Dell Precision M6700 Owners Thread

The fact that we are seeing different variations of temperature is interesting as you say. Clearly the heatpipe and radiator are OK or we would see this across all cores so it does seem to suggest either the heatsink is warped(unlikley but I will check when I repaste), not tightened down consistently, and/or the thermal paste has been poorly applied.

I am going to do some research and order some good thermal paste which will may arrive by the weekend so with any luck next week I will be able to post a set of temperature readings like I did before after the same workload for a post re pasting comparison.

I do like to play a bit of Rise of flight, this is the only thing I use windows for. I have the K3000M. It's interesting that dell clocks this chip at ~324Mhz - half the speed nvidia rates it at. I have pushed this up to 550Mhz with +150Mhz on the memory clock and it still only just touches 60C at times usually hovering around 57-58C while playing rise of flight with the CPU is around 70-80C. During my normal work in linux it clocks the memory back from 1400Mhz to 324Mhz as its not doing anything demanding and hovers around 46-48C.

 

I wouldn't go that far. But a closer examination of the venting and/or fans on both machines would be necessary before making such a determination. Don't you think?

That's not the scientific method. For one there's no way to know what the conditions were or the accuracy of the equipment when all these temperature were determined. Without that, these temperatures are anecdotal at best.

Ok, but could you please try and recreate the conditions exactly between test?

Incidentally, has anyone contacted the OEM? I should be interested in hearing what they think about all these variations in temperature.

 

Yes I agree.

My intention was not to put too much time into this as I am time limited. My tool for measurement is i7z_64bit. It is recording temperatures measured by the chips built in sensors. What I see consistently is that while idle the four cores tend to stay very close temperature wise. While under load cores 2 and 3 tend to be about 10% hotter often hitting 104C and also I have noticed the CPU does seem to reduce speed slightly while this is occurring. What I was going to look at specifically was the difference between cores 2 and 3 and cores 1 and 4. If these were noticeably closer and the peak temperatures of cores 2&3 were closer to the current peak temperatures of cores 1 and 4 I would consider that an improvement.

I can probably improve on this somewhat with the time I have. I can modify i7z_64bit so it traces the average and peak temperatures of cores 1&4 and 2&3 during a run. I can boot my machine to the console in a room with a reasonably controlled ambient temperature and leave it to warm for an hour. I can then run i7z_64bit while doing 100 make -j 20's of the kernel source. I can then note the peak and average temperatures of the two pairs of cores. The peak temperature should be seen to hit 104C and the average should show cores 2&3 being about 10% hotter then 1&4 during the test based on my observations so far.

I can then repeat this exact process after the re-paste. With any luck the peak and average temperatures of cores 2&3 will then be much closer to 1&4. That's about as scientific as I can get with the time available but should do a reasonable job of controlling the many variables.

If I do as described above and a careful job of re-pasting and refitting the heat sink it should give us a fairly good idea as to if these variations across cores can be the result of poorly pasted or fitted heat sinks.

 

I have i7_3920XM, and running Prime95 my cpu temps max out at 98,104,101,103. It runs very cool in normal operation though (like 60 or so).

 

I have been looking at IC Diamond thermal compound.

Anecdotal data is sufficient for me... The fact that different people are experiencing temperature differences over different cores is also telling.

I have considered this but I have read some horror stories about how Dell techs handle the machines when doing repairs. I think I could do a better job and use a higher quality thermal compound, but I am a little apprehensive about delving into the depths of the laptop; however, it seems that accessing the processor is not a major operation. I am also a little concerned about how repasting the heatsink could affect the warranty.

For me, I have noticed that core 1 seems to be about 10% cooler throughout the usage range (even at idle).

I agree. If I could get all the cores to perform closer to core 1 temperature wise, I would be happy.

Agreed.

 

It's not the heatsink alignment/installation, it's the heatsink's limitation. It can't handle the heat of your XM chip. You can reduce those peak temps by a few degrees by applying a higher quality thermal compound but it will still hit the limit and throttle down occasionally. That was one of the deal breakers for me when choosing between business class and power laptops. I wanted a M6700 so desperately but after getting one for a couple of days, changed my mind. I wish Dell would invest more money into cooling these beasts. I don't see any point of making a "business grade" machine and equipping it with a XM CPU when it can't cope with the heat. HP is even worse, they put both the CPU and GPU on a single fan and when both are stressed, the CPU would throttle down even faster. So, those of us who need CPU power for our work are screwed big time.The best bet would be grabbing a 3740QM and getting a XM heatsink separately from Dell. That way the CPU will at least work without throttling which on the long run would mean better performance.

 

I am getting extremely frustrated waiting for the M6700 with Wacom multi-touch. This was supposed to be available now. Was it held up by Windows 8 issues? I'm about ready to go buy a used desktop computer instead, as an interim, because I need a Cad/rendering computer. Can anyone tell me when the M6700 with Wacom multi-touch wil reallyl be available?
Also, if I buy a Wacom pen, will I be able to use it with the M6700, as with a normal Wacom tablet pc or digitzer pad?
Thanks, Mike Reilly, Architect

 

You are certainly a hard man to please. But consider that this present class of workstation laptops are the most powerful ever built. They work at speeds and capacities comparable and surpassing all but the very top class of desktops.

That kind of computing power would have been unheard of in a portable device even a few years ago. Not to mention that cramming that much computing power into a 17" form factor is a monumental achievement. Particularly when you consider only a relatively small segment will ever purchase or need that much power. Just compare a 4 year old 17" laptop to one of today's and you'll see what I mean.

To imply that that heavy CPU users are "screwed" is unfair. And you should also keep in mind that most buyers want a compromise between size/weight/power/functionality, etc. Nevertheless, everyone had his preference and opinion, and you most certainly have the right to voice yours.

 

I am not sure that this is the case. I don't see why the entire chip could not be operating at the temps of the coolest running core(s). If the rest of the cores ran 10% cooler, the temperatures would be more than adequate.

 

I think perhaps my use of the word throttle was not a good choice. At the end of the day it is a 3GHZ processor. Under load turbo will take it to 3.7GHZ on four cores. It seems at least with the XM processor it will stay at max turbo - 3.7Ghz all the way to 105C. A better way to have phrased it may have been 'come off max turbo' rather then throttle.

It does seem that the system is within a hairs width of sustaining max turbo continuously(at least at the ambient temperatures I tend to work). I dont see why 2 cores should be running consistently about 10C hotter under load. If we assume that under the situation where I see one core hit 105 and the processor comes off max turbo for a second that the heat pipe and radiator are at maximum capacity and cannot move and dissipate any more heat it is still conceivable that if 3 or 4C were taken off the two hotter cores and added to the too cooler cores the highest temperature of any core would not exceed the 105C point at which it comes off maximum turbo. Also if there was a better thermal interface between chip and heatsink the total amount of heat that can be dissipated would likely increase slightly also. If there was less gradient between the chip and heatsink the heatsink would reach a higher temperature, ultimately the radiator would reach a higher temperature(very slightly) and heat would be dissipated to the air at a slightly higher rate. As I say it is very close to sustaining max turbo. Its important to differentiate between the chip coming off max turbo because a cores hit 105C and doing so because the processes running on it have become momentarily IO bound. That is a much more common occurrence normally.

I don't understand the logic here. Assuming the 3740QM is not more efficient(why would it be) it would not be running faster all other factors being equal. It may be able to sustain its max turbo indefinably(assuming it exhibits the same behavior of holding max turbo to 105C ) but that will still be 200MHZ slower then the XM's max turbo. Not going fast enough to need to come off its max turbo is not going to make it do more work per watt available(whats available being constrained by both power in and heat out). The XM would still spend ~99% of its time running 200MHZ faster dropping back to about the same speed for a moment as one of its cores hits 105C.