Z820 CPU fan does not automatically speedup when CPU temp raise at high load

[BambiBoomZ]

Horsem,

Yes, I am puzzled by the z820 liquid cooling design as the fans and pumps have seperate connections and that may be an opportunity for sensor communication problems.  I too would have done this differently; as an all in one similar to the z420 design. There is enough room on the z820 that it could have had a radiator mounted on the pump base with a push and pull fan arrangement directly on the radiator.  Also, the CPU's are too nearly transversely aligned, such that the front CPU is feeding warmed air to the rear one- a temperature differential? 

The fan controller location is not shown on  the system board diagram, whereas the CPU 0 and CPU1 pump connectors are shown located along  the top edge of the motherboard.   This is what leads to the possibly erroneous conclusion  that there a two separate but linked controllers- somewhere. 

I use a z420 liquid cooler in a z620 -which has an extremely similar motherboard and uses the same z420/620/820 BIOS controls.  It is a very good all in one design and when connected to the same 5-pin CPU fan / heatsink connector, immediate ran very well without any adjustments.  I run an E5-1680 v2 overclocked to 4.3GHz on all 8 cores and the idle temperature is between 28 and 32C and even under a long CPU rendering slog, never over 72C.  The rating for the 1680 v2 is 85C.

As the z820 is running on the same BIOS, as the 420 and 620, it seems odd that one BIOS can accommodate three configurations of CPU cooling:  fan /heatsink, all in one liquid and separate radiator, fan, and pump.  This is why I keep imagining that the problem is related to be the separate interconnection between fan and pump control. 

Fan controls are not linear as they operate on a temperature to fan speed curve and still it's not clear in the way the pump is  regulated. On the z420 liquid cooler, the pump and fan both clearly change speeds with temperature.  However, I noticed that when I tried advancing the setting in Advanced > Thermal, the cooler would "hunt" a fan and pump speed- oscillating RPMs. Only the default setting produced a steady idle speed and normal response to load/temperature. I keep returning to the idea that the key has to be in the details of the what seems to be separate fan and pump controls.

Checking around a bit I noticed that there have been quite a few questions regarding both z800 and z820 liquid cooling systems.

BambiBoomZ

[DGroves]

the z800 and z820/840 cooling systems were given a lot of thought during the design phase of the systems many engineers who make their living desiging things like this  were on the teams that came up with the designs for all three systems and not being a design engineer i will not chime in and say that their decisions were wrong since i have absolutly no experence in this field

what i will say is that the design of the liquid cooling option leverages the existing air cooling setup for use with the liquid cooling option quite nicely as it uses the existing air cooled fans to now provide the nessary cooling for the liquid raidators

while still keeping the existing fan shroud which cools the motherboard/cpu's and ram and directs it to the rear exaust this cooling system is very quiet (be it air or liquid cooled) on all three z workstations and is one of the lowest noise cooling systems i have worked with over the years on any workstation class system.  it compares very closely with a corsair H80i on a asus x99 motherboard using a 150 watt xeon in a HP z6200 case in terms of noise

there is a fan header on the motherboard, it connects to the large black fan shroud that houses 5 fans two of which fit flush with the air or liquid radators when the shroud is installed. the shroud also ducts air in such a way that neither cpu will get any hot exaust from the other cpu and directs the hot air directly to the rear exaust. this shroud also cools the motherboard memory and VRM's using the 3 remaining fans. it's really quite a complex setup in how it directs airflow while mataining low noise levels doing so.

as to the tempture sensors, just about every system now made uses two diffrent tempture circuits................ one for the cpu's using the cpu's onboard thermal sensor and other sensor(s) that the motherboard maker adds for the case cooling fans..... nothing new or exotic there

as for the z820/840 liquid cooling headers LC0/LC1, all they do is supply power to the pumps, nothing more and they allow a neater method of powering the pumps instead of having molex/sata based power taps.

they also show that liquid cooling was considered during the motherboard design phase and not a afterthought after the board/system was released

[horsem]

Thank you both for the long detailed response. A very good intro but I hate to say that the problem of overheating is still not solved. I agree that it is a quite workstation if the idle fan speed is set low.

Do you guys have access to the people who designed this model? Could you ask them to see if the radiator cooling fan speed is designed to be variable in response to cpu temp? If yes, then we probabaly are missing the driver for it in windows 10 system.

If it is not designed as variable in response to the cpu temp, then I have to say it is really a dum design considering it being a powerful workstation capable of high load work.

By the way, BambiBoomZ, how did you overclock you CPU in the BIOS? I didn't even see an option for this in my Z820 although this CPU is overclockable. Is there a secret way to enter the advanced BIOS setting mode?

Thanks,

Zhibo

[DGroves]

both thermal sensor circuits are variable and controlled by the pre installed algothims of each circuit

furthermore the case fans can be manually controlled via the bios settings overriding the default fan profile settings

the cpu fans that cool the liquid radiator or air cooled heatsinks are directly controlled by the bios using predefined settings that will ramp up/down the two cpu fans in the cooling shroud based on the cpu tempture

since HP has full control over the system (unlike aftermarket cooling systems) they can tune the systems cooling to a much more acurrate level because they know the cpu temp, the size of the radiator, the cooling pump's flow rate the radiator size/thickness/fin placement and fan airflow as such a properly installed/working watercooling kit from hp for the z820/840

is designed to work well within the stated tempture range of the workstation.

you might have to take the system to a HP authroised repair center and have someone look at it if you are unable to resolve the problem yourself

[BambiBoomZ]

Horsem,

I assume, but not  not 100% certain, that the specification for the z820 originated with HP, and probably the motherboard design, while the chassis is by BMW Design Works. If someone knows better, please correct it.  By the way, there was both a  Z8 car-  the BMW Z8 roadster that evokes the late-50's BMW 507, and of course the current HP Z8.

Relevant to this discussion, while I adnire the chassis design intentions, I am not convinced that the aerodynamic solution as regards cooling in the z820 and the current Z8 is optimal. This is because in both cases- figuratively and literally, the z820 CPU air flow has multiple deflections and in the Z8 there appears to be convolution and possibly  some static pressure points.  The 6-fan z820 shroud is a 3D masterpiece, but each change of air flow direction is a loss in speed. One might argue that the air flow is traveling through the massive shrouds may have some heat sink capacitance- a couple of pounds of hot plastic is adding to the case temperture. The basis  that design  is that it's quieter as the fans are buried in the chassis volume rather on the exterior surface as in gaming systems.

The CPU cooling fans are most certainly designed to respond to CPU temperatures and that is a BIOS function. I've had five HP zX20-series systems, both single and dual CPU and have never had a cooling problem.  When I ran CPU renderings 100% on all 16-cores of a z620 (2X E5-2690) , the temperatures

Many users have tried software cooling control in worstations, e.g. Speedfan, only to find that HP prevents mucking about with their well-laid plans.  As a proprietary, integrated design, that is logical. 

In this discussion I'd thought  to consider adding a 5.25" add-on fan controller,e.g.:

https://www.nzxt.com/products/sentry-3

As that kind of controller has mutiple, independently controlled fan channels and repsonse curves However, whether that could interface properly with the z820 withough ocnstant- CPU fans missing errros that shut down the system is a fundamental question. Trying to modify very carefully integrated proprietary workstation designs is difficult.

 Another possiblity for the z820 is an external liquid cooler.  I'm preparing to install one of these so as to increase the all-cores; speed of the z620's:

https://www.alphacool.com/shop/neue-produkte/21360/alphacool-eiswand-360-cpu-black

The good feature of this i sthat it has a thick Copper radiator with 6X 120mm fans in push/ pull at a constant medium speed, plu two ceramic pumps also to reduce noise, plus as an enternal unit, all the air heat tranfer is happening outside the chassis. It is possible to design and build a similar configuration using the immense range of stock components marketed to gamers.

the simples possible using an ordinary closed loop Corsair. Shown with a z620, to avoid having to cut a cpu block-sized hole in the rear panel.  That design is prefereable with a seperate resorvoir and pump a fluid tube PCIe slot cover pass through so the CPU waterblobk can be installed an dremoved  by using quick release disconnect connectors. Also, that could be a 360mm- 560mm radiator.

Speaking of modications, yes, there are certain Xeons E5's that may be overclocked using the Intel Extreme Tuning Utility (XTU).  This is not a highly refined and detailed overlcoking though and one ends up putting more voltage into the CPu than is typical with unlocked consumer CPU's on non-proetary bards. For the zX20 series these are however all 6 and 8-core  E5-1600 v2 - single CPU only,  e.g. E5-1650 v2, 1660 v2, and 1680 v2.  The benefit in the higher clock speed is imporntant for single-thread applications. I changed from the 2X E5-2690 z620 to the 1X E5-1680 v2  8C@ 4.3GHz speciifaclly for 3D CAD and correspondingly changed from pure CPU rendering to CPU + GPU accelerated.   For compute and pure CPU rendering systems, more cores will still produce more lock cycle persecond; the z620_1 with 2X E5-2690 with 16-cores had a Passmark CPU benchmark of 22625 whereas z620_2  with the E5-1680 v2 scores 17418. However the important single thread mark- 90% of my use was z820_1= 1815 and  z720_2= 2373.  Our friend Brian1965 is running an E5-1680 v2 at 4.7Ghz, having a single thread mark of over 2500.   For comparison, the current Xeon W-2145 8-core (@$1,300) scores an average of 2522 whereas the current cost of a used E5-1680 v2 is about $330.

BambiBoomZ

[DGroves]

this is a bit off topic, so i'll keep it short

the z820/840 shrouds do not soak up heat, they don't even get warm after days or running,.. the plastic has a very low heat rentention compared to the systems metal case in soaking up and disapating heat

no cooling system is perfect in all aspects, what counts is does it meet the design goals in reguards to noise, cooling capacity,  and cost constrants the mfgr set as design goals.... and i think HP meet these goals with the z820/z840 systems

the HP cooling shroud is a low flow/volume rate device, as such the loss of airflow by the ducting from the shroud appears to be minimal, and HP has rated this setup for all 130 watt systems (air cooled) and the 150 watt systems (water cooled) and it appears to be quite sucessful as it was carried over from the z820 to the z840 with minimal changes

can a user put a bigger higher capacity liquid cooling system on a existing system?....sure but at what cost or tradeoff?

while this approach is  possible for a home user it's not something i want to see or maintain on a corprate users desk and these "z" systems were designed for the corporate setting not the home enviroment

i personally like the idea of personal z620 with a custom radiator that mounts to the outside of where the dual case exaust fans are,.... and using a custom water cooling setup to cool the cpu/gpu,...... however i don't want to ever see such a system at my work site i have more issues than i have time allready....and one off non standard setups make me want to cringe!

we currently plan on testing the HP z4 (and possibly the z8) systems sometime towards the summer of this year untill then i will have no personal experence on these platforms, however for those who are interested here's nice writeup on one

https://www.anandtech.com/show/11838/hp-updates-z8-workstation-up-to-56-cores-3-tb-ram-9-pcie-slots-...

[horsem]

Hi DGroves and BambiBoomZ,

Thank you both for trying to help! 

Actually, my Z820 with dual E5-2687w does a not bad job on cooling when the BIOS idle fan speed is set to level 3. Under 100% cpu load, after 15min, the core temperature is about ~82C. Comparing to most 150W system, this seems not too bad.

With BIOS idle fan speed set to level4, the temp is about 73C under the same test condition. The only drawback with level4 idle fan speed is the noise. 

I guess increasing the idle fan speed increase the overall air change rate, which helped lowering the temp.

The only expect of posting it here is to see if there is any configuration in the BIOS or issue with the windows 10 driver that is causing this problem.  I just purchased this workstation from a computer recycle store. Not very familar with the hardware setup, specially the cooling design which still doesn't make too much sense to me. However, it works very well on noise and cooling if you hare not running very heavy work.

The reason that I have concerns about the temp is that I accidently saw that the Tcase of the CPU is only 65C. If it is 85C like most of other CPUs, then I won't have any problem about the cooling.

Do you guys know how bad it will be to the CPU if my CPUs runs in the 75-85C range? I haven't seen any thermo throttling of CPU. Would it be safe to say as long as no thermo throttling, it would be safe? My understanding is running at temp that's higher than Tcase would greatly reduce the lifetime of the CPUs. If it won't quickly ruin my CPU at this 75-85 on heavy load work, which is not very often. Only a few hours per few days, then I will just leave the idle fan speed at level 3. If it significantly hurt my CPU, then I guess I will have to sacrifice the quiteness and set the idle fan speed at level 4 or higher.

Anyway, Thank you both for the help and long response!

Z.

[BambiBoomZ]

Horsem,

The Intel TCase temperature rating ("TCase Temperature is the maximum temperature allowed at the processor Integrated Heat Spreader (IHS)") for the E5-2687 is 67C and I would consider  82C for very long to be worrying.  These ratings appear to be somewhat conservative to protect what was an expensive ($1,890) processor when new. 

However, if the problem was serious, there would be signs of thermal throttling and thermal shutdowns.  Still, it's not a good long term solution to have to adjust the fan speed for particular use and also be pushing the processor limits unnecessarily. One of the reassuring aspects of workstations is the idea that they are designed to function at 100% load for very extended times without fear of damage or data loss.  This is the reason that mnearly all Xeon CPU's have locked multipliers.  I have one of the few that is unlocked (E5-1680 v2) and overclocking has led to several (4 or5) somewhat mysterious ( = not thermal protection) shutdowns over two years, one of which did lose some unsaved work. 

Consider trying to ioptimize the situation by a careful remounting of the coolers using a high quality thermal paste such as Arctic Silver 5, Noctua NH-1, or Thermal Grizzly, ensure that the case and all fans are dust free all through, and verify that all the case fans are running properly.

BambiBoomZ

[Scoobis]

i have the exactly same issue under windows 10.

I have tried every combination I can think of in bios and system settings, and I am not new at this.

I believe there are driver(s)/Software that's needed to support the fan speed changes under windows 10.

Personally I'm running 2 x e5-2667 v2's  and air-cooled... system will thermal throttle, fan speeds won't ever raise, though you can change the idle speed in bios and it will stay at whatever that setting is.

[BambiBoomZ]

Scoobis,

The number of recent posts regarding z820 cooling problems has been surprising and equally surprising is the lack of consensus as to a clear solution.

The Xeon E5-2667 v2 is rated at 130w with a 74C  TCase rating (= maximum temperature allowed at the processor Integrated Heat Spreader), so the thermal load even t sustanined 100% should not exceed the design parameters.

As the fan speeds are BIOS controlled and changing the BIOS setting does raise the fan RPM, the only explanation I can think of is that the CPU temperature sensor(s) has/have failed. I am not aware if the tempertures used by BIOS and the figures displayed in a CPU monitor derive from the same source.

If the actural cause can not be resolved, the soltion might be to change to the z820 liquid coolers, accept a permanent higher BIOS setting fan setting, and consider relocating the system to minimize the affects of the extra fan noise. 

What are typical temperatures  the system is displaying?

This issue seems to be common and important enough that It would be extremely valuable if someone at HP would comment.

BambiBoomZ