Clevo Overclocker's Lounge

is static voltage working on the DM3 or only on Premas bios? I'm just curious i can only check with the Power Usage and -75mV offset only gave me about 75w with Throttlestop 32M test and static 1.25v was about 87w.

 

Its not static all the way, but still helps at higher clocks.

 

The fans on DM1 is 5v vs. 12v for the newer models

 

I ended up using the Intel Extreme Tuning Utility for CPU changes and MSI Afterburner for GPU changes. Monitoring it all with CPUID and HWINFO64.

Benchmarking with 3Dmark (and a few of the features I paid for like firestrike extreme), Heaven to generate a constant graphics load and Cinebench

Even with all of that I still was popping off intermittent GPU and CPU crashes when Running FO4 with every setting at ultra and at 4K. So I had to start dialing back some GPU settings to resolve those problems and a minor tweak to the CPU as my temps started to get up around 92c after an hour or so of gaming.

I know that there are folks who benchmark and that is pretty much it, to achieve a score. I just want to squeeze more performance for gaming and some CPU-intensive tasks for work (engineering simulations).

It all runs good now unless I accidentally block off a vent

 

>_>
<_<
You're an overclocker, you should know overvolting those 5v fans means more air moved

 

I put 15.3v on my 12v Vardar fans (U3 mod). Works great But push 12v on these 5v are perhaps a bit heavy

 

Static voltage is more static VID (the voltage the CPU requests).

 

Are you familiar with Throttlestop? That may work better for you than XTU, although I don't know if it plays nice with Kaby or any issues with the new chipset. Also, did you buy from a Prema Partner shop to get his custom bios later?

Welcome to the lounge, cannot wait to see your numbers. Also, do you use a laptop cooler, have you delidded the CPU, played with the TIM and thermal pads, checked contact between the IHS and the heatsink, etc.? What you gave is a start, but knowing what you've done for your system to date, what the ambient temps are, etc., will help to get you that extra performance.

Edit: ignore above except about the Throttlestop part. Did XTU screw up your fan tables (had a negative impact on prior versions).

Anyways, shame it is not a Pema Partner ( https://biosmods.wordpress.com/). You will never fully optimize it without that.

You did the standard recommendations. I'd say to get some pressure sensitive film and remove the screws, then test contact and pressure as you screw it down. I went with going straight screws and counted rotations per screw to the 1/8 turn while testing with the pressure paper. I should have got a bicycle torque wrench with inch pounds, but...

So, you could try a different program or setting the OC in the bios.

Edit: @Meaker@Sager - Did you ever figure out optimal LLC settings on your beast?

 

You can get away with supplying a higher voltage to something like a fan but you want to keep that to be less than 20% or so.

This is not a CPU where "mo is betta", it is a discrete electromechanical device and you will burn out the fan windings with too much of an overvolt.

------------------------
I measured clearances with something known as "Plastigauge". it looks like a thin plastic string of a very particular dimension. When you put it between two objects and tighten them down the plastic string gets flattened. There is a scale on the side of the package of Plastigauge that shows you what that means in fractions of a mm of clearance. You can find Plastigauge on eBay and it costs about $6 for a six inch long piece that can be chopped up in to tiny 1/2" long test segments. The paper package looks like a soda straw wrapper with a scale printed on the side.

I checked clearances between the IHS and the chip die to make sure that they were minimal, before I used liquid-metal TIM inside of the CPU. I also used it again on the installed CPU between that and the heat-sink to make sure those clearances were minimal.

I have to travel for a few weeks but when I get back I will break out my K-type thermocouples and the four channel logger. Then I will temporarily glue down thermocouples on;
1) edge of the CPU IHS
2) top of the heat sink
3) where the heat pipes go on to the radiator
4) the radiator elements on the back of the computer
0) what the computer reports as package temperatures

Then I will stress test the CPU up to around 94c and document the temperatures at the four points. Let it run for a half-hour to heat-soak everything and put the data in a spreadsheet.

What that will tell me is how efficient every one of those interfaces is at removing heat. In a perfect world #0, #1, #2 and #3 will match exactly. We know that it is not a perfect world so there will be a drop-off at each point. #4 has to do with how efficient the fins are at converting that thermal energy in to the air (horribly inefficient, air really sucks as a way to remove heat).

What I will be looking for is some major discrepancy between #0 and #1 (bad mating between the CPU die and the IHS), #1 and #2 (bad mating between the CPU IHS and the heat sink), #2 and #3 (horrible performance of the heat pipes), #3 and #4 (very piss poor airflow or fin performance).

I will document the results and post them on the forum so you can all see the results.

(BTW, this is the type of work I do professionally so I am very familiar with this type of engineering process)

 

Nice thorough methodology, but I'm not sure how applicable to others since the QC on the mating surface on clevo/foxconn heatsinks is so infamously variable.

If you can reproduce the method in assessing lots of individual heatsinks, maybe resellers can start offering a "copper lottery" as well lol

 

I had lapped both the IHS (that had a high spot just above the lettering) and the heat sink so I know that they are extremely flat. Unless those screw locks are preventing the heat sink from getting closer to the chip I do not think that the problem has to do with mating between the chip and the IHS.

Even though I do not believe that is the problem I will still test it to rule out the possibility of poor mating surfaces. Proper scientific method means you question everything, particularly your own opinions and assumptions. Since I have a four channel thermocouple recorder and it is just as easy to glue down an additional bead on the IHS I will test there as well. Then I have five test points between the chip die and the air blowing out of the back of the computer. Somewhere along the line there may be a bad interface.

I need to pop off the heat sink anyway because I am going to re-gauge clearances and take a more detailed look at how the screw locks might be contributing to a proper heat sink fit.

More to follow once I get back from my trip and have some time.

 

Well I tried a little stock test in Witcher 2 with Ubersampling on for a bit. Holy meowmix, this thing needs some serious work for temperatures. Putting my ridiculously-high-at-stock CPU voltage aside (which I'll properly begin working on once I shove W10 on it, probably today or tomorrow) I think I probably need to hurry up and get those thermal pads and run some LM on the GPUs, as well as a shim and LM on the CPU, and to connect the CPU heatsink to the grid.

These fans even on max don't seem to push that much air. I don't know why. They're loud as hell but doing very little, apparently. I'll have to test with propping up the machine soon too. The bottom cover design being curved upward at the front is nice to suck in more air I'm guessing, but it makes padding underneath rather difficult, which I consider a big problem. I might simply need to invest in a modded U3 cooler, but that's going to be another level of hell to find.

I'm talking 92 max CPU temps on 3 cores only pulling 54W max not even using 60% total util. If I wanted to stream this thing would legit be worthless as-is.

 

Try running 50 under the top entry for load line calibration. 0 for the one below and +25mv dynamic at 4.6Ghz.

 

Sorry if I gave any offense with my previous post. I hadn't read back far enough and have gotten quiet in this thread as I've been focused on my desktop and new build as of late.

I am interested in that straw, but do see the pressure film as having the advantage of showing pressure variances across the plane, meaning you could diagnose if a certain area of the ihs to heat sink is making a weak connection. Also, some have mentioned that the thermal pads may cause an impact on contact of IHS to heatsink by lifting it up slightly due to it not compressing enough, or needing to go thinner on thermal pads containing a higher W/mK and lower impedance.

Here is the link to the ultra low and extreme low pressure films:

https://www.amazon.com/gp/aw/d/B004...ssure+film&dpPl=1&dpID=51jzhiPd4uL&ref=plSrch

https://www.amazon.com/gp/aw/d/B004...X236_SY340_FMwebp_QL65&keywords=pressure+film

At somewhere in either the P770ZM or P750ZM lounge, I wrote up an estimated PSI on the socket allowance. I believe Intel puts the socket pressure in terms of torque, though, in their white paper about their sockets. Just trying to generate helpful ideas. But getting above 4.8, many people use laptop coolers or portable ACs for these chips. But I definitely support your efforts and look forward to your write up.

Sent from my SM-G900P using Tapatalk

 

Maybe you have a girl friend. Undress her and use her nylon stockings to something sensible

 

Haha that would be interesting. Really though, I'm just trying to find the materials is more. It's $$ that's the problem hehe

 

Yeah thats one of the added costs of these systems i guess. When someone asks me how much the system might cost, I end adding the cost of the repaste and U3 with 4 fans + Controller to it.
And maybe even better thermal pads.

 

I make do without the extra cooling fans but Rockit88 and CLU are necessities for me.

 

True, initially i was also planning on a single 1080 build, since the cooling with that is phenomenal using the VC HS. But Phoenix happened.

A lot of times i forget to turn on the U3 fans as well xD

 

Single GPU is a dream. Don't think I could overheat that if I tried (I tried). mGPU especially in the daytime is another story, that may need some proper work done on it. But if you've connected your CPU heatsink to your VC (hell, even if you didn't it still lends some cooling it appears) then you're set, basically.

Either way I'll tame this thing. It'll take a while, though.

 

The 1080s in my machine are mostly just short of voltage now for clocking higher.

 

I have not had a problem with GPU temperatures on the 1080's and it appears that by tuning the undervolting I have dealt with the CPU temperature issues. I will still do the five point monitoring to look at heat sink efficiency because I believe that would be of interest to all of us.

My limitations right now are just in how the silicon lottery of CPU chips shaked out for this machine. I do not know if I am ready to go spend ~$400 on another 7700K just so I can get another 200 MHz.

I guess that I am not a hard-core overclocker, I only wanted a bit of a performance boost. This is a notebook and you just end up fighting too many other compromises in the design of the cooling system, fan capacity and heat sink efficiencies. If this was a desktop then it might be a different argument for me.

 

That 200Mhz is not going to be the be all and end all of the experience at the end of the day.

 

Well, you already have a 1 in 4 shot of getting a 5.1GHz desktop overclocker (now 28% of chips in this segment using 1.408V). So, it is likely your chip is decent as is. Spending to get the top tier for a laptop is ridiculous. Leave those for desktop LN2 users.

Meanwhile, in other news, pretty sure I'm gonna get an 1800X chip.

 

Yea, the heat transfer characteristics of materials is a fairly straightforward engineering challenge. It is easy(er) to model and to make some projections on what a change might do, one way or the other.

I looked at the pricing on the pressure sensitive sheeting. I do not think I will be spending $120 for a foot long piece of pressure sensitive film. I am pretty confident that the small tooling and machining skills that I was taught as a young engineer have provided me with flat surfaces on the IHS and the heat sink. My only question in that regard has to do with attachment downforce for the four screws and if those seeger rings (ciriclips) introduce additional spacing. If so that can be an immediate improvement for heat sink mounting for those still concerned that the carrier PCB thickness on the 7700K is affecting spacing.

There are several things that can affect spacing between the CPU and the heat sink;

• the thickness of the 7700 PCB
• The physical thickness of the IHS and whatever additional spacing is lost when you mount it down hard against the PCB after removing the silicone adhesive
• Manufacturing tolerances for the mounting ears on the heat sink and if they leave enough room for the heat sink to be torqued down to the board
• The length of the standoffs that are on the PCB and if they are bumping up against the seeger rings that hold the four screws to the heat sink assembly

I think that people might be too concerned about #2 on the list, changing the PCB thickness is an extremely minor change and if the entire assembly is that critical then it is a piss-poor design to start with.

Sort of a better design would of been to use springs on the top of the heat sink and the shoulders of the four screws (or washers) so a constant downforce could be provided. You could set the loading of the heat sink by adjusting the hardness of the spring material

Getting crazy and tourquing down would be a bad thing too; you can end up smashing the CPU socket, ripping apart the motherboard (if it does not have a metal backplate to spread out the mechanical tension) or strip the screws or standoff threads.

 

It is not 100%, but if one is using a paste based TIM, you could put down a layer of saran wrap, put on the X, *, pea or favorite manner of application, cover with 2nd piece of saran wrap, and finally tighten the heat sink.

Remove heat sink, and check the coverage. The spread pattern should give you a rough indication of some of the contact.

 

I'm getting there, i'm getting there 16k WILL be mine, soon, 4.5Ghz no throttling when benching really feels good, but when going up to 4.6Ghz i somehow shifting the power balance, it doesn't throttle the CPU but my GPU throttles more.

Firestrike Link - GPU +120Mhz MEM +400Mhz
Firestrike Link 2 - still squezzing same clocks but with Speedshift

edit:
Firestrike Link 3 - GPU +140Mhz MEM +425Mhz

 

The mounting arms are attached to the backplate, which seems plastic but may be plastic coated metal. Earlier I discussed using something, like non-conductive thermal pads, etc., to use between the backplate and the board to allow for a standoff to lower the arm height and remove those from the equation, but then the downward force on the socket needs remeasured to make sure it isn't exceeding spec.

We've considered the manufacturing tolerances, but a shim is a lot easier for many than removing the board to space the plate on the back. Also, we've complained incessantly about QC issues on the heat sink contact plate itself. Many come bowed, meaning they need to be adjusted or lapped for better surface contact. Meanwhile, the uni-heat sinks of the P7 series also require an awareness of azimuth due to the connection between both sides effecting full contact (why I went through the downward force calculations, pressure film, lapping both IHS and heat sink to mirror shine and flat, devised the leg offset, etc.).

Sent from my SM-G900P using Tapatalk

 

@Tishers @Mr. Fox @Johnksss@iBUYPOWER


Ran into a weird problem with my current BGAbook listed on the signature below.


As usual I put the laptop on top of the AC unit at work, I set the coldest temp and saw a 28c idle load temp on the GPU while on P0 (15-30w).

But whenever I run benchmark the laptop sometime blackscreen and whatnot, really dunno what's the problem here. I need to shut down by power button and reboot to get normal control again.

Condensation maybe?

 

No, I doubt it is condensation. What you are experiencing sounds like a weak GPU problem. When I was having trouble with weak 980M cards I had exactly the same problem. Only happened when overclocked. After having Khenglish beef them up with the missing MOSFETs and higher capacity VRMs that went away. Some drivers are worse than others. You might try using different drivers and see if the behavior is the same. Go back several versions. If it is the same with all drivers, your BGA GPU is having an issue. Back off the core overclock a little bit. It's the core overclock that induces it, not memory. If you go too high on the memory overclock you'll more likely see artifacts or TDR errors than black screen lockups.

 

What if it isn't overclocked though? xD

 

Double check if the thermal pads are making proper contact. There is a gap somewhere between some GPU component and the heat sink.

 

the voltage isn't bumped up, it's the same 0.8v at 115w

I don't think I can add more vrms, I'll double check on that heatsink contact

what component should I look out for?


this didn't happen with the gs63vr and the g752vs (stock thermal pads)

the only component not covered is the chokes (little gray squares)

I'll open up the heatsink again tomorrow to check on GPU VRM

CPU VRM is covered with K5 Pro though (gummy thermal pad)

 

Those actually need to be cooled. But not that critically.
I tried K5 pro with my modded heatsink and the 980 in the DM1, and ended up black screening even at stock clocks.
I think it's more effective with very tight fits.

 

Hmm, VRM overheating then...

Very weird that it happen during very short bursts of benchmark (below 40c), even in my extended gaming sessions and CPU OCCT+GPU FS Ultra loop on ambient it doesn't usually black out like this (68c).


I'll leave the memory door open next time, that will allow some cold air to directly touch the VRM area of the heatsink.

The VRM / choke area for the GPU has one small heatpipe covering the general area.


I'm pretty sure the AC is the same, I don't feel that the air is any significantly cooler than usual.



HP didn't feel that the CPU VRM needs to be cooled, it didn't come with a thermal pad. That's why I opted for K5 Pro, seeing that the gap is only around 0.5mm from CPU VRM to the heatsink.

 

Doesn't really matter. The problem is the same. In my case it only happened with overclocking 980M cards that were weak. Some could overclock higher before a black screen more than others, But, It could happen at stock, too.

Either something is getting too hot or a component is weak/defective. Could be either one. It could be the inductors (the gray square things) getting too hot if they are not being cooled. As lame as the cooling system seems to be from everything you have posted, it's probably got something getting too hot. That may lead to component failure, if it has not already.

 

Test with the nvidia driver which is verified by the OEM for your lapbook.

 

I'll try to get the WHQL driver, atm I'm on the latest hotfix.


Really weird when the computer can do 10+ hours of burn in without trouble on ambient.




K5 Pro will be applied on the chokes, lets see if that helps.




Edit: oh yeah, I burned the stock thermal paste in for the first 12 hour of receiving the machine, possible break something with that large of a gap in the fan recirculating hot air all over?

 

Have seen several had black screen problem lately. For some it helps with another driver. Start with this first, then try with new paste - pads.

 

Yeah, I'll revert to WHQL tonight.

 

Hello. W10 is trash. That is all. Goodnight.

 

Mine fixed up after I replaced the tired looking pads on the fets. I had to use 5mm pads on one of them as the alignment was out and 4mm wasn't enough after I bent the heatsink to have better core contact.

 

Some of us already knew this, but there is always room at my table for people that refuse to pretend it's a good product.

 

Any love for Windoze X ?

 

Oh no. You misunderstand. I always knew it was trash, but not THIS much trash. I had to reinstall my OS 3 times yesterday for various reasons AND STILL each install behaved differently. Down to I had microsoft remote support on my PC for four hours last night to do a crapton of work JUST to tell me that my under-6-hours-old install of W10 was entirely missing things from the registry and unable to configure the update service properly

Sent from my OnePlus 1 using a coconut

 

So true. Having same problems my self.
Thinking of just going 8.1/7 + Linux now.

 

8.1 on this machine (kbl anyway) I have found to be a real detriment, but I won't even joke.... I think it might be worth it if these W10 issues are par for the course. I had to start another reinstall last night around 2:30am because the OS was already so broken.

It kept. insisting. on installing drivers via update too; regardless of what I set. Reverted my 376.60 desktop mod and replaced with a 376.54 notebook driver, completely against my will. It also updated a few others too. I couldn't even stop it. Even the MS techs kept checking winver and system properties because they're like "it... says 10 Pro... but it's not letting me choose a thing... even the key is valid...". I think it's ok " now" but I am not even touching much else until I can burn a drive on a backup image I can restore. This is some ridiculous crap, I have never needed to store OS images and quite frankly I don't really have the spare space. But hooo boy if it isn't apparently necessary as all hell right now?

Sent from my OnePlus 1 using a coconut

 

I reinstalled a few weeks ago and things seem to running ok once I got all the drivers installed. I do prefer windows 7 though but never tried out windows 8. Ended up skipping that operating system.

 

Has anyone found that auto fans don't actually ramp up to Fn + 1 Max fan speeds on their own?

 

Not really. On mine they go up to about 80%