Clevo Overclocker's Lounge

Sorry, but nothing cool-ish that doesn't look like an Apple? Looking for a new 14" system for family, which I can get 'premafied' and benched before gifting it away. Looking for a mobile hexa-core with 14" and 1050-1060, but Clevo has planned nothing of interest either...

 

Ok then i have crap chip then...

 

that but also theres no LLC/vdroop in bios as option. its basically a combination of things i gotta try to get idle voltage now, which i have now.

@Johnksss @Mr. Fox @Prema i pat myself on the back with new bench score and settings !!! this is with bloated services and 2400mhz 16gb sticks, barely throttling, 51x core 49x ring. with hwinfo turned off.

load voltage is just barely under 1.3v also got rid of that stupidly high constant idle voltage.







edit: so apparently with 51x core and 48x cache i throttling a bit less than 49x ends up with higher score lol.

 

It should be there on Prema mod... I know I have access to it.

 

you're using 870tm or km? care letting me know the name of that option?

 

I'm talking about the drop between the core turbo counts from 1 core to 2 core to 4 core. Used to be 100Mhz each time but now is 200Mhz.

 

they are slowly lowering the standards and make people accept it. someone that buys a computer say 5-6 yrs from now will only understand thats the norm and be okay with it. we are slowly getting killed off. there are already individuals are brainwashed they cant see the truth anymore, such is danishblunt

 

Oh’well. Have you seen this Kaby BGA in use? Don’t follow the line.

 

Intel release so many products now there are always exceptions but I was talking about the mainstream chips you will find in a Clevo chassis, just need to adjust my mental maths going forward

 

To be fair 100Mhz jumps don't make much sense from an engineering point of view, it does not give much leeway to extract performance from the lower core counts.

It's a bit odd from a design perspective to think that if you can do 3.5Ghz on 4 cores then with only 2 active you can only do 3.7Ghz, if anything pushing that up means that you might actually get a decent bump on lower threaded tasks.

 

If there isn't an LLC option, then you need to mess around with the IA AC DC Loadline setting in Intel I/A Domain.
The reference value for this is 1.80 mOhms for Kaby Lake (probably also for Skylake). For 8700K, reference value is 2.10 mOhms.
Pay very close attention to the divider shown in the help for this setting. Usually a value of 180 is 1.80 mOhms, and 0 (Auto) uses the reference value.

If your temps are skyrocketing at load and the VID is fluctuating massively at idle, try reducing the IA AC DC setting to value of 1. This will remove absolutely ALL "VID boosting" but vdroop will be massive (and i do not know if your clevo has a VCORE sensor, if you do not have a vcore sensor, you won't be able to measure true live vcore, as "VID" is simply voltage target before vdroop is applied.)

On desktop 8700K systems, from what i read on overclock.net in the 8700K thread, from people's vcore readings, this IA setting is ignored when using STATIC (manual) voltages (VID will still report higher than vcore with the Auto setting or if its set to 2.10 mOhms, and VID will match vcore at the "1" (0.01 mOhms) setting, when loadline calibration is used to stop vdroop), but vcore and temps will not change vs an "auto" value or a value of "1" (0.01 mOhms), but it's may not be ignored on kaby Lake (you'll have to test for yourself).

I found a value of 5 to 10 works nicely on my MSI laptop, you'll have to test it on your Clevo.

Please, do NOT, under ANY Circumstances whatsoever, go ANYWHERE NEAR the upper limits for this setting!!!!!!!!!!!!

 

i do see some ac/dc options so i'll give those a try. previously i had it set static and my understanding was its suppose to just use what i set, as thats always been with my ivy/sandy laptop. doesnt seem to be the case anymore unsure if its integrated vr on cpu now rather than on mobo, idle fluctuating like crazy with both static and adaptive.

the lowest i could get to is what i have got now, 1.31v idle and 1.21 underload at 4.9ghz and basically i had to lower the idle voltage with both cache and core voltage offset. problem with that is lowest i could do is 1.31 because lowering cache/core offset also affects load VID, if lower anymore offset then load VID will hit below 1.2 and crash. however cstate finally working the way i wanted it to be after doing this setting with adaptive voltage instead of static, this drop actual idle voltage to like .6/.7v when its only 800mhz. previously even with cstate enabled it was still at like 1.35v idle beyond retarded.

 

Damn, did we just jump 28 places into almost the world top 50 with hwbot v7's new way to calculate score?!:

Spoiler

http://hwbot.org/league/teams?offset=41

 

Not bad in national ranking either there

 

That ranking system is nowhere remotely close to being done....

 

@Falkentyne so confirmed ac/dc settings do work to lower idle voltage however it also lowers load voltage, though i dont think its direct ratio 1:1.

i had to up my voltage offset for both 50x and 49x overclock settings in order to stay stable but overall the idle voltage is lowered by quite a bit more so i think the trade was worth it, maybe 1:1.5 or something like that lol. ac/dc set at 186, havent try going lower but should try. though i can confirm this is not good for bench and overclocking cause it raises load voltage (offset) as i can't lower it in order to stay stable.

 

I was more poking fun at there being only 3 entries for the registered country. ^-^

 

LLC and what you're doing makes it more true to what you set. If you have it aggressive and you set 1.1v it'll SIT at 1.1v. Unlike now, where I have it at 1.05v and load voltage is usually around 1.07v-1.09v on average when stressed. So that'd explain why you needed to bump your voltage, as far as I can tell

 

yea, i can probably lower the AC/DC value more and just up the regular VID setting like the old days. like i mentioned earlier this will help towards idle voltage, but while load voltage readings seem to stay the similar if not higher, in reality i think load voltage is going higher. benching at higher frequency 51x is no longer possible throttle like crazy due to too much voltage.

normal use 5ghz and 4.9ghz is perfect though.

 

Don't underestimate Micronesia. Did you check the other teams score:

Spoiler

I would actually deactivate the entire country setting or set it to 'World' if possible in order to make everyone feel welcome. So instead we have chosen a place with a minimum of polarization.

 

@D2 Ultima @Falkentyne @Johnksss @Mr. Fox
did a few runs involving lower idle voltage and i got these.
lower AC/DC lowers idle voltage but also load voltage, have to compensate by upping offset votlage to keep it stable. in all of these runs load voltage keep going up as idle voltage goes down, but power usage overall seems to be the same. but just like overclocking theres optimal settings.

AC 186 50x 47x -50.0mv 129w
DC 186 49x 46x -85.1mv 114w

AC 165 50x 47x -35.2mv 129w
DC 165 49x 46x -75.2mv 114w

AC 140 50x 47x -10.1mv 129w
DC 140 49x 46x -62.5mv 113w

look at 50x overclock getting affected, from 186 to 165 AC/DC value only upped 15mv but going more will need 25mv+ which increase load voltage by quite a bit. so this can only mean that (LLC higher settings with higher AC/DC value) we get higher idle voltage but lower load voltage.

johnksss mentioned 1.3v limit imposed by clevo on 870TM so we can up AC/DC value say to 250 to lower load voltage under 1.3v to keep away from throttling as long as temp allows it, this is only really good for benching but regular usage gotta find the best settings for best frequency.

default setting auto i believe falkentyne was mentioning 210 for 8700k

 

HWBOT is still seriously messed up. Apparently not through recalculating things. Ranking for everyone is totally wrong now (team and individuals). Every time I sign in it is different and varying by 100+ places on the leaderboard, up and down. I have seen my leaderboard ranking swing from #25 to over 200 in the last 48 hours and some with hardly any points at all are near the top of the leaderboard, LOL. The link to the leaderboard from my profile is wrong and if I manually browse the leaderboard I cannot even find myself.

 

The auto reference setting came directly from Raja@Asus. Although he was referring to the desktop chipset boards for the 8700K, it's still an 8700K chipset.

 

what would be considered as high LLC value for 8700k? if default is 210 then would 265 count just normal?

i did a test at 265 and as expected, i could run 51x 49x without throttling at all and bench higher score, i could up LLC say ac/dc value at 285 for 52x bench at under 1.3v but im scared as it'll damage CPU. when i took a look for 265 ac/dc value, idle 1.44v load 1.26v for 51x and idle 1.48v for 52x so im assuming a bit more room i could go but thats already not healthy. assuming limit is 1.5v then i could do a 275 for example and bench 52x.

@Johnksss whats the highest ac/dc value you tested lol need your experience

 

I can't even answer this unless your laptop has a vcore sensor. Does it have a VCORE Sensor or just VID?

if it does NOT have a VCORE SENSOR, then you won't know exactly what Vcore is going to the CPU Because VID does NOT show "vdroop."

And remember that IA AC DC loadline is not "LLC" (loadline calibration).
Loadline calibration works directly on vcore itself.
IA AC DC (according to Raja @ Asus) boosts the VID signal based on current load.

 

are you sure they are unrelated? llc and ia ac/dc. in some way they work similar no?

also heres what i found. i try to find out what voltage needs to be stable so i set AC/DC value to 10. my 4.9ghz now needs VID at 1.285 to be stable to just run CB15. compare to when i had 165, it was 1.22v load and 1.295 idle which i can make it use 0.6v by turning on cstate. honestly going with 1 value of ac/dc loadline doesn't seem to be worth it. ultimately it comes down to every cpu has its optimal setting and each of its frequency.

 

Yes. Raja said that the IA AC DC loadline setting boosts the VID signal based on CPU current load (almost exact quote).

Loadline Calibration (LLC) flattens the voltage vcore droop slope.

On 8700K desktop boards, the IA AC DC loadline setting is supposed to get IGNORED when using MANUAL (static) voltages, with respect to vcore. Example: VID may show up as 1.4v VID on 8700k with 1.3v manual override at 5 ghz, but true vcore will read 1.31v (medium Loadline Calibration used, and IA AC DC Loadline set to "auto"). Then changing IA AC DC Loadline to "1" (lowest value) or 0.01 (depending on Bios), VID then reads 1.32v VID and 1.31v vcore. Temps and power draw identical with both.

This setting is "supposed" to be used for Adaptive voltage boosting.
But on some laptops, it gets applied to static voltages too (probably because Loadline Calibration is not available). But it's not the same.

On the 7700K desktop boards (Asus), it was recommended to change IA AC DC loadline to 0.01, because the auto setting interferes with static voltages and causes v-rise at load (Vcore linked to VID).

 

in this way no real way to know how settings work with one another because dependent on the bios. i set static in bios but ts is adaptive so i THINK TS is override bios. because if static is set in TS, then even if i turn on c state my idle voltage remains high, not like those .6v at 800mhz. basically it keeps the clock that high even on idle with cstate on which is messed up.

but for 870tm, putting ac/dc value to 1 renders laptop unable to use at high clock i end up needing more voltage than i had ac/dc value to somewhere in the 160s, that is a good balance but i guess since each chip is different one would have to test and find what clock they wish to run at, at what temp and what voltage.

 

TS will override because it adjusts the MSR Registers in windows, rather than the BIOS. It's why it conflicts with XTU because XTU adjusts the BIOS in real-time (and can make changes the BIOS doesn't allow/accept, even). It's like adjusting changing signals from two different points, one will generally override or conflict with the other at some point.

 

I really do not understand how it is you can throttle. Especially if that part has been removed from the equation entirely and all you are over clocking is the cpu.
I don't get it.

And I'm also not getting involved in the ac/dc line debate either. We all seem to see that in a very different respect.

3dmark11 5.2Ghz - 3200 Mhz - 2100/1401
https://www.3dmark.com/3dm11/12616376

 

^ Oh my, I need sleep, for a minute I thought it said "Gigabyte" on the motherboard page of CPU-Z. I must be absolutely insane.

 

Am in Taiwan right now.

Home of the crooked heatsink design and bad bios design choices.

 

@ole!!!

We did all testing with the PremaMod and BIOS only, no XTU, no TS, no CCC, in order to ensure there is no throttle left whatsoever...using to many things at once just messes things up.

 

i have no doubt theres prob something else im missing but when im hitting that 150w mark, theres also temp gotta worry about as well. but i never seem to be able to hit like over 155w for w/e reason, though regular usage i will never be able to hit that high.

 

See if you and @bloodhawk can find them maglevs for the P870xMx series

And send me them

and say merry christmas

a really late merry christmas

but do it

 

Second of topic. Sorry brothers. But My 4 nice * computer* friends is now camping in my garden Both yesterday and today's night. Nothing goes unaffected when you have a good watchdog. He will let me know if something happens outside the door and I have to take a look what happens. No ****y thieves this time but still fun




These here are a lot more courageous than thieves who just run like the devil when they see Ivan *the* Terrible come out

 

so that's where you live and thus you can have crazy overclocks! that's not fair I live in a desert!

 

I'm sure you can manage better temp with the correct tool No exuces, bruh. Run and buy
Works out to about $75 per year. I'm sure J00 have the money

 

This is really so nice, as @ Phoenix said, we are living in the desert, what do you think we will feel after seeing this , hope that I can live in such beautiful place someday.

 

That's awesome brother Papu!

 

The biggest we get are some urban foxes.

 

I LOVE THOSE CREATURES

CAN I VISIT AND PLAY WITH THEM PLEASE

AND OVERCLOCK TOO

Sent from my OnePlus 1 using a coconut

 

Professional overclocker finds a way to cheat benchmarks like 3DMark TimeSpy


Pro overclocker Ronaldo Buassali, working together with GALAX published a video in which he shows a way to apply altered driver settings shortly prior to the actual benchmark run. This results in false higher end scores. The issue is not isolated to the 3DMark suite.

Pretty much most benchmark software used in competitions, like the hwbot, can be cheated on.

How fun is it play with your hardware and know by yourself that the performance ain't correct? What a waste of have it fun in your free time.

 

Im surprised it took people this long to notice this trick. Also there is no need to do that timer thing.

 

As long you know what scores you should get from different fixed clocks you can see if they cheat. Still a waste. The problem is money involved. And do this against yourself is stupid. Benching is meant to be fun.

 

Yeap i think tested this a few months back for the lulz with a GPU score of 40k+ or something, ill dig out the screenshots once im home from work tonight.
But i didnt know that 3DMark was literally oblivious to this kind of thing. It just makes one doubt so many top scores over at HWbot.

 

I wonder how screwed up next 3DM version will be. I hope they test it properly and that we don't have to live with broken bench software +4 months down the road as in 2016

 

Is this the same timer BS people did 18 years ago, occasionally with Soft FSB, but wtih some old windows 95 timer editor, to make games like Quake, old CS (won.net) 0.4 beta and Diablo 1 run in super turbo speed?

 

btw did a few more tests
ac/dc at 1 vs ac/dc at 0 (auto, i think you said 210 for 8700k?) both 49x
at 210, load is 1.22v during idle with cstate it'll peak to 1.36v at times
at 1, load is 1.29v idle is the same as load voltage

strangely with ac/dc valued at 1, it runs noticeably cooler even though load voltage is much higher, im guessing after load the spike in voltage causing spike to higher temps. after running dota2 for like 2 hrs on each settings peak temp is like 79c vs 72c.

with 210 ,load of 1.22v will only show when its stressed hard enough like 95-100% usage on core. seems harder the usage lower the load voltage, and if light workload voltage remains higher which could also be the cause of overall higher temp, still unsure at this point.