Clevo Overclocker's Lounge

OK, thanks. I am on beta 18. Maybe that will help. I will let you know. Back in a few.

Edit: Ugh! Still has the sinfully butt-ugly Metro UI like beta 18. Makes me want to puke, LOL. What on earth was the person that designed that tacky RTSS skin thinking.... super-ugly.

 

What version of MSI AB? Link is no good. I am on 4.4.0 beta 6 right now.

 

It's Beta 7. Try clicking it from this page http://forums.guru3d.com/showthread.php?p=5419768#post5419768

 

Nope... exactly the same as beta 18. Here is where it it makes the game halt. MSI Afterburner is not even running. I don't ever use it for OSD output. HWiNFO64 and AIDA64 both do a better job of that than MSI Afterburner.

 

OK, will try it just for kicks, but doubt it is going to help since I don't have it running and never use it for OSD data output any more. Thanks for the link though.

 

Ah, I see. Maybe try closing HWiNFO and using MSI AB for readouts for now? If it doesn't work, I can link you to the thread where you could report the issue.

OH MY WORD YOUR VOLTAGE

 

Nope same problem.

I won't use MSI AB for OSD output. It sucks and always has sucked in my personal opinion. It's missing the information that I care about most and what it does have makes for a a terrible screen pig taking up too much of the left side of the screen. I like my data arranged like shown in the screen shot. Well organized and hardly any screen space is used. I just won't have OSD for that game and run the fans on max all the time.

Meh, the voltage is fine. There's no load and that's why it seems high. It's 1.300V static for 50x4 and when the CPU is under load, that's about where it is (about 1.312V). High idle voltage keeps Skylake and Kaby Lake stable. Low idle voltage makes them unstable and causes too many lockups and BSOD at higher overclocks.

 

I see. I was able to get my 50x4 run at 1.28v, but I don't know if it's stable enough for 24/7 use. I certainly can't currently cool it, haha.

 

If you got it to run at 1.28V in Cinebench without a crash, your CPU is better than mine. I can run it at 1.28V but it's not enough load voltage for mine. I need at least 1.300V for 5.0GHz to complete Cinebench. But, when you stop and think about it, we're only talking about 0.02V difference, LOL. Not a big deal.

I have my system set to boot at 5.0GHz (50x4, 83 core max, 83 cache max, 1.300V static) and use ThrottleStop to slow it down for normal use. I did not have ThrottleStop running.

Here, maybe this looks less scary to you, bro. 47x4 at 1.100V. This is with ThrottleStop running. This is also what I normally use for gaming. Stays cool enough and most games don't really benefit from more than 3.5-4.0GHz anyhow.

 

Maybe tonight I'll try lowering my static voltage by 0.05v and running cinebench over and over and see how well I can manage for 50x4

 

Well, I exported the new GoW4 SLI profile, got rid of the new driver and imported the new profile into my preferred driver and GoW4 actually works better than it does with the fancy new driver, LOL. Another fail for the Green Goblin. @D2 Ultima

 

Hey Mr Fox

How are your CPU temps with your new shim?? I am thinking my cpu heat sink might be bent and not making correct contact. I took the C clips off but made very little difference. if is the contact the problem any advice u got for me other than solder shim?

 

I am not using a shim now that I have a Bitspower IHS. (Only buy the version for 6700K since it is 1.0mm taller than the one for 7700K.) I no longer need one. My temps are respectable.

My advice would be to add a shim now that the c-clips are eliminated. See if the temps improve. If they do, get the shim soldered and they will improve even more. If they get worse or stay the same, you probably need a new heat sink. I would recommend using a 1.0mm shim. The extra thickness won't hurt anything, but it might help if contact pressure is way too low. Remember to adjust the thermal pad thickness. Check to be sure there is not an air gap after adding a shim. The stock pads are too thick without a shim, but they are perfect when the shim is added or the 1.0mm taller Bitspower IHS is used.

Have you used a straight-edge to see if the copper heat plate is bowed in the middle? This seems to be too common. The copper plate is too thin and weak, and I believe what we are seeing is they may fatigue and lose their shape over time. While you are checking that, also look for a twist. Are the 4 corners parallel, or would an imaginary line drawn across them in either direction have an intersection?

What I am not sure about is whether soldering a shim, even if I do not need one, might be good step to take to improve the structural integrity of the flimsy copper heat plate.

BTW - I sent the heat sink that I soldered the shim to back to @Ted@HIDevolution @Donald@HIDevolution and @thattechgirl_viv for further examination and testing. If it works as well for them as it did for me, that could become one of the value-added services coming from HIDevolution in the future. I appreciate how they and @John@OBSIDIAN-PC are taking the lead in setting the bar that their competitors will be measured against.

 

I wanna try the bits power eventually. I am running @ 4.5 on all cores at voltage of 1050 (static) and while gaming (heavy games) My temps stay around high 70s and low 80s but they do Peak to 85c seems high to me for just gaming

 

Yeah, that is a bit high, but not horrid. I am cooler than that at 4.7GHz, so there is clearly a lot of room for improvement from where you are now. I think your voltage is right about were it should be, too.

 

Can someone point me to where to get and how to install @j95's drivers?

 

Why do you think I say nVidia profile inspector is the reason to use SLI at all? You didn't even need the profile from the other driver; you could have manually added it if you were willing to fiddle, or found someone else who had it somewhere. Glad it's working on your preferred driver

 

Yes, I have been using Inspector for years. It's a great tool. I had been fiddling trying to get it to work correctly for GoW4 and had found nothing that worked well. Stuff I found posted by others sucked for me. Single GPU was better using the suggestions I had found.

NVIDIA did good on the GoW4 SLI profile and (as they often do) dropped the ball with another performance-impairment driver. Since I had the new cancer driver installed (before I knew it sucked) I thought, might as well take that and use it elsewhere. Saved me a lot of time. Used to do the same thing with newer drivers that buggered up 780M SLI on my Alienware M18xR1/R2/18 machines. Used SLI profiles from new drivers with drivers that were a year old and had no throttling problems. Worked better than new drivers every time.

 

Nice.

What do you suggest I set PL3 and PL4 to by the way?

Sent from my OnePlus 1 using a coconut

 

It may not make any difference. I am not sure PL3 and PL4 get used. I set mine the same for PL1/2/3/4. Leaving PL3/4 set to zero doesn't seem to affect anything, but if those are reduced performance levels (maybe higher number works like c-states and means lower performance) that would make sense. My impression is that setting them all the same helps with stability above 5.0GHz. Right now I have all of them set at 300000, but 160000 is enough.

 

Yeah I just set all mine to 300,000 just now. I was getting random edp other throttling in occt at 5GHz. I was checking for @Mobius 1 since he said he throttles 100% of the time. I am able to get it to work *sometimes*. Literally once every couple runs it will run fine until thermal (which is what it should do). But other times it just eDP throttles to between 4.2 and 4.4 (fluctuating) and sits there. 4.7 was working perfectly too 100% of the time so I was confused.

@bloodhawk showed me some other limit settings to raise though so I will be checking those next.

Also when it works it seems 1.28v works for 5 with occt

Sent from my OnePlus 1 using a coconut

 

Well the thing about 5ghz is I need a lot more power to sustain an OCCT run, I can do 5GHz on simple bench like firestrike.

Don't suppose I can raise the power limit any higher, need prema on this @Prema

 

5GHz on OCCT didn't even use 150w for me tho

Sent from my OnePlus 1 using a coconut

 

maybe vrm overheat?

 

Don't use OCCT. Use Cinebench and wPrime. OCCT is silly. It presents a workload you will likely never encounter in overclocked benching, so using it to test the effectiveness of overclock settings doesn't make any sense. Plus, it's really harsher than it needs to be. In my personal opinion it is as pointlessly brutal as using Furmark on GPUs. Brother @Johnksss@iBUYPOWER doesn't use it either, in case that means more than my advice.

KM1 needs power limits fixed. Already fixed in DM3 by @Prema if you use the right settings.

Turn off the setting for BCLK aware adaptive voltage. If you're not using BCLK to overclock, no reason for that to be enabled. Also disable RSR and leave AVX offset ratio to zero.

If you have the cache ratio set too low it may help with temps for normal everyday use and gaming, but at 5.0GHz and higher that is going to drag down your core clocks and give the impression of throttling. For 5.0GHz and higher, set max core OC ratio to 83 and cache ratio max to 83 and let the core and cache voltage be in sync. To avoid throttling from cache being too low (or crashing) at 5.0GHz or higher the cache ratio should be 300MHz less than core ratio. If you set the cache ratio to 83 on the same screen with the multipliers it will take care of this automatically for you. It will always be 300MHz less than core ratio. 50x4 will make the cache 47. 47x4 will make the cache ratio 43. You'll never have to touch that setting.

If I set my multis to 50x4 and the cache is manually set to 42, it will not hold 5.0GHz in a CPU-heavy benchmark.

 

Set Bloodhawk's settings. 5GHz basically stable in OCCT. Seems the system agent current limit needed bumping.

After a while on 5GHz I get these, though:

Spoiler: pictures

but on 4.9GHz I never get them. Voltage issue maybe? Nothing crashes on the PC. I'm not overheating when it happens (below 95c). Same settings were used for Cinebench and it seemed to work.

I don't care for OCCT really, but I decided to check for Mobius and see what the problem was.

 

Yeah, Brother @bloodhawk and I collaborate on overclock settings. I collaborate with Brother @Johnksss@iBUYPOWER as well. Chances are good I am using the same settings, or something similar.

Again, why even screw with OCCT? What are you accomplishing? Run Cinbench or wPrime, or 3DMark 11, Fire Strike, Time Spy or Vantage. If the clocks hold all the way through and you get a nice score, you're good to go. No more fussing necessary. What happens with OCCT doesn't matter.

 

See the bottom of my post xD

 

Cool.

Brother @Mobius 1 knows what he is doing, but he is fighting a different battle than we are with the stock cancer BIOS on the KM1.

Brother @Prema has not arrived on his white horse to save KM1 owners from Clevo throttling hell yet.

 

I cant remember the last time i ran Prime95 / OCCT / LinPack / IBT on anything other than my desktops. They are jut way too brutal for anything portable, these systems cant dump heat fast enough. Heck even some desktop cooling solutions cant as well.

 

Yup! Though I'm more and more interested in how 4.7GHz runs. If I can get voltage down lower it might suffice for daily clocks.

Also, he's so annoyed with it you don't even realize the half of it

 

I've never understood the rationale for using those tools. Sometimes they are too harsh even for stock clocks for air-cooled desktops and laptops.

We have our own little private thread going and I totally understand the mess he is faced with. Not only cancer firmware, but I think Clevo may be cutting even more corners on cooling with KM1. I notice the vapor chamber is missing a heat pipe compared to ours and the plate that goes across the rear plane to tie everything together is gimped now. Nice that they opened up some BIOS menus that were hidden before, but this other stuff is for the birds. But, to be fair the stock Clevo firmware has always been worthless feces, so that's nothing new. I say to hell with an extra M.2 and Optane if the overclocking potential is gimped. Those are nice features, but they're not job #1 in my book.

 

I get it in real-time on plebcord xD. Hell half the things he randomly shows up here and asks is because I told him something xD

 

I have to ask...How is it you can not run 4.7 ghz?

 

Yes, the KM1 stock bios is beyond broken.

5820K 4.3ghz 1.296v 200w under occt

NH D15 with dual fan 2500rpm not able to keep cool under 85c

 

<3

4.7 works for me, under 68c

ambient 25c

He have ambient high in the seas of Caribbean full of pirates

 

Yeah my 6850k under a custom waterloop @ 4.8Ghz with 1.445V maxes out at 78C with Prime95. (26C Ambient)

 

This room runs in the 100F and can still run 4.7 ghz.

 

Environment hazards are tough to deal with. Pirates make it even worse. ARRRGGH, matey.

Even playing games for 6 hours and it's still OK at 4.7GHz in a 100°F room? Mine handles 4.7GHz all day long just fine, too. But, I'm in a 70°F room. And, when I start doing something like that it can get a bit warmer than I like it. Not dangerous, but hot enough that I have to keep checking things for my own peace of mind. The GPUs running are actually much harder for me to keep cool than the CPU is at 4.7GHz in an extended gaming session.

 

anything above 1.3 is said to degrade for hwe tho

 

Careful now. Unless you need another time out, LOL.

 

I'm surprised someone didn't die from the heat in that room

100 degree is the boiling point of water, right?

 

F not C. Yes, 100°C is your corpse is turning to BBQ and you're already dead. That is 212°F, LOL.

 

That is to say that one does not need to redline a cpu at 4.7 to play a game.

F and C are different.

 

I could, just toastier than I would like. It hasn't shimmed or bitspowered yet or anything, though. But it can get rather hot when the GPUs are stressed a good deal for me

Sent from my OnePlus 1 using a coconut

 

Indeed. Just about anything more than 4.0GHz playing most games is wasted mojo that does absolutely nothing to make the game any better. I usually clock down to 4.5GHz and that's still way beyond overkill.

 

Remember...I had a factory Sager bios with no mods and was able to do all this, but then I was not trying to "optimize" it like it's a desktop which can not be done. The results will always be different. The harder to try to make your system run like someone else's...The more problems you run into thinking your system is not running right.

 

And, if I remember correctly (which I might not be) Sager had actually fixed a couple of things that Clevo had totally screwed up. It wasn't your generic Clevo Taiwan cancerware, was it?

Even so, having all of the extra stuff not related to overclocking hidden would still suck six ways to Sunday. The absence of menus for things that have nothing to do with overclocking blows and they deserve a lot of hate for that.

 

Correct. They readjusted somethings after I asked them too. I did some testing for them to get it closer to where it should be without them violating any agreements with Clevo. Since Clevo was the one doing the re adjusting.
When I first got my laptop. They didn't do anything remember. Mine and Meakers which came from Sager were not totally thrashed in the power department.