Clevo control center 2.0 is the one causing throttling.

-Notebookcheck

The issue of the P955ER throttling was caused by the Clevo control center after all.

Examples for the throttling :

A 1060 throttling in a Clevo laptop(1190mhz for 1060)



A 1070MQ throttling (923mhz for 1070MQ)

It wasn't the 150w PSU or the bios.

@XMG , @Danishblunt

 

I want to see if I can fix that using it CCC, but... excuse me while I lel and repeat that people should not use CCC and instead settle for having to FN + 1.

 

Right, CCC v2 will cause an additional GPU throttle at lower temperature.
But not sure if it'll also do that when set to performance mode...
(The combined load and power throttle are in the firmware).

 

Yeah.

So this isn't a bug ? It's intentional ?

 

Pretty sure it has to do with Max-Q certification temps. Easier for them to do that on OS level rather than requesting a new vBIOS...

 

Yes but the normal 1060 is throttling to 1050 ti performance too despite running at 65c. And once you kill the CCC 2.0, bam you're back to 1060 performance.

The same thing happens with the 1070MQ . Kill the CCC and you will go from 923mhz to ~1300mhz.

 

Whoa.

 

By the way,

Can one install the older Clevo Control center, before v2?

I wonder if that would work on this generation of systems.

 

Here's what I'm going to do: I'll uninstall the control center and re-run the benchmarks.

Sure, I may not be affected (based on previous results) but at this rate, it's good to confirm.

 

Seriously, Clevo: Why do you suck at making decent S/W for your machines?

CCC has been a cancer for ages. Fix it, or nuke the **** out of it.

 

@ahmad hendeh one day I will perhaps be able to explain everything about this topic. There are so many factors which 1/ we know are intefering and 2/ could be intefering depending on the scenario. There are many things I can't post in public because they are private discussions with Clevo and internal to our company - even if I post a prooven solution, you will still see differing results from other resellers because they are not having the same discussions that we are on this topic.

Don't know if this ^^^^^ actually makes sense, but in essence if you take what Prema says below, add in the vBIOS topic, add in a number of discussion points which directly affect load clock speed results, DPC issues, multiply that by 1000 - you get somewhere close to what we (XMG) is working on at the moment.

And here's exactly one fo the points. Running stress tests in AUTO fan profile whilst in Entertainment mode does not work out well. That's why Performance mode exists. Similarly (I will post a link to this in the P950ER thread) running stress tests in MAX-Q profile mode does not go well. It would be a little like:

- setting a phone to maximum battery life and having poor performance results
- me putting my car into sport mode and complaining that it revs too much and doesn't change gear until the revs are higher than I want

The problem with the laptop is that (in my personal opinion) they are expected to perform at their maximum level no matter how the control panel is configured. The MAX-Q setting is exactly that and we all now that this limits the fan speeds to under the Nvidia noise level for MAX-Q certification. It also has other effects which I won't go into here, but you can see the results in the OP. That's a whole different topic as to if people think MAX-Q should even exist or wether you should performance test at that setting and not be surprised with thr results when they are very low.

To clarify the PSU topic, I should point out that the default option is in fact the 150W on both the 1060 and 1070 MAX-Q P9%*ER. There is a 180W available, but the 150W does not limit performance. As I posted several weeks ago, max power draw that we measured was 157W. This was at stock settings and max system load. 157W is fine for the 150W "rated" PSU. If people want a 180W PSU so that they can overclock, I can only advise that there are better chassis available to OC which are not physically designed to be slim and under a specific dB level.

I can say things like "the test results in that independant review do not match our internal results" and it will me shot down violently by some members - I understand why because I am in no position whatsoever to 1/ post screenshots which directly conflict with the review (this can ruin relationships with review sites and more importantly force forum members to choose sides between me, a company rep, and an independant reviewer) 2/ as above, every scenario is different, settings in CCC DO MATTER and affect performance 3/ there are other things going on in the background which we have actively been working to improve for a long time now - for example some of you will know that some drivers and firmware limits power draw when certain software is detected. There's no real surprise to see low CPU and GPU clock speeds in Prime/Furmark, purposeful performance throttling is rather difficult to differentiate from what can look like (no offence, but to the untrained eye sometimes) as inadequate physical design and incompetent R&D/testing.

Please post your findings here or feel free to pm me anything you don't want to put on the forum. To be honest we're a lot further down the road on this topic, but as you can imagine I am limited to what I can post in public until we have reached solution.

 

@XMG It's Saturday and time to go off-line and ENJOY! Don't let the topic spoil another football match!

 

It's acctualy quite less complicated.

Open task manger
Kill CCC 2.0

you will go from this



To this




Thankfully the solution is a 100 times simpler than I expected. So no need to worry about it any more. I thank you for being the only reseller here who cared about the issue.

Enjoy the WC and don't stress it

 

@ahmad hendeh he he - yes indeed, thanks The difficult part (I was refering to) is fixing CCC functionality and performance. But if you're fine with just throwing CCC out the window - which I don't object to in any way whatsoever ;-) - then go for it!

Happy weekend to everyone too - weather is rubbish so my bike has to stay indoors, going to join the collective walking dead in Ikea for reasons I definitely can't talk about in public

England v Tunisia tomorrow is what I have most vested in, win that and we'll be guaranteed second in the group so will go through no matter if Belgium bring their A game in the third match or not. Will see if Germany remember how to play football against Sweden, could be tough to watch (English guy living in Germany!!).

 

Or better, uninstall CCC and install this: http://forum.notebookreview.com/threads/rlecviewer-custom-clevo-fan-control.818858/

 

Nai wa. Doesn't exist. Not from Clevo, and certainly not from anyone else in that price bracket, in the 15" category. I could make a big list of the competition but realistically... we need a P6xxHx unit remake. PA71 and P955 don't count, and is all we can buy from them.

 

Yeah, but I'm not saying any of that am I ;-) Don't buy a slim 19mm laptop with limited TDP components and targeting a certain dB level if you want to overclock - the headroom is reduced compared to a chassis designed outside these parameters.

 

I definitely agree, I just don't think there is much choice in non-bulky 15" models. The CPUs are locked anyway, that's out of the question. The GPUs are power limited anyway, we can't raise them... we're only getting 90W spikes with an average of 80W drawn from that GPU no matter what one does really. The only real difference is heat management and the lack of artificial throttle. Raising CPU power limits, OK. You got me there, but that's not really overclocking... more Intel is stupid with every single 8th gen run

Sent from my OnePlus 1 using a coconut

 

Was interested in upgrading to Control Center 2.0, buuuut maybe now I won't lol.

 

Raising power limits is overclocking, it's running the chip faster than spec. Just a different lever you are pulling.

Its a bit like saying you did not mod your engine ecu, you just removed the rev limiter.

 

I literally just admitted he got me there in that... but it also is to be argued that 8th gen CANNOT function at 45W. There is nowhere near enough power, and many notebooks actually come with raised limits out of the box (surprisingly so!) as well.

 

It can function at 45w just fine. Its just going to depend on how much you push it like the 15w chips.

 

So if you do kill the control center, do you have to control the fan speeds manually?

 

Function, yes. Provide its rated speeds, no.

I won't slaughter an OEM with words for leaving a CPU at intel spec for power limits, but... that chip is a LIE at the intel set power limit. Both the -H and -U lines.

At least the i5-8300H is quite a good chip, being a slightly overclocked 7700HQ. Now if more places would actually offer it... the P955ER would likely have a fantastic time cooling if that SKU was available from more places.

 

It's not a lie, it's people not understanding how it works at the end of the day.

 

It's intel BS way of screwing the manufacturers and customers. They only say 45 TDP at base speeds. So they only gurantee the base clockspeeds at 45W, so in a nutshell you can expect from any notebook that has a 8750H the base frequency (2.2ghz) if you do something very CPU intensive. Anything else is considered a bonus, the MSI GT models with a 8750H have unlocked the 45W limit to 200W, hence those things run permanently at 3.9ghz on all cores even on full load.

Basicially this is marketing BS again, if you buy a thin and light 8750H then that doesn't remotely mean you have the same performance as a real 8750H notebook such as the GT 73, GT 75, Chimera or Acer Predator. It's just another way of giving the customers a middle finger like BS naming of some I7 U processors which are not remotely as powerful aas regular I7 processors.

 

GT73 doesn't come with 8th gen, and the Chimera and Predator throttle all the same.

Chimera:


Helios 500:

 

Oh dang they didnt make the 8750H models yet? FeelsBadMan.

Well I9 is a different story, the reason for those throttling is more likely the temp rather than the power limit.


Here is the GT 75 holding the clockspeed while playing AC origins on ~75% usage

 

Either the power limit is raised by default on that GT75 (most likely) or AC: Origins doesn't exceed the 45W power limit, letting the 8750H stay pegged at 3.9GHz there.

The GT75 can keep the i9 at full boost as well in CB, but obviously power limit is raised:

 

Origins does go far beyond the 45W power limit. and yes, power limit of MSI GT notebooks is typically raised to 200W. Even on my 6700HQ model for some reason.

 

Yeah so that's a special case. Other notebooks have stock PL1, most with no ability to raise it, so they power throttle after one run of CB and scores drop.

 

So that officially unofficially confirms the issue i had after i powermodded my 1070 too, before i set performance mode in CCC i had relatively low performance when comparing to @Kittys 1070 hmmm. @D2 Ultima i everytime thought it's typically CCC in a nutshell lol

 

Wow. I wondered why I was seeing so many complaints of throttling on this model and never had the problem with my own unit... it doesn't throttle if you go to the "basic settings" page in CCC and set "Performance" mode, which is of course the very first thing I did. This is exactly what I would have expected from experience with other laptops of various OEMs that have a built-in power profile system besides the one in Windows. It would be pretty sad if professional reviewers are missing this.

 

According to Notebookcheck. It keeps throttling even in performance mode. The only way to stop it was out right killing the CCC in the tadk manager.

 

Wrong, it always throttles on load. The question is how hard it throttles. You'll never see stable 3.9ghz on all core under load and you'll never see stable GPU clock on load either.

 

OK, technically correct in the sense that there are limits that will be hit regardless of anything that CCC is doing.

You sure about that? I see it every game of Overwatch...



This is after letting it get up to temp for about 20 minutes. Those clock speeds are rock solid, except that the GPU speed will go down into the 1500s sometimes if I'm standing still and staring at a wall (i.e. no-load throttling). I'm even in a poorly air conditioned room without a cooling pad right now. With a slight overclock, and with CCC running. Without the overclock it just sits at a slightly lower GPU clock and temperature as you'd expect. It is also able to maintain full clock speed through almost the entirety of Time Spy, with only some CPU throttling towards the end of the CPU test as it approaches 90c.

I suppose in theory a game could get the CPU to throttle due to temp limit but I haven't found one that does (and Overwatch beats up the CPU pretty hard in my experience).

 

yes I am very sure about that. As you can see the CPU usage is extremely low.

Go ahead, make a small youtube clip of your stable CPU clockspeeds instead of a screenshot where they reach 3.9ghz for a very short time. I'm willing to bet that isn't going to happenn.

 

I mean, honestly, no it's not going to happen because I can't be bothered, and if you want to believe that I have some nefarious reason to lie about this oddly specific thing, that's fine. I didn't think I was going to change your mind anyway because your post history on this subject shows that you've already reached your own firm conclusions. But there is my refutation, for anyone else who's interested.

FWIW, it's the same in synthetic benchmarks that generate 100% load and push the CPU up to 45-54W indicated TDP, with the exception (that I already mentioned) that certain benchmarks like Time Spy CPU stress test step and repeated Cinebenching can get it to hit 90c and throttle slightly -- which to me is perfectly satisfactory cooling performance and does not represent any kind of unusual throttling. Anyway, I thought we were defining "under load" as normal load that would be expected during playing a game, not a synthetic power-maximizing benchmark that is by definition going to drive the CPU to throttle at some point.

Anyway, this topic is about whether there is throttling due to the CCC, and my point was that there really doesn't appear to be unless you have it set to something other than Maximum fan/Performance power.

 

As already proven by everyone else, the CPU doesn't run stable, your refusal basicially confirms your lie. I don't know why you're lying nor do I care really, I just wanted to make sure people who don't know about this topic don't get misguided by people like you.

Nobody cares about your made up story, so please refrain from posting lies, thanks. We already have a user named @Dennismungai here who tested the unit with no end and we know how it runs. I won't allow you to ruin his work by spreading lies.

https://www.youtube.com/channel/UC3QxpHkvvqkYAvRctGa3tLA
Several games, all throttling differently depending on how demanding they were, but all where throttling and stable clockspeeds isn't happening ever.

 

What happens when you unlimit the FPS?

 

He just made a screenshot when the CPU ran at low usage. He basicially tells us his 8750H runs at full turbo with 35Watts

 

CPU and GPU usage are both low in his screenshot because FPS is capped.

 

Good catch, I did not realize I had left that on when I was playing around with stuff. In this case the GPU is seen to get down to 1392 at some points as shown below in the worst screenshot I could capture -- however still achieving a higher and fairly stable framerate. Importantly, however, this does not vary whether CCC is running or is killed so long as CCC is set to Performance mode, which was the point. I'm still pretty sure that's all there is to it.



CPU usage is still not 100% in Overwatch -- I'm seeing it max out around 52% and 40W typically -- but FWIW I did see that TDP reading get as high as 44W and at that point I think one core did flicker down to a low 3.x GHz speed. Again, though, the point is that that's another kind of throttling that doesn't have anything to do with whether the HkeyTray.exe is running.

When doing Cinebench I generally see it temperature throttle down to 3880 all cores later in the test -- again absolutely no difference whether CCC is running and set to Performance or killed.

edit: Here is pretty much where it wants to sit while running Cinebench. Again, CCC is running and this is in a hot room with a preheated laptop where I am not even matching my best Cinebench score by a good couple dozen points. I can get it to throttle worse if I basically repeatedly hammer Run in Cinebench with the laptop sitting on a surface that impinges the airflow. Sorry for not getting the temperature in the screenshot but I think we can assume it is bouncing off of 90-95c like we would typically expect under PROCHOT throttling conditions.

 

Note the CCC bug is only happening on certain models with specific BIOS version. The Q6 from Dennis did not get gimped further than EC with CCC.
Using cinebench to prove your "CPU" is running stable is useless, since we all know that notebooks "cheat" on the PL1 state, thats why notebookcheck always run cinebench in loop and record multiple tests because the first test always score much higher than the rest.

As you finally realize yourself. the throttling which is not caused by the CCC is caused by your EC firmware to throttle down the CPU to 35W, otherwise you'd reach thermal throttle. And this is why the system is not stable.

In a nutshell, running a game like Assassins creed Origins is a trash experience imo.

 

I'm sure you'll tell me that this was only achieved for a split second before the throttling kicked in and why don't I post a video and etc. and that's fine, nobody is here to try to change your mind. But we are getting further and further from the point which is that having the CCC running in the background doesn't induce any additional throttling by itself. You seem to want to discuss the more general issue of laptops being typically thermally limited by nature and your concern that this laptop in particular might be somehow more limited than competitors, which wasn't my point here and IMO doesn't seem to be borne out by any of the tests I can do anyway.

 

I believe the PL1 "cheat" you are referring to is the Intel turbo power limit behavior around "package power long" and "package power short" where the processor is able to achieve a higher power state for some limited amount of time before shifting down into a lower power state. The Long limit on this laptop is 70W for 28 seconds. This is exactly borne out in benchmarks where it will achieve the 63W state for about 30 seconds before going to the 45W state. This isn't unusual, different vendors set different limits based on their cooling design goals and I'm pretty sure these limits are typical for the 8750H.

There is never any 35W wall like you're talking about. In fact the post you quoted shows that it is hitting the thermal limit and not some other lower limit.

By running Cinebench 3 times in quick succession I am able to get the last test to end up stuck at 45W the whole time -- yet it achieved within 3 points of the previous 2 runs. This highlights the important fact that the CPU power state isn't directly proportional to the performance achieved on a full-load workload, since different workloads engage different working subunits of the processor. Nor is either necessarily proportional to operating frequency.

 

Exactly right.

FYI that poster didn't seem to understand turbo boost months ago when flaming Clevo resellers over "lying" about coffee lake power consumption when he just got caught up in the confusion of the multi core enhancement on by default thing (apology to the wrongly flamed was not forthcoming)

A feature present in Intel CPUs for a decade, specified and described in public white papers, is a "cheat"... K.

 

Yeah. Like I said, I'm providing this information for the benefit of those who are able to make use of it, but I saw danishblunt's posts in this and other threads and I'm not going to spend a whole ton of extra effort trying to keep jumping through moving goalposts for someone who has already made up their mind what the conclusion should be. My attempts to respond to the specific arguments that don't really make sense are more for the benefit of any other readers who might be confused by the muddling of waters that is going on here. I've attempted to provide test cases that can easily be reproduced by anyone else who has the laptop, specifically to see whether they have extra throttling due to the CCC, so I'm not really asking anyone to take my word for anything here.

 

The "wrongly flamed" by that poster would have been me ;-) But as per that thread, thanks to you @Papusan and @ole!!! for your intervention and stamping out the flames so vigorously that the thread didn't make it past the first page.

Here's a translation of the full text explaining notebookcheck's conclusion, until the English version of the review is uploaded by them:

"Update 23.06.18: As has been proven by prolonged trial and error, the new version of the Control Center is guilty of GPU throttling. If you "kill" the program via Task Manager, the GeForce GTX 1070 Max-Q behaves normally in The Witcher 3 and works with active Turbo (~ 1.291 MHz), which is very well recognized on the graph and screenshots. Since the XMG A517 and the XMG A707 are affected in addition to the Key 15 , the software bug that causes the GPU throttling currently seems to apply to all Clevo barebones with a pre-installed Control Center.


Active Control Center (GPU throttling)

Disabled Control Center (no GPU throttling)
Update 07.07.18: Schenker has provided us with an updated Control Center, with which the GPU throttling no longer occurs (tested on the XMG A707). We therefore exclude the original devaluation."


I should just clarify that notebookcheck didn't discover this issue, it was something we were working to resolve after the first review went live and in part that's why I had to go a little bit quiet on the forum as I wasn't in a position to start posting incomplete information until we got further down the path of understanding what was going on. CCC shouldn't throttle on Performance mode, as stated - you'll remember that one of the nbc reviews was carried out in Perf and one in MAX-Q mode - same chassis, same spec, different clock speeds as a result. As has been noted, the version of CCC used in the reviews will limit performance in certain applications but not all, plus the EC of course will control performance, power draw and so on at the same time. The crossover between these varies and so there's no clear global line between the two, it has to be taken on a case by case basis.