Clevo Overclocker's Lounge

Is that what you want for lunch when you come up to the bay?

 

ON A POSITIVE NOTE!

I was battling high NVMe temps (960 Pro and BPX SSD) in the Tornado F5... always in the mid to high 80's even after throwing my money away on the nice looking black M.2 cooler on NewEgg (worthless product). Well, I ordered these on eBay...

Silver-White Heat Sink 150x20x6mm LED Heat Sink Aluminum 150*20*6MM Cooling Fin

I cut it to length for the BPX SSD, used ICD with a copper shim over the NVMe controller processor, then used those extra fragile sticky blue 0.5mm thermal pads that tear up so easily on the other chips. I used a pair of o-rings to hold it on the M.2 drive, and WOW... 20°C temperature drop. The Tornado F5 has been in use for 6 hours and it used to get the NVMe drive this warm in 10 minutes. Very happy with how that turned out. Plus, my metal palm rest is not getting hot like it normally is. I used one of the two fancy black heat sinks (that did not work well) on the bottom side and used the other one with a pair of o-rings and ICD on the top side of the X400 and it's much cooler now as well.

Edit: I had already peeled the labels off despite the "warranty void if removed" threat on both 960 Pro and BPX. Seems like those labels had little or no effect on the NVMe temps. I saw no difference after removing them for direct contact with thermal pads. I thought they might be insulating heat, but that was not the case as far as I could tell.

 

you know it!

Damn. Weirdly enough i haven't had any nvme/m.2 drive overheating issues yet on the DM3.

 

You can also mix Liquid Ultra and Conductonaut. Shouldn't be a problem. If you have the older more dry Clu, this will be a big win. More applications from what you have and the metal will be dryer and not so wettery.

 

OK, then that is not my imagination. I noticed exactly the same and thought that maybe I messed up somehow. Maybe it's the copper against copper. Following what Obsidian-PC is doing, I am going to be soldering my shim directly to the heat sink, so it will be interesting to see how that turns out. I will probably do that in the next week to 10 days, so more to follow.

 

Oh, me neither. Only the Tornado F5 has that weird issue.

I have a fat thermal pad sandwiched between each M.2 SSD and the motherboard in my DM3. I used one of the stack of 2 that was installed under the 960 Pro that came with the Tornado F5. I placed one of each under the M.2 drives in my DM3. The thickness was perfect for the DM3. Just enough to have to forcefully smash down the spongy thermal pad. The 960 Pro runs as cold as ice--not really, but you get the picture--when it is installed in my DM3 with that pad sandwiched between it and the motherboard. The 960 Pro is usually in the mid- to upper-40's and lower 50's except when it is under severe load, and even then it's only reaching low- to mid-60's now. It was hitting 90°C in the Tornado F5 with the stack of pads between it and the motherboard.

Not sure why they get so crazy hot in the Tornado F5 unless it's only a matter of poor air circulation where they are located. Even the X400 was getting way hotter than normal for an ordinary SSD before I used one of the crappy Alphacool thermal plates on top of it. It almost has to be because they are stuffed into a corner and cannot breathe properly in the Tornado F5. But, problem solved... working great now. I think the fins did the trick. The crappy flat Alphacool thermal plate was probably getting saturated and trapping/storing heat.

 

Yeah i have all the parts as well, planned for the next week end

That is indeed pretty freaking high. I think one of the reasons why the DM3 M.2 temps dont get too high would be because of a decent amount of space for air circulate under the keyboard and that fan hole.

 

I'm sure that helps, too. The M.2 drives in the DM3 are on the cooler side of the motherboard as well. The M.2 drives in the Tornado F5 are in close proximity to the CPU. I suspect that does not help matters in terms of having an endless supply of radiant heat "feeding" it plenty of hot and stale air.

 

If cache is at 40x, is it safe to lower it to 38x if the core clocks are at 42x?

 

Yes. It should be safe to set it to whatever ratio you want to. It should also be stable at 38x. It might be a better approach to use ThrottleStop to lower the cache voltage and leave it set at 40x if all you are trying to do is improve temperatures. If the core and cache are set to the same voltage it will run hotter than it will if you lower the cache voltage.

 

Yeah, the F5 need HEATSPREADER MOD

I tested ATTO In comparison vs. stock F5. And the P870 models have good ssd cooling as @bloodhawk said. I don't know how the temp is with the newer Samsung NVMe Pro disks.
It's damn sorry the reviewers don't test how the ssd's works and what temp can be in the paper thin JokeBook's!!

ATTO Disk Benchmark Is a better ssd benchmark than ASSD - DiskMark if you want to see how your ssd works.

As bro @Georgel have temp problems, I suggested he lower the cashe. If lower voltage doesn't help much... I will still suggest him trying to lower it. Only fixing his heat problems is the real fix. But if he don't want... What else can I suggest.

 

http://www.3dmark.com/vrpbr/16746

So who is getting the 3k for us? @Mr. Fox @Johnksss@iBUYPOWER @bloodhawk @Meaker@Sager

 

You're closing in on that 3K!

Don't worry, the delid will happen sometime in the future, just trying to prolong the time before I do that

This and there are no delid tools available anywhere near, gotta take it up with Romanian postage if I order one and it'll be very unpleasant

That sounds interesting! Thank you!

 

Very nice Yeah, the same thing happened after @Diversion made his ssd heatspreader mod. 20C degrees lower ssd temp vs stock. This is just very yummy SSD's gets boiled in the ALIENWARE'S I expect the *nice* SSD temp in these thin AW Echo models, will rise if they overclock the cpu/gpu. I really mean the Reviewers who get test projects, should test out the SSD's vs. throttling. It's not enough to just show the nice numbers from the testing. They should check the max temp as well. And also push out the ugly numbers.

 

I do not believe any of the ODMs or OEMs properly test any of their products. We could pretend they are using us as free R&D crash dummies, but that would be a bit of a stretch since they do not pay attention and seldom correct their mistakes. They do not even make an effort to pretend that they care to try to fake us out.

Well, Alienware pretends to care, but actions (or lack thereof) speak louder than words. That adds insult to injury. Perhaps that is for the benefit of their man in the mirror... helps them sleep at night. After pretending for so long, some are even likely to believe their own lies. This is especially true of the trained yes men with very little technical expertise.

 

All to much people think they know tech Who say yooo have to - will run unlocked 7700K stock?

 

I don't think people expected just quite how hot the SSDs would become. However so long as the temps are not silly normal use wont see any impact even in a warmer setting.

 

no big achievement for you 1080 guys, but quite a achievement for me, i just tinkered a little bit with premas bios + memory OC to 3100, GPU +130/+500 and finally cracked the 16000 mark CPU ofc still @ 4.5ghz only + it's the old 376.33 driver, the newer ones 378.xx are **** for me, no brightness control anymore + gsync black screen flickering



btw is it normal that i can't use 1T Command Rate, even at Underclock to about 2400mhz it won't work, hope it isn't a big perf hit like back in DDR1 times.

 

Yes, that's right (tim-buying days are over ). Made a compare with clu; not a spot of difference between them. Actually did that test using an earlier 30g batch from Amazon, but that was about the same price as the 80g deal on ebay, so ... things got even better. Someone could start a thread to gather interest for combined buys and splitting the 80g in multiple batches (one syringe of clu measures only 1g).

As for longevity; been using it for a year now on a lot of different systems and it's still very much liquid beneath the heatsink. Really, there's no way to distinguish between galinstan and clu once they're out of their respective containers.

 

Nice job, bro. Looking good.

Yeah, the 1T thing doesn't work. It just resets itself back to 2T for some reason. Does it on the P870DM3 as well. But, at least we have memory voltage control. That did not work on the previous models and that is actually more valuable. I can set 1T for the Tornado F5 and it sticks, but memory voltage resets back to 1.200V like the previous generation Clevos did. For what it's worth, 1T is less stable and did not have an appreciable effect on any of my benchmark scores on the Tornado F5. 1T and 2T are more or less indiscernible in terms of performance, so don't worry about it too much.

 

Another day, another bench:

http://www.3dmark.com/vrpor/92659

Stop talking and start benching guys. Wanna see 12k!

 

Keep it going, @Prema !

Just a daily reminder that this is one of our culture's sign of happiness! The logo is beautiful!

 

good to know thanks, looks like i have to overvolt my memory a bit, got some crashes in Overwatch regarding memory read errors, when using 3100 instead of the stock 3000 mhz, what tool did you guys used for memory testing again? Plus i have to triple check some settings, Speedshift locks my CPU most at only x42 multis, have to use 0 EPP efficiency to work at x45, and Speedshift is only doing x40 multis have to start Throttlestop once to get the x45 working.

 

Anyone who went past 4.5 on a 6700K in P870DM - did you get ICCMax throttling cutting between 200-500Mhz off your core and cache under stress e.g. IBT or wprime 1024M or AIDA64 cache+memory? And how did you deal with it?

Bios 1.05.03PM

Hwinfo shows a 20Amp limit for SA/uncore ICCMax and I can't find any way to tweak it.

I get it a teeny bit at x45 (which my CPU can do at -15mV adaptive)
bad at x46 (+25mV)
terrible at x47 (+75mV)
such that the throughput on IBT on all of them is lower (110-115Gflops) than x44 (-60mV, ~123Gflops)

I have some thermal headroom it's 92C max running a 120W+ load in IBT at x47 in a warm room today prob ambient 25C

Tried setting static volts through TS, it was even worse, constant throttle showing even at idle

Tried undervolting everything as low as they'll go, cache to x42 at -150mV, SA to -175mV, analog I/o to -150mV... SA undervolt is needed to minimise the throttle on x45 outside of IBT but it doesn't seem to impact bench scores in any meaningful/detectable way vs stock SA voltage.

I am changing these settings using TS on the fly in most cases. Using bios and rebooting it every time would have taken a year all the settings tweaking I've tried...

 

TM5 and Memtest86 are the tools I use.

Keep your current settings and bump the memory voltage to 1.300V in the BIOS and that may take care of the crashes.

Disable Intel SpeedShift Technology. Set the Windows Power Profile to 100% for minimum processor state on battery and AC power. Intel SpeedShift might be OK for a CPU running stock under Windows 10, but it's not very useful for one that is overclocked. Having SpeedShift enabled lowers all of my benchmark scores, so I view having it enabled as being detrimental.

 

will do thx, funny tho is that even when setting min/max to 100% on high performance profile results in 4ghz, i set it to 10% min/max for the lulz and guess what 4ghz, starting Throttlestop fixes it even with a completely fresh TS install. I reset my power plan settings, wouldn't be the first time windows 10 kills itself or do some borked up stuff.

edit:
reset power plan, bios, EC yada yada, still not working without TS, but i looked through task manager and found an XTU service running, DAFUQ? uninstalled Hotkey drivers cause you know, XTU CCC bundle etc etc and service is still running.



have to get my flamethrower and exterminate that, i guess that could be my culprit, and for what is that ICCProxy Oo where is that "feature" bundle coming from?


Long Post but interesting
edit2:
Can it be that a game alone has some sort of disable turbo feature? I just started GW2 and WoW, both pretty cpu heavy games and without tinkering, my CPU was locked to 4.5Ghz permanent, i then restarted my PC again and reset Throttlestop to stock and both still 4.5Ghz without TS 8.40. Then i started Overwatch again, boom in Menu and Practice Range 4Ghz max, for curiosity sake i started TS 8.40 again, yeah guess what 4.5Ghz. Good, restarted PC again, reset TS again and started Overwatch ... like before 4Ghz in Menu's and Practice Range, i then started TS 8.40 again and get welcomed by the "first start" error/warning, what i found amusing was that the disable Turbo was checked , to be sure again i restarted the PC again, set TS back to stock again started GW2 and was good, reset TS again started WoW was good, reset TS again (last time i write that i promise) and started OW, well disable Turbo was checked again lol

 

I am not sure what is going on there Brother @Scerate. Both interesting and odd behavior with the turbo getting disabled.

I do not know what the ICCProxy does. You should be able to uinstall XTU and CCC for testing. Will be curious if the issue is the same or better with XTU and CCC uninstalled. Also disable the XTU interface in the BIOS menu to see if that changes anything.

 

So, you have now been hit like me and jaybee were. We were lucky, the new owner was not...

Sent from my SM-G900P using Tapatalk

 

very odd indeed, i mean it's ok i can fix it with starting Throttlestop but still, i want to know WHAT causes it, right now i would point my finger at M$ and nV drivers

 

Yes. I am not going to ship computers with liquid metal paste applied any more. Have done it numerous times in the past without issue, but one time is enough to not want to see it happen again. I just wish CLU was thick like it used to be. Seems like every time something is truly excellent, the manufacturer feels compelled to do something to ruin it, LOL.

Edited my post while you were replying... added this...

 

already did that brother, even disabled Cstates, disabled the "efficiency settings" but still nothing. GW2 started to do it now too, it looks like that old "throttling" bug when GPU get's involved but it's definitely not or TS wouldn't work, even HWInfo doesn't have any limit reasons, it's just that the turbo boost get's deactivated.

I will just reset my bios again and do some DDU driver uninstall and install the latest "working" 376.60 if that's not working either i will try out the latest 378.78 and see what happens. thx for listening @Mr. Fox soo far

 

When doing this, remember most heat pipes use a 150C blow out safety. So pick a lower temp solder with good thermal properties (doesn't mean you cannot go past 150C, but you don't want to be above for long). Also, the thinner the better, as always.

Sent from my SM-G900P using Tapatalk

 

haha i just fixed another bug, i installed the 378.78 driver via device manager instead of the installer and changing brightness seems to be working.

 

Now that is interesting right there. I wonder if that would restore brightness controls for Brother @Johnksss@iBUYPOWER?

 

Already did.

Will try this in a few minutes.

 

what i forgot to mention tho is that i had 376.60 installed first then i just upgraded via device manager way.

 

I'll have to try that on the Tornado F5 as well, since brightness control was broken with the 120Hz screen upgrade.

 

Damn thank you for pointing this out. Brightness control isnt working for me either using the 120Hz Panel on the DM1. I think it was working fine with the KBL BIOS, but not with Skylake. Was working fine with the 4k Panel.
Tried both modded and stock versions of - 378.78/.33/.60.

 

No you didn't: http://www.3dmark.com/spy/1376277

I'll better count my seconds of joy until you get to it...

 

old driver that is not certified
http://www.3dmark.com/vrpor/92681

@Prema
That score might sit there for a while.

@Mr. Fox

Brightness is still broken on mine.

 

Jesus Christ, how are you guys even crossing 2050 Mhz. AC ?
That can't be on air.

 

He might have one hidden that we don't know about.

http://www.3dmark.com/compare/spy/1376277/spy/1066382/spy/911958

all of the above are at or above 2050... AC is air... just cold air, LOL.

 

Sigh. That 3rd place was nice while it lasted.
http://www.3dmark.com/spy/1282817

 

Bummer.

Did you also try DDU'ing it first to kill all of the old registry crap?

 

Honestly that run even temp throttled a lot:



Not sure how it came out on top (it was literally the second run I ever did with TimeSpy), but I have done a lot of fine tuning with the CPU/RAM.
These KBL CPUs seem to have a huge efficiency margin even while holding identical clocks without throttle the end-results differ a lot.
Also for some odd reason setting the GPU lower will crash the system. Had to push it 100Mhz higher (core & vRAM) in Inspector and suddenly it was stable. Never seen anything like that before...

Trying get a feeling for the system & establishing a baseline before getting to vBIOS...

 

LOL imma try this tonight.
Haven't been able to go over +180/+500 without crashing.
Problem being even at those clocks and the temps below 45C , the GPUs like to run at 1939-1942 MHz. Under full load. Don't understand why. On the desktop the GPUs go full throttle under load.

 

bummer,

will reinstall windows 10 anyways, but still not sure if i should wait till the new update or just reinstall the current LTSB again, it's so weird still didn't found out what disables cpu turbo when starting some games i even uninstalled all intel programs and drivers and deleted the service itself via cmd, re-enabling turbo with TS it stays @ 4.5ghz like it should hmmm...

 

I have a quick question for some of the heavy hitters, @Johnksss@iBUYPOWER , @Mr. Fox , @Prema . Are you guys in an Air conditioned room or do you have an AC duct actually blowing into the laptop cooler?