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.
Not unhappy! Just curious on how its stacking up to the other people with 9550 i7 laptops.
I realise that 'but I'm having troubles in this thread on i7s' was a bit badly worded. What I meant was that I was having difficulty finding people quoting i7 temps.
P.S. Super interested in your results with the XPS 15 + Razer Core.
Quick thoughts for you based on 9550 postings here:
**Did you adjust "offset voltage" of CPU core, CPU cache and Intel GPU?
**Did you replace the defective VRAM pads? (custom90GT opened ~10 9550s and said every one had this defect)
**Are you using ThrottleStop (and XTU not running in the background as that causes some odd power limit throttling and conflicts?)
**My laptop guru buddy suggested disabling turbo for these tiny laptops as the cooling system is a limiting factor
**You might also run the Asus RealBench stress test and look at thermals/throttling
Thermal improvements of other 9550 i7owners:
So it looks like you got about 10*C via undervolting and 3*C via repaste, for a total of 13*C delta.
I did a quick informal "survey" of thermals improvements before buying the 9550. The survey is not particularly comparable or reliable due to self testing, different configs, small sample size, differences between chip type/quality, etc. But your improvement seems quite good.
if you want to compare your results, I think the following people reported some i7 thermal improvement statistics via undervolting and/or repaste and/or new VRAM thermal pads, so feel free to take a quick search of recent 9550 posts by:
-mrpeaches
-Bullhonkie
-superdavemaine
-Splitframe
-loopty
-R#ph
-Eason (I think he owned both the i5 and i7 but I didn't see his thermals improvements)
-custom90GT (I think he owned both the i5 and i7)
-GoNz0 (has owned at least one i7)
* I adjusted the CPU core and cache but have not touched the Intel GPU at present
* I did not replace the VRAM pads because I didn't have any spare - I did notice mine only had 3 out of 4 actually on the chips (one seemed to be missing??) but the ones that were there looked to be making good contact.
* I'm using ThrottleStop only, XTU has never been installed
* Turbo after undervolt and repaste is actually on 100% time during the prime95 tests so I can get full speed 3.1Ghz from all cores, no throttle. I am very happy with that.
* I will look into ASUS RealBench as per your suggestion
Most my delta was from the undervolting as you say, the 3 degrees improvement from repaste was a bit underwhelming. However I am mostly concerned about the VRAM pads (and lack of one!).
I'm going to look for those posts you mentioned now - thank you.
EDIT:
Looks like my temps are in the same ballpark as other guys here, possibly a bit higher but we are using different paste and it's all self tested/recorded.
So, I guess i should probably be replacing these VRAM pads...
I had forgotten about parkcontrol from bitsum, you don't need the app but you can enable it in power management via a reg tweak https://bitsum.com/parkcontrol/
scroll down to How to Show Core Parking Options in the Advanced Power Options
and download the reg tweak, set core parking to 0% so all cores can park when idle.
it will drop the idle temps by a few degrees all being well and keep the fans off a bit longer.
I can't give temp change because I never bothered to test the unpasted temperatures. I did hit something like a max of 78C CPU and 85C GPU under heavy extended gaming loads, with the 960m overclocked and turbo-boost on full the whole time.
I don't know if you will see any difference in benchmarks but may in extreme circumstances and that is where things fall apart. I also don't know if the GPU ramps everything down when just one VRAM chip starts to heat up. Regardless, this is one of those limit problems whereby one small pad can ruin everything so I think it is worth correcting.
The fix is well documented here, easy and cheap. You also get to see how good your thermal paste job was and redo it.
Once the bugs are worked out, I think these are excellent little laptops!
That ratio of improvement was similar to my experience and seemed to be in the range of other i7 systems.
I have a ton of questions about under volting. I know how to undervolt with throttle stop, but I don’t know the procedure to undervolt.
What programs/combo’s should I use to test undervolts? Is there any way to control fan speed with the dell xps? How can I stress the Intel GPU? How long should I stress the system? Has there been any progress on being able to undervolt the d-GPU? What power plan setting should I use? Do you have any general tips with undervolting? What is the average undervolts I should be able to achieve with the intel CPU and the intel GPU?
what are the ways to lower gpu temperature mine hits 90celcius during dota 2 but my cpu temp is 7X celcius
repaste 2 times already with mx-4 and gelid extreme maybe my repasting suks need advice
also change vram thermal pad to fujipoly 1.5mm 11w
Sadly there is not a ton you can do to keep temps down in the 9550. It's a slim notebook so the cooling is one thing that suffers. If you haven't already undervolted your CPU, I would do that to help reduce the heat soak. Wish we had a vbios that had lower voltages on it...
sad 2 know
i think i will redo my thermal paste next time dont 1 to keep redoing the process
i underclock my gpu using software so at least its below 90 celcius now
Race War!!! Thanks for stepping in mod. Gonzo has helped me out in the past. I can vouch for him.
@custom90gt I would also love to see a vid on repaste. I'm sure a lot of users here respect you as well for all the input you provide to the forum. Any recommendations on what pasting to use besides IC diamond? I have some spare thermal pad. Any advice on what I should do with it? Place it on top of the heat sink? Lol. Serious question. I think I'll just repaste and not mess with UV. I don't feel comfortable playing with xtu settings unless someone can show me exactly what to do to UV. Ironic I know, I'll rather open the thing up to repast. Silly me. I have i7 SIG. Thanks. P.S. I've somewhere else that disabling turbo boost is supposed to help with temps, etc? Is this true? Anyone have a few cents they want to throw at this?
When I'm playing Just cause 3 on my i7 9550 on medium settings, mi GPU hits 92°C and CPU around 89°C.
The CPU is undervolted (-150mV) and repasted with mx-4.
Has anyone else played this game and how did the temperatures look like?
They do look a bit high in comparison to mine, but my gpu% also doesn't seem stuck up at 100% at all times. I don't have Just Cause 3 to do a proper comparison.
Did you check your VRAM pads when you did the re-paste? Some other guys in this thread have been commenting on that.
Might be worth re-pasting again to make sure you have no air bubbles and not too much applied?
I replaced my VRAM pads with Arctic 1mm thermal pads and the one VRAM that was not making contact with the heat sink, with 2mm, so now all VRAM's make a good contact.
My re-paste is good and there should be no air bubbles. It also looked good, when I removed the heat sink after the first re-paste.
When dell sends me a working laptop I will repaste it. Probably can't do a video, but I'll try to do step by step instructions and put it in my initial post. Also I picked up some Kryonaut from Thermal Grizzly that I want to try. Really anything is better than the what dell had (mostly because of their crappy pasting job).
In terms of UV, you'll likely get much better results from it and it's much easier to do than repasting. You really only have to change one setting with it. I haven't used throttle stop yet, but that's what I will try when my laptop arrives.
Sorry, but what setting is that exactly? (my apologies in advance for my ignorance) I know this entire thread is dedicated to this very topic but I have not seen instructions on "how" to uv using the Intel utility yet. Many thanks in advance for any help. Cheers!
Under advanced tuning and core you will see a core voltage offset. This is what you lower to undervolt the CPU. My i5 did a -180mv undervolt while my i7 seems to max at -160mv. It is totally dependent on the cpu, so your mileage may vary.
Ok, last thing. The numbers everyone are using are only three digits so only bring it down to the first three decimal points when you say, for example, -160mv.23398492830 or -125.234334mv? Because there appears to be a whole lot of number galore. I am lost and don't want to fiddle with numbers till I know what I am doing exactly. (lol)
Like Gonzo said, ignore the decimal places. It's not where it only works at specific voltages. It may be totally stable at -180 or -160, it really is a silicon lottery. If it's stable at -160 then it's stable at -159, -158, etc...
Nice, as far as monitoring and testing I know that there are a lot of options. It be great in someone summarized the best ones in the following manner:
Prime95: Super tester (pushes temps to max)
XTU: best monitoring + UV
CPU-Z:
HWMonitor:
Aida64 . .etc.
I don't know what to use as a baseline with so many options out there. Anyone have a guide on how to properly assess your machine to test for throttling, monitoring, tuning etc?
is there a way to control the fan directly?
i undervolted and the temp cores are idling around 37/38.
noneoftheless the fans start blowing after 5min and dont stop for 2..
thats annoying. especially, because they are not gentle but pretty audible in the lowest level..
when you extract it, there is a folder "dptf_cpu.inf".
installed that (try an unknown device in device manager) and the fans keep totally silent in idle/light load here (only unplugged in battery mode..
.
i cannot install "esif_manager.inf" though. service couldnt start. its related to unknown device acpi/int3400.
anyone knows what this is related to and which service it needs?
usually its only set to hipm (win7 AND 10). you will instantly save 1w in idle and your ssd will be much cooler.
get batteryinfoview, look at the discharge rate (mine was about 7.5-8w).
then watch the dip when you activate hipm+dipm (mine now 6.5-7).
Have people really seen a difference by switching from RAID to AHCI? Or is this just a few people on here going the extra mile for something that might not be really worth doing or going through the hassle of doing. Thanks.
As Gonz0 and Bommel87 noted, the 9550's CPU/GPU voltage regulator module (VRM) is poorly ventilated and gets hot. So I would like to fix that.
Your comments and advice are welcomed.
BACKGROUND
The VRM is located at the back of the laptop between the two fan vents. I speculate the biggest heater offenders are the MOSFET chips: 3x Alpha Omega Z5019QI for the CPU and 3x TI 87351 for the GPU (I think as there are 6 ferrite core chokes in the VRM area (Magictec? labeled R22)- and at least 5 other chokes littered around the laptop presumably for other tasks).
I have some stereo gear with MOSFETs; they have high max temp specs but as the MOSFETs get moderately warm they get more inefficient so work harder and get hotter and hotter. This is a death spiral from heat which is probably some exponential function. The 9550 has a minimum number of MOSFETs so each cycles frequently (in a poorly ventilated case) so I can envision the same heat problems, bad performance and early failure.
Desktop overclockers seem to have MOSFET problems as they optimize cooling for the CPU / GPU (often resulting in little to no ventilation around VRM.) Their solutions seem to be
gluing aluminum heat sinks to the MOSFETs and/or simply adding a small fan to blow on the MOSFETS.
- Thermal pads are out as the distance between the MOSFETs and the case is too long
- Gluing alu heat sinks might not help much after the first few minutes if the heat has no where to go
- So I thought improving ventilation slightly could help. As this is an exponential problem, a small change can make a big difference. Presumably (?) Dell's thermal engineers optimized airflow so we don't want to be worsening the situation via poorly implemented mods
A.
AIR INTAKE - The metal case bottom already has vent holes running the width of the case; but a lot of those vents in the middle were crippled with a sheet of mylar/copper. The open side vents allow the two fans to suck cool air from under the laptop and force cool air past the two radiators and out the back of the laptop. I am not sure if any additional air gets circulated around the case by design or oversight.
B.
AIR EXHAUST - I see the vent at the back end runs the full length of the laptop. I also see that there is some grey fabric between the VRM area and the back vent that seems to prevent (some) airflow.
PRELIMINARY IDEAS FOR YOUR COMMENTS
1. To provide VRM air intake,
punch some holes in the mylar at the back/middle section of the case bottom.
2. To provide VRM air exhaust,
move that grey fabric (or remove it or put some holes in it) at the back of the laptop.
3. I imagine frigid artic air from under the laptop being pulled in, cooling the MOSFETs as the hot air from the VRM storms out the back vents.
4. If the convection dream #3 fails, I can think of other options:
4a. Add one of those
big cooler fan units from Amazon that go under the laptop to force some air into the AIR INTAKE (for which Bullhonkie shows some test improvements of 4-5*c, but unsure of VRM impact)
4b.
Add one of those ultra tiny fans from SEPA Japan (10mm x 10mm x 2mm). This is my favorite solution but I would need to figure out where to splice continuous 3.3v without screwing up the laptop
I think the fact the vent runs the length of the laptop is to please the eye, the copper plates designed to pull excess heat faster so the air coming in cools the bottom panel better (but not good enough to start sticking thermal pads to the back of the heatsink as I soon found out)
The same applies to the exhaust vents, they stuck the foam strip there to stop warm air going back into the laptop, this will cause the base plate to get hot and possibly go back into the fan, this will soon get out of control and overheat the chassis, motherboard and stop the cooling process dead. Then everything will throttle.
As the VRM's only go critical when we are daft enough to run furmark I think they are still within specs, not sure how hot they get under heavy gaming, maybe someone wants to check?
If I thought I could/need to cool the VRM's I would have looked into it but I don't think they need anything extra, I doubt that little (expensive) fan will be able to do much more than move the heat an inch anyway?
Heatsinks otoh would be ideal as they pull the heat away from the VRM's and create a larger cooling area, in theory it would radiate to the chassis and cool via the air intake, how much that will impact cooling due to suking in warmer air, well go back to the part about sticking heatpads to the chassis.
So probably best left alone as making holes in the base when heat rises doesn't add up unless you can get the get out with the farty little fan?
I would buy a 2nd hand base off eBay so you have something to fall back on if it doesn't work out.
These all a lot of great suggestions to cool down VRMs, but I might raise the question of what exactly is the aim of cooling off the VRMs?
Yes, they get hot, but they are designed to operate within spec so they shouldn't overheat (there shouldn't be a functional failure). As long as they're <100'C, there's no performance degradation (as per datasheet:
http://www.aosmd.com/res/data_sheets/AOZ5019QI.pdf).
If we're concerned with skin temperature (temperature at the laptop case), the heat has to go somewhere (as you pointed out), so you're really just moving that power from the VRM to somewhere else (whether convected out by fans or conducted out to the case). So we won't really be able to reduce the total thermal output of the system.
If we're concerned with temperature-dependent power consumption, that's a more complex issue. Pulling up the datasheet from the AO VRM, they don't specifically call out temperature-dependent efficiency losses, but there is a plot for VRM losses vs. output current on Page 6. These losses at reasonable current levels for our system are in the 0.8-2W range (assuming 5V output @ 0-10A = 0-50W). The sheet says that the plot is generated using an eval board, naturally convected at 25'C Tambient, so it likely implicitly accounts for temperature-dependent losses (it's not holding VRM temp constant), in addition to VRM efficiency losses. The interesting point here is that for low current levels which you'll see for idle/web browsing, the loss is roughly constant (~0.8W), so you won't get any improvement on your battery life at low power levels. For high load scenarios, you'll get more losses, but again the slope of the curve isn't very steep. If even if we assumed all of the delta between the zero-load case and the high load operating point was temperature-dependent loss, we're still only talking about less than 1W -- and from experience, most of this is likely efficiency loss in the VRM, not temperature dependent power, so cooling wouldn't change this by very much at all.
Gonz0 & runnerguy780 - Thank you very much for taking the time to educate us about the 9550's VRM. Super helpful insight!
It was fun to learn how VRMs work in general and I appreciate you outlining the strengths and weaknesses of the 9550's VRM solution in real depth. Whilst this may have been a "dead end" investigation, I learned VRMs are a strength of good gaming motherboards, so the exercise has some practical value for those (like me) interested in building gaming rigs.
The Cooler Master U2 Plus seems like the correct size for the 9550 and the 2 movable fans are a neat idea. I think the feet should have been less clunky & foldable and a less restrictive/lighter weight grille would have been better industrial design. The mod Papusan refers to is replacing the 5v fans with 12v fans. Also some shrouds to prevent hot exhaust air from being sucked in by the fans might be helpful.
Next time I am in Akihabra (Tokyo), I think I will drop by those pick-a-part electronics stores and build a thin, portable notebook cradle that takes advantage of the two long rubbery feet on the 9550...
Are you based in Asia, or do you travel a lot? One of the best things about being in the region is the assortment of tech parts you can look at in person.
Eason - I am fortunate (?) enough to travel throughout Europe, the US and Asia frequently enough that owning a lot of stuff is not in the cards!
Akihabara's electronic specialist stores are slowly being taken over by maid cafes and cartoon stores but Tokyo is the largest city in the world and still has a significant engineering contingent. So there is a lot of electronics gear and expertise on the ground. The Japanese still are very good at combining hardware and software...
I have heard about the staggering parts supply houses/shops near the Foxcon facilities in China...
Problems? See this thread at archive.org.
