The Official MSI GT73VR Owners and Discussions Lounge

A reason you have to test with different software. FYI Long or daily prolonged stress tests will degrade your chips (It will need more voltage vs. clocks vs. before). Best tests... Your daily use/workflow + benchmarks. On top you will get a score

 

Is it later Today already?

 

Papusan I am not sure, but you are right, like usual !
I think AVX has a different VID table
I tried 2.9 ghz (2900 mhz only) and did AVX test. At the VERY Beginning of the test, core 3 was instantly hotter than the other cores. Even at just 60C. Cores aren't even hot enough to show heat imbalance. This can only happen if VID is set differently on one core than another core.

You see how in the AIDA Test, if its core 0-3, core #3 is the COOLEST, core #2 is hottest and core 0 and 1 are inbetween?

That is really...like...whatever.

Now you know why I like small FFT prime 95 without AVX. As you can see, lower temps and core balance is very strong. Even @unclewebb likes prime95 (not AVX) because of balance. The best test for "general use" system stability with prime 95--which many overclock.net people use is:

1) disable AVX.
2) disable FMA3
3) prime 95 custom FIXED FFT: 1344K

This will test vcore stability without testing your heatsink max performance.

 

No, I mentioned disabling AVX several times before, and now as it sinks in, enjoying the moment.

 

If you want test the core balance... Try Wprime 1024M. In worst case... Run it without stop several times. Just click start again, and there we go. And you get a score, bruh Same also with Cinebench R15. AIDA64 is a nice test for AVX/FMA if this is what you opt for. But as I already stated... Not all run workflow with this type load daily. P95 with FMA/AVX is for desktops!!

 

Ok well I'm playing league of legends again while my taptop is doing 1344k fixed FFT prime (no AVX). I'll do Wprime later. Wprime is easy though.

 

And Ryzen sucks on Wprime as well

 

1344k fixed fft. next is wprime

 

wprime 1024M @Papusan 2 runs. cuz didnt get hotter

 

those temps ... i will achieve them only without hs.. and at idle btw great temps hopefully ill be much closer after sanding with in between temps

 

Well AVX did make a difference. I disabled it and then ran small FFT prime95 and it ran at 3.93GHz as it should be, temps were also lower and more balanced @3.93GHz. 81/78/78/75 which I find interesting because there is usually an 8C spread between each side.

I also had to drop my undervolt to -95 because apparently when AVX is enabled it increases the voltage 0.30 or something. I got a BSOD 1 minute in @ -100, so I found it interesting that I had a sort of false positive running @3.9GHz with a -100 undervolt due to the AVX increasing the voltage anad making it more stable.

 

i went back to 4.1ghz -100mv.
if just cores were all same as 2 and 4 id be more than happy. well hopefully temps stay at leats like this till end of the march when ill be sanding off hs

 

Did you see my result?
AVX enabled (assuming cores 1-4 instead of 0-3), core 3 is hottest and core 4 coolest.
AVX disabled: all cores within 1C, even at 100 watts!

It's almost like AVX enabled boosts all cores VID by at -least- 30mv, with average VID 30mv higher but core 3 an even hotter vid than the others and core 4 the lowest boost?
really bizarre. If there were a realtime vcore sensor then I could see what was happening. but VID is not vcore....VID does not show vdroop at all.

 

Yeah. I think a 0.1mm (ZERO POINT ONE) mm reduction in heatsink thickness on the bottom side (cooler cores) will bring all cores much closer and then stop pump out.
Before I did the sanding, I saw core temp differences even with LM, although it was usually 3-4C. But that's 3-4C too much for my OCD.

You need to understand just how little 0.1mm is. When you are trying to sand fix a heatsink, you go SLOWLY. it's always better to do too little work than too much. Because you can always re-paste.

That's why really really cheap and inexpensive paste that you can buy a lot of like Arctic Ceramique 2 in those huge 20+ gram tubes, is really useful for stuff like this. You apply it, attach, check core temps, if good, then remove it and move on to the good stuff.

the nice thing is, once you get experienced, if you get normal core temps to within 4C maximum with regular paste (without AVX load and using prime95 small FFT for perfect balance, or AIDA without stress FPU), when you move to Liquid metal, you'll get 1C.

@hmscott @GENOCID @Papusan
When I first used Kryonaut repaste, even Phobya Nanogrease Extreme, and then tested with prime95 small FFT (no AVX), I got 4-5C core temp difference, 6-7C with phobya, 4-5C with Kryonaut.
It was great, but after 1 week, temps degraded. Always first and third core getting hotter. I had sanding kit from 15 years ago, but had FORGOTTEN about it. Moving to LM dropped all my temps but core temp difference was still there. I did multiple repastes and eventually got the core temp differentials to 1C on a FRESH LM application. However THIS also degraded and the core temp differentials rose to 3-5C. Then I was unhappy again. Then I removed the LM and saw exactly what was happening (this is MUCH easier to see what is happening with LM, than with regular paste):

1) lower end of CPU (cooler cores) looked like there was barely any LM.
2) oxidation (hardened/reacted with copper) LM On the hotter (top side by VRM's) cores.
3) LM seemed pushed out towards the top (still on the cores).

So there clearly was a pressure problem.
I thought maybe I didn't clean the heatsink properly enough so I tried polishing most of the silver stuff off. this helped at first, but eventually the temps increased a little!
so basically, with fresh LM, the temps were close. But it seemed like when some of the LM got absorbed by the copper, the temps got worse and always core 1 and 3 (first and third core) getting hotter! So that's a sign of weak contact. Using "more" LM would of course fix this but only temporarily. So I decided to find out the real problem.

So I started doing some polishing (not sanding) AND Switched to 0.5mm pads.
After seeing core 3 being cooler too much, I decided to try a little 400 grit in the bottom right corner area (core #4 area).

I then noticed that now, instead of cores 1 and 3 being hotter than cores 2 and 4, I was getting strange results, like core 1 and core 2 being hotter and the other cores being cooler. But core 3 was no longer the much cooler core now, it was now the same as core #4. So now I sanded the entire right half of the heatsink with 400 grit sandpaper, just enough to wear off a VERY tiny amount of the copper (again, little is more), then I vigorously sanded the entire heatsink down with 2000 (?) grit polishing sandpaper to remove absolutely all of the silver stain from the absorbed LM, even though it felt smooth fully without that, then now you have my results of my OCD hard work.

 

Not the big change in temp vs. your 4.2 Clocks.
41x


42x

 

thanks for this helpfull info

 

I edited my post btw.

 

I would say that is a system engineering blunder or manufacturing defect rather than a CPU problem. BGA aside, there is no excuse for such high temperatures at such low clock speeds if a proper repaste has been done. The ODM should have invested more effort into making a good cooling system, or the heat sink is defective. Seems that notebook engineering mistakes and manufacturing defects are becoming status quo for the industry. It's pretty sad, and that is why I am back to desktops as my main systems. I have a laptop, but generally only use it when I need to travel. The desktop is too superior to use it the rest of the time.

 

i ran it after few hours on 42 and temps went up to 81

 

The biggest Cooling quality/QC change in notebook history is from Dellienware Camp. They had super HS quality and QC. Nothing left. Only a... Is it possible sinking twice?

 

@Mr. Fox has a very good point people should be paying more attention to.
I know that at least in USA, MSI taptops have actual thermal paste gobs on it, but it's pretty clear that at least in some regions (like @GENOCID 's), they're still using the "Alienware" Thermal STAMPS. Yes, stamps.

And why would a company use thermal stamps?
Because of defective or badly designed heatsinks.

How do you cover up (i'm going to be a little nicer than Mr. Fox here) weak engineering?

By throwing on so much thermal goop that the defects get hidden, because fixing them would....um.....cost money? Or get someone fired?

MSI clearly has uneven heatsinks. That's just shoddy QA, complete incompetence or defective machining here.
The heatsinks are well DESIGNED, but just like in some of Clevo's older versions, you can design something well and screw it up with incompetence so bad it would make monkeys asphyxiation on mars make them look like our best historical geniuses while they frantically scribble on the ground with their feet while they are being blown apart by space empty air pressure.
Alienware (ever since the devs left them and the true Alienware became Origin PC, although for a TIME, Alienware was still making kickass laptops, but with flaky components that might need replacing, e.g. motherboard, etc, then crippled later by Dell) is just incompetent. A nice metal chassis with tripod heatsinks and thermal pads of the wrong thickness. That's just bad.
MSI on the other hand, has actual good design , but some serious lack of QA going on (91W TDP power limit on their 16L13 that ignores Bios overrides, cancer EC current protection shutoffs on the MSI GT73VR and GT75VR (set EC RAM register D7 from a value of 00 to 01 and see what happens)--MSI never thought it would ever be possible to pull 110W of power on their BGAbooks without reaching Prochot, badly machined heatsinks that have even less pressure because of VRM cooling, and yet they top it off with the best keyboard you can buy on a laptop.

I mean, my god....

 

Just wait and see. Dell's engineerst put in a power cap on last AW18 (No more possibility of use 2x 330w psu). This after seeing this could be done with previous AW18 models. They will continue working hard on how to cripple their produckt. Just wait and see. We haven't seen the start how bad it can be. A nice looking keyboard can't replace performance.

 

I can't wait to see what Mr. Azor has in store for us for the next Alienwares, the enthusiast dream brand of taptops that he crapped all over

PS: you know you can turn on spelling mistakes checking in chrome right? you no speak Americano brah?

 

My temp after playing Assassin Creed origin for more than 2 hours.

 

Everything written is Pooposan's amazing English. Americano is popular, you know. Maybe the reasons for the many likes


Regarding Dellbooks... Maybe this will be the new BIG from Azor? Nothing beat thinnes

 

Very nice

 

@Papusan

 

Do you guys know any trick how to fix plastic sound when I grab my laptop. I can hear it from bottom cover on the sides. This issue exist after I opened laptop laptop. I have not broke any plastic parts.

 

They stamp the heatsinks. - Refer to this video of a MSI Laptop Factory Visit - @4:40

 

You may have not aligned the case halves accurately and it's not snapped together completely in a small or large perimeter of the laptop. Look around the edge at the gap where the 2 halves meet and see if you can find a larger gap in an area.

The screws should pull the halves together well, but there can still be a misalignment in a long run between screws.

Loosen the 2 or more screws around the gap, and pop it up a little and then wiggle it back into place with no gap, then put the screws back in.

Be care not to strip the soft heads of the screws, they are designed to only be installed, not repeatedly taken out and put back in. You can also strip the thread, often it is only a plastic sleeve - metal sleeve's are few - and it's easy to over tighten and then the screw will keep turning even though it is all the way in.

The watch word is "sensitive", be sensitive to what you are doing, feeling the force and feedback as you work the parts.

Good luck

 

Yup, standard stuff, the work of my life from the summer before High School, through all 4 years of High School, and 3.5 years of College, after which I went software + hardware systems design.

The pre-applied thermal compound is standard across many applications, not just CPU / GPU thermal interfaces. It's a good way to have a measured amount of paste consistently applied over millions of applications.

Sometimes the mating of the heatsink to the device will go awry - shifting the target spot, but that is rare and should be caught by QA. But, when you do 100's a day, it's easy to miss a blind spot.

Now that we know the AVX instructions are skewing the core temperature differential, with the consistent core temperature results @Falkentyne experienced and posted after disabling AVX, it's clear the GT73 doesn't have a design problem at all, and the temperatures are nice and low. Which we've known / observed since it was released.

Hopefully everyone will notice our posts about disabling AVX with Prime95, and we won't see any more panic posts with false positives for core temperature differentials.

I didn't see any on the GT73 I set up either, 1c-2C differentials, within the error of measurement and run variances.

Now, I hope people will listen when I suggest undervolting the CPU, tuning the OC, and tuning the fan curves, investing some time in software tuning, before wasting time on non-essential re-pasting.

Really, you don't need to re-paste, and if you find it needs it, return it before the return period ends (usually 7-14 days, check your receipt), and get another unit.

 

But when someone disables AVX and has a 14C core difference between core 2 and 3, that's definitely a problem. @GENOCID was just running Cinebench, and getting absurd temps. Jack (Forgot his name) also had outlandish core temp differentials on his GT75VR, and having to send an entire laptop to MSI when they can just say "ok here's a heatsink, here's how to replace it" is just...I have no words for that.

These things are NOT easy to ship, and if they don't supply prepaid shipping labels and pay for shipping both ways, well, it's easier to just buy the heatsink directly or sand it yourself.

Yes you're right that heatsink has to be RMA'd, but not if you have to send the entire laptop to MSI for 2-3 weeks or more and pay $70 shipping on top of it (since you know how FedEx wants things packed perfectly and refuses to ship unless you pay for their authorized packing material). They should just agree to send the user a replacement heatsink. As you saw in the Clevo section, more experienced users just opt to fix the problems themselves rather than be without a system. But MSI charges you for a heatsink replacement even if it's under warranty. That's if you can reach Tom (was it [email protected] ?) or someone who has access to the parts. They don't charge for replacing broken fans, but heatsinks...um...you saw how MSI treated @Shehary who had a clearly defective heatsink.

I do disagree with you on not having to repaste. On my stock original heatsink and original paste, I could not even run prime at 4.2 ghz , undervolted (which was your original suggestion for undervolting) without having core temp differences of 12C and the hottest core reaching 90C (!). That's 4.2 .ghz and -100mv. That I have a problem with and that I find personally unacceptable. And yes that's without AVX.

I then decided to do a real world test using a chess engine, 8 threads, at 4.2 ghz, and the core temp differences (almost 10C) was what made me give up and repaste. Stockfish chess engine has a balanced load, similar to non AVX prime. I personally could not accept seeing core temps of 86/78/88/81. That's just too much for me. I can't just accept that. I had to fix the problem.

 

There's nothing wrong with those temperatures

You could have run with those without re-pasting, and for the huge amount of time and trouble you've gone through it's just not worth it for most people to do it.

I wish I could get you to see that for most others you are suggesting a large burden of work for unnecessary improvements.

If you could see your way to helping new people load hwinfo64, get good baseline temps using non-AVX benchmarks and then help them undervolt using XTU or TS using -100mV and then run the tests again so they see the large drop in temperatures at 100% CPU load, and then let them decide what to do.

The problem is you start right out of the gate telling them they must re-paste, and that's just not true for 99% of the owners. That's way too much work for them to commit to - even if they don't know it - it's going to occupy their lives for weeks, it's just a waste of time.

That's the problem I have, you are burdening everyone you reach out to with way to much, instead of working from a small ask through to a resolution that fits 99% of the owners, you jump them to the 1% solution.

I'd appreciate it if you could see your way to amend your method and start people out with small asks and work them up to being done quickly - without any case cracking hardware mods.

You'll find you'll help more people in the same time, and not chase away all those that read your posts and get scared off right away due to the commitment of effort they see required.

Then you'll still have those OCD people that need to feel good about getting involved in re-pasting and extreme tuning, but they will speak up for themselves and ask for it, so to speak.

Even then, I try to make sure they have experience doing these things and a real motivation that will carry them through to completion. Nothing is more disappointing that having someone you are helping give up and either stop posting or need to RMA and stop posting due to feeling like they failed.

So far I've seen these guys you are helping spend far too much time re-doing this over and over and still getting the same results, and at least one RMA'ing to get things back to stock + improvements from MSI.

You can avoid that, filter out the easy fixes first with undervolting, fan tuning, and general use advice to get them running and off playing games or whatever they wanted to do when they bought the laptop, and what is left will be the guys with enough talent, experience, and desire to ask for help re-pasting it themselves.

Something to think about

 

day after repaste -100mv custom fans i will be ordering sanding kit soon but will do it in march ive contacted msi about spare heatsink waiting for reply

 

Update: I just noticed your hwinfo runtime is almost 8 minutes... that might include running other stuff in between that can affect the peak readings. If you click on the runtime icon on the bottom of hwinfo64 it will reset the readings to that moments values, and that's what I do before I start a single benchmark run to get an accurate reading for just that run. Can you please do that? Take a snapshot before you start the benchmark after you click reset, and then again a second or two before the end of the run so you can capture current temperature during the end of run.

Given your average temps are pretty close together, you might try turning on logging in hwinfo64 and see how the temperatures track per core during the Cinebench run.

With the log you can see the data second by second and not just see an overall peak temperature.

I've helped people do this a number of times and what they find is those peak temps may only be reached a single time for a second, while most of the time the real instantaneous temperature throughout is much lower.

This is mostly helpful for gaming benchmarks that vary the load more than Cinebench will, but it would be interesting to see with Cinebench if the peak temperature grows over time (waste heat is building instead of disappointing via the fans), or if the peaks are intermittent with longer periods of lower temps.

That 11c apparent core temp differential is right on the border of too high, and since I haven't seen it on GT73's myself, even one from a few months ago, and this thread hasn't gotten such reports up until recently, I'd like to know WTH is going on.

Maybe @Donald@HIDevolution can ask his techs if they have been seeing recent problems with core temperature differential with the GT73's they test - re-paste - and service.

Hopefully your's is an extreme isolated case.

 

yea, i was lokking at temps at games as well and constantly it was 10c difference. Ive contacted msi about spare heatsink this reply ive received
Dear *

"You’ve requested an out-of-warranty repair. Please note that all the laptops that are completely out-of-warranty are shipped to our main repair centre in Poland. The process for an out-of-warranty repair is as it follows: Before a quotation is issued, our technicians first need to examine the unit. After this, we will give you a quotation for the repair. If you accept to pay the amount quoted, we will proceed with the repair. If you refuse to pay the amount quoted, you’ll still need to pay €93.60 for workmanship, admin, VAT and shipping costs.

Please note our service centre will need to order additional parts for the repair, and if our factory in China discontinued those parts, then we won’t be able to repair your unit. However, seeing that this is out-of-warranty inspection, you’ll still be required to pay €93.60 for workmanship, admin, VAT and shipping costs.

Let us know if you agree with these terms so we can collect your laptop. "


IVE ASKED ABOUT SPARE HEATSINK IF ITS AVAILABLE FOR PURCHASE not to sending my laptop

 

Yeah, you'll need to respond with a *brief* specific request - not all in caps either

They have boiler plate to paste, and likely handle high numbers of responses a day, so it can take a couple of go around's before getting a focused response.

In games there will be core temperature differences anyway, as games load cores differently and intermittently.

That's why you want a test that uses 100% the same CPU on each core / thread.

That's why I use prime95 without AVX, others like other tests, but anything that uses 100% consistently on each thread / core - without rising / falling loads like most tests - will be useful to see core temp differentials.

Check my last post again, I updated it with a request to post a shorter run hwinfo64 result - just from 1 Cinebench run instead of the almost 8 minutes monitoring time that hwinfo64 captured.

 

i never send that what i wrote here in caps its just for here. i explained them clearly whats the problem is ITS THEM who conctantly dont care read properly to try see the problem . the thing with hwinfo ive did it X times watching temps its constantly 8-10c difference aida, cinebench, gaming, no matter what, of sometimes when i was watching temps in assasins creed origins as its demanding on cpu ive noticed core 1 and 3 constanly hotter about 10 sometimes 12c. so its constatly ofc sometimes cores goes for example like now 67/68/61/61 but when gaming or cpu on full load always temps difference there. and then it leads to dry out and causing throttling and super hot temps. in idle only temps are same

 

So your last hwinfo64 monitored run of Cinebench shows 67/68/61/61?

That's completely normal

The first cores are going to get more load than the two other cores, which is showing in the results.

There isn't anything wrong... really starting to look like that.

Again, you can *show* core temperature differentials easily, in fact it's the normal outcome of using the computer for various games, apps, browers, benchmarks, etc. That's how it works.

You need a special application or test to run a consistent enough CPU load at 100% across all cores to see the true differentials. That also depends on all 4 cores set to the same multiplier. I am assuming you have your CPU OC set to balance all 4 cores the same and you increased power limits so there isn't any power throttling.

You can't do the test easily with BGA power limited locked CPU's either, they will downclock and power limit during sustained CPU loads on all cores longer than the TIMEOUT set, usually 8 seconds to 28 seconds.

So Cinebench with the result of 67/68/61/61 on a single run doesn't look like a good candidate for checking core temperature differential. The desired output is 4 core readings the same, that output looks more typical of a non-balanced load application.

I don't think you have anything wrong, and never did out of the box.

If you had tuned the undervolt, tuned the fan curves, and tuned the OC to stable operation at 4.2ghz, and limited your hwinfo64 monitoring to only accumulate values during the actual runs instead of always on I think you would have been ok without repasting too.

The original paste wouldn't have pumped out if left alone, I'm not sure what happened when you got a sudden increase in temps, but that just doesn't happen without some change to software or a hardware interaction / problem to cause a sudden change.

Your testing, monitoring and reporting skills need tightening up, as they are indicating problems that don't show up when they are done properly.

 

no those temps were just an example when i was browsing or runnin some programs that recorded on hwinfo i wish it could be like that in cine i told u im getting constantly 10-14c difference on full load and no there was no any interaction when i bought it for last 7 months i had bsods out of box! i tried to factory still same problem then i found out it was killer driver. i was running on stock speed and clocks and fans for last 5-6 months!

@hmscott for example core 3 is always hottest even in gaming or tests look at the difference and yes i know it doesnt count as much when its not on full load but still it still happens on high uisage as well jumping on hottest and sometimes 17-20c difference which i think is too much and in games 8-10-12 constantly jumping

at least ive got reply from msi CS saying china factory is on vacation from 08-02-2018 until 22-02-2018 so i will contact them afterwards.

 

You've gotten caught up in thinking something is a problem that isn't a problem.

Wide swings in temperature that match wide swings in usage / load on the CPU are normal.

You've overloaded the words "core temperature differential" to apply to far more than it's supposed to. You are taking normal asymmetrical load - differing loads on each core - and seeing those wide core to core temperature differences and calling them a problem, and they are not a problem.

It only applies to a specific test mode, 100% CPU across all the cores with the same consistent / constant load delivered to all the cores over the period of the test. A very controlled example, specifically designed to show any problems in cooling between cores, core temperature differential.

You can't apply that naming to any and all asymmetrical differing loads on each core / thread at different times, like normal computer usage.

So far I can't get you to see the difference, and that misconception is going to continue to haunt you until you do. You are going to see the problem everywhere, because that's normal computer usage with differential loads showing different temperatures between cores.

That's a long time for a vacation, those guys at MSI have it great...

Please let us know what they say when they get back. Try to have fun until then, maybe tone down the OC if you don't like the peak temperatures, but so far it looks like a reasonable temperature range well below the thermal throttling point.

 

The hwinfo you posted, was that for the hwinfo run? That's a hwinfo64 monitoring run of over 1 hour and 29 minutes, anything can happen in that time.

Please post an hwinfo monitoring display for core temps, with the runtime showing, reset the values when the Cinebench run starts and take the screen shot / snipping tool grab a few seconds before it ends. That should be an hwinfo64 runtime of around 1 minute maximum.

 

I agree that undervolting and a more agressive fan profile is sufficient for the vast majority of owners for out of the box use, and overclocking to 4.0-4.2GHz as well. In this case, he said many times, that his CPU was overheating with an overclock of 4.1GHz or less. So yes, the safest and recommend option is to RMA, but some people don't want to be without their notebook for several weeks and opt to do it themselves, like he did.

 

I'm trying to help him get past his misconception of core temperature differentials, and get his testing, monitoring, and methodology to show what's really going on. I don't believe we know that yet.

Based on his misconceptions and inability to give a valid test run monitoring image, I don't think his OC'ing / undervolting / fan settings and hwinfo64 monitoring interpretation were adequate to assure the laptop paste was at fault.

His temperature peaks could have been anything, look at the last hwinfo64 image he posted, it was for a 1 hour and 29 minutes... a lot can happen in that time

One of the problems with laptops fan curves is that they don't respond quick enough to instant high load situations that continue - the fan curve doesn't want to ramp up too fast for short high load situations so it holds back ramping up the fans high enough to stop an instantaneous high peak temperature.

That's why I either run 100% fans while benchmarking / testing, or if I am testing on normal fan curves I click the hwinfo64 reset button after a few seconds of the test to get the fans ramped up, after that happens the hwinfo64 reading is representative of the real run.

If he doesn't do those things, has normal fan curve running, and starts a high load application / benchmark then hwinfo64 will record that peak temp - never to be seen again after the fans ramp up - and that will obscure what's really going on. And, if you run hwinfo64 for 1 hour and 29 minutes, it will record the highest temp of all the runs attempted during that 1 hour and 29 minutes. That's really not a valid peak temp to quote and hinge your complaints on.

He needs a good solid test monitoring run for the duration of the actual single test, with 100% fans or clicking reset after a few seconds of the test run to clear the peak temps reached before the fans kicked in.

 

wwhen on load all cores difference 10-11c and thats there ive done test x hundred times. yes temps looks ok for now but i know they will degrade soon.. as from a latest repaste as well.. because if they would stay like this i wouldn mind it a lot but i know it will go up again because heatsink is unbalanced
here is another one only easus recovery running backround but i stated it it doesnt matter !! what i do! its always that high well on full load max 12c...


@hmscott ive done it for u reseted hminfow and straight test on full fans

 

Update: Ok I see you updated your post with the 50 second run, and it is showing lower temps, but still with a core temp differential.

I wish I had been in on the original problem report before you re-pasted, it's pretty rare to run across a real vs perceived problem on GT73's.

I guess we now need to wait until you get a new heatsink from MSI, if that's it that's great, I hope it's not something else...

Thanks for posting the hwinfo64 image, it was good to get a confirmation hwinfo64 reading that wasn't for far too long, one that showed just the test run.

If you have time a similar short hwinfo64 run with prime95 with AVX disabled small FFT - for say 5 minutes @ 100% fans - it would be nice to have as a good before reading - before you replace the heatsink.

 

Core temp differential aside, I am curious about why his VID is so high at 4.1GHz, when he said he is running that with a 100mV negative offset. I don't have the GT73, but a SC17 with the same CPU and I lost at the silicon lottery, so I can't run at voltages like falkentyne's golden chip or others here with good voltages. Mine with a -40mV offset at 4.1GHz is at 1.095mV. I underdand falkenstyne explained about the VID reporting, AC DC loadline stuff, and vdroop.....so I guess it would be easier to compare his actual reported VID and undervolt with others with the same model.

 

I edited this request out of my other post, but maybe that's another good reason for @Falkentyne to do the same run as @GENOCID 41x x 4 on Cinebench with hwinfo64 for just the duration of the test @ 100% fans to compare.

 

Of course this has nothing to do with the core differential issue, but I got the impression that the 100mV offset is not being applied correctly.....so I got curious about that.
EDIT: maybe I misunderstood and he his running without an offset.