Sign in with
Sign up | Sign in
Your question

I7 980x throttling

Tags:
  • Intel
  • Intel i7
  • CPUs
  • Overclocking
Last response: in Overclocking
Share
February 2, 2013 11:18:03 PM

I have a i7 980x running at 4ghz. It has apparently started throttling execution when I run multiple primarily cpu bound programs. For example, running singly the program might take 18 min. Twelve together are taking 36-48 min. My prior experience has been that there is only a small increase in execution when running 12 at a time, perhaps 1-3 min.

Cpu temps will rise to 70-73C with a peak of 76C, then fall to 49-54C, and then rise again after several minutes. Both fans are running close to max. This has happened occasionally in the past, but seems to be more frequent now. I've had the system for three years and have probably driven it at full capacity for 2-3 hrs/day.

I'm not especially familiar with the issues. I bought a custom configured system.

Is this a sign of portending cpu failure? Is there something I can do to alleviate the problem? Is this typical behavior? I read in another thread that throttling wouldn't occur until 100C and that the low 70s are safe.

With this level of slowdown, it would be faster to just run 6 at a time instead of 12.

More about : 980x throttling

a b à CPUs
a b K Overclocking
February 2, 2013 11:40:03 PM

Well, according to Intel:

http://ark.intel.com/products/47932/Intel-Core-i7-980X-Processor-Extreme-Edition-12M-Cache-3_33-GHz-6_40-GTs-Intel-QPI

The Tcase is only 67.9 degrees Celsius. It looks like you are seeing Tjunction (a guess) temperatures. Are these being reported as Core Temperatures and you are seeing as many of these readings as there are cores ? If so, I would think you are reading Tjunction temperatures. No one knows what max Tjunction temps are for Intel desktop processors, so you could be reaching one of the first throttle stops.

The CPU won't fail due to slow increases in core temperatures, because it can react fast enough to throttle then shut down. If you want to test the CPU to see if you are really seeing thermal throttle, try using IntelBurnTest on max, and watch after each run (a few mins each, more ram=more time per pass), what the Gflops number is. If the thermal throttle kicked in, you'll have Gflops numbers between runs that vary by 1Gflop or more (or if you have background processes using the CPU).

What is the CPU cooler ? The first thing to do is to open it and see if it is collecting a lot of dust, then that narrows what your next actions are:

If dust => Clean first, see if problem goes away.
If no dust => Check computer intake and exhaust fans for dust/wear/seized bearings.
If no wear and no dust anywhere => might be time for you to upgrade the CPU cooler.

The STOCK intel cooler is notorious for collecting a TON of dust/fibers between the fan and the top most portion of the extruded heatsink body.

Also the obvious should be pointed out just in case, move the computer away from heat vents in the house, make sure the computer intakes and exhaust have room to breathe, etc, etc...
m
0
l
February 3, 2013 2:55:41 AM

The cooler is an Asetek 1366 Liquid cooling system. The temperatures are per core as reported by HWMonitor. I cannot find any HWMonitor doc so I don't know whether they are Tjunction temperatures or something else.

I opened the case and dusted the cooler's radiator & fan. There was some but not as much as the last time I dusted. If that doesn't work I'll check the fans more closely. Somewhat unexpectedly HWMonitor reports the aux fan as running but the load is virtually zero. In the past it has only kicked in at high loads.

Thanks for the suggestions.
m
0
l
Related resources
a b à CPUs
a b K Overclocking
February 3, 2013 1:03:13 PM

fbell50 said:
The cooler is an Asetek 1366 Liquid cooling system. The temperatures are per core as reported by HWMonitor. I cannot find any HWMonitor doc so I don't know whether they are Tjunction temperatures or something else.

I opened the case and dusted the cooler's radiator & fan. There was some but not as much as the last time I dusted. If that doesn't work I'll check the fans more closely. Somewhat unexpectedly HWMonitor reports the aux fan as running but the load is virtually zero. In the past it has only kicked in at high loads.

Thanks for the suggestions.


If they are per-core, and you used HWMonitor, then these are indeed (to my understanding) Tjunction temperatures.

Those Asetek closed loop cooling systems does very slowly evaporate (if sealed properly), just FYI. Check radiator surfaces (both sides), you may have to remove the fan. If you are still getting poor temps, you can try doubling the fan on the radiator, making the setup a push-pull (if it is not already) with one fan on each side of the radiator.

Can you post a picture of your computer on the inside ?
m
0
l
February 3, 2013 4:07:06 PM

Maxx_Power said:
If they are per-core, and you used HWMonitor, then these are indeed (to my understanding) Tjunction temperatures.

Those Asetek closed loop cooling systems does very slowly evaporate (if sealed properly), just FYI. Check radiator surfaces (both sides), you may have to remove the fan. If you are still getting poor temps, you can try doubling the fan on the radiator, making the setup a push-pull (if it is not already) with one fan on each side of the radiator.

Can you post a picture of your computer on the inside ?


The Asetek is a 550LC, http://www.asetek.com/desktop/cpu-coolers/550lc.html. It is a single fan setup to exhaust. Dusting appears to have lowered the max temperatures by 6-8C, but I think I'm still getting throttling, just not as much. Times have improved by 15-20%. That's good, but I think it could be better. I can post a picture if the Asetek web link doesn't provide what you are looking for. The fans seem to be running ok. I cannot find much on evaporation rates or how to check liquid level.
m
0
l
a b à CPUs
a b K Overclocking
February 3, 2013 6:49:48 PM

fbell50 said:
The Asetek is a 550LC, http://www.asetek.com/desktop/cpu-coolers/550lc.html. It is a single fan setup to exhaust. Dusting appears to have lowered the max temperatures by 6-8C, but I think I'm still getting throttling, just not as much. Times have improved by 15-20%. That's good, but I think it could be better. I can post a picture if the Asetek web link doesn't provide what you are looking for. The fans seem to be running ok. I cannot find much on evaporation rates or how to check liquid level.


Okay, I see the cooler. It seems to be a standard one that everyone except for Corsair (who sources CoolIT made water units) and Swiftech (who is legendary in making their own water units) carries.

These coolers have very dense fin layouts, to maximize surface to air contact. As a side effect, the dense fins trap more dust fibers, and due to the density of the fins, there is a drop in air pressure between the fan, and the outside of the case where the exhaust air is dumped. This drop in pressure deposits a lot of dust/fibers/etc to the surface of the fins. This is typically between the fan and the radiator if the fan is pushing air through the radiator to the outside, or the opposite side of the radiator, if the fan is pulling air through the radiator. It is a good idea to clean this fairly frequently, especially if where you live experiences cold winters that requires some form of forced-air heating (that stirs up dust and dries the air).

So, it is good to clean these water units fairly frequently, especially in the winter.

Beyond this, if you are still seeing thermal throttling, my next advice is to install 2 fans on the radiator, in a push-pull configuration, so like this CASE||FAN_1||RADIATOR||FAN_2||INSIDE OF CASE, with the both fans circulating air in the SAME direction to the outside of the case. This is known to lower your temperatures further (since dense fins sometimes necessitates a lot of air pressure, and the 2nd fan provides additional air pressure across the fins).

If this is STILL not enough, and you don't want to upgrade your cooler, you can try a better thermal interface compound/material, see:

http://skinneelabs.com/2011-thermal-paste-review-comparison/2/

I love their TIM reviews, they are very thorough (I can't think of any other sites that actually has data points for differing amounts of contact pressures/contact fitments, and ALL recent/relevant TIMs in ONE review).

PS. While you are at it, check where the pump head is connected to. The pump typically draws power from the motherboard on a 3 pin header. If this is the CPU header, you may want to check in the BIOS to see if you have some sort of temperature controlled fan speeds enabled for that header. If so, you may want to disable it. Also double check for the fan that is connected to the radiator, where it is plugged into, and what the settings are for that fan port in the BIOS. If the FAN/PUMP are connected directly to the PSU, then never mind on this.
m
0
l
February 3, 2013 11:30:46 PM

Yes, you are right. I should schedule a regular cleaning of the radiator. It's easy to forget. Tuniq TX-2 was used as the TIM, which is right in the middle of the review you linked to.

Do you have a recommendation if I decide to replace the cooler? This machine is designed to run quietly without using one of those custom heavy duty cooled gamer systems. Even with quiet fans, the noise is noticeable when under load.

I'll pay closer attention to this stuff with next system I get.
m
0
l
a b à CPUs
a b K Overclocking
February 4, 2013 12:48:36 PM

fbell50 said:
Yes, you are right. I should schedule a regular cleaning of the radiator. It's easy to forget. Tuniq TX-2 was used as the TIM, which is right in the middle of the review you linked to.

Do you have a recommendation if I decide to replace the cooler? This machine is designed to run quietly without using one of those custom heavy duty cooled gamer systems. Even with quiet fans, the noise is noticeable when under load.

I'll pay closer attention to this stuff with next system I get.


Well, I'm not a fan of closed-loop systems. I remember reading a publication from Thermaltake on their water 2.0 systems saying that their systems exhibited low evaporation. So I started to dig up and found some numbers, I think in Computer Power User magazine, they quote Thermaltake test data showing that these systems evaporate about 6 grams of cooling fluid over 3 years, and another source quotes 2.5 grams per year. These systems usually contain about 2 to 3 times that amount as working fluid. So over 3 years, I would expect to notice a performance drop of closed-loop coolers.

That's why, I would recommend, for a (relatively) quiet, high capacity liquid cooler, get the Swiftech H220 (about 140 bucks) http://www.swiftech.com/h220.aspx that has been just announced and will begin shipping at the end of February of this year. It is similar to a closed-loop setup, but has all the components of higher quality, and interchangeable. And most of all, it is user-refillable. Of course, check to see if your case can accommodate a 2x120mm radiator.

Otherwise, you can try a really high quality air cooler that is meant for high powered CPUs, but these aren't cheap either...
m
0
l
February 4, 2013 8:12:11 PM

Thanks for all the help. Maybe it's time to take a look the latest Intel chip and see if it makes sense to get a new system. I'm certainly learning the importance of a good cooling system.
m
0
l
a b à CPUs
a b K Overclocking
February 4, 2013 10:56:09 PM

fbell50 said:
Thanks for all the help. Maybe it's time to take a look the latest Intel chip and see if it makes sense to get a new system. I'm certainly learning the importance of a good cooling system.


No problem! Your system is still very good by the looks of it. I would wait until Haswell, which is supposed to make improvements in multi-threaded performance. Seeing as how you are using a lot of cores, I suspect you'll see some improvements to your liking then.
m
0
l
!