High Temp.

[Grolch]

Heya all,

                  Like a cpl of other folk's am having some high temp's with my 2nd hand Z600

SPEC'S

              2x Intel Xeon X5550 @ 2.67GHz (standard heatsink & fan)

              Hewlett-Packard 0AE8h (Motherboard)

              8gb Ram (4gb per CPU - Soon to be 8gb per CPU)

              2x WD VelociRaptor 160gb HDD

              Saphire Radeon HD 6770 (post Quadro FX 380)

The grahpics card is only 2 days old in the system and has made not neglagible inpact on temp's...

Upon arival home,  windows 7 Pro x64 bit installed, heatsinks removed, CPU and Heatsink cleaned, Arctic MX-2 thermal past applied, heatsink re-set, boot up and .......... no noticeable change in temp's, heatsink's have been re-checked.

Right, so i'll wear the CPU's being at 45-50 C, but the motherboard being 45-50ish does still worry me a bit...

my main question is -

                                           Is it woth me pulling the motherboard heatsink and re-pasting that and is it a sound idea - will supply a pic to highlight what i'm talking about

(and to make sure i'm talking about the right thing 🙂  )

will be trying to source a cpl of the 3 copper pipe heatsinks as replacements..... possibly the water cooling from the Z440 or Z840 might fit....

Also, is there a better thermal past i could be useing? Am more of a tinker than a tech i have to admit 🙂

[gabornico]

what program did you use to read the temperatures ?

Download HP Performance Advisor and see what temps it reports.

[Grolch]

Heya,

          I used "Speccy" a Piriform product, will have a go with HP Perfomace Advisor, and see how i go 🙂

[Dan_WGBU]

Grolch, a couple of comments:

The heatsink you circled is over the IOH (Intel IO hub).  This chip is unique to the Z800/Z600/Z400.  It connects to the CPUs, and provides the PCI bus for Slots 1, 2, and 4.  I strongly advise to NOT remove this heatsink.  IIRC, the IOH package is a bare die, with no metal heat spreader, so there is a risk of damaging this thin, fragile glass die removing and reinstalling this heatsink.  The IOH heatsink has its own thermal pad, it does not use the normal CPU thermal paste.  The IOH does not get too hot to worry about cleaning and reinstalling thermal paste. 

Parts of the motherboard being 45 to 50 deg. C is not a worry.  CPUs can dissipate over well over 100W (new Workstation CPUs are up to 160W) each.  The onboard CPU power regulators powering these CPUs are 85 to 90% efficient, meaning these regulators can put ~20 to 40W of power directly into the motherboard.  That warms it up.  But don't worry.  HP does extensive thermal testing of motherboard components, they can take the heat. 

Likewise, high core count Xeon CPUs can get hot, too.  That is by design.  I have a longer explanation here, but to summarize:  

Many are worried if their CPU temps are over 60 deg. C, but the CPUs can handle it.  Intel specifies a maximum junction temperature for each CPU.  To see this, run CoreTemp, and select distance to Tj max as an option.  There is a wide variation of Tj max from one CPU type to another, from 85 deg. C to 105 deg. C.  CPUs can run indefinitely at 10 deg. C below Tj max.  I have personally run a dozen systems, with 100% CPU usage (using HP test programs), continuously for months, without any problems.  The CPUs will get to about 10 deg. C below Tj max, but that is OK.  

Workstation CPUs Xeon CPUs are designed to run hotter than what some expect.  Higher frequencies run hotter and higher core counts run hotter - higher system performance in general causes higher temperatures.  Intel tells us that CPU maximum performance is obtained when they run hot, since they are running at the highest frequency and active cores.  

I know many like to have lots of margin between Tj max and their CPU temperatures.  But think of a car engine - it has more power when run closer to redline, but that makes the engine run a lot hotter.  A cool CPU is usually not doing much, or is purposely designed to run at low power, like the slow speed, low core count laptop CPUs that run on batteries.  Intel CPUs will thermally throttle frequencies or shut down if they run too hot.  As long as your Tj max is not exceeded under 100% CPU usage, don't worry about harming your system.  

[SDH]

I want to second Dan's recommendations.  Heres a bit of added info:

For the Z600 the BIOS is like for the xw6600.... they include access to fan speeds and important temperatures straight from BIOS itself.  That is under the first "tab" of BIOS.  I have seen some programs that you can run from the operating system that probe for fan speeds and temperatures but give incorrect information when compared to the BIOS approach.  So, make sure to double check your numbers with those reported by the HP BIOS.

Second, the fan you circled is attached to the "southbridge" of the motherboard, and some workstations don't even have a fan for the heatsink there (the xw6400, for example).  I've worked to optimize my workstation performance and reduce noise levels (medical environment), and thus have experimented with that fan both on the Z620, Z600 and the xw6600.  

Noctua has in-line 4-wire fan speed reducers that fit perfectly for this fan/motherboard connector... they're about 4" long.  I have their Low Noise Adapter, and their Ultra Low Noise Adapter.  I also sent off to Noctua in Austria and got a few of their even higher resistor value adapters (I call it Ultra Ultra), and always run at least one of the Ultra Low Noise Adapters to reduce fan noise from that little high rpm fan.  I've watched the temps for each value (stock, LNA, ULNA, and even the UULNA) and don't see heat issues from doing this.  

I personally feel comfortable recommending that you plug in a LNA, which is easy to find on Amazon.  So, don't worry about the temps for that heatsink/fan if you're running stock speeds (HP already programs in some speed reduction via the PWM features of that fan versus running it at 12VDC without PWM).  Here's some results:

HP Z600 "chipset" fan RPM at stock with HP PWM from motherboard:    2530

Same, with ULNA adapter added in (82 Ohm resistor):                                    1925

Same, with the UULNA adapter added instead (100 Ohm resistor):           1800

I went straight to a ULNA and then UULNA on this build first so I don't have the RPM value for a Noctua LNA, and am happy with the noise level and cooling performance using the Noctua 4-wire adapter with the 100 Ohm resistor).  We're not pushing these workstations hard, so if you're doing processor-intensive work and want to drop the RPMs this way just watch via BIOS for how things might heat up above what you'd see normally.  You'll get accurate readings with a quick reboot into BIOS right after you push the workstation.... things don't cool off instantly, and if you find a program that gives you accurate readings from the OS then you can monitor real time.

The motherboard header and stock chipset fan plug is the conventional PWM type of morphology, as are the ends on the Noctua adapters, so it is pretty much plug and play.