Clevo Overclocker's Lounge

Thank you for posting this. I was about to make a video explaining this, since @Phoenix or someone was asking about it.
Super convenient for me , specially to able to boot into W7 for benchmarking only when I really want to. Otherwise it can just go to W10 by default.
But I think.my video is also set to UEFI. Going to double check.

 

Yup... made by turdbook lovers, for turdbook lovers. One size fits all is seldom an elegant solution. They wanted one half-assed OS for all platforms, so they chose the low road with a smartphone-grade OS. They seem oblivious to the notion that not everyone appreciates the flavor of a digital urine and excrement cocktail.

 

So it's pretty much confirmed now, my primary (formerly; slave is primary now) GPU is basically dead, and unusable for games. Looks like I need to find me one of these machines and join in some OCing with you all!

is dying inside

 

Sorry to hear that, bro. We've got room at the table for you. Would love to have you join us with a DM3 monsterbook from a @Prema partner. You'll be glad you did.

 

I was possibly to actually change the Video to UEFI too, but only after W7 has been installed, drivers installed etc in Legacy first on the DM1.
Can you see if that still works @Mr. Fox..?

 

I'll check. I don't know if UEFI video will work for DOS booting though. Will have to see if that works. I like having "set it and forget it" boot options.

 

heh, I know. I want one pretty bad, honestly. Just no idea how I'm going to get one. I'll almost certainly have to sell this too.

 

So, as usual, @Johnksss@iBUYPOWER is right. (I know, my surprised look is not on my face about that.) My 6700K will validate 47x4 with stock voltage and no changes to anything except multipliers.

http://valid.x86.fr/2bvn6f

I goosed the power limits to 270W and added just 5mW and it completes 3DMark 11 Physics test. Max core temp was 73°C... with auto fans!!! LOL! That's awesome. Thanks, Brother John.

 

Tested and it's a big fat 10/4 negative!

 

Guess I'll have to go back and try mine again at 47x4 with stock voltage... Last time I did it was set to -10mV, lol. Guess that why I was able to run wPrime 1024m once, and the next time it didn't run all the way through it locked up... Not enough voltage Still learning the ways of the overclock, haha! What offset are you running to hit 4.9/5.0GHz?

And of course I don't have the @Prema awesomesauce BIOS yet and dealing with the stock BIOS/EC

 

Still massaging my 4.9+ so more to follow on this.

Yes, you will need more voltage to bench at 47x4, but go slow. Try in +5 increments to offset. Watch the load voltage amount once you find the stable value it needs, then you should be able to change from adaptive to static voltage with zero offset for a similar load voltage value. Once you find the sweet spot, static is always better than adaptive (IMHO). Adaptive can be more forgiving sometimes.

 

Awesome, thanks for the tips! Yea I hadn't even thought of going static yet... But I guess I've reached that level now that it's more crucial to get precise with the voltage!

 

You just applied stock voltage, and 47X multipliers and it was stable with those temps on auto fans?...

I mean, what is the secret?... There must be something!?

 

@Mr. Fox Do you have any opinion about what I talked about in my previous post #6404? http://forum.notebookreview.com/threads/clevo-overclockers-lounge.788975/page-641#post-10387095

@Georgel (You just applied stock voltage, and 47X multipliers and it was stable with those temps on auto fans?... I mean, what is the secret?... There must be something!?)
Good cooling and good silicon bro Aka delided processor and Liquid metal makes a difference (better cooling) Read https://biosmods.wordpress.com/2015/10/21/sky-x9/5/

 

Negative as in you get the Windows boot load screen that's it?

 

Yes. I agree with you. Shared cooling is not a good idea. I have never liked that. It's possible there may be some benefit for the GPUs to have more headroom due to NVIDIA ridiculous Pascal throttling that starts at crazy low temperatures. One of the things @Prema aims to correct with the vBIOS mod is to remove that stupidity so the GPUs can run at full boost up to the same temps as Maxwell. There is no legitimate excuse for the GPUs to throttle to less than max boost clock if the temps are below 91°C.

P775DM suffers from unified heat sink just as my P750ZM. This is not a good design for any laptop in my opinion. You are going to need to delid and use CLU to get the best temps that machine can offer with the unified heat sink. I think you can expect nothing but higher than acceptable temps with stock paste.

Yes, see above about my opinion. I agree with your comments to @Georgel as well.

It has never worked in the past for me and it does not work now. With UEFI video Windows 7 freezes at the logon/welcome screen.

 

As in you don't see anything....A black void.....

 

Could you monitor max power draw from single graphics vs. the new RazerBook? http://forum.notebookreview.com/thr...16-owners-lounge.797308/page-10#post-10387820

 

Maybe later. Working on tuning now. Almost time for bed after a few more tests.

Turns out my G.SKILL 3000 has problems. Until I can get them exchanged I am running my 4x4GB spare Kingston 2400 RAM.

 

This is with a desktop 1080 with limited boost clock. (on a desktop)



and with stock OC BIOS :

 

This is with the 2400MHz Kingston RAM.

http://www.3dmark.com/3dm11/11741018

 

This was the above 3DMark 11 run... stock vBIOS.

 

For testing my memory I followed @iunlock 's method and used occt. If that memory is unstable it will fail with in the first 5 seconds. If you can get past the first few minutes it usually mean you should be ok with clock settings. I'm going to use your @Mr. Fox software to test mine as well.

 

With the TM5 test the G.SKILL 3000 produced over 100 errors in the first 5 seconds and rebooted the machine the first run. The second run had a BSOD and almost 200 errors in the first 5 seconds. The Kingston 2400 completed the TM5 test three times without any errors. I'm no longer having Cinebench crashing or 3DMark 11 and Fire Strike exit to the desktop with the "workload exited unexpectedly" error message from Futuremark. I was fighting that for a good while, since before @Prema BIOS testing started. I thought it was drivers based on the STOP error messages. I suspect it was causing my frequent Time Measurement errors in 3DMark 11 and Fire Strike as well. About 1 out of 5 runs produced that error, even with BIOS defaults.

I never had any issues at all with the G.SKILL 3000 RAM that I purchased from HIDevolution for the P870DM-G that I sold @Rage Set. It was rock solid.

 

Well that sucks on the ram. I just tested mine to be sure and it's working as intended.

 

Yes, it was a bummer to see that.

Oh well. At least it's easy to swap and good that I had the spare Kingston 2400 to use. And, it is great that I know what was making my machine misbehave now.

 

it's a pain the ass having such these problems, hope you get your change sticks asap man.. the fist time I had a stick gone up I was having random bsod/freezes and thinked many times it was a software problem. After the stick change, all came back to function like a charm

 

I love REOL but we gotta admit that this video especially looks like a video driver having a crash or a video card that is dying...

Makes me like their videos and music more tho..

 

Unified heatsinks are great for day to day use in terms of noise and for enthusiasts with a good eye can be corrected for if you want to push it most of the time.

As for my CPU heatsink/VRM I might just take a physical look at the pads/paste to double check before playing further.

 

http://www.3dmark.com/fs/10791313

 

Nice score bro Fox
And the the GREEN MESS continues It just gets worse and worse.
http://www.guru3d.com/news-story/nvidia-geforce-375-86-whql-drivers-have-major-issues.html

 

Man. Those are drivers, they can go wrong. at the same time Nvidia has over 10.000 people writing those 300-700 MB drivers.

They could re-write the entire driver in binary in less than a week with the man power and resources they have.

 

sure...IF they can organize themselves properly

jealous of your cinebench scores fox! still cant get a handle of this mysterious current limit throttling... i found out that its not TDP, but rather voltage bound. i was able to reach 130W on the cpu without throttling, but once the vcore breaks 1.39 the clocks go down instantly. also, i keep having those clock dips here and there while running gpu + cpu, in firestrike for example. that happens even without reaching that voltage barrier... very unnerving!

in any case, cuz of this im limited to 4.7 ghz on the cpu, even though 4.8 is stable CB11.5 gives me 11.33 points at 4.7, not too shabby

Sent from my Huawei Mate 8 NXT-AL10

 

JB, Try to disable RSR in the Mod OC section.

 

thanks for your input, but already had that disabled for a while now... still no dice. i get the feeling that this is something in the EC maybe....

just as a side note, i have all wattages and currents in the bios maxxed out just to be on the safe side.

btw, any chance whatsoever to get static voltages working properly on the DM? activating it in either xtu or the bios doesnt change anything, still on adaptive with fluctuating voltages...

Sent from my Huawei Mate 8 NXT-AL10

 

I am quiet sure I already gave you the modded EC...how much is the system pulling from the wall at max load?

 

yes you did, definitely no offense intended there buddy.

ill have to check and see, but it shouldnt be too much, especially on cpu only benches. when only the cpu is stressed, the current limit activates once vcore 1.39 and above is reached. when both cpu and gpu are stressed, those dips happen even at lower voltages, no idea whats happening then...

Sent from my Huawei Mate 8 NXT-AL10

 

ok just checked: ran occt large data set on all cores and checked the wattage at the outlet.

4.6 Ghz -30 mV adaptive = 150 W, no throttling
4.7 Ghz +35 mV adaptive = 170 W, no throttling
4.8 Ghz + 100 mV adaptive = 155 W, current limit throttling

 

Well, the throttling may possibly be due to +100 being too much voltage for 4.8. The benchmarks I posted about 5 or 6 hours ago was with +55 for 4.8GHz.

 

As far as I know, skylake doesn't static. You can influence it with manual voltage (targeting the baseline) and offsets, but the CPU will do whatever the hell it wants with the actual voltage.

As for the stability impact of setting a good manual voltage prior to an offset or simply using "default" combined with offsets? Beats me.

 

I see the most stable (not static) voltage with @Prema BIOS by setting Adaptive for Core and Ring then making the offset the same for core voltage, extra turbo voltage and offset, then mirroring the same offset to Ring voltage and offset. For 48x4 I have 55, 55, 55 for CPU and for Ring 55, 55. There is still some fluctuation in voltage, but seems to be less erratic than with static voltage for some reason. Maybe @jaybee83 can try that.

If you make Ring voltage less than Core under heavy load the Core voltage will eventurally drop to whatever the Ring max voltage is and you'll either crash the benchmark or get a BSOD. It is OK in shorter tests, like wPrime 32M, but in 1024M or Cinebench testing that runs a lot longer the Core voltage will gradually get lower and lower and eventually it will be too low, and then the crashing or BSOD comes. That's why Ring and Core need to match. At less aggressive clock speeds setting Ring voltage less than Core seems to be fine.

 

yes and no, since the actual voltage is what matters in the end. i get the following during load in occt:

4.6 Ghz = 1227 - 1336 mV, avg. 1270 mV
4.7 Ghz = 1278 - 1367 mV, avg. 1304 mV
4.8 Ghz (extrapolated, since it throttles) = ~1330 - 1470 mV, avg. 1370 mV

what are your total voltages? ballpark figures should suffice, just wanna check and see where i stand with my cpu here if its indeed just too much voltage, i guess theres nothing i can do btw, those spikes at 1.47V are not frequent and the average VCore is definitely below 1.4V (my personal limit for VCore).

also, here are some pics of my day to day bios settings, if you see anything out of place there concerning that current limit throttling, please let me know!

Link: http://imgur.com/a/Hd50e

 

this post is worth GOLD! thanks so much! so now i finally KNOW for SURE that Ring and VCore are INDEED coupled and CANNOT be separated from each other in Skylake, even if the PremaMOD and ThrottleStop allow for separate adjustments

so far, ive only set either offsets OR extra turbo / core voltages, but never together. this is perfect! i will go ahead and try those settings right away, lets see if that influences my stable settings and also heat in any way

edit: question - isnt the "core voltage" setting the actual total voltage for the core instead of an offset? so in your case, setting "55" would result in a core voltage of 55 mV? also, wouldnt the extra turbo voltage increase the voltage range even further? since it adds another 55 mV to the regular voltage fluctuations... these are just my naive thoughts on this

 

NOICE!
@Papusan really needs to tell us how he optimizes his windows installs.

Just ran this for kicks. Obviously unfair to compare ..

Spoiler: :S

 

I really need to get a skylake machine. I feel bad knowing more about Skylake machines than most of the people who own them without ever touching one (my sister's 6700HQ doesn't count).

Edit: Since Prema liked this, I assume this is exactly how Prema feels about the dozens of Clevo models he's never touched when people are asking things about them as he unlocks their BIOSes

 

yeah, im still in the learning phase most of what ive learned for my previous 4790K doesnt apply here anymore, so its a bit of a trial and error thing. especially with contradictory reports when asking different people / reading different articles online!

on another note: im facing a very strange conundrum, namely the apparently quick degradation of my kryonaut paste! as mentioned before, both my gpu and cpu heatsink sit absolutely perfectly flat, i havent seen such perfect fit in stock heatsinks before! when i take off the heatsink, both cpu and gpu show perfect, thin spread of their entire surfaces. still, when freshly applied i get awesomely low temps on the cpu, but only a few days later they suddenly jump up ive already repasted a second time and the spread was still perfect after heatsink removal, soo....wth is going on?

 

Looked at your screen shots. Under the System Agent, Core/IA and Ring Domains try changing your VR Voltage Limit from zero to 2000. On the CPU Configuration menu I disable CFG lock and disable c-states. Try setting your Power Limit 3 and 4 values to match your other power limits if you have not already. It may or may not make any difference, but I make all of the power limits exactly the same. I also set my Turbo Time Window to 1. Also try enabling IMON Scaling Support and see if that helps, hurts or makes no difference. It seems to help me at 4.8+ to have it enabled even though I cannot find a definitely answer on the web about what it does.

If you set static voltage, yes. The machine will crash with a max of 55mV LOL. With Adaptive it appears to be "added" voltage. Again, there is no reliable or authoritative written documentation that I can find that is worth squat on these settings, so I am speaking based on my trial and error impressions.

 

Yes, I'd like to know that as well. Killing background garbage with Autoruns helps a lot, but there has to be something Brother @Papusan is doing differently in Windows to reduce overhead and I don't know what.

Congratulations on getting a heat sink that fits well. Kryonaut works well but does not last very long for me either. It's good for GPUs but on my overclocked CPUs not so much. I know you had an experience with CLU getting off the die, but that's really the only stuff that works excellent for me. If not CLU, IC Diamond is the most durable option. It lasts me longer than anything other than CLU. I won't use Conductonaut because it is too watery for my comfort. Given your experience with CLU moving with your rough and tumble computing habits, I would never even consider recommending Conductonaut. While it works amazing, I have seen three examples of Conductonaut moving from where I placed it; thankfully only on the heat sink copper plate, not onto anything critical. I've never had that happen (touch wood) with CLU. For that reason I am no longer using Conductonaut for anything.