Clevo Overclocker's Lounge

I wonder if the IMC needs a slight voltage bump to deal with the dual memory rows.

 

I had no idea that it's bad to keep a higher clock on two cores... It came like this from the factory and I thought this is how it is supposed to work......

I'll see if I can 4.3 on all cores, and if not, I'll try 4.2

I guess that sometimes the mistakes we make are the most silly ones

 

From the factory its 42/40/40/40 so 200mhz higher on a single core only.

 

ah no, its totally fine, i was actually thinking that u might have some important programs / tools that only support up to two cpu cores

when tweaking hardware its always best to keep the number of variables as small as possible, so that you know which component gives out on you if a setting is unstable. limiting all four cores to the same multi decreases the voltage fluctuations and thus the offset required to get a multi stable plus, it can potentially decrease your peak temps

ive tried that already with +200 mV on the system agent, but no dice on getting 3000 Mhz stable. cant even get the machine to properly boot at the XMP setting, i have to loosen up the timings to even get to 3000 Mhz...

 

I hope I'm a lucky outsider... Did the TM5 tests and got no errors, ran it in windows XP SP3 compatibility mode with administrator priveledges.... I'll have to run a long run of memtest86 and see how it fairs overnight

Spoiler

[

 

hm, strange though that your test only lasted 3 minutes, for 32GB it usually takes at least 10...

 

Immersive VR:

https://imgur.com/a/XNbOr#5i13KQL

 

Definitely seems fast compared to the others... Ran it again and it was a few seconds faster though

Spoiler

 

@Mr. Fox @bloodhawk @Papusan @Johnksss@iBUYPOWER

Guys, Phoenix logged in to my taptop and overclocked my CPU using ThrottleStop (which was already pre-setup as per Mr. Fox's recommendations) and had the power limits maxed out and properly adjusted the voltage, but be it a 4.2 GHz or 4.6 GHz overclock, the CPU throttles to 3.9XX GHz the moment the stress test starts.

I was under the impression that a single PSU would be enough if I only have a single 1080 GTX but apparently not. Phoenix told me that even if he disables SLI, his system would throttle if he uses a single power adapter. So shall I get another PSU or do you recommend anything else to change in the settings?

Note that I didn't even install the new CCC which contains the Intel XTU virus, I have 5.xxxx.50 from the get go and overclocking was only done using TS.

 

did you check the throttling section of HWInfo? that might give you a hint as to what kind of throttling were talking about here.

 

h3r3 j00 g0:

 

interesting. so as i understand correctly, the moment that @Phoenix switches from one PSU to two, the throttling stops, no matter if hes got SLI enabled or disabled?

 

my system is not usable with one adapter under any circumstance. if I use one adapter, the power adapter light will switch off and I would have to unplug it / replug it for it to briefly come on then do the same thing again. This happens the moment there is any load on the system

 

hm, that doesnt sound like tripping though, more like an automatic switch off....

 

So, I've been playing with my desktop and figuring out my M8E. Turns out my board is limited to 3733MHz (haven't found stable above that and XA at hwbot evidently had to bin MBs to find an M8E above 4000).
My Intel 600P and 6x Noctua iPPC 3000 rpm fans and controller arrive tomorrow. I've got 4.8GHz P95 stable all day (should be able to get 4.9 after the fans, using the aio thermaltake water 3.0 ultimate). Even though a desktop, after I get it fully setup, I'll post some benches for a desktop comparison (still trying for a clock for clock win on cinebench against @Papusan). My 980 Ti Classified ASIC is 70.6%.

@Johnksss@iBUYPOWER - I may be PM'ing you for advice after I play with it some more...

Sent from my SM-G900P using Tapatalk

 

What are the desktops expected specs? Update your signature with them!

 

I'm waiting to get my 24/7 overclock setup for final specs put in my signature. I've done [email protected] CL14 17-17-280 CR2 32GB (4x8GB), 4.8/4.8GHz@74-71-71-72 degrees per core pumping 1.42V for P95 stable (wish that was a bit lower, but...). I probably will lower the ram voltage for my 24/7, but I'm playing with the settings to learn/have fun at the moment...

Sent from my SM-G900P using Tapatalk

 

Well, after a day of tinkering with BIOS/EC flashing/setting and Windows reinstalling, I think I made my system stable as much as it can be until Prema finishes his magic BIOS.

A message to P775DM3 owners: The latest main EC for our system is 1.05.03, but I personally don't recommend it. There is a bug in EC 1.05.02 which Clevo "fixed", however I can not consider that fix a real one. The issue with EC 1.05.02 is that some panels (like the FHD panel I have) may show purple/pink lines flicker at Windows login during system boot. It is not that annoying since it does not always happen and even if it does, the lines only flicker once, so they can be tolerated. Now Clevo "fixed" this with a funny method...they did not fix the problem itself at its source, instead they implemented a timer in the latest EC which turns of the screen for 2 seconds after xy seconds of power on, right around the time where the flickering would happen. I see this more annoying than the lines themselves and will happen even if I boot into pure dos environment, so it is not connected to Windows. Also, hiding the problem is not a way to fix it in my opinion. So I for myself flashed back EC 1.05.02.

As master Prema confirmed all current Clevo laptops will throttle to 4.0 Ghz (6700K) because of the factory BIOS. They will hold the increased multiplier when only CPU is stressed, however when the VGA starts to draw higher power, the CPU will throttle back to the default x40.
However with the default power settings when the VGA was used my laptop did not even hold 4.0 Ghz all the time. This problem was easily triggered when I set "FAST" as vertical sync in the NVCP and run Fire Strike test.
A test I have made (I don't remember the BIOS version, it was either 1.05.01 or 1.05.02):

As you can see the final points were are good, no sign of any issue, however checking the monitoring results, you can see that the CPU frequency throttled during graphics test, but was okay during CPU test.

I believe this is unacceptable, so I decided that I will not give up until I finally stabilize it so it wont go under 4.0 Ghz.
My goals also were to somehow undervolt the CPU and increase the default fan profile without any external program and in a way that the settings would remain even after Windows reinstall.

#1 lock 4.0 Ghz so it would not throttle lower
First I thought that the XTU app (CPU OC utility) in the Clevo software caused some bug which resulted in a throttle. I have seen such an issue with the P775DM1-G where if I changed any setting under Windows with the XTU app, the CPU would throttle immediately to either 800 Mhz or 3.0 Ghz when stressed.
First I have tried uninstalling Clevo app and doing a factory settings restore in the BIOS (this usually solved the similar issue with the DM1) -> still throttling
Okay, I then tried reflashing either the same or an older BIOS (I have tried 1.05.01, 1.05.02 and 1.05.03) and reinstalled Windows. On a fresh install I have tried the same test with the mere basic drivers and no Clevo app ->same result, throttle under 4.0 Ghz (which as I said, mostly was triggered when I set FAST as vertical sync in the NVCP, in other words I let the VGA to run wild)
Okay, I have tried 369.56, 373.06 and 375.95 NVidia drivers too (used DDU to remove older one) -> throttle
I first tried this with EC 1.05.03, however same result when I downgraded to EC 1.05.02.

So here is what did it for me:
- first I have installed the latest BIOS from PC Specialist, which is based on the latest Clevo BIOS 1.05.06 (PCS BIOS has a version number of 1.05.06TPCS)
- I have reflashed the main EC (1.05.02) and 2nd EC (1.05.02)
then

#2 Undervolt CPU
I wanted to keep my cpu undervolted without any additional programs, however in the BIOS I can only set positive value for Adaptive and can not set negatíve.
The CPU OC app in the Clevo software has the ability to change BIOS settings when you reboot, so what I did was to install version .81 from the Sager website and set -170 mV undervolt in the CPU OC app (not changed anything else!).
However before reboot I also

#3 changing fan profile
wanted a better fan profile for the laptop. As to my experience there are 3 fan profiles stored in the EC:
- maximum: which is loud even when I browse the internet or watch a movie
- automatic: which is quiet when I am not gaming, however will not turn on maximum fans during CPU or GPU intensive processes so major components may overheat
- overclock: the perfect gray spot for me. Quiet while watching movies, but turns up the fans when CPU/GPU intensive apps are run so no overheating

Unfortunately with the FN+1 button I can only switch between maximum and automatic. Overclock profile can only be accessed with the Clevo app. So what I did was after undervolting I set the fan profile in the main Clevo app to "Overclock".

After this I rebooted the machine. Now the undervolting has been stored in the BIOS (despite the BIOS not stating it as a negative value) and the selected fan profile in the EC has been set to "Overclock".

After reboot I have uninstalled the Clevo app.

After uninstalling I rebooted and entered the BIOS where I set multiplier for all cores to x40 and also changed the default 95W max power limit 1 to 100W (may be pointless since even in the CPU killing test which I have showed Phoenix the CPU did not draw more than 85W).

Now to make sure that all settings have been stored in the BIOS and EC and will remain even after format, I again reinstalled Windows and can confirm that all settings which I wanted remained even after an SSD wipe.

Only cons I see so far is that when I want to change the undervolt values or if I accidentally press FN+1 and select another fan profile, I have to reinstall the Clevo app to change settings. However until I dont want to change these, they will remain the same.

Now, here is Fire Strike with the new settings (also set to FAST in NVCP) and BIOS:




The 4.0 Ghz is finally stable! I have not tried Overclocking, since it would probably not hold and downclock to 4.0 Ghz.
Also found that the calculating speed of the CPU has also increased. With older BIOS when I have undervolted with the Clevo app I have got 195 Gflops @ 4.0 Ghz, now with the same amount of undervolt and same temps, I get 200 Gflops @ 4.0 Ghz. Nothing I can see during gaming, but still an increase under the hood.
I don't know if the cache is also undervolted this way or not, however the temps are the same as when Core+Cache was undervolted with Throttlestop.

Another interesting thing is the RAM. I have 2x16 Gb 3000 Mhz Corsair RAMs which when used with XMP 1 produced the same problem as the above mentioned G.Skill RAMs. Even when selecting XMP 1 in the BIOS, the laptop would not boot and only show black screen and I had to remove AC+battery for 10 min in order for the laptop to skip the error check in the bacgkround and allow the system to boot.
Now with BIOS 1.05.06TPCS I have set XMP 1 several times and each time the laptop booted without the previously experienced issue.
Doing a Memtest86 right now under DOS environment, only 3 hours have passed so far, but no error yet.



Will post the final result tonight.

On a side note I am ready for another round of repaste and this time I will be changing thermal pads too!



Too bad my Rockit88 delid tool is still MIA...USPS tracking says it has arrived in Budapest, but local post office says it has not entered the country yet :/

 

I can run a single 1080 with 1 PSU an no issues , when you PSU quits providing power it will be around 440 watts from the wall and it will be tripped by a breaker and it will shut down completely. In HWINFO I would look at the IA: limits to verify the EC isn't getting into the way which I would suspect it is.

 

The older .50 version have the OC fan profiles for DM1

You use high performance in Win power plan ? No home made crippled power plan? Have you tried put 1023 for pp0. And you don't need to put clamp on power settings.

 

d00d, the maximum allowed value in the BIOS is usually 300A so when I tried to put PP0 to 1023, the moment I save, it set it back to 300 again.

 

makes me suspect even more that the fault lies somewhere in the BIOS concerning the Corsair DDR4-3000 16GB sticks...

btw, im guessing your rockit 88 is stuck in customs, thatll take a while before they release it to your local post office

 

It is pushing the bus, maybe due to the topology of the layout.

 

Well my friend...For starters.....I would actually try raising the power limit. 60W for 4.6 ghz is no where near correct im afraid. It should be more like 80W to 85W.

 

I thought I was the only one getting that pink lines / pink screen flash thingy...

 

where to do that in TS? I can't seem to find it brah

 

Yep that's a bug in P7xxDM2/3 EC v2, you need to use v3 to fix it...

 

So since I didn't find any way to increase the power limit in TS from 60 to 85, I installed CCC 5.xxxx.90 and maxed out the power limits and VR Current limits, clocks hold stable @ 4.6 GHz

the moment I launch AIDA64 stress test they all go down to 3.998 GHz

what the heck is going on?

@Prema kindly check my PM to you. thx

 

@Prema

This is my BIOS/EC version:

 

Sorry to ponder this question but will it be solved, or the display is just shut down when the problem would occur, and also, where do we get new EC? They don't seem to be on Clevo site nor on Eurocom site.

 

Usually from your reseller.

 

I thought EC v3 is here = https://repo.palkeo.com/clevo-mirror/P7xxDM23G/

 

My reseller has not informed me about providing those things and at the time it did not come to my mind to ask about it...............

Thanks! I never heard of the site but I'll look into it.

Sorry to pester, but I've no idea how to do this

2. Run ECFlash.bat under pure DOS to start EC flash procedure,

It doesn't want me to just double click it, but something more complex?...

 

@Prema @Mr. Fox @fortune515 @bloodhawk @Papusan @Johnksss@iBUYPOWER @DRevan @GTVEVO @EmberV

As you all know, I've been complaining that my 4.6GHz or any overclock for that matter would start throttling down to 4.0 GHz the same instant the GPU is involved. All power limits, current limits, Turbo Power max are maxed out in the BIOS. ThrottleStop didn't help. Some of you said it's an EC problem.....

I just ran a Fire Strike test and the clocks helped stable not even a brief second downclock like with the previous Eurocom Sky X9 I had.

the only thing I changed today is:

1) I set XMP Profile from 1 to 2 as 1 was giving me random crashes and eventually Windows stopped booting altogether.

2) I uninstalled the old Control Center AP v5.0001.0.50 which I preferred as it doesn't have XTU bundled and installed the latest Control Center AP v5.0001.0.90 but with the OEM switch to NOT have the XTU crap installed

WOW!! I am super impressed now!

Results (still can't get my head around why my 3DMark score is so low despite me setting power management to high performance and texture filtering to high performance but I'm glad the CPU clocks are sticking

 

Yes, this is excellent info. But, it only worked after tweaking with manual settings. Ideally, this would be plug-and-play without user intervention... and, hopefully it will soon be.

Interesting that Corsair 16GB 3000 sticks has same or similar problems.

Update: Brother @Prema is all over this, like white on rice. The problem could soon be in our rear view mirror with respect to the 16GB 3000 sticks. I am now over 1 hour with Memtest86 v4.3.7 with zero errors using a test BIOS on XMP1, so it is promising. Before it would have errored out within 10 minutes. Be encouraged... one more thing to add to your list of things to be thankful for on this Thanksgiving Holiday.

 

Wow! this would be like Prema gifted me a new set of RAM sticks! Thank you so much for diagnosing this issue. Without you, I wouldn't have even known that I am running on a higher timing!

 

Thank you. It is a combined effort. The real heroes are @Prema and @Johnksss@iBUYPOWER. John because he bought a pair of 16GB sticks to help with testing, finding a temporary workaround that helps, and @Prema for finding permanent solutions.

All testing passed with Memtest86. Zero errors on XMP1.

 

I am running it once more just to be sure it did not pass by accident. This time I am running it in UEFI mode. Memtest86 Pro (purchased) is dual-boot with version 7 for UEFI mode that includes a couple of additional tools and a memory benchmark. The DOS version is 4.3.7.

 

please run a Fire Strike test also. Remember when my TM5 tests passed, I would crash/reboot the moment Fire Strike was run both on overclocked and stock CPU settings.

 

Factory Bios / Factory EC 1.05.06 / Nvram Wipe / No Thaiphoon Burner / No changes what so ever in bios. / Settings are from TS
And this is what I get.

Edit: First picture is wrong, but working on Premamod V3

 

TM5 says, of errors is not detected!

love the English

 

Ok, got the result for Corsair rams with the latest BIOS, took more than 10 hours lol





Will run the Windows mem test too just to make extra sure.

@ Johnksss@iBUYPOWER : Running memory testing under Windows is not really reliable ...

 

So Prema Mod is now the factory BIOS, I like that factory...

 

omg, i would LOVE to have a similar solution for my machine as well i really hope its the same issue here

 

Sorry about that.
That one was showing that it could work under the premamod bios.

 

Well, if everything in the world was as it should be, it would be.

It took me a bit to realize you meant factory settings, not firmware.

So, the good news it, or as it appears to be at this time, there is nothing wrong with the 3000MHz G.SKILL or Corsair 16GB sticks as long as the firmware on the machine isn't goofed up.

 

Had the wrong picture up first.

 

Fixed my earlier post.

 

Same for Corsair rams with P775DM3-G. But it seems the latest factory BIOS fixed that (will do some more testing to confirm)

 

Yup, it certainly seems so. Now we need the @Prema magic for it to be all it can be... i.e. the retarded EC throttling fixed and all of the other configuration settings accessible so we can actually "own" our machines and decide how we want things to work (full control of UEFI/CSM, SATA options, ACPI, full TB control, etc., etc.)... all of the stuff that let's the owner actually be the owner, not control freak idiots calling all of the shots and reducing functionality for everyone.