Maxing CPU heat levels

Post Reply
Will.Rubin
Posts: 27
Joined: Mon Apr 09, 2012 7:08 pm

Maxing CPU heat levels

Post by Will.Rubin »

Just received an IPC2 and am having trouble with the CPU temps. The idle temps are in the low 50C range. A bit higher than expected but not too bad. But when CPU runs at 100% the core temps very quickly hit the maximum 100C mark and the cores start throttling..

Should I be seeing these high temperatures this quickly?

I have an IPC1 and it idles about 45-50C and maxes at 80C. This seems more "normal" to me.

--Will

irads

Re: Maxing CPU heat levels

Post by irads »

How quickly does CPU temp reach 100C (seconds, a minute, 10 minutes?) and what seems to be case temperature compared to IPC1 - similar or hotter?
We need that information to determine if there is a marginality with thermal coupling of the CPU.

Will.Rubin
Posts: 27
Joined: Mon Apr 09, 2012 7:08 pm

Re: Maxing CPU heat levels

Post by Will.Rubin »

A couple/few minutes to hit 100C when CPU is running at 100% from cold. Once the PC has "warmed up" it only takes 20-30 seconds or so to go from 60C back to 100C.

The case maybe gets a bit warmer than the IPC1, but not enough to prevent you from touching it or anything. (My infrared thermometer died so I can't give exact numbers.)

irads

Re: Maxing CPU heat levels

Post by irads »

This profile indicates a problem with thermal coupling of the CPU to the heatsink. I will ask our RMA department to replace the unit.

Will.Rubin
Posts: 27
Joined: Mon Apr 09, 2012 7:08 pm

Re: Maxing CPU heat levels

Post by Will.Rubin »

I received my RMA'd IPC2. Unfortunately, this IPC2 exhibits the same thermal overheating as the previous unit. Though over a slightly longer time.

I've attached graphs of the CPU temps when running LinPack via OCCT. The graphs are from a completely cold start. First boot after machine turned off and everything at room temperature.

Note that I set OCCT to stop the tests when the cores go above 90C. The temperature does continue to increase to 100C whereupon CPU throttling kicks in. I just didn't see the point in repeatedly hammering the CPU.

Subsequent runs heat much faster: five minutes or so to thermal maximum. Running Prime95 on all cores heats even faster than OCCT, but it doesn't make the nice graphs.

I'm running a fresh install of Windows 8 and no BIOS modifications from the default except to display the POST screen for 3 seconds before continuing. I've installed 16GB memory and a SSD.

Here are links to the three pieces of software I'm using for testing.
Prime95: http://www.mersenne.org/
OCCT: http://www.ocbase.com/
CoreTemp: http://www.alcpu.com/CoreTemp/

Q1. Am I seeing expected behavior for IPC2s?
Q2. Is the I7 IPC2 designed to be run full out (both physical and both logical cores) for longer times?
Q3. Any chance you could run either OCCT or Prime95 on a sample there and see where the temperature maxes out for a known good sample?
Attachments
Core 1 Temp vs. Time
Core 1 Temp vs. Time
2015-03-05-15h11-Temperature-Core #1.png (39.6 KiB) Viewed 11270 times
Core 0 Temp vs. Time
Core 0 Temp vs. Time
2015-03-05-15h11-Temperature-Core #0.png (50.59 KiB) Viewed 11270 times

irads

Re: Maxing CPU heat levels

Post by irads »

Thanks for the concise feedback.
Based on analyzing the behavior I believe what you are observing is normal IPC2 behavior and is by design.

The Haswell CPU has a very flexible power consumption profile - under load it may consume power beyond the stated TDP. It also has a smaller die area than the Ivy Bridge used in IPC1. These two characteristics lead to higher heat density under load that result in larger deltaT between die and casing.

Under continuous extreme load as in your tests it is possible to reach throttling temperature with IPC2. The profile in your tests indicates gradual heating of the case used to dissipate heat and corresponding heating of the CPU.

If such continuous workload is the normal profile you are using it may be that IPC2 is not the ideal platform for your needs. It will work reliably, but will not provide full performance most of the time. You may return it and choose an actively cooled PC.

If you wish to use IPC2 for this workload I suggest you ensure case temperature remains below 50C. This can be achieved with maintaining low ambient temperature or better yet providing any kind of light airflow over the cooling ribs. It is important to note that for the majority of usage scenarios such measures are not needed - only for continuous full workload.

Compulab continues to refine thermal interface materials, thermal coupling and power management throughout product life cycle. Your input is very valuable in that respect. I hope you find an adequate solution for your needs.

Best regards,

Will.Rubin
Posts: 27
Joined: Mon Apr 09, 2012 7:08 pm

Re: Maxing CPU heat levels

Post by Will.Rubin »

I'm bummed to see your response, but glad you responded. Thanks.

I've been playing with the IPC2 for the last few days to see where it can fit, and based on your reply I honestly don't know what I could run on the machine to keep the temperatures in the desired below 50C range?!

Booting Win8 and doing nothing but monitoring Temp with CoreTemp and CPU with Task Manager. Power is 2.5W, CPU utilization runs <1% most of the time with occasional jumps to 3%. Core 0 temp runs about 45C and core 1 runs about 40C.

Open up a browser to, say, Yahoo, MSN, or YouTube doing nothing but leaving the browser open on the page and the temperatures hover around 50C. Surfing, watching videos, Windows updates, etc. all regularly push the temperature into the 60C range.

Running various non-graphics applications that causes average CPU utilization of about 30% and spread out across both cores will put the power at 11-12W and push the temperature to the 75C-80C range with occasional spikes to 90C.

This is in a room with ambient temperature of 20C and normal airflow around the IPC2, including allowing air to circulate along the bottom too.

My goal is to replace all my actively cooled systems with passive silent no-fan systems. I've been purchasing a couple Fit/Intense PCs each year and swapping machines out with little regard for their underlying purpose. Streaming, analysis, light duty server, general desktop replacement, and others. They've all worked really well so far. But I also want to keep as close to the most modern technology (CPU, wireless, USB, SATA, etc.) as possible.

How about the uSRV machines running the i7-3555LE? Will they run cool enough to withstand continuous load? They're a generation behind, but they're a bump up in performance from the IPC1s.

I'll play with the IPC2 machine for a couple more days to see where I might fit it, but I suspect I'll have to return it.

Thanks,
--Will

Post Reply

Return to “IPC2 Hardware”