-4824-96c1-47b60b740d00\ 5d76a2ca-e8c0-402f-a133-2158492d58ad

3. Change Attributes from 1 to 0

- This will add a new entry into the power schema in your power options under control panel=> power options. 

4. In Windows’ Power Settings (I have mine set to maximum performance, change the advanced settings of your plan, find Processor power management -> Processor idle disable, and set it to Disable Idle. 

Bingo noise gone, without even needing to do a restart!! Woohoo!!!

[MagikSeven]

[edit]

[MagikSeven]

Perhaps I spoke too soon.

The above solution did in fact stop the coilwhine / dot matrix noise immediately.

However by disabling processor idle in the control panel power scheme task manager is reporting 100% utilisation across all 8 cores when little to nothing is being processed. This in turn means the cores are running at the max turbo frequency on all 8 cores of 2.79GHz. The max turbo frequency when 4 cores or 2 cores are in use is higher. 

So basically my turbo boost is limited to the max boost of all 8 cores and being that it is saying utilisation is 100% across the cores i must be using up more power. Some benchmarks I've run also report lower CPU scores such as 3dmark. 

Hmm... if someone can figure out how to disable a particular c state that solves the problem yet does not impact the fore mentioned issues. Pls let us know!

edit: I've checked aida64 and even though task manager reports 100% utilisation across all 8 cores aida64 shows the correct CPU usage where it's sitting 4-5% CPU usage while im typing this, but the temparture is another issue. With CPU idle disabled temps go up about 10-15 degrees almost instantly and then down when turned off. I was told that power usage would only be dertermined by the usage of the CPU as it will only pull power when it's processing and that if the cores  werent in actual use then it  wouldnt be using too much power. But if the temps are going up instantly upon disabling idle it's definately pumping a fair amount of power in to the cores. 

[JderulWoodsend]

Hi Magik!  Thanks for digging for some solutions. I appreciate you further tested&feedbacked us on it.

Your theory that Windows does not properly manage the excess of electric energy and that is being badly dissipated onto the hardware parts does get closer to the problem I think.

I do not know whether it does prematurely wear out the hardware parts or should we go on with it(as if it is safe or not) ?

[MagikSeven]

Hey so I found another solution (sort of). I downloaded throttlestop and unchecked C1E.  This takes away the coilwhine but unlike disabling core idle the temps dont go up significantly and CPU usage is windows task manager is accurate. The draw back here though is that the maximum core speed will be the max turbo speed achievable on all cores (which is a little lower usully than the max turbo speed, i think it will be 3.4GHz on e5 2690 rather than the max of 3.8GHz).

I think it may be able to disable a specific Cstate too but i haven't explored throttlestop yet.  

==> https://www.techpowerup.com/download/techpowerup-throttlestop/

[DGroves]

while power chokes are the most common source of noise,ceramic capacators can also  produce noise, along with surface mount caps.

you can use a straw, and some aquarium air line (with one end held to your ear) to try to pinpoint the source of the noise

a bit of hot glue on the part will usuall quiet down the noise in most cases, but if the frequency of the noise is right it can propagate into the traces/circuts on the motherboard and removing this source is very hard to do

[MagikSeven]

A long time ago there was a nahlem issue (symptoms being CPU coil whine - though it might not actually be  the coils resonating) that microsoft submitted a fix for:

they said : 

"This issue occurs because of an error in the firmware implementation of the ACPI C-states (processor idle states).  This firmware incorrectly maps the "core C3" processor state into the ACPI C2 state.  Therefore, when the system enters C2, the platform goes to C3. This behavior causes the local APIC timer (LAPIC timer) of the processor to stop counting and to stop delivering interrupts.  However, the Hyper-V hypervisor uses the LAPIC timer as the hypervisor’s internal time source.  This behavior causes the problem that is described in the "Symptoms" section."

Microsoft's recommendation back then for a temporary fix until manufacturers could release new bios revisions was to either disable ACPI C-states in the BIOS or for systems that could not disbale C states in bios to do a regedit to disable C2 which was reg add HKLM\System\CurrentControlSet\Control\Processor /v Capabilities /t REG_DWORD /d 0x0007e066until  

Now this doesn't work for this issue but i suspect the issue is very similar. Therefore if someone at HP cared to they might be able to release a BIOS that addresses the issue. 

[MagikSeven]

I can't imagine why if it was a firmware problem it hasn't already been resolved over the years via a bios fix. So i dunno. But perhaps Some Z620 owners might have used different OS's - perhaps only some CPU's are impacted, perhaps certain motherboards have different small components where the firmware impacts them differently (i dunno bout that one).

But i definately know that C1E and the way interrupts are occuring are to blame. If it was just plain coil whine then even with c1e turned off the coils should still resonate where the cores have similar load to when c1e is turned on. But alas that is not the case.