***The Official MSI GT83VR Titan SLI Owner's Lounge (NVIDIA GTX-1080's)***

Hmmm, only the one core, Core 0 is getting that hot, 5c-6c higher than the others.

Please, turn the fans up to 100% - wait a minute for the heat to be flushed, leave the fans at 100% while you run it again. Let's see the temps at 100% fans.

 

Cooler boost titan was on, cpu and gpu fans were at 5k RPM each for that test. Is that 100%? Or should I be using the basic fan @ 100%?

 

Ahh, ok. That's too bad that one core is ruining it for the rest of them

It might be the heatsink isn't making good contact at the core location. Typically askew contact from uneven pressure from unevenly tightening the CPU heatplate screws gives 2 cores worse cooling - or 2 cores better cooling depending on how you look at it - due to an askew fitting of the heatsink.

What is the temps during a run with auto-fans?

Also, we only know the single peak temperature during the run this way, you can start a logging session going to log 1 sec or 2 sec (or any interval you like) to log the temps at each sample interval during the run so you can see if that peak temp is a constant or a rare event during the run.

That's good to know because if it's only occasionally tripping the high temps then normal use should be fine.

Maybe try running a gaming session, reset the hwinfo64 readings before hand, start the logging and run for a while, 10 minutes to whatever, and then see how often peak temps and thermal throttle point 93c is reached.

If you can - take back the laptop - I would if I found this on the first day, which I would - for another unit, as this one has a physical issue requiring opening up the laptop to repair it. I wouldn't personally waste my time fixing it, instead putting it back on the maker to fix it themselves. Paying as much as we do for these laptops just doesn't justify taking the risk and voiding the warranty to repair such things, even as easy as it might be to fix it.

It's going to either take a tightening of the screw nearest Core 0, even a 1/4 turn might do it, and perhaps a slight loosening of the opposite screw, but this might also mess things up good.

I wouldn't pull the heatsink to try to reseat it, often as not this traps air bubbles in the paste which creates a void that causes raised temps on that or another related core. Going further to repaste is a pain, requires practice to do right, and might end up being a seemingly never ending task... best to avoid wasting time that can grow from minutes to tighten a screw to hours, days, weeks - I see it happen to people all of the time.

See if you can swap for another unit, and this time do what I do, bring a set of tests on a USB flash drive for on site eval of the new out of the box laptop before leaving the building, and a buy a 32GB USB 3.0 flash drive to create the MSI BurnRecovery boot drive if you haven't already done it - a good thing to purchase the flash drive at that retailer too - makes them happy.

So annoying that only 1 core is the problem, right?

 

Hmmmm, based of what you said about the heatsink not making good contact at the core location, Does it make more sense to do a repaste to ensure good contact?

Unfortunately I can't take the laptop back. I bought it april 1st and my window for return on ebay has expired. It's also not registered to me so I can't send it in to MSI for servicing even though it still has 4 months left on the warranty(which is probably void now since I punctured the warranty sticker in order to pop it open for the upgrades).

I think my best option at this point is to take this to my local computer repair shop and have them repaste.

 

Darn, well you can solve all of this by asking the seller if they knew about the problem. Ask if they re-pasted too. Just simply ask.

Then when they answer in the affirmative, report this to ebay, and tell them you want your money back, to return the laptop. Even if you are out of the return period this is an extraordinary situation they should accomodate your return / refund.

Contact Ebay or paypal if you used them, or your credit card company, a polite request should help you recover your money, be patient with them and persistent at the same time.

A good example why testing immediately is important, it's good to learn this, unfortunately it usually takes one event like this to realize it.

The other thing is to ask the seller to open a ticket on this serial number - they registered the laptop so they can ask to have the ownership transferred. That way you can use up the rest of the warranty, or at least RMA and pay for a fix that isn't covered.

The problem with using a "repair shop" is unless they have worked on your laptop model before, it's going to be the first time for them too.

That means even if you have experience doing laptop repair / re-pasting, every new model presents unique problems to learn and overcome.

A paid repair person giving an estimate on time / cost may likely run out of time before they perfect their technique and therefore not get the best results on your laptop - they have to stop when their paid time is up. So, I'd ask if they have re-pasted this model before, ask them to ask their technician if they don't know. The sales person may say anything, ask the tech yourself if you get that feeling. Also, you could be forced to then pay more to make sure they have enough time to get the job done right; it's best to work this out up front before they pull apart the laptop and put it back together.

Backup all of your stuff, use the MSI BurnRecovery app to create a restore flash drive (32GB USB 3.0 needed), and /or make a Macrium Reflect image backup of your boot drive - and other drives you leave in the laptop when you give it to repair.

I hear lots of sad tales from people that get their fixed laptop back with a "clean install" of Windows, provided helpfully by the tech who couldn't be bothered to swap in a boot drive of their own - also, don't give them your account(s) password(s)!!

I play it safe and pull the drives, put in a spare drive and restore the recovery image to that drive before handing over my laptop for RMA / repair. Many vendors as a matter of course will restore the original OS image to your boot drive to do their work, that's why they tell you to backup your stuff before RMA'ing.

You could also try setting the Core 0 drive to 38x and the rest of the cores to 40x to see if that is good enough to drop the thermal throttling, if it is, try Core 0 39x and / or bump up the other cores to 41x.

Balancing the performance to make it work without repasting or returning is still possible.

 

Seller got the laptop from someone else. Warranty is still in the original owners name most likely. Seller says the person he got it from is not responsive. I'm wondering if this thing was stolen lol. When I called MSI, they said I would have to have the original receipts from an authorized vendor/dealer to be able to get the warranty transfered over. Do you know if MSI would repaste if I paid them?

 

Yeah, that's why I always ask for the serial number and check with the vendor before buying a used laptop.

You can sign up with MSI and open a ticket / request to transfer ownership to you, attach the emails from the seller telling you they can't do it themselves, tell MSI no one in the current ownership chain has the receipt, and of course ask them to check if this is a stolen laptop first.

If MSI tells you the laptop is registered to someone else, ask MSI to reach out to that owner to ask them to transfer the ownership to you - it may turn out they will be happy to have their stolen laptop found...

Ebay / Paypal / CC can help with this, and you can get a refund if it is stolen - include the ticket / email responses from MSI.

I would get this straightened out before paying for laptop repairs.

 

@hmscott while I know you mean well, but you know how I feel about MSI's heatsinks.

MSI CPU heatsinks are notoriously convex, while the GPU heatsinks tend to be nice and flat. Jeanlegi (I belieev) found out. Try to see if the RMA department will agree to ship you a free replacement heatsink for the cost of shipping (like $10 or $20), under defective heatsink, which will save a lot of expense. Replacing the heatsinks on these laptops is quite easy, far easier than alienwares.

If you live in USA, try giving them a call and see if they will agree to ship you a replacement heatsink and you pay shipping charges. Make sure you claim the original is defective. If you don't, the heatsink will probably wind up costing you $70 instead of shipping.

And even then you may have to pay for it anyway. The problem is, unlike FANS, heatsinks are not kept in USA warehouses and must be ordered from the China factories. That's the problem.

Otherwise if they won't do that, you can either choose to repaste (REPASTING DOES NOT VOID YOUR WARRANTY) or do the RMA.

 

That's because from your POV you have no problem re-pasting on an hourly basis.

For most people it's not within their capability, and they shouldn't be cajoled into doing so, it's not nice. Too many chances of ending up with a basket case, many of which I've fixed over the years.

Really, a laptop is like any product you buy that fails to meet your expectations. You'd return it, not fix it.

For those that make a hobby out of DIY laptop work their POV isn't a good fit for those not so inclined, the answer for them is return a broken laptop, not fix it.

And, in this case the ownership provenance is unclear, and should be cleared up before trying to fix what might be stolen property.

Whenever someone tells me they don't have the original receipt, and further tell me they can't transfer ownership because they never registered it, I stop right there and politely tell them I am not interested in buying a laptop I can't prove ownership or register myself.

That proof of ownership is the first hurdle when I am interested in buying a used laptop, and if they can't login and manage the laptop to download the uploaded receipt they used for registration, that's a red flag I don't ignore.

 

Proof of ownership is important. Unless something is disposable and not valuable, you should never buy an expensive item used, without knowing how the ownership transfers over.

 

Good news. I did some tuning.

Here is the latest benchmark test. 4.0 OC, .085 Undervolt, cooler boost titan on cranking fans to 5k rpm each.

http://www.userbenchmark.com/UserRun/8620560

According to Hwinfo64 I did not thermal throttle at all. According to the website I used for benchmarking, I throttled at 99% usage. According to XTU, my CPU temps topped out at 91c and says I did not thermal throttle. Cores are within 6-8 degrees of eachother, things seem to be looking better. May not even have to do a repaste.

 

Hey, you can't argue with success.

What tuning did you do to accomplish this?

 

Also, Here is my custom fan curve for the CPU. i'm currently idling around 43c as I type this while being OCed to 4.0.

 

Are you on the High Performance Power Plan (Windows)? If so you can reduce idle temps and temps overall by using the Balanced Performance Plan with Advanced settings CPU Min/Max set to 0%/100%. That lets the CPU downclock when not under load.

You lose a little edge of performance, but in regular use you won't notice. When benchmarking to get every digit higher you can switch to High Performance CPU Min/Max 100%/100%.

 

I tried skinning the cat like you suggested. Unfortunately, it would not allow me to adjust it to 39/40/40/40 or 38/41/41/41 like you suggested. Dragon center let me do it, but after I made the changes, XTU showed an even 4.0 or 4.1. XTU also wouldn't let me set uneven core ratios. So this prompted me to test 38 all around to see if it wouldn't throttle, then 39. Throttled both times during benchmark test. Then I started undervolting at .040, then .060, then .075. Temps were not throttling during benchmark test. Then I decided to go 4.0 all around and figured I should jack the undervolt to .100. This was a mistake as the laptop BSOD during the benchmark and restarted on its own. I decided to back it down to a .085 undervolt. Test completed, Hwin + XTU both say i'm not throttling(XTU said my highest temp was 91c while Hwin said 84c) but the stupid benchmark results are saying i'm thermal throttling at 99%. I'm going to guess the website is wrong right?

I'll have to do a lengthy Crysis 3 session now to see if this is stable in game as it's probably the best game for testing since it's the most CPU & GPU intensive game(or at least one of).

 

I'm currently on the balanced plan with min/max set to 5%/100%.

 

Dragon Gaming Center can interfere with XTU settings and visa versa, as DGC uses XTU for it's tuning.

I don't use DGC myself for this reason, and rely on XTU + MSI AB (GPU tuning) + RTSS (frame limit set to display refresh rate) + Silent Option for fan tuning or I just use auto fan for normal use + full fan for benchmarks.

XTU alone shouldn't limit even/odd multiplier, but I think I recall a weird dependency on "higher/lower" core number for setting multiplier.

So if you set Core 0 to 40x you can't set any other cores higher, like to 41x, Core 0 needs to be the highest multiplier. Weird right?

It's pretty rare to come across a hot Core 0, other than other applications running during a benchmark running up the temp on Core 0, there usually isn't a situation where only Core 0 is hot - maybe that's what is happening, some other app is running at the same time as the benchmark and heating up Core 0.

Try disabling the DGC service in Windows first and reboot - and download DGC from MSI to make sure you have it before uninstalling - check the version numbers too.

 

Yeah, default Min is 5%, I set it to 0%, IDK if it really helps, but I feel better making the change myself.

 

I had the following applications open(all were minimized into the system tray except HWinOF64-I left up the hwin monitoring window so I could see the temperatures in real time while testing): MSI SCM, MSI Dragon Center, XTU, HWinOF64, windows defender, Nvidia Settings and synaptics pointing device(program that controls the trackpad). In general, I like using dragon center because it allows me to use MSI Shift via Fn Shift key for when I want to shift gears quickly. I keep SCM open because it's the one program that controls the "P1" Fn key. What would happen if I disabled start up for Synaptics Pointing Device? Only time I've used the track pad is when I ran into problems installing the ram and had no usb mouse usage when the computer booted into auto repair mode or w/e. If I disable synaptics on start up, will I have issues using it if I ever get forced into the auto repair mode where it gives me options such as troubleshooting, recovery, uefi etc etc?

 

The monitoring software takes cycles, and you have 3 running

DGC, XTU, and hwinfo64 all take cycles to do what they do.

I exit the XTU systray app when I boot up, unless I need it, as it is collecting data. Same for the DGC service - stop it before benchmarking.

And, once you are happy with the data seen in hwinfo64, run your benchmarks for best numbers with that app exited too.

The monitoring data collection tools take CPU even when minimized and you need to stop / kill / exit them all to get the best benchmark results.

Defender and other anti-virus can fire off a scan while a benchmark is running, so exit those too - or use the pause feature if they have one.

Don't minimize apps, exit them.

And, Windows 10 has a plethora of background services always running, plus scheduled reports back to MS, plus monitoring your actions.

Use a good tool or tools to disable and stop all monitoring by MS, and disable Windows Updates as well.

Not only benchmarks are affected by this, these background annoyances can cause game lag, skips, hangs, and other annoying interference.

Nice series of events, you are helping document all the things to do and not do when getting a new laptop

 

Welp, so the undervolt was enough to pass the benchmark but .085 undervolt crashes games within 2 minutes. Had to bump it down to .050ish to avoid crashing issues but i'm thermal throttling again. I did some research on google and found this. Curious if this is true and if it is a possibility as to why the temps are going so high. If it is, how do I disable HT and what type of impact will this have on performance?

 

Any undervolt at OC for a 6820HK is good, I only get -15mV @ 4.5ghz with 7820HK, and typically there is no undervolt available much beyond stock before this, the 8950HK seems to have a much larger undervolt at OC - given one example so far, -140mV @ 4.5ghz, so it can vary widely.

Again, your cores are all fine except for 1, Core 0, and you said you tuned and got no thermal throttling, my guess is you are right at the border when benchmarking, but should be fine with your custom fan curve while gaming if you also use RTSS to limit FPS - unlimited it's going to be rendering frames not needed past refreash rate of the display.

The 6820HK for most OC's to 3.9ghz/4.0ghz stable with no overvolt, but it's much more rare for higher stable OC, but we've seen 4.1ghz-4.3ghz a few times.

My GT80 5950HQ would OC to 4.3ghz for benchmarking with a +100mV overvolt, but I usually ran it at 4.0ghz with no voltage offset, and it didn't thermal throttle even under load.

I'd be happy with what you have, spend some time seeing if you can OC your display to 100hz - 90hz-100hz, and set the RTSS frame limiter to (refresh - 1), that should help with thermals on games that render at a higher FPS - saving your CPU / GPU excess load.

And, explore uninstalling DGC and use XTU / MSI Afterburner / RTSS / Silent Option to manage your OC and tuning with finer controls.

And, don't forget to contact MSI to get your laptop registered in your name.

 

For benching/stress tests It's able to pass without throttling, just real gaming it throttles when i'm clocked higher than anything other than MSI Shift's Sport Mode setting with a .075 undervolt. Core 0 max temps on those games with these settings has been no higher than 83c with stock fan settings when playing starcraft 2(single core heavy workload), csgo(multicore load, gpu light) and Crysis 3(heavy GPU/CPU load) all on ultra settings which is good.

I know I probably sound like a broken record at this point but I just really want to be able to utilize this laptop to it's fullest potential and the way it's advertised(oced to 4.0). As it stands right now, I feel like i'm missing out so I'm taking it to the local shop near me and having them repaste this week for sure. Someone else in this thread repasted a gt83 model with thermal grizzly conductonaut LM and max load temps dropped 15c, it's worth a shot here. I'm mildly annoyed at this situation, especially when I was at my friend's place seeing his i5-7600 with liquid cooling + decent performance paste idle at 30c with 7 chrome windows open and when at max load it topped at 55c(whole entire thing cost him a grand, it has a gtx 960 that suits his needs). I almost wish I had built my own desktop at this point. I know that desktops have much better cooling due to better ventilation etc but that's just nuts. I love this laptop so i'm going to give it one more shot with the repaste. If that doesn't solve the issue i'm just gonna dump it on ebay or craigslist to cut my losses and build a desktop.

I did try turning off hyper threading and it did not do much at all to improve temps. i've turned it back on after several different tested configs. Damn that core 0 ><.

I spoke with MSI and there is nothing they can do to get the laptop registered in my name or warranty transferred into my name unless I have all the original receipts from an authorized seller. Seller's previous seller has been unresponsive so chances of me getting the paperwork is less than 1%. ><. I live and a I learn I guess. Wasn't really educated on what to look out for when I bought the thing. If I was shopping for a used one today with what I've learned over the last week, I'd have asked for screenshots of several different benchmarks/HWinfo64 stats when OCed + warranty information before pulling the trigger. O well, at least I know now and in case anyone is reading this, they also now know some of the proper precautions to take before purchasing used laptops. TIL

Anywho, I definitely appreciate all your time spent giving me guidance. I'll make an update once it is repasted and share the results.

I just noticed I have what looks to be a "hot" pixel on my screen. It is lit up white. I assume it would be a bad idea to OC the monitor at this point right? Are there any real fixes for "hot pixels"?

Edit: Actually going to take it to the shop today, hopefully they can repaste today so I can share results right away.

 

Just so you know, I have done my best to steer you away from re-pasting to get that last "nit" of performance.

I have helped lots of people over many years, and have seen and helped people through the aftermath of going just a bit too far, and it's not fun for me, far less fun for them... so I've done my best to prepare you ahead of time for the potential problems, I hope you've pulled your data drives, backed up your recovery image, and done an image backup of your boot drive - pulling your data drive is good too.

Really, it's only 100mhz-200mhz that is motivating you to get knee-deep in hardware modding. That's a 2%-5% clock rate increase, which won't even translate into a noticeable difference in every day usage. And, even in benchmarks will only be a few points, which matters to no one.

The pixel may be able to be massaged out, I've done this myself a number of times with client hardware:
https://www.google.com/search?q=how+to+massage+out+a+stuck+pixel

The display OC'ing isn't concerning to the pixels, it's the driver hardware that will "fold" and not support higher than a certain OC refresh rate. The 18.4" display and many 17.3" displays will OC well, not so much for 15.6" displays - but that could change as new display's release.

Good luck

 

Great news. Just got the call back from the repair shop. Liquid metal applied to the cpu. Runs 35c cooler now under stress test @ max load. Max temperature during cpu stress test with cooler boost titan fans @ 5k RPM was 57c!(Edit:This was on comfort mode. Max temp whileOCed to 4.0 during stresstests/benchmarks was 75c, 20c decrease down from 95c and no more thermal throttling!) I highly advise everyone who owns a gt80, gt80s, gt83 or gt83vr sli model to repaste with liquid metal ASAP, even if you are buying it brand new. It makes a huge difference compared to stock toothpaste They asked me if I wanted to do the LM on the GPUs but I was a bit nervous about that(i'm fine with risking a $180 processor but the GPUs are substantially more pricey) so I just had them repaste the gpus with whatever performance paste they had on hand(i think it was coolermaster thermal paste if I recall correctly). I'll be repasting every 2 years from here on out(or as I see fit) and OCing to 4.0 for daily use. Next thing i'm going to do is measure to see if there is enough room to put heatsinks on top of the ssds to lower ssd temps by about 15c-25c. Time to fill my m.2 sata 3 slot and get a better laptop bag for this .

I certainly appreciate all your help here. I know I went against your guidance but I had to see for myself and I'm glad I did. Again, I highly recommend all owners to repaste the CPU with liquid metal assuming you do not have an aluminum heatsink.

Update: I found some red heatspreaders on aliexpress for my m.2 drives that will fit under the hood. Yay. I'm trying to find red heatspreaders for sodimm ram. I'm having a hard time finding them in the color red ><. If anyone knows where to find them, please let me know. This brings me to my next modification . I want a custom 3d printed transparent hood or glass transparent hood for my gt83vr. I'd like to display the hardware just like I would with a transparent desktop tower. This way I could display the m.2 drives w/ red heatspreaders and the 2 ram cards w/ red heatspreaders once I find them. I'd also cover the the blue ray drive + stock hdd with red or black non-conductive material to keep the msi red/black theme throughout the entire hood. I've contacted one company so far to see if they can do it. Does anyone know of a person or company that could make this happen for me? I want this monster pimped out

 

I have a question about the dual gtx 1070s in my gt83vr-6re. Throughout my 4.0 oc tuning, I've been monitoring my graphics cards temperatures and core clock in MSI afterburner. Temperatures have never gone above 70c on either card when playing Crysis 3. I currently do not have these overclocked(I probably won't OC them either since i'm not using a 1440p or 4k external monitor, waiting for my packed pixels monitor to arrive). I noticed in the MSI afterburner monitoring graph shows that my coreclock on each card goes as high as 1924(stock is 1443, boost is 1645) on its own while playing games like Crysis 3 on max settings. Just to reiterate I have not overclocked either gpu. Is this normal or is this good? Why is it doing this? Does this affect the lifecycle of each GPU? Is there a way to make it stay at the advertised stock clock(1443)/boosted clock(1645)? If there is, should I be doing that? Also, I noticed that I cannot change the core voltage, power limit or temp limit for these GPUs in MSI Afterburner, how come?

I'm also curious about the SLI rendering mode settings in Nvidia control panel. What is Nvidia recommended settings? What is AFR1? What is AFR2? What is single GPU? In what scenarios should I use each mode? In case it matters for these SLI settings, My games library consists of the following: Crysis 3, Starcraft 2, Starcraft 1, Diablo 3, League of Legends, Counterstrike: Global Offensive, Heroes of the Storm, DOTA 2, Hearthstone, Warcraft 3, World of Warcraft, Overwatch and Final Fantasy VII.

Also, If I am playing @ 1080p @ 60/90/120 hz(90/120 OCed) with vsync on(I use vsync to limit how how hard my GPUs/CPU work, I don't see a purpose in producing more frames than my monitor will display), What should I have my graphics power management at? Optimal, adaptive or maximum performance?

Lastly, looking to test the absolute most GPU & CPU intensive game on the market. What is the most system intensive game that taxes both the GPU and CPU the most?

 

Probably PUBG for CPU, but there are a number of other recent STEAM games that are CPU hogs.
For video card, you don't need a particular ultra recent game. All you need to do is to enable 8x Supersampling antialiasing (i dont know whether this can be enabled in the NVCP or if this requires Nvinspector or other tools), and then watch the video cards scream for mercy. In overwatch, you can do this by using 200% render scale, with maxed out Ultra and in game AA settings. Enjoy the pain.

 

I totally forgot about PUBG. I'll have to try that one out, especially since I heard the optimization is sub-par.

Overwatch was cake at 150% rendering scale @ maxed out Ultra/AA settings with MSI's ECON shift mode. I held 60 fps the whole time(this was before I oced my monitor) and gpu/cpu temps never went above 60c. I didn't bother trying it on comfort, sport or OCed as ECON handled it just fine(I was really surprised by this, ECON wouldn't handle CSGO maxed out but it handled Overwatch just fine). I'll have to try it out @ 200% render scale like you suggested.

 

Oh 60 FPS.
I forgot that you have a 60hz screen.
I have a 120hz screen, so dropping down to 60 FPS( with vsync off) means that I'm crying and sad. Means that's very bad.
If you only have a 60 FPS screen then trying to put enough GPU load to drop under 60 FPS (without something like Firestrike Extreme/Ultra benching) becomes MUCH harder to do.

 

Yeah unfortunately 60hz @ 1080p is the max a GT83VR will come with. Luckily the displays seems to overclock fairly well. I've overclocked the display to 90hz without any issues, I'm going to shoot for 120hz in a few.

 

0% chance you will be able to reach 120hz without extreme artifacts. Many of the 60hz TN screens can reach 100hz however with absolutely no problems, since they are overspecced (probably are actual higher hz screens used in other systems to begin with ). But I know nothing about IPS. Nothing.

 

MSI Afterburner comes with RTSS which you can use for frame limiting, and many find it better than vsync, give it a try.

Did you get the display OC to 120hz?!? Wow, that's awesome.

I find that the higher the FPS the hotter everything runs, no matter how you slice it those extra FPS cost in thermal load.

I ran at 100hz and that was a 66% improvement over 60hz, not sure if a 120hz or 20% improvement over 100hz is worth it... you might try setting to 100hz and limit FPS to 100 FPS to see if you like that visually and fan noise wise.

As far as the other questions, the SLI modes are best on Auto, I didn't have any games that improved by swapping around different modes, Nvidia and game developers seem to have worked things out, and those modes are mostly at this point historical - of course maybe someone has a favorite game that they think works best in one mode or another and can post their findings.

The GPU's are fine tuning themselves, and that's normal, and you have good temps so not to worry.

 

Well, I hope it works out for you. LM has taken out GT80's and I'm sure GT83's over time we've seen a few.

That's why tuning using stock paste is recommended, no sense getting 20c more cooling you don't need. If you were thermal throttling at load, it's like that's rarely reached, and tuning the CPU undervolt and games FPS was enough for me and many others to keep from thermal throttling.

LM is great until it isn't, and you mentioned you didn't mind risking a $180 CPU but didn't want to risk the GPU's. You need to know that the CPU is soldered onto the motherboard, and so your entire motherboard would need replacing should the LM leak and screw up components on it, usually it's not the CPU that gets hit by the leak, it's the motherboard, but even if it was only the CPU it's soldered on... so motherboard replacement again.

The GPU's are MXM so they are replaceable.

Since the GPU's are no where near thermal throttling for anyone out of the box, it makes no sense pulling them apart to re-paste, LM or not. Usually the LM makes less improvement on the GPU's anyway.

Again, undervolting and tuning to reduce load is enough for most, LM or re-pasting with a non-conductive paste is a vanity mod, bragging rights for the most part.

There have been 10's of thousands of people that make it work out of the box without re-pasting, LM or non-conductive pastes.

It's an expensive risk that some have rolled the dice on and lost.

And, even if you get it done by a shop or at purchase, it's gonna need re-pasting at some point, or maybe leak - eventually the warranty you purchase will run out, and you'll want to sell it to someone else.

Have fun explaining LM to a prospective buyer and not scare them off... I know I won't buy a modded laptop, and don't recommend others buy them - unless you have that owner promising support - and who knows if they will honor their promise down the road. Will you?

 

It's not a problem to try, but just to give you a heads up, the most I've seen is 104hz, turned down to 100hz just to be even.

There was only one report of a guy that got to 120hz, but I guess it's worth trying.

 

I'll have to check out RTTS, I wasn't aware it came with afterburner. Hopefully I can use that to limit frames instead of vsync since vsync adds a little bit of input lag when mousing(makes a huge difference with mainly first person shooter games like csgo while other games like sc2 seem unaffected).

I ended up deciding that 90hz on the monitor is the max i'll OC it to since I want to keep everything running on the cooler side. 60hz gpu temps top out between 65c-70c while 90hz topped out the gpus at 75c. This is with auto fans. I might just go back down to 60hz for better cooling as I can't tell much of a difference between 60hz and 90hz. It's probably game dependent though. Who knows, I might just try for 100hz since you said that was an improvement over 60hz, maybe that extra 10hz makes it noticeable. CPU temps are also slightly different when the monitor is oced, anywhere between 5c-9c hotter with 90hz.

As far as the LM goes, the shop made sure to be extremely cautious. They put the clear coating on the surrounding parts that needed to be protected and also used thermal tape. I wasn't aware the CPU was soldered into the mobo, in that case I probably should have just had the GPUs done with LM as well.

I'm not sure if you mentioned this in the past but what is your stable undervolt when @ maxload OCed to 4.0 on all 4 cores? I'm sitting at an undervolt of .055 without any crysis 3/starcraft 2 crashes(i noticed crashing when undervolting @ .060 without vsync on and if I undervolt any higher those 2 games won't open). I take it i've found the sweetspot for undervolting? I'm windering if adding a second monitor will affect the undervolt(I ordered one of those "Packed Pixels" monitors to maintain portability, all the FHD monitors are limited to their own fold up monitor stand and I like how the "PP" monitor just hangs with my other display).

Not too long ago in this thread I mentioned I was going to be looking at storage configurations for optimization purposes. Currently nothing is in raid.

This is my current setup. Is this optimal for performance and durability? Is this optimal for backup and/or disaster recovery?

256gb Toshiba NvMe(Utility)-boot drive, software programs(nothing for rendering).
256gb Samsung NvMe(Gaming)-purely for gaming
500gb Crucial m.2 sata 3(Media)-Movies/Music(it's empty atm so I could use it for w/e)
1tb HDD-general storage & backup storage(currently storing a recovery image of the utility/gaming NvMe drives & all my downloads).

Currently not replicating anything at this time. Is an SSD or an HDD better for backup? Is a system recovery image enough or should I be replicating my Utility NvMe and my Gaming NvMe with raid 1 to an SSD or HDD?

Is there such a configuration that will automatically update a recovery image with the entire contents of both of my NvMe drives on a regular basis that way I don't have to create new images manually?(Or should I just replicate both drives to whatever drive I use for backup-assuming replicating will also copy windows etc)? still trying to wrap my head around the whole concept. Basically what I want is a configuration for a backup drive whether it's my 500gb m.2 sata 3 or my 1TB HDD that my BIOS can automatically use said drive as a secondary boot option for windows that will always contain the latest contents of all my other drives that way if for some reason my NvMe boot drive fails I don't have to worry about manually taking time to boot a recovery image.)

In general, From what i've read, Raid 0 is fast but in the event the drives fail, i'm screwed. I've opted against raid 0 because of this and also because it shortens the lifecycle of my NvMe drives(not that this has to do with backup/recovery at all lol).

 

Sometimes AB comes unbundled from RTSS, but if you download from MSI it's part of the package these days, check out your installed programs list to see if it is already there. If you don't have RTSS installed with AB you can also download RTSS from guru3d, and other sites.

There is FUD floating around to sell LM and extreme cooling product that isn't necessary for the long life of your CPU's, and that trend has made everyone skittish about CPU temps.

As long as you aren't thermal throttling you are ok. Even then you are losing performance only for the duration of the temperature load at the thermal throttling point, usually around 93c.

We've been running laptops for many years running much hotter than they run today, and they are still running.

99% of the time you can drop the temperatures of the CPU out of the thermal throttling zone by undervolting -100mV or more at stock, and gradually less undervolt as you OC.

The 8th gen CPU's are undervolting even better than Kabylake at stock and OC'd, which undervolted better than Skylake, which undervolted better than Haswell.

In fact as the processes progress we are seeing more and more headroom to use to undervolt, which means more extreme cooling measures are actually more and more unnecessary.

Getting temperatures down lower with LM and more extreme methods aren't going to extend the life of the CPU within your ownership time, or the 2nd, 3rd or 4th owners time - because the CPU's are rated to run at 100c maximum, and have built in protections to reduce thermals when they get hot - thermal throttling.

Below thermal throttling is where you want to be, below that is a waste of time and money.

Actually the GPU's have a much larger possibility of leakage due to their heatplate size and surface, and most vendors that were doing CPU LM at the time of the GT80 wouldn't do GPU LM for that reason.

On the 7820HK @ 4.2ghz -125mV and @ 4.5ghz -25mV but I run it with no undervolt, on the 5950HQ @ 4.0ghz -15mV and at 4.3ghz +100mV - and reports for the 8950HK are -140mV to -170mV at OC, so you can see the progression of undervolt improvement at each process improvement.

Every CPU is unique, some undervolt more than others, so you need to spend a little time pushing the limit of undervolt till you BSOD and then back off until you are stable at 100% load, and finally at Idle - you may need to shave another +5mV or +10mV off the best undervolt to be Idle stable.

The undervolt applies to CPU only, and if you've already tested at 100% all core load and Idle for stable operation, you are done. Any other variances in use won't be to blame on undervolt.

That's why I recommend making one change at a time, and don't push through any change until you are completely satisfied it is stable unto itself, before adding more variables.

It's tough to explain to some, but the storage speed only matters in a few instances, otherwise that super fast storage is mostly sitting idle not helping FPS or other compute CPU / GPU operations at all.

Think of it this way. Most gaming operations take CPU, GPU, but only need disk at the beginning and during saves / level loads.

So even if the disk IO is 2x as fast, the actual wall clock time is very small, so halving it's time of a few milliseconds won't be noticeable.

Only for copies, or long reads / writes will there be a difference, and usually the load animation is enough to distract from those times too.

HDD's have the nasty side effect of parking heads, and powering off to save power and reduce wear, so if you are noticing lag or stutters during level loads or resource loads during a game, consider turning those HDD features off using the vendor tool, or something like:

QuietHDD
https://sites.google.com/site/quiethdd/

Or, replace the HDD with an SSD and stop those side effects altogether and improve speeds.

500MB / sec vs 2000MB/sec - it's hardly noticeable unless your app uses a lot of IO, and you'll know that if you are a serious user of that app.

So if you know you need the load / write speed improvement and it affects your workload handling, then it's worth getting 2x more expensive NVME storage over SATA M.2 / 2.5" storage which is much cheaper, and runs cooler.

For most people, including me, SATA 500MB/sec is fast enough. Given that it's 10x or more faster than HDD, SATA SSD's have already given us the lions share of time savings for normal app / OS usage, 2x further improvement from NVME is only beneficial for applications specific use, otherwise it's just expensive *hot* storage.

For backups HDD is cheaper, but then again, it takes a *long* time to backup at 50MB-100MB/sec, so if you can afford it get a large external SSD running at 500MB/sec instead.

 

It's a good idea, but in practice having multiple bootable drives in a laptop, or computer, can be problematic. I've done it, but found I've needed to physically disconnect drives other than the drive I am updating or upgrading to avoid blowing away the bootable options of other bootable drives online - taking them offline solves that, a bit more difficult in a laptop than a desktop.

You don't need to fail-safe your laptop like that, there isn't enough free sockets in the laptop to make it feasible and still have free space for use.

If you like, you could clone an M.2 drive (or 2.5" for that matter) in an external case, and have that as a bootable hot backup, I've done that before - for important work machines. I can boot from the external drive, or pull the drive and swap it into the laptop.

In the end I didn't need it, so YMMV, I guess if you end up needing it it's worth the effort.

Now I image the bootable drive with Macrium Reflect Free and keep the image on an external drive to use with a bootable thumb drive to restore in case of a drive failure in the laptop, but even then it's only bad installs or Windows updates that required me to do a restore, not a drive failure.

You could keep the images on your internal HDD, but what if the laptop gets "lost", then you'd better have another backup at home, or your data is gone.

Anyway, keep good backups regularly, and you don't need to worry about fail-over in the field, unless you have critical work, then carry a clone drive to swap in should a failure occur.

 

Are there any shortcuts to keeping backups regularly? Is there a storage configuration that will do this(replicating?) or windows 10 software that will automatically create a new backup on a scheduled basis? I'm hoping for some sort of "set it and forget it" type of solution/configuration.

 

Yeah, I wouldn't trust automation to be 100% there for you. I do this for servers of course, but for a laptop that may be here and then there, online then not online, I find the assurance of doing it myself to be well worth it.

It's really not that big of deal, you open the software, you push a couple of buttons, and the backup happens.

I manually delete previous backups, and in fact migrate them after backup to a spot I can be sure they won't be deleted unless I want them deleted - even a different folder on the same drive partition is better than leaving it in the default backup folder, in case the automation goes wonky -> first time starting up the backup software in months and it see's backups older than 30 days / 90 days, and automatically screws you by deleting them.

Automation is good when things don't move around, and are always online and powered on, so you can count on them working, but for anything else I want to have control and confirmation it really happened as intended.

 

Gotcha. Sounds like my current configuration is pretty good as long as I add your recommended bootable USB drive loaded with windows. Thank you.

By any chance do you know of anyone that sells red 1mm or 2mm(not sure if 2mm will fit with .5mm thermal pad) heatspreaders for SODimms? I can't seem to find red anywhere ><. I'm working on getting a custom transparent acrylic hood made(still trying to figure out options for the buttons) that way I can display the ssds w/heatspreaders and i'd like it if the SOdimm heatspreaders matched so that I can keep the MSI black/red theme going.

 

Given there isn't much air movement in that area, at least on my GT80, I didn't bother with trying to add heatsinks to the M.2's as they overlap physically when you load up all the slots.

The RAM is pretty well flush to the board... IDK if there is any room for effective heatsinks. You need air flow over heatsinks to be most effective.

If you are going through the trouble to put finned heatsinks on M.2 and RAM, I'd look at splitting power for a tiny fan(s) to pass air over those fins.

 

I totally wanted heatspreaders/heatsinks with fans but only the m.2 sata slot has an outlet to plug something in. The NvMe slots don't have the outlet to plug anything in. Ultimately, The limiting factor ended up being space. There is about 3-3.5 mm of space available from where the installed ssd lays and the hood. This left me with only one option to guarantee the heatspreader doesn't touch the hood, 2mm heatspreader + .5mm thermal pad. The available 3mm options were more aesthetically pleasing but I felt like I would have been cutting it really close. As far as a fan goes, Maybe there are some tiny independent fan options floating around somewhere that I can at least install underneath the m.2 sata III ssd. I went with these since they match the msi theme.

 

I was thinking of splitting off the power from the 2.5" slot, others have reported doing this, and still fitting their 2.5" drive, I don't have a post to refer you to, but look around for a adapter that can fit a bare 2.5" SSD internals, as I recall that's what they used. There are probably other methods that would work too, a 2.5" M.2 carrier with exposed fan power taps and an M.2 SATA drive... etc.

 

I had a fun thought. I noticed 4k is pretty much locked in at 60hz due to the nature of 4k monitors and cabling at this time. Would 4k DSR with my monitor oced to 90 or 100 be on the same visual level of a 4k monitor or better since i'm at 90 or 100 refresh rate? I don't quite understand the difference between using an external 4k monitor and using 4k DSR on my laptop screen. Would I actually be at 4k with more frames than is normally possible because of my monitor being oced and how DSR works?

 

What kind of 4K monitor do you have? 4K monitor is just a monitor and 4K DSR is a technology used on a 1080P monitor. So since you have a 4K monitor then you are already benefiting from it.
https://www.geforce.com/whats-new/a...-improves-your-games-with-4k-quality-graphics

 

Oops I meant to say I use 4kDSR on my 1080p monitor. Is 4kDSR on 1090p roughly the same as using a 4k monitor?

 

It looks slightly better, but it's hard to tell the difference. 4K will still show more details due to higher resolution/pixels available.

 

Time to talk about upgrades. I'm considering upgrading from 1070s to 1080s in my gt83vr6re. Would I have to upgrade my power supply for this or would the two 230 watt supplies suffice? Also, I noticed the 1080 card prices are ridiculously high for laptops(there is no way they are worth double the amount a desktop 1080 costs). Would it be a better financial decision to purchase the cheapest 1080 laptop that has two 1080s in it, swap gpus and then resell the extra laptop?

 

20-30% benefit in raw terms

But 1070SLI is such overkill for 1080p already and the MSI 1070 which I've played with A LOT will handle the overclock to stock 1080 levels of performance if you increase the power limit (modded vbios).

If your GPUs are at 1924mhz they are at full boost because they're not being fully 100% used. The 1070 is severely power limited at 115W, if they are using 115W they will be closer to 1700mhz but can go as below 1300mhz under furmark. There is a lot of performance to be gained from a higher power limit and the cards are very robust and the power phases can easily handle 200W, cooling and power supplies permitting.

Whatever next gen that comes this year would be worth waiting for if you're desperate for more powah.

 

Update on OCing the monitor: 100hz stable is as high as it will go. The laptop hanged when I tried for 120hz. Crysis 3 crashed a couple times when OCed to 90/100hz. Not sure if this is related to the monitor OC or if it was because my graphics card was OCed at the time or if this is a similar situation to how you can undervolt higher for benchmarks but the same undervolt won't work for real gaming usage.