Q9550 E0 - stock cooling

bgiddins · Post by **bgiddins** » Thu Oct 30, 2008 4:30 am

Here's some interesting figures - I'm running Ubuntu Server 8.10 on an Asus P5Q-VM. The CPU is an E0 stepping Q9550 quad core - I thought I'd try and push the CPU at default BIOS settings with the stock Intel cooler to see how hot it would get - so I've installed VMware Server, and added 3 VMs each running Ubuntu Desktop 8.04 64-bit, with 2 CPUs each and 2GB RAM (so 6 virtual CPUs in a 4 core sysem!). I added the SMP Folding@Home client, and fired them all up.

Using lm-sensors and gkrellm, I'm getting the following:

Idle: 32/32/30/30
100%: 49/49/44/46

The first core spiked to 51 momentarily, but is sitting at 49 about 95% of the time. The VMs have been running for about an hour now, quite stable.

So looks like the stock Intel cooler works quite well with this CPU at default BIOS settings. Still thinking about OCing, but given the machine is handling 6 virtual CPUs with aplomb, I'm quite content

After posting, I realised I should add that this is in an Antec Three Hundred case, with 120mm and 140mm case fans running at LOW, and a pair of Arctic Cooling PWM 120mm fans mounted in the front of the case as inlet fans. The side vent is open, and is drawing a light vacuum, so it's still a negative pressure system. BIOS says the AC fans are running at about 1150rpm.

bgiddins · Post by **bgiddins** » Fri Oct 31, 2008 3:32 pm

Repeated the test - after about 3 hours I was getting 51/50/48/50.

EsaT · Post by **EsaT** » Sat Nov 01, 2008 9:21 am

Intel Burn Test or Thermal Analysis Tool would probably give entirely different readings...

bgiddins · Post by **bgiddins** » Sat Nov 01, 2008 3:04 pm

No doubt - they're geared towards stressing the CPU - I'm not sure for example if the combination of VMware and Folding@Home utilises all of the available L2 cache etc.

What it does show however is 100% clock cycle utilisation across all four cores - and as this is a machine only used for development and running VMs, it's a close approximation of what my intended usage will result in. If running siz virtual CPUs had shown temps hitting the 60s or 70s I would be adding aftermarket cooling for my needs. As it is, I can barely detect warmth in the Intel heatsink when touching by hand (the Antec Three Hundred has a 120mm and 140mm fan within cms of the CPU - lots of airflow up in that corner!).

Mind you I'm running Ubuntu, and both of those apps are Windows, so I can't try them for comparison.

leifeinar · Post by **leifeinar** » Sun Nov 02, 2008 2:58 am

runing torture tests gives to high temps, u can never go that high with normal use. Folding is better, stil 1005 on all cores but temps are normaly 5-7 celcius lover

try undervolting ur cpu, it probably doesent need any more tha 1,0V

bgiddins · Post by **bgiddins** » Sun Nov 02, 2008 3:42 am

Undervolting is on the cards - I'm waiting for a power meter to arrive (want to test stock power consumption of PC), then will bump the processor to 3GHz (faux-Q9650

) and see how low I can undervolt for efficiency.

leifeinar · Post by **leifeinar** » Sun Nov 02, 2008 1:27 pm

I got a QX9650, its primestable @ 1,04V, and does 3,66ghz @ 1,2V. u may can do the same

bgiddins · Post by **bgiddins** » Sun Nov 02, 2008 2:51 pm

Thanks for the figures - they should be a good starting point. Using lm-sensors, VCore varies from 1.08 to 1.2V - I need to check the BIOS to see what it displays as in there.

Anyone using the P5Q-VM board - the "out of the box" BIOS version of 0202 only had two OC profiles that could be stored - in 1406 at least a dozen different OC profiles can be saved.

bgiddins · Post by **bgiddins** » Mon Nov 03, 2008 10:00 pm

Interesting - I just cranked up the two fans (120mm & 140mm) on the Antec 300 to see what impact there would be on system temps - the NB temps dropped about 6 degrees, but the core temps went up anywhere from 2-4 degrees each! Dropping the case fan speeds back from high to low saw the NB temps rise again, but the cores fell back to previously encountered temps.

I think that the massive airflow disruption in the case interfered with the stock cooler's ability to push air through the cooler - hampering it's cooling ability.

leifeinar · Post by **leifeinar** » Tue Nov 04, 2008 8:30 am

thats probably wat happens.... airflow overkill

1.08 to 1.2V

this might be speedstep work, its "EIST" and "C1E" in the bios if i remember correct

Nick Geraedts · Post by **Nick Geraedts** » Tue Nov 04, 2008 11:45 am

This is what's currently powering my beast...

I've got a TRUE with a 500RPM Slipstream on it along with everything else in my case... highest load temps I've seen so far are 50C. Can't complain too much.

leifeinar · Post by **leifeinar** » Tue Nov 04, 2008 1:49 pm

You just got to love these 45nm cpus

and the TRUE, I got a noctua 5-600 rpms on mine

bgiddins · Post by **bgiddins** » Tue Nov 04, 2008 4:31 pm

Hi Nick,

Thanks for posting that - have you set any other voltages besides VCore? I have only dabbled mildly with OCing before - this board looks to have a dozen different voltages alone that can be set, let alone RAM frequencies, timings etc. I'd like to bump mine to even just 3GHz if I can do it without increasing power consumption (my voltages are all set to Auto at this point, with Speedstep enabled).

I'm looking to tune this server for efficiency, so I might end up undertaking an exercise in undervolting rather then overclocking. Once I get my power meter I'll play some more - I would love to know how many watts it uses at idle - I don't want to turn say an ~80W idling machine into 120W by overclocking when I'll rarely be pushing the CPU's capabilities. I tried setting some explicit voltages at 2.83GHz (like 1.05V for VCore), but the machine would freeze when I fired up six virtual cores of Folding@Home

Nick Geraedts · Post by **Nick Geraedts** » Tue Nov 04, 2008 9:48 pm

I actually ran into a minor stability issue with that setting. I currently have my Vcore set to 1.225V in BIOS, and my NB voltage set to 1.37V. My RAM voltage is set to auto at the moment, which the BIOS has decided will be 2.0V. So far, things look pretty good.

bgiddins · Post by **bgiddins** » Thu Nov 13, 2008 8:53 pm

bgiddins wrote:Idle: 32/32/30/30
100%: 49/49/44/46

Well after a few hot days (27C ambient) the CPU temps began to creep up at 100% - I was getting the following (or higher) on the stock cooler:

Idle: 32/32/30/30
100%: 59/57/56/58

Even saw a couple of spikes to 60 very occasionally on some cores. Still within limits, but dammit, I want this box to run cool!

So today I fitted a Xigmatek HDT-S1283 - using pushpins, as backing plates were out of stock. Without too much regard to cleaning etc (I'l be fitting a backing plate in a week or two when available), I fitted the cooler with the stock thermal grease, and with ambient temps around 27C again, and now getting the following!

Idle: 32/32/30/30 (no difference)
100%: 39/39/34/36 to 40/40/37/37

Sometimes all 4 cores are under 40C. That's almost 20 degrees less on each core.

Very impressed with the performance of the S1283. When the backplate comes I will be properly cleaning the CPU, applying OCZ Freeze, and adding a NB heatsink at the same time (requires mobo removal to fit as well).

At these temps I might move my AC fans back into the PWM daisychain - it will shave a couple of db off the sound levels.

Q9550 E0 - stock cooling

Q9550 E0 - stock cooling

Re: Q9550 E0 - stock cooling