9510 CPU temps

Hi all,

I've just taken delivery of a new XPS15 9510, and it's running hotter than I'd expect. I've spoken to Dell & they're sending an engineer round to replace the motherboard & fan assembly, so we'll see what that does. However while I wait for that to happen, I wondered what other people's experiences were, in case it's just a generic problem with the model.

When very idle (CPU ~5% usage) it sits ~55C
2 threads of prime95 (CPU ~26%) and it hits ~85C, with some cores at 100C.
4 threads of prime95 (CPU ~60%) and it hits ~92C, with some cores at 100C.

Adding more CPU load doesn't change things much, so I'm guessing this is where things top out. Compared to my old XPS15 (9550), this seems to be overly hot. With a few basic apps running, it's easy to hit 35% CPU and 85C temps, with the fans running pretty loud. If I switch to the 'cool' setting in dell power manager this will come down to ~70C, but I don't know what else that's changing.

Are these kind of temps normal?

Thanks!

 

Seems out of spec. The TJmax of those 11th gen Intel chips is 100c, so you're actually throttling with 2 p95 threads running. If you hear the fans running, it might be a wonky seated CPU heatsink? I did see something that stood out about your testing though:

Prime95.

That program puts more stress on a system than you can if you tried with multiple programs running at once, except it does it by itself! Especially on laptops, where 4 or 5 ounces is shaved off every place it can be (including heatsinks), higher temps are really expected, but p95 is a rather high-end-of-extreme method of temp stressing. I know the program has a lot of support for system stability because of the inherent complexity in calculating Mersenne prime numbers but laptops run hot anyway, and as a stability test it is mostly used to verify stability of overclock settings. For temps, you could download IntelBurnTest and run a quick linpack to spike the temps and see how it responds.

When you say "few basic apps running... 35% CPU and 85c temps," are you seeing 100c cores at that time too? I would imagine so, but just checking because if the insane stress p95 puts on CPUs.

 

Thanks for the reply, glad to hear I'm not imagining things. I've seen the odd 100C core while just running standard office apps. I realise prime95 might be extreme, but it's simple way of forcing different CPU loads. I see similar temps under the same CPU load from other apps prime95 is just a more controlled way of testing it!

 

What are your system specs?

Prime95 will send even many desktops into the thermal stratosphere. Some laptops can handle it but usually end up adjusting something down (clocks for one) as it just isn't a realistic workload especially if you have AVX enabled across the board.

Try running Cinebench R23 or R20.

What profile are you running? Dell software/bios gives you three options. If you're running Ultra Performance, then those temps are right as you're basically telling your system to not limit power as much as possible and you will hit a thermal wall quickly and after a few bounces off the thermal ceiling, the system will drop to ~75w and if it is still hitting the ceiling it will drop down to stock 45w TDP to keep running cool.

If you're running stock 45w TDP and still overheating then there's a problem.

That seems to be a drastic response by Dell to replace the motherboard and fan assembly, but they could be thinking you have a particularly hot running chip (binning) and address both potential issues at once.

 

Running the Cinebench R23 single core test sent temps to ~80C (from an idle of 60-65), with a CPU usage ~30%. Running the multicore sent it instantly to 95-100C. This is on the 'optimised' power profile in dell power manager, which I assume is what you mean by 'stock 45w'?

If I set it to cool power profile, then the fans go into overdrive and the multi-core cinebench test gives temps ~70C, but that's probably largely because this turns off turbo boost.

Meant to say, system specs are core i7 11800H, 64Gb RAM, 2TB SSD.

 

Load up HWmonitor or HWinfo and watch the wattage pull when running it and see if it starts off 75w+ before tapering down to 45w (or close to it).

Cool prioritizes fan speed to keep the CPU as cool as possible so you get the full (or close to it) fan blast.

If any of the profiles results in thermal throttling all core @ 45w then there is a problem.

If you're thermal throttling at ~75w+ all core running CB23, that's about right.

 

OK, so running HWInfo and Cinebench R23, on the 'optimised' power plan.

The instant I start the multicore test the CPU shoots up to a maximum of 71W and 100C, then thermal throttling instantly kicks in and it drops to 45W over the course of a few seconds. It then sits at 45W, 90C, and thermal throttling is off, while the test continues.

I'm not clear if you meant that throttling to 45W was a warning sign, or if you meant that throttling while at 45W was a problem. It looks to me as though it's OK once it's at 45W (temp ~90C), but it won't even get to 75W before throttling kicks in.

 

If it is thermal throttling at 45w which is the actual TDP rating of the CPU, that's a problem
If it thermal throttles down to 45w then doesn't overheat while sitting at 45w it is running within spec.

Dell (and most makers now) design their laptops to run within spec (45w), but the system is designed to run as hot and fast as possible if able and will bounce off the thermal ceiling for a bit looking for the optimal frequency that doesn't throttle. Take the
same 11800h in your system and see it running in a GE76 for example at scores are much higher and it doesn't throttle even at 75w.

For example, my 9710 (in the sig) starts off at 90w but throttles down to 75w and stays there on ultra performance. The other settings it doesn't throttle at all as it will peak pull ~75w max vs ultra performance.

If I purposely hinder the air flow I can have it thermal throttle down to 45w on any of the settings.

I have a bunch of adjustments and mods I have planned to optimize cooling and get it to stay at 90w when I get the time as I was able to take the bottom off and sit it on a cooling pad and have it hold 90w and score 12700 in CB23 so
I know it is a cooling issue.

 

OK, so it's operating within specs, that's good to know. In that case the question is whether you'd expect a laptop like this to run at 90C when at 45W. I.e. is that the kind of temp that's expected?

 

By way of an update, I was thinking of cancelling the engineer, but let him carry on in the end. The new motherboard/fan does seem to have made a difference. With Cinebench multithread, I see the same behaviour in terms of ramping to 75W before dropping back to 45, however when running at 45W it's now about 80C instead of 90C.

Subjectively I wonder if that's because the clocks are ramping down more aggressively than before, but it's the same CPU, the same BIOS version, and the same OS, so that must be my imagination!

On that basis, I guess there was an issue, hopefully it's now fixed.

 

Maybe just better thermal grease application.

You should check out the ThrottleStop thread. There are some tools there to help improve thermals and monitor performance.

http://forum.notebookreview.com/threads/the-throttlestop-guide.531329/

 

It's high but within spec, but sometimes a new thermal application and/or system board with a better binned chip can bring those temps down some but.....

As you can see your temps are better but it still throttles down to what is acceptable 45w TDP spec but runs cooler than before which is always nice overall which is good for system health and fan noise. If I had to guess, I'd say it was more the new thermal application / pairing but still you shaved 10C off even if it ends up in the same 45w state.

As long as it isn't throttling (and appears to be running even cooler now) at 45w you're running within spec it is just that the turbos can be very aggressive and try to extract as much performance as possible as long as it doesn't overheat. Dell's BIOS will give it a few chances before it backs off. The 9710 has a larger heatsink and a vapor chamber which lets it sit at 75w after backing off of 90w+ but the same principle is applied.

Can't go wrong with TS that's for sure.