The Notebook Review forums were hosted by TechTarget, who shut down them down on January 31, 2022. This static read-only archive was pulled by NBR forum users between January 20 and January 31, 2022, in an effort to make sure that the valuable technical information that had been posted on the forums is preserved. For current discussions, many NBR forum users moved over to NotebookTalk.net after the shutdown.
To reply to your ideas... Yes in a way it can definitely reduce performance by heating the area overall via the thermal pads...but also keep in mind I have super crappy 3w/mK pads. The whole ideas wasn't to dissipate heat via the thermal pad but more to create additional pressure down on the heatsink for proper heatsink contact on the die's. It works quite well and I haven't noticed any major issues with the case overheating.
I agree and can second this. That was one of my combos when testing and it did not do very well.
There is a fine balance that needs to be had for optimal cooling. Quite frankly, it's turning out to be that what we think is best is not really the best...strange I know, but this is our garage engineering in the works here haha...
We're getting really close.... GPU stabilized temp under full load is now 76C. (Ambient temp is ~22.5C).
Although you're using a low rated thermal pad, it is enough to bring the overall bottom lid to its maximum thermal retention, however, due to the large surface area, the large area does increase the overall internal ambient temps. If you pay attention to the PCH temps, along with the hwinfo64 ambient temps (sensors inside on the mobo) you can see the changes in temps pretty much in real time.
I've also been using an infared heat gun and lifting the lid to peak through to get some shots in to measure temps....(to collect data and to validate the sensor readings).
If you're wanting to apply pressure over the CPU and GPU heat plate, I would use thicker rubber pads and/or thick electrical tape.
I've started seeing perfcap "power" throttling for the first time during benchmarks. VRM temps are fine (low 70's). My GPU just goes to 400mhz randomly and then goes back up, saying the max power has been used. This never used to happen before. Could this be related to GPU temps? It's a new mainboard
edit: just found my new mainboard came with bios 1.0.2. I will update it and test again.
Same here. I didn't disconnect the battery either before removing the shield. The sparks came out quite violently and one side of the shield was burned out with a hole on its edge. (I am not sure if the shield already had a hole on it? I couldn't believe the voltage could burn a hole on that metal shield...)
Holy cow guys... that's crazy! lol. I haven't always disconnected the battery when I'm working on the inside but that's usually when I'm doing simple stuff like adding more thermal pads... when I took off that EMI shield I definitely had the battery unplugged...
where'd you hear that? I'm running the latest drivers...took forever to install...like an hour to detect the hardware but it installed with no problems and I've been gaming on it ever since...working great.
I think it is the ampage that did that m8, looks like you found power and earth fortunately without anything serious in it's way.
Also no issues, I wonder if it was from an older driver jumping the next one as the laptop installed with 378.77 then inline upgraded to 378.88
The issue I had was the frigging intel HD driver updating itself back to a lower version, that caused a BSOD and I had to disable that specific driver from being updatable in group policy in the end.
Laptop cooling pad just arrived...hello 2GHz GTX 1050! You can do it! Lol jkjkj...
It will be interesting to see how well this pad can push air up into the chasssisiiisss and lower the fan RPM also. I noticed from GoNo0 how I believe he said he's on fan #3 for his 9550... hopefully I don't tempt the same fate...
Here's Unreal Engine stress test "realistic rendering" which is just a 3d room. Graphics on it are great and it stresses my GPU like nothing else...if I can run it stable in this...it's stable.. Here's the temps with a +260 memory +100 core overclock.
Pretty good I'd say. Cooling pad only helps out marginally...I'd say it might help lower the fan RPM slightly when its not on full load all the time...definitely helps keep the chassis cooler...
OMG! dude that's SOOO much liquid metal on the heatsink...like 4x too much... I'd be scared to carry that for worry that you'll have 1 drop of liquid metal go flying around in the case and land on the motherboard or worse the aluminum frame...
It'll peg your GPU at 100% the entire time the benchmark runs. It's pretty much like Furmark... I can run it because I know my machine has some major mods on it so it's not going to melt down...I just wanted to check for max stability on the GPU...and 0 throttling in worst case scenario...
Many games load the GPU 100%, squeezing out as many FPS as possible. But the GPU still doesn't really work all the time, because it spends a lot of time waiting for data to be loaded from the memory. So the heat released depends on the code efficiency. What I wanted to ask was if this this one drives the GPU temperature higher than Furmark?
Also regarding the padding I guess it is a bit overdone, though I think we haven't got ultimate answers yet. I think padding from the heatpipes to the case bottom is likely counter-productive, harms cooling of the VRM area long-term. Eventhough there doesn't seem to be much air movement in the VRM area, I'd leave more air space. Also the taping of the heatpipes, which hasn't been repeated enough times to tell... I'd only seal heatpipe-to-case, leave the fan-to-heatpipe gap open, because it likely lets some cooling air leak towards the VRM area. But I'm interested in the results anyway.
But better lift the heatpipes an check for LM leaks before you proceed. If you see a leak, remove the excess. If you don't, the gaps between the chips an the copper plates are likely too wide, and you should sort out the alignment somehow (pad thickness etc).
No worries here I used a pin drop as suggested on both sides it spreads very good. looks like a lot but its a pin drop on both sides. On my second test, I lifted the heat pipe to check the spread everything looks good. Thanks for the concerns. But temps are worse without the thermal pads.
with pads installed CPU runs cooler.I ran occt for 1 hour and it doesn't drop as much as it did (800mhz). Still throttles below 3.5ghz to about 2.5 doesn't go any lower as it used to.
I get lower cpu temps but vrm stays the same remove the pads cpu temp goes up VRM stays the same.
Just repasted my precision 5510 (An xps 15, but with an i7 6820hq and a Quadro m1000m) with AS5, made all the difference in the world. CPU rendering in Inventor used to give me 95+ degree cores and thermal throttled to hell, but now I ran it for 10 minutes at 75-80 degrees pulling 40+ watts at 3.21 ghz. Like magic! Thanks for giving me some background.
I've managed to reduce my VRM temps while playing games on 4k from 85C to 75C using this padding configuration:
http://imgur.com/a/Fb5ZJ. CPU and GPU are about the same (70C CPU and 77 GPU), but I still experience some throttling, although to a lesser degree since I drop from 60 to 30 now instead of 60 to 20. It's probably GPU throttling since it only occurs on 4k and not 1080p. My padding config before was just on the two mosfet areas on the left and bottom from the middle-center pad. I've also followed the padding config from Eason85's website, but it was same temps as the previous one. 5mm pads also led to higher temps than 4mm pads, not sure if anyone else had the same experience. For reference the game I used was ESO on 4k/medium settings.
I know I blocked the air flow, but I'll see if there's a significant difference in temps while light browsing.
My CPU speeds doesn't drop as often with the now lower VRM temperatures (it still does from 3.2k mhz to 1.1k mhz, but less often, and doesn't drop below 1k mhz anymore), but my GPU clocks does ocassionally go from 1600 mhz to 900 mhz which causes FPS drops from 60-->25 fps. Don't know if that's normal for GPU that it can't maintain 1600 mhz. Temperatures range from 60C to 78C, and 78C is a guaranteed drop in FPS although there are times when GPU is 60C and core speeds still drop.
Edit: I should also note that switching back from 1mm arctic pads I placed under the VRAM back to stock pads reduced my CPU temperature by ~7C, although it could be attributed to the paste job I did (line on each die). Could be that stock pads were squishier and had a tighter fit.
I suppose it would be best if neither your GPU nuor your CPU throttled...
Interesting - that is a big CPU temp drop. Is the only difference the VRAM pad change?
Overly thick dense VRAM pads might prevent the heatsink from properly contacting the GPU, particularly given the new 3 screw configuration. Not sure the impact on CPU temps should be so big.
~7C CPU temp improvement could a better thermal paste job or more consistent heatsink screw pressure, but there are so many factors who knows...
Not sure why CPU speeds is not a consistent 2k+ mhz since both my CPU and VRM are now around 70C range. I disabled BD PROCHOT/C1E through Throttlestop while on AC power if that matters. I'm unable to game on battery power for more than a few mins or my laptop will freeze and make a beeping sound sometimes even though my undervolt settings are the same on both battery and AC. (-140mv CPU/-120mv GPU)
My thermal paste method is the same (one line on each die), but maybe there was more uniform spread when I put on the heatsink the second time. My screw pressure is the same too, just that I changed the pads. I was surprised to see the CPU temps drops from 78C max to 71C max as well. Unless the pads on the VRM somehow lowered the CPU temps.
Can't seem to find my Prime95 stress test screen shots, so the only info I can go by is my temps while gaming ESO on 4k/medium settings with bloom off/shadows on low with everything else on default. I don't really play the game, it's just the only game I have currently. But I do the same cycle every time and record max temps.
That is an aggressive CPU undervolt. Try -120mv and see if no more freezing.
Are you running SpeedShift EPP=0? That might help lock your CPU speeds.
If temps hit ~78*C anywhere Dell might be throttling system.
Every time you repaste, thermal results will differ. As a general rule less is better. Some people here recommend pasting, bolting down the heatsink, unscrewing the heatsink to see how paste spread and if it is thin enough. Repeat.
As the VRM area runs so hot, good VRAM thermal pads might be taking heat from the motherboard, through the VRAM and into the heatsink. That would be good for the VRM temps but leaves less cooling capacity from heatsink for GPU and CPU
-145mV undervolt on CPU would cause my 9560 to crash while on AC power. I don't game or do anything intensive on battery power anyway, just that sometimes I forget to plug in my power cord while checking max temps and only notice it when my laptop freezes. Does undervolting cause CPU throttling?
My SpeedShift EPP is 0 on AC power. Saw a post from unclewebb somewhere for these tweaks for maximum performance.
Everytime I reapply paste, I noticed the previous paste job on the CPU die goes completely to the sides and there is only minimal residual paste left on the CPU die. The GPU die is completely fine though in terms of paste layer. I didn't check if the paste job was better on the CPU after I changed back to stock pads, just the same method was used. In total, I've reapplied paste 5 times; 4 times with 1mm Arctic pads on VRAM; 3 times with Arctic-MX2 paste, and 2 times with Grizzly Kryonaut.
Only temps hitting 78C now is my GPU while gaming. The second highest is the second ambient temperature on hwinfo64 at 75C, but not sure what that is. Third highest is PCH at 72. CPU/VRM at low 70s, and everything else is below 65.
I did intel xtu benchmark and 5 min cpu stress test, and the highest cpu recorded was 61C for benchmark and 58C for stress test.
On ESO 1080p / very high settings I'm able to hit and maintain the 100fps cap easily without any reduction in clock speeds. But it's pretty blurry on 1080p. Don't know if these fps/clock speed drops are "normal" since the card is not that strong for 4k.
If you can get -145mv to crash your computer, you need to reduce your undervolt significantly below -140mv. That undervolt is probably causing your crashing. Try -120mv and see if it still crashes. You maybe able to sustain -130mv but try to end the crashing for now at -120mv.
Undervolt should not cause CPU throttling. However the bigger the undervolt, the cooler the CPU can run and potentially reduce throttling. Until the undervolt gets too big and crashes your computer. Then you need to reduce undervolt...
XTU stress test is worthless. Use the following to start:
Prime95 small ffts (15 minutes)
ROG RealBench StressTest (15 minutes)
I don't have any Kaby Lake benchmarks for Prime95.
With a Skylake i5 6300HQ Prime 95 temps are around 59*C max on my computer with GeLid Extreme repaste (and a -170mv undervolt). Your Kaby Lake i5 will run hotter as it is clocked faster by Intel and can't run such a huge undervolt so your results don't seem unreasonable.
I don't think you have SpeedShift enabled as I don't see the letters SST in green on the main page of ThrottleStop. If you click TPL, then check the two SpeedShift boxes, then hit OK that should enable the green SST in the main page.
If you want to veryify SST is running you can look via HWiNFO64 (summary only-run, SST will be green). That is not necessary as the green SST text in ThrottleStop is reliable.
BTW - I don't think SpeedShift should impact temps on such a brutal test as Prime95 small ffts
You're right, I didn't have SpeedShift enabled. I've read that the green text means that it's enabled, but couldn't find how to activate it through Throttlestop. I have it enabled now, so thanks a lot. I'll run more tests another day to see what difference having Speedshift on will make. As for the Prime95 test, the only person that I see on this thread (which is quite long) that I can compare to is Philaphlous with an i5 as well. We have same CPU temps, but his PCH temp is lower while my VRM temps are lower. I didn't pad my PCH though not sure if he padded his or modified his since page 102. I've padded my PCH in the past, but decided against it for a cooler surface temperature. I previously though PCH>45C activated fans, but that didn't seem to be the case. Even at 40C PCH and 40C CPU, the fans would still go off. Not sure what causes the fans to actually trigger during light browsing.
Philaphlous is about the only one with a lot of 9560 i5 results so if you are in the ballpark of his figures.
If you want quite fans when web surfing you can try dell command-power manager "quiet". But you should be running "ultra performance" for gaming. That is too much work for me so I just adjust performance with SpeedShift EPP=0 or EPP=78.
Heatsinking PCH gives you less cooling capacity for critical vrm. That alu case bottom can not cool very much...
I know I am going to ask something obnoxious but can someone please post screen shots of exactly what to do on XTU to undervolt? Like use Paint to draw an arrow that points what settings to change? Or just simple write instructions. Step-by-step. I don't know what I am doing but I want to undevolt. What bar to I slide? And do I slide to the left or to the right? Seriously, someone help. Thanks
I've been keeping track of what's been going on in here for a while, just registered to ask for some advice and post some thoughts.
I repasted yesterday with some cooler master stuff that came with a heatsink with disappointing results. Temps up from low 80s to over 90 on the CPU. Figured it was old, cheap paste and ordered some Zalman ZM-STG2 as it was well regarded, with a decent price.
I've since repasted who knows how many times with exactly the same results. Any advice? I've attached pictures below. One oddity I have noticed is that the CPU package and individual core temperatures seem to have little bearing on the CPU temperature (in HWinfo), so the fan doesn't ramp up despite the package and cores hitting 90 because the "CPU" can be a lot lower, any one else experienced this?
Some benefits have come out of this; I used Gelid 0.5mm 12 W/mK pads for my VRAM, as well as MOSFETS and chokes. I avoided the heating of the bottom of the case by putting a double layer of insulating tape over the top of the pads on the R22's. A few of the "ambient" readings are down a few degrees and anecdotally I think the fans turn off faster than they used to.
How long does it take to heat up to 92C? Looking at your prime95 seems like it's almost instantly?
I would say use 1/2 as much paste as you're using. It also looks like the heatsink/heatpipes may be bent with the non-uniform appearance of the thermal paste on the heatsink. You may check it out. I will caution you that if it is bent be VERY careful if you try to bend it back because it's super easy to bend the heatpipes and pinch them off...
Any paste should give you a lot better results than the stock dell stuff. Was your CPU throttling with the dell stuff and making it seem like it was running cooler?
Your core 1 and core 2 temps are 10*C different. That might indicate your heatsink is bent or thermal paste needs to be reapplied. Or see VRAM padding comment below. Delta of say 0-4*C degrees is typcial
A few guys had problems with replacement VRAM thermal pads on the new 9560. The old 9550 had thermal pad fitment problems but the 9560 factory VRAM pads are fine. The new 9560 VRAM to heatsink is really close so your dense high-quality replacement pads may be forcing the heatsink away from the chips. Sounds odd but a couple of guys put their original factory fluffy soft thin (low w/mk) VRAM pads back and thermals improved.
Let's rethink your VRM cooling strategy. I think your idea was to put thermal pads on the R22 chokes to bring heat from the vrm to the case bottom. The electrical tape between the case bottom and said thermal pads defeats that. Try removing the electrical tape and see what happens.
Finally, the 9560 heatsink screws are poorly placed by Dell. I think the best you can do is try to screw it down as evenly as possible. I use a cross pattern that mechanics recommend for lug nuts on a car wheel, tightening every bolt a bit over iterations. Don't overtighten or you could strip the screws or rip the mating out of the motherboard.
Package and cores jump instantly to 70C then climb a little slower until they stabilise around 90C, but its still pretty quick.
I'll try using a bit less paste, but given the truly dire job that Dell did I was certain anything I did would be an improvement. It was never thermal throttling with the Dell paste, and was barely TDP throttling after I undervolted. The only reason I did this was for the sake of reducing fan noise!
It's getting late today, I'll report back after I've tried some careful bending of the heatsink tomorrow.
That's a good point. Given how many times I've reapplied the paste with negligible difference (and I do have experience doing so in the past) it's looking like a bent heat sink. That would make sense with Dells paste job doing better, I guess that much paste compensates for the bend somewhat.
Sorry, my mistake on the VRAM; I had reduced contact with my pads so I replaced them with the stock ones for all but one where there was a visible gap. With this one I just stacked some of the Gelid pad on top of it.
I was really only trying to increase the surface area for the R22's as well as a heat spreading effect. I did try without the tape and the bottom of the case got noticeably warm, I don't really know how it effected thermals because this was all after I did my first repaste... I do plan on doing some more experimenting with this later on, but I need to get some more reasonable results for CPU temps before I do that.
I tried to do the cross pattern like you mentioned on my later applications (or as well as I could do it with seven oddly placed screws!) and it gave a bit more of an even spread of paste but didn't help with my temp problems.
Thanks for the replies so far, looks like I'm in for some metal working tomorrow.
Problems? See this thread at archive.org.
did some tweaks
Last edited: Mar 16, 2017![[IMG]](images/storyImages/33432004336_2cc574e7c7_o.png)
