XPS 15 9550 temperature observations (undervolt + repaste)

Thanks for the $20-worth piece of advice. There're plenty of people who believe that stacking thermal pads can destroy a computer...But I guess I can't really go wrong with the 17w/mk pads, carefully stacked.

Great idea. I was also wondering if 1.5 mm pads could cover for the RAMs as well. I'm using old aida64, and as gonzo pointed out in an earlier post, the DIMM sensor is not what it says it is. How's the temps on your RAM chips? I'm one of those guys who would love to have a free 15" heatsink under XPS, as my desk doesn't really complain much about the hot back cover lol.


BTW, I searched around the forum but didn't find much related to undervolting the dGPU (GTX 960M). I'm really interested in turning down its heat since throttling brings it down to 400Mhz and renders it basically useless... I did some researching and tried to mod the VBIOS of 960M myself, in this case the VBIOS should be integrated in the main BIOS, which is the case in some other models that I know. But the after I extracted the bios files from the exe BIOS update from DELL with PhoenixTools, MaxwellBios Tweaker couldn't find any VBIOS... I'm no pro at this but I'm sure DELL couldn't just power 960M without a VBIOS, it's got to be somewhere in the BIOS. Does anyone know what's the usual practice at this point?

Another guy got interesting results with modding the P5 state specs via NVIDIA Inspector. Sort of undervolted the 960M but also lost some performance. Also he has no real idea as to how the process works, so kind of risky....

But whatever method, modding VBIOS / undervolting 960M is sure to produce fruitful results. Really hope that someone would explain how to proceed from here, or that I just overlooked a thread and it's already worked out.

 

Dell doesn't read the ram chips, either way Aida64 now reports the correct DIMM sensor (it's under a DIMM not on it) Dells own BIOS was never wrong it was just Aida64

The ram doesn't get hot so no need for more pads tbh. I had my hand against a DIMM under stress test and it didn't feel warm.

 

Thanks a lot for posting your observation. I'll leave the RAMs alone.

 

I would love to undervolt the 960m but I haven't had time to look around at that. I thought about messaging Prema and seeing if he would be willing to work on it, but again too busy with school to undertake that project.

 

Yeah work should come first. I know the 960M on many models could be undervolted, but I tried the same method with no luck. The work really takes some expertise and experience... Do you have a link to Prema's previous mods? I'd love to take a look see if there's anything that I missed and if necessary contact him myself. The work is either routine or unexplored territory, should be easy to get a perspective on how this might work out.

Again, many thanks to everyone contributing to this thread, undervolting CPU and thermal modding is a huge improvement from light use to heavy gaming!

Edit: Just tried a search for "Prema" (another member of notebook review actually...) combined with 960M, and a couple things are clear:
1. VBIOS could be locked in various ways... making it difficult for us to learn how to mod the XPS without a complete skill set...
2. Prema claimed that he could "unlock VBIOS of all kinds"..hopefully not an exaggeration, and that he might not consider working on an AW13 model with 960M if it's vBIOS is in the sBIOS until the system reaches its end of life, lest having to release BIOS mods everytime DELL updates its BIOS.

My preliminary understanding of our GPU undervolting issue is that:
1. Modding the vBIOS for certain models (XPS 15 9550 obviously included) is indeed difficult, and possibly takes work since we might need to mod the BIOS everytime dell gives an update.
2. If we were to seek help from Prema or other pros, best thing I see us get is he taking a look at the 9550 BIOS and possibly teach us how to mod this specific BIOS. We XPS users could then open up a thread updating the modded BIOS if necessary so that Prema wouldn't worry about having to follow up on the XPS BIOS releases and doing extra work.

Love to see some more discussion on the 960M undervolting.

 

Yeah I would love to undervolt the 960M, my 980TI runs nice and cool undervolted, I would imagine the 960m could greatly benefit as well.

Here is a link to some of Prema's work
https://www.techinferno.com/index.php?/forums/topic/2442-biosvbios-mods-prema-mod™-stock/
https://biosmods.wordpress.com/

Also, I think he extracts the bios so you can flash just the vbios on it's own using NVflash.

 

Thanks. Didn't know there's a separate page for mods... It seems that at least in that topic Prema does mostly Clevo, which doesn't really lock vBIOS... No sign of 960M... When I get some free time I'll do more research on existing mods. To me, it's great new already seeing that some people seem to know their way around locked vBIOS, which means hope for the freaking hot and throttling 960M in XPS 15...

I've posted a discussion in another forum (different website) about how DELL could've or might've adopted the new 965M (GM206) instead of 960M(GM107). Mind the anachronism but the TDPs are actually not much different, with some sources putting 960M even hotter than 965M, and vBIOS readings from CLEVO laptops putting 965M at 50W and 960M at 43.5W. Anyhow, just a fun thought since the new 965M beats 960M by a blazing 40%.

Also, I'm not really the type of person who follows NVIDIA's new unreleased series but just saw multiple posts claiming that 1060m is much more powerful, even equaling 970m, and runs only at 960m's level of power consumption. Would be nice if Dell decides to release a new version of XPS 15 later this year or early next year, solving the screen issues and actually putting in a decent graphics card. Guess I'm just tired of 960M throttling and knowing that it's only a tuned up version of 860M.

 

WHat is your temp when your memory goes down to 400mhz? Is your core still at 10xx mhz? It sounds like maybe you're running with some kind of TDP or other throttling. Make sure "PCI link power management" is off, and in the control panel the GPU is set to "maximum performance".

 

Run GPU-z while playing and when it lags drop out and see what reason it gives on the history graph.

 

Core is throttled to 400Mhz when I run Furmark for a while. It's about 70+ degrees Celsius when the throttling happens. Seen some others struggle with throttling in GTA5, too.

 

only 70C on the GPU? That's impossibly low for furmark. What's the limit reason, according to GPU-Z?

 

Seems that when it goes over 75 it starts throttling and before long the core clock is around 400Mhz. I didn't check GPU-Z, thought it was TDP throttling or something temp related. How did you do on Furmark regarding temps and throttling?

 

The cusps in temps correspond with the core clock throttling, thought that was the reason...

BTW, I saw you recommend MX-2 MX4. I've been using MX-4 myself for the good specs and affordable price. But MX-4 has a poor durability. With my old laptop, 9 months after I applied MX-4 the temps went up 20 degrees Celsius. Retailers say that usually MX-4 lasts 1 month, GC-Extreme and XI-3 3-6, and TX-4 a good 6 months. I'll test on my laptop over the months after I re-paste, but it does appear that MX-4 might not be the best choice... Just wanted to let you know.

 

I hit about 85C after 10 minutes of furmark on the GPU.

If you want to solve the problem then you need to determine why it's throttling. GPU-Z is the first step.

Again: are you running on high performance mode, with PCI-link power management off?

 

Yeah. No throttling during your test?

 

None, never.

edit: also make sure if you have the dell power management utility installed that you're set to ultra-performance.

 

\

That's quite encouraging, thanks I'll take another look at my system later, probably improperly configured somewhere..

 

You shouldn't run furmark on a laptop

Sent from my SM-G920F using Tapatalk

 

Yea I heard people burn their chips with long Furmark runs... never heard 10 min runs would cause permanent damage though...anyways, better to refrain from heavily relying on Furmark

FYI, I decided to get GC-Extreme for its positive temperature reviews and better durability compared to MX-4. I already have more than enough MX-4, but I really hate to re-paste every other month. With that purchase I also got a few Laird T-flex 740 1mm pads, rated at 5W/mk, a cheap but effective solution for the loner VRAM chip.

 

To be fair using any of those pastes would be good for years as it doesn't degrade like cheap crap, also furmark is more for stress testing overclocks with good cooling, one thing a laptop is missing. I never understand why people think it is ok to push laptops so much, it equates to sitting with your car at redline on the drive with the rad covered and wondering why it blows up a few months later

Yes I use furmark, and never run it for more than about 30 seconds as all I want to do is make sure I did a good repaste.

Now my GTX980 with it's EK waterblock plumbed into a 140.3 copper rad overclocked and overvolted will quite happily sit with furmark running with a core of 39c to 45c & I wouldn't hesitate to leave it on for 24 hours.


Even running Defence grid 2 at 4k throws up a performance and volt cap straight away as the chips got a very limited thermal envelope.

 

Haha, good analogy. And it makes perfect sense to me now. Any benchmark software, e.g. aida64, could readily determine a system's cooling capacity with comparison, no need for Furmark here. I guess the only place I'll be using it in the future is if 960M gets a vBIOS mod and I test it for undervolting stability.

But I disagree with u on the TIM durability. Besides the common wisdom of the forums, I myself have a laptop that I upgraded and converted into an NAS, 9 months ago. It's an old model, Vostro 1088, but nonetheless a perfect example because I noticed that it gets really hot now. Right after I did the paste, I ran the AIDA64 stability test and the CPU stayed nicely below 80. Yesterday I did another test and now it throttles at 90. Everything else went up 10 degrees Celsius as well, WITH the CPU throttling to keep itself within the safety range. Basically no better than the 6-year-old factory job, which keeps the CPU around 85. It's not a perfectly fair comparison since I replaced the factory T6570 with P9700 (lower TDP), removed the ODD, added 850 EVO, and other stuff. But it's safe to say that at least the MX-4 seriously degraded/dried up over the last 9 months.

I've always hesitated to re-paste the Vostro 1088 since this old model require that I take everything apart before I reach the CPU. But I guess for my XPS's sake I'll find time to re-paste it and see in a few months how it goes. I haven't had enough mileage on the 9550 yet so it wouldn't make a good comparison, and I'm not planning on opening the back cover every now and then since it seems that the clips are easy to break.....urgh...

For now I'll say that MX-4 is definitely good for its price tag, but could fall a little short on durability. But just to be sure I'm going to test it myself. I'll keep you guys updated!

 

Used MX2 quite often but the best I have seen from personal experience is good old AS5 as I went back to a CPU I did about 5 years ago and it was still doing good temps when I cleaned it up.

As for those clips they do not do much to be fair, I got better results attaching the chassis stand off's with a dremel to make the case mate better.

 

Interesting. Could be that AS5 has silver in it, less "volatile" than mx-2 and mx-4? It sort of runs with the liquid metal family. (Though I think as5 is non-conducting?...weird...) I wouldn't use liquid metal since it's corrosive and dangerous If leaked.

I don't quite follow... Had to look up "Dremel" and the only explanation from urban dictionary is confusing... I get the idea though

 

It's been discussed previously in this thread, the problems with the xtu undervolt resetting and different solutions have been suggested. Those solutions include adding XTU to programs launching on startup, but my undervolt also resets when the laptop goes to sleep. Has anyone succeeded in finding a working solution to this?

Edit: The batterylife on my XPS is a bit lower than I expected, what is the typical CPU package power at idle for the 6700HQ? HWInfo reports mine at around 5W, is this to be expected?

 

1.5 w or something. use throttlestop.

 

Okay, thanks. Power usage is now close to 1.5 W, don't know what caused the value I had earlier. I'll try throttlestop.

 

Use ts to check your c7 core status %. It's more important than cpu usage% according to task manager. Just close the apps one by one that are taking up the cpu until you are above 90% c7 state

 

Okay, thanks for that. Just to be clear: You didn't suggest to use throttlestop for getting the undervolt to not reset in your previous post, just to find out why my value was so high?

 

no I suggested you use it for everything

 

Finally gotten around to finding out how ts work, and the undervolt seems to stick - thanks!

 

I just ran AIDA64 stability test on my old laptop with MX-4 again. Didn't re-paste when I took it apart because there was literally piles of dust in there, only 9 months after I last cleaned it up... So I ran AIDA64 again to give the 9-month-old MX-4 a fair test, and guess what, the temps are nearly identical with the ones I recorded when I repasted the laptop 9 months ago. Only the idle temps are higher but I wouldn't blame MX-4 because I was still using win XP 9 months back...

So, MX-4 is definitely up for the task, at least for low-pressure laptops like Gonzo suggested. And guys, do remember to clean up the fans and heat sinks twice a year, or at least once a year if not heavily used. While you're at it, repaste the chips and you won't have to worry about performance loss.

BTW, 4 grams of MX-4 can be used on a dozen chips. At only a couple of degrees hotter than the top greases and much smaller price tag, MX-4 (or MX-2 for even higher cost-effectiveness) is the way to go.

 

wow, just wanted to say how this helps so much. I noticed Overwatch would eventually result my PC in throttling to below 2Ghz, resulting in slowdowns during more intense portions of the game. Since the game is heavily multi-threaded (it seems to use all the cores, though the manual says 6 max), it heats up more than other 3d games I play. I did a conservative undervolt of -125mV and it doesn't seem to drop below 2.6Ghz now (it actually stays turbo boosted for a good portion, and this is all I would ask for in such a thin notebook). I'll have to play for a few hours to know for sure long-term, but the short-term tests look positive.

edit: ignore the above. it is better but it eventually goes below 2Ghz after 2 hrs of gaming AND with a cooling pad, even though the temps for the CPU core actually dropped to 60-70 because of CPU cap, and the GPU stays in the upper 70s/low 80s. I think something else is in play here; almost feels like a bug. When Pascal comes out for notebooks I'm going to buy that...

 

Last week I undervolted the core/cache by -170mv (and -125mv for the Intel GPU) using XTC. Those voltages are stable for my i5 9550. Yesterday I did some re-pasting and re-padding. Results were excellent overall, with one major question mark...

Yesterday's work:

- Cleaned the CPU/GPU area with ArtiClean 2 part solution and 50+ sheets of dry camera lens paper. Key goal was not to scratch surfaces by dabbing with a clean surface every time and not rub with pressure, much like a pro photographer does with his lenses.

- Repasted the CPU/GPU with Gelid GC-Extreme. Did a test run, cleaned up and repasted a second time.

- Replaced the 4 vram thermal pads with Fujipoly 17 w/mk (3x0.5mm and 1x1.0mm for under the heat pipes).

Last night I tested it under Prime95 small fft's for 15 minutes and looked at the temperatures the last few minutes as indicated by XTU:
Stock - 75*C
Undervolt by -175mv - 62*C
Repaste and undervolt by -175mv - 56*C

XTC benchmark (score & max temp):
Stock - (854) 68*C
Undervolt by -175mv - (855) 57*C
Repaste and undervolt by -175mv - (859) 51*C

So the undervolt was good for ~12*C. Plus the repaste/repad was good for another ~6*C per these tests. For sure, the fans are running slower all around.

However, the 15 minute RealBench stress test was a bit disappointing. First, in all scenarios, the CPU seems to run at 2.8ghz for the first half of the 15 minute test, then throttles down for the second half; I have eyeballed and estimated average throttled-down CPU speed for that second ~7.5 minute period. Second, CPU temperature average last few minutes of test:
Stock - (1.3Ghz) 86*C
Undervolt by -175mv - (2.0Ghz) 76*C
Repaste and undervolt by -175mv - (2.2Ghz) 74*C

*** Why are the RealBench results post-repaste "disappointing"? Not sure. I tried to use the thinnest thermal pads possible; they make contact with the heat sinks but are definitely not being crushed like the soft-fluffy factory ones were. I will try 3x1.0mm and 1x1.5mm for under the heat pipes and see if that makes a difference.

_______________

For those keeping track of vram solutions, I believe that Custom90GT used the thinner thermal pad set (3x0.5mm and 1x1.0mm); GonZ0 used the thicker thermal pad set (3x1.0mm and 1x1.5mm); AMD_i7 did similar. Also Eason noted the TWO left chips on his unit were not contacting the heatsink.

Primer for removing thermal paste
http://forum.notebookreview.com/thr...ratch-your-cpu-die.752647/page-4#post-9662435

Primer for applying thermal paste
https://www.pugetsystems.com/labs/articles/Thermal-Paste-Application-Techniques-170/

Primer for thermal pads
http://www.overclock.net/t/1468593/r9-290-x-thermal-pad-upgrade-vrm-temperatures

 

Nice summary for the VRAM padding! Looking forward to the tests on thickness.

I think RealBench stresses both CPU and GPU, so the temps are higher than stressing CPU alone, ergo throttling. GPU+CPU stress almost inevitably lead to throttling, but that kind of load is way beyond realistic even for heavy gaming so I wouldn't worry.

 

Quick question: In ThrottleStop should I be only undervolting the CPU Core, or both the Core and Cache?

Also, can anybody recommend a program for GPU over/underclocking? And is there a way of manually capping or controlling the fan speed?

Cheers

 

Core and clock are generally the same UV. Igpu can be undervolted more but it's more sensitive.

I use nvidiainspector to OC, and use task scheduler to run the shortcut to overclock on login + wake

 

'Core and clock', do you mean 'Core and cache'? I don't have a clock option for undervolting in ThrottleStop.

Also, I don't think it's actually being applied. Using XTU with -160mV at 900MHz it's around 620mV. With ThrottleStop it's around 750mV at 900MHz and toogling the voltage offest on and off doesn't do anything. Is there some option I need to enable for ThrottleStop to work correctly?

 

Sorry i meant cache.

You can't run ts and xtu at the same time. It gets all ****ed up because they try to set wrong values. Reset to defaults in xtu and uninstall it then use ts. Make sure you set the uv values for all the profiles you'll use.

 

XTU and ThrottleStop don't play well together. I also note that ThrottleStop does not always activate changes.

Perhaps unrelated, but upcoming ThrottleStop v8.20 will have some Skylake bug resolutions:

Finally, for new users, to get ThrottleStop to work, I needed to install XTU first; perhaps XTU unlocked the tuning controls.

 

I will try uninstalling XTU. But I already fixed it - by activating cache and core undervolting at the same time in ThrottleStop it's working properly now.

Thanks for the help

 

That's right. On the 6th Gen CPUs, Core and Cache voltage in ThrottleStop must be adjusted together to the same values.

 

Yesterday I repasted and repadded my 9550 for a second time (to put thicker pads under the VRAM). First repaste/repad results are summarized above. Second repaste/repad results are summarized below:

QUICK SUMMARY

===> ~17*C total thermal improvement by undervolt/repaste/VRAM repad (as indicated by XTU in Prime95, XTU benchmark)

===> First & second repasting & repadding (thin THEN thicker pads) results were virtually identical (YMMV based on how "warped" your heat sink unit is)

===> Given flakey behaviour of XTU in GPU-intensive tests, you might consider [EDIT- DELETING XTU ENTIRELY] and testing ThrottleStop undervolts for scenarios of unexplained "Thermal Power Throttling" (e.g. video gaming)

YESTERDAY'S WORK (DETAILED)

- Continued to undervolt the core/cache by -170mv (and -125mv for the Intel GPU) using XTC

- Re-pasted a second time (same Gelid GC-Extreme)

- Re-padded a second time with thicker pads (4 vram thermal pads with Fujipoly 17 w/mk (2x1.0mm and 2x1.5mm for left side by the heat pipes)

- The Prime 95 small fft's, XTU benchmark, and RealBench Stress Test results were virtually identical for the first & second (thin & thick) repads

- Per XTU, the undervolt was good for ~12*C. The repaste/repad was good for another ~5*C in tests 1 & 2; (test 3 results were closer to 2*C but I think XTU issues invalidate test 3 results--see below)

- Special thanks to GonZ0, Custom90GT, Eason, SuperDaveMaine, etal. for taking the time to document their upgrading processes and to provide some great intelligence to the community.

ISSUES NOTED WITH XTU

NormenKD noted a odd behavior relating to "Power Limit Throttling" last weekend on his laptop. This caused me to look deeper. Undervolting with XTU I observed some odd "Power Limit Throttling" at 100% on most RealBench Stress Tests (15 minutes).

I think what happened was the NIVIDA card hit 80*C really quickly (as every other temperature sensor was mid-60s, or less), triggered "Power Limit Throttling" and dragged down Package TDP & Core TDP (bringing core frequency towards 1ghz). But maybe not.

However, when I reset volts to factory offset, closed out XTU, and undervolted with ThrottleStop, the RealBench Stress Test (15 minute) results were excellent: package power creeped down from 25W to 20W but the CPU stayed pegged at 2800mhz and the NVIDIA card stayed pegged at 1163mhz. As a sanity test, I tried running the stress test again with ThrottleStop undervolting (with XTU open solely for monitoring purposes and not for undervolting). Given that package temperature was ~70*C in all tests, I am confident the ThrottleStop undervolt was applied.

FYI - I found it difficult to make XTU and ThrottleStop voltage settings stick 100% of the time, particularly after shutting the laptop down [EDIT - KILLING XTU SOLVED THIS ISSUE]

PRELIMINAY CONCLUSIONS

===> Excellent thermal results from: undervolting, repasting CPU/GPU, and repadding VRAM

===> Undervolting with XTU increased throttling significantly in RealBench Stress Test

===> Undervolting with ThrottleStop provided much better performance; un-throttled CPU speeds in said tests

===> Given flakey behaviour of XTU in this test, you might consider DELETING XTU FROM YOUR SYSTEM and trying out ThrottleStop for scenarios of unexplained "Thermal Power Throttling" (particularly CPU intensive)

 

I always recommend using ThrottleStop OR Intel XTU but I would avoid running both programs at the same time. There is a single register in the CPU that controls pretty much everything that is important. Full voltage regulator control including all of the various CPU voltages, the multiplier when overclocking and voltage monitoring is all controlled through a single register. It is a clever design when one register can control so much but this elegant design is a problem waiting to happen if multiple programs are all trying to access the same data at the same time. Let's just say that ThrottleStop is well aware of Intel XTU but Intel XTU is oblivious to what ThrottleStop is doing. XTU seems to have no idea when another program or the bios makes changes to the CPU via this register. XTU goes on assuming that it is the one and only program on the planet that has access to this register. Sometimes the data and voltages XTU displays may not be correct and the voltages that Intel XTU claims your CPU is set to may not be correct at all.

When comparing programs, I challenge users to do some proper benchmark testing. Run only Intel XTU to control your CPU and see what sort of benchmark scores you get. Next, reboot, completely exit Intel XTU and any of its helper programs in the Task Manager and only use ThrottleStop to control your CPU. Use the same voltages in both tests. I have yet to see a test where XTU outperformed ThrottleStop.

You need to run ThrottleStop every time you start your computer. ThrottleStop does not interact with the bios so it will only apply the voltages that you have previously saved after ThrottleStop starts. It is simple enough to add ThrottleStop to the Task Scheduler so it runs every time you start up your computer.

While only running ThrottleStop, can you post an example of your voltages not being applied correctly? Open up the ThrottleStop FIVR window and post a screenshot so I can see the voltages in the monitoring table at the top right. The voltages should be applied consistently as long as you previously saved them in ThrottleStop. When entering Stand By mode, most voltages are automatically reset to their default values. ThrottleStop is designed to restore your CPU voltages as soon as you resume and it continues to check your voltages on a regular basis. The FIVR monitoring panel should show this.

In the FIVR window, if you know what you are doing, I would use the option, "OK - Save voltages immediately". If you ever accidentally screw something up, delete the ThrottleStop.INI configuration file, run ThrottleStop and start again. As long as you are slowly adjusting your CPU voltages and settings and putting some thought into what you are doing; you should rarely need to delete all of your settings but this is a simple fix if all goes wrong at 3:00 AM.

 

Is there a need/use for undervolting anything besides GPU/CPU/Cache? There are options like I/O etc..

 

The CPU Core/Cache and Intel GPU are the most important ones. After that, maybe the System Agent might be of some use but the other voltages rarely need to be tweaked. They were included for special situations.

 

Thanks for all of the info and of course for publishing ThrottleStop!

 

Hello,

I've been redirected here from another thread. I seem to be having Power Limit Throttling issues when gaming. I've used XTU to monitor my laptop and have not changed any settings through XTU. While gamine, I've noticed that when the Package Temperature reaches 79C in particular, my gaming performance drops drastically. In XTU, I can see the Power Limit Throttling up at 100% and my Core Frequency down to 0.80 GHz. I have tried to address this problem once before by following what someone wrote down on the dell forums which seem to help a little, but the problem is still there and I could not find any answers regarding this matter there. In the latop's power settings, I've set the Minimum Processor State to 5% and the Maximum to 80% (By default they were both 100%), from what another user has done and said it worked for him. Since I've been redirected here, my guess is I should download ThrottleStop to fix this problem?

I am using a Dell XPS 15 9550 (Core i7-6700 2.60GHZ, 16GB RAM, GTX 960m 2GB RAM, 512GB SSD).



Thank you for your help.

 

Hi guys,

I re-pasted my XPS 15 i7 6700HQ with MX-4 as per suggestions here, and I have also undervolted by -155mv

Running prime95 on the small FFTs test (top one) renders my CPU to a temperature of ~79 degrees C. Before the re-paste but still with undervolting it was roughly 82 degrees C.

A lot of people seem to be getting better temps than this from prime95, but I'm wondering which option they're using to test? On a 'blended test' mine runs a lot cooler in the 74 degrees C range

Running without undervolt the temperature sky rockets and I see some quite harsh TDP throttles. (prime95 small FFTs, 90+ degrees.)

I see a lot of i5s getting temperatures much lower than this, but I'm having troubles in this thread on i7s. Is this considered 'normal' for the i7 XPS'.

Side-note: Haven't touched undervolting on the iGPU - should I?

Thanks

 

That's a great prime 95 temp. Not sure why you aren't happy.