920xm throttling at 20x multiplier

I'm using ThrottleStop to mess around with my multipliers. I set the CPU to idle at 24x and go down to 20x when all 4 cores are in use.

Idle temps average around 51. Once the cores hit an average temp of 83 degrees (average DTS of 16) the multiplier drops to 10x.

I am using TDP 90 / TDC 75. These seem to be the standard from what I've been reading.

My CPU seems to run unusually hot compared to what others are getting, but shouldn't the CPU be allowed to run hotter than this? I've read that it should not be throttling until the temperatures hit around 100 degrees and 0 DTS.

I repasted my CPU twice with Arctic Cooling MX4 and got the same numbers after each application. I have a cooling pad but first want to figure out why the CPUs are throttling at such a low temperature. I've done a lot of reading and haven't found anybody else with this problem.

 

can you post a screen shot of your settings, and make a tstop log running wprime 1.55 @ 8 threads, the small test should be fine.

 

The log is attached [and a little lengthy] but I can post it on here instead if you want. You can see the numbers changing significantly at 19:51:56, and despite the log saying the multiplier is 20x at that point it actually drops to 10 and stays there until I end the test (I'm using HWiNFO32 to check this as well as temperatures). I only ran the wPrime155 8 thread 1024M stability test until 40% completion.

In the first screenshot you'll see the FID readings are fairly low. At idle, according to ThrottleStop they fluctuate slightly in that range while HWiNFO32 shows them staying steady at 23. I'm inclined to believe that HWiNFO32 is displaying the correct multipliers in both instances.

 

well it looks like its running the 20x just fine, also make sure to check the 2 clock mod boxes too.

do you have the 1024 test log?

 

During a 1024M test a CPU can become pretty hot, the load is at 100% all the time. That's nothing unusual. In applications like games you normally never have such a load and the CPU can cool down before being loaded heavily again.
Throttling during the 1024M test isn't a sign for a system that throttles during everyday use and gaming.
The general consensus is, that if you can run a 32M test withouth throttling issues, your settings should be more or less fine for gaming as well.

That's not true. Throttlestop is the only tool I know of which is accurate at displaying the clock speeds and multis in (almost) real time.
HWinfo32 isn't accurate in this case. (Either is CPU-Z)
TYour CPU is designed to automatically clock down if there's not much load. This saves a lot of energy and produces less heat. It's totally normal that the multi isn't at the max value all the time. (Actually something would be wrong with the system if the multi was all the time on the max.)


Edit: Ah, and I forgot.... like Inap said, check the first two boxes (actually csimply checking chipset clock mod will be enough)

 

You might want to upgrade to Throttlestop 3.0, and click on the DTS box so you can see the temps. I'd also check bpth the clock mod boxes. There may be some other settings to change - this behavior seems strange. Double check that turbo isn't disabled in the bios, and that Stealth is not on (I doubt the latter - I think it would be slower).

Joe

 

If you click the DTS button, it will turn the readings into degrees C which should help. Run prime95 on all threads and keep and eye on the temps.
Edit: JD beat me to it.

 

You're looking at a classic case of thermal throttling, implemented at 85C through Chipset Clock Modulation. The 85C limit is set in the BIOS and is what ThrottleStop is designed to overcome to permit the CPU to operate at higher frequencies and the temperatures which accompany them. You fix it by putting a checkmark in the Chipset Clock Mod box in TS and then clickng Turn On. That will take you to about 100C, at which point Intel's internal protective routines will cause it to throttle by reducing the multipliers. If it should reach 105C, it will automatically shutdown to prevent damage.

This works pretty well with ThrottleStop, but you're 920xm is running way too hot too early. You should not be hitting 85C. Here is an example HW32 temp shot of my 920xm at stock clocks with the 5850m at 800/1050 after running 3dMark11, with the CPU having been recently re-pasted with MX-4. You can see both the idle temps and the max temps are well below yours. You need to re-paste to get those temps down so that idle is around 45C or less. I will rerun with multipliers at 24-23-20-20 and see whether there is any significant difference.

​

Edit: Re-ran 3dMark11 with 920xm at 24-23-20-20, which increased CPU power use to 65W and raised CPU temps about 5C, but still below 70C.

 

Code:

  DATE     TIME    MULTI   C0%   CKMOD  CHIPM  PWR_mW  DTS
05/23/11 19:51:56  20.00   74.2  100.0   50.0       0   20

Here's when things first go to crap. See the CHIPM column? That stands for Chipset Clock Modulation and as soon as that shows up in the log, that means your CPU is being throttled and is running at approximately 50% of its rated performance. Both CKMOD and CHIPM should always show 100.0% in the log.

The Revelator is right. Control the heat and you will be able to maximize your performance. The 920XM is capable of much more but you need to keep the GPU temps under control to run it at full steam without the GPU triggering CPU chipset clock modulation.

Grab version 3.00 from my sig. Lots of minor fixes and improvements.

 

Wow, lots of useful stuff in here, thanks guys. I've definitely made note of everything said above.

As for my insane temperatures, I guess I'll try pasting again. And GRIEF Revelator I am jealous. I'm seriously wondering what the heck could be wrong now. I'm glad to see someone who has the same paste and CPU to measure up against but the difference between our temps is insane.

What do you guys think the heat issue could be? I'm pretty sure I'm not failing that bad at pasting, lol.

Edit: Testing my CPU on stock clocks (to accurately compare my temps to Rev's) I'm hitting 79 max, which is still a drastic difference from his. And oh yeah, I can safely go past 85 degrees now, awesome. The only problem I have now is trying to make sure I don't reach that on 20x or above.

 

With the M15x, I think it's very easy to do everything exactly right with re-pasting and then screw it up attaching the heat sink. Because quarters are so tight on the CPU side, it's very easy to smear the paste trying to get the HS on flat and aligned, resulting in partial or uneven coverage. What works best for me is starting by gently inserting the radiator/heat pipes/fan assembly with the HS cocked enough to slide the radiator/pipes section in place, then lowering HS while nudging the pipes/fan section into place but raised enough to avoid contact by the HS. Once the fan,etc are properly positioned, the HS will lower straight down into place without smearing the paste. It's hard to describe very well, but if you're familiar with the process, you'll understand. You can still mess it up. You may as well just stop and start over. I've tried telling myself it probably won't matter, it probably didn't screw up the paste, etc. Inevitably, I have to re-paste in a month, if not sooner. Maybe no one else has this problem, but I think it's probably why you see M15x temps well above what you might expect given the cooling capacity.

Edit: Forget using the 1024M test of wPrime 1.55 to check CPU performance or temps. The extended test is murder on turboboost CPU's; most CPU's throttle during the test. Try 3dMark11, which is gentle but stresses the CPU enough to provide a fast check. 3dMarkVantage puts much greater demands on the CPU, more than you will encounter in normal gaming. With decent paste, a 920xm at stock settings can finish the 1024M test, without throttling and without temps exceeding 80C, but cranked up to 25x or 26x, it will suffer.

 

Exactly, having everything connected to the heat sink does make it difficult to press the heat sink straight down to get it evened out. I probably wasn't as careful as you are and will definitely try again.

 

Just looking at your numbers made it clear that I need to repaste. My idle temps are in the mid-40's, and they rise well into the 80's - even with multipliers below the max. Will open it up, add MX-4 again, and post.

Thanks for your good advice, Rev.

Joe

 

Joe, there's nothing wrong with mid-40's idle temps. My paste job is brand new, and it looks like a good one, but it will deteriorate over time and idle temps will edge upward. I do like to stay below 45C. The high end temps depend entirely on what you're doing and how hard you're overclocking. Cranked up to 25x-26x or above with enough TDP to let it run free, a 920/940xm is going to hit max temps in the high-80's or low-90's in high stress apps and benchmarks. It's all relative, so it's hard to have any hard and fast temp limits.