Anybody got a Skylake HQ/HK CPU yet? I want guinea pig-- I mean testers.

I've been following this thread hoping for some insight into rather or not the new quad-cores are worth it. I'm looking for a new laptop and trying to determine is a duo core would work or if I need to drop the cash on a quad.

Is there any speed testing that's been done comparing the two? I live and breathe in the Adobe apps, Photoshop, illustrator, indesign and dream weaver.


Sent from my iPad using Tapatalk

 

You want a quadcore

 

quadcore all the way with your application usage scenarios

Sent from my Nexus 5 using Tapatalk

 

I've got a Clevo P650RG with the 6700HQ, and XTU does have the option to adjust sliders, but every way I've tried (over 100 combinations lol) gets the same message "The settings for 1 Active Core, 2 Active Cores, 3 Active Cores, and 4 Active Cores could not be applied" when I reboot.

I am able to undervolt, and sit stable at -75mV.

 

hmmm iirc a user mentioned a bios setting for enabling cpu overclocking. search through the bios for a setting related to cpu performance scaling... hope im not mixing that up with the gpu scaling tho

Sent from my Nexus 5 using Tapatalk

 

Nope, that's for the GPU sadly.

 

But can you raise them? Lowering them makes sense.

You may need a Prema sBIOS mod to get full access if you didn't get your machine from Eurocom/Mythlogic/the-other-prema-partner-shops-whose-names-I-forget

 

Other makes with the 6700HQ aren't able to move the multiplier sliders.

It may be is that the Clevo BIOS you have enables the right BIOS bits to allow/enable OC, and XTU is responding to that - showing moveable multiplier sliders - but the 6700HQ doesn't support changing the multiplier - hence the option that does nothing.

Or, maybe a new version of XTU will enable the change multiplier function with XTU...

@Prema ?

 

I have the Clevo 650RE with the 6820hk.
My experiences - significant current and tdp throttling. Highest settings I've managed to avoid any throttling in (in the XTU benchmarking program) is -
x1 core = 37x multiplier
x2 = 35x
x3 = 35x
x4 = 34x
Core voltage adjust = -160mV
Turbo boost power max = 400W
Turbo boost short power max = 400.625W
Turbo boost power time window = 28s
Processor Core ICCMax = 600A

Putting the x4 cores up to 35x is stable at that voltage but current throttling kicks in depending on which stress test is being done, and I haven't fully played with the other multipliers as they're less relevant IMO.

I can run considerably higher settings before heat becomes an issue without it throttling with Prime95 - -50mV voltage adjust & x37,38,39,39 before heat becomes an issue even on this slim laptop but it will throttle instantly on the XTU benchmark.

Interestingly TDP was the initial throttling mechanism until increasing that maximum dramatically but increasing current only partially alleviated the new issue of current throttling. What's more, the reported package TDP never exceeds 56W despite the much higher limits I set. Current throttling will kick in even on those settings as soon as I e.g. run Furmark as well - knocking the multiplier down to below 30.

So currently torn on which settings to use day to day since it depends a bit on what I'm doing/which game on whether throttling kicks in..

It would be nice to send all this package power throttling rubbish to the dustbin and have the CPU use temperatures to manage throttling..

 

I received my Alienware 17 R3 with the 6820hk.
I can run it at a base frequency of 3.8Ghz and Turbo 4.0Ghz with no throttling or instability, so far.
x1 core = 40x
x2 = 39x
x3 = 39x
x4 = 38x

 

Actually, if I run TSBench, I see no throttling. If I run Prime95, I see some Power Limit throttling. Power is about 77W

 

That means that current limit slider does nothing. 256A is enough for a CPU to avoid current throttling on any program.

That also means that the TDP is limited for you. It's not limited so low on the Alienwares with 6820HK, so that likely means a BIOS update might fix it for you.

That makes me sad for you, honestly.

@Prema I know you wished to hold back with releasing sBIOS mods for a bit, but if you could let us check with a little mod magic for the sake of testing this CPU, that'd be fantastic. I understand if you don't want to, however. It is good business for your partner shops.

Are you able to adjust this setting in XTU or your BIOS? Setting your turbo boost power limit and turbo boost short power limit to something like 120W should be more than sufficient to complete tests like these. It isn't stressing the GPU, so even with a 180W PSU it should not be a problem.

On the other hand, if you have already raised the limits, it might be a board power limitation. The M17x R4 had a 68W board limitation to the CPU, so I don't put it past them to do the same thing here. Hopefully a couple more people reply and do some testing.

So far it looks like the skylake chips are much better at holding their performance, even though the overall performance has gone down from the QM/MQ days. I'd like to find out everything.

If the HK chips turn out to have an overall 77W TDP limit, that means their overclocking is worthless in larger machines that can cool them, like the GT72. Yeah, sure, you'll get maybe a solid 3.8GHz out of it or something, but if you stress that baby it won't hold a nice plump 4.3GHz like is common between 4930MX and 3920XM users on this forum.

 

I first tried adjusting settings in XTU and couldn't get much at all without lots of throttling. I then found some overclocking settings in the BIOS and tried them and that was much more successful, so something in the bios changes allow more power, I guess. The bios has 3 or 4 settings. I tried the highest one but had some instability. I backed off to the 2nd highest and it's been very stable during game play, Lightroom usage, etc.

 

With Sandy Bridge, Ivy Bridge & Haswell lower end (non-extreme) mobile CPU models could officially be OCed 2-4 extra multis on top of their factory Turbo.

Now on mobile Skylake you can't even OC a single multi on any mobile chip except the "HK" model(s)
As usual this story was twisted around and sold to you as the "Skylake being the first mobile chip to feature a K-series CPU".

Translation:

"For the first time we blocked OCing on ALL mobile chips and only allow you to OC on the extreme CPU, which we now named the "HK" to divert your attention."


The Prema Mod has all options to play with the "HK/HQ" chips (aka no throttle), but you can't OC a "HQ" except over BLCK.

 

Is there a Prema Mod in the works for the P65xRA? Please, please, please...

I'm a data analyst and only need lots of CPU power and memory, but still want a fairly portable machine with decent battery for travelling. So, getting the P65xRE/RG would just be a waste of money.

 

There is always a "Plan P"...

 

Yes it is sad that performance is handicapped like this when it's an overclockable CPU and thermal constraints would allow higher clocks. I'd be interested in a BIOS tweak to raise throttling limits if this becomes available.

 

Yeah. Such is the way of mobile CPUs these days. Thankfully Prema is here to help. I wish many times I had something I could offer him for the help he does for the community.

 

beer/coffee money! ive gotten into a habit of sending him some small amounts of cash here and there through paypal. helps me have a good conscience and i can also deduct that money from my taxes, win-win

Sent from my Nexus 5 using Tapatalk

 

I really don't have the money, that's the issue.

 

im talking about 5 bucks here and there...wouldnt consider myself rich either, not even middle class poor-ass PhD student but not for long anymore.... *mwahaha*

Sent from my Nexus 5 using Tapatalk

 

Can't change them at all. They aren't greyed out, and I can physically move them, but no matter how I set them, when I apply and reboot, I get the message that they could not be changed. Very disappointing. I was hoping I would be able to bump them after I read the 4-core turbo is limited to 3.1 Highest I've ever seen my package power draw go was 41.5 though.

 

yes, so I've heard. Seems that the way to combat the TDP lock is to make the chips run so slowly that the lock never shows up. This way it's technically not an issue for TDP, but at the same time the chip is very weak.

I can't say I have a problem with that design though, as I did and do say design a product to work within its specs. If it can't deal with power limits, then adjust elsewhere until it fits. Don't give it speeds it can't sustain.

The HK is the new wildcard though, and I wish to know how the 6920HQ chip works, as it's a 3.4GHz chip on 4 cores, which means it's much more likely to hit its limits.

 

I know my mobile CPUs~

 

i7-6920Bga vs i7-6820BgaK...
Intel takes well paid for Intel® Trusted Execution Technology for safe computing and Intel® Stable Image Platform Program (Intel® SIPP) supported with i7-6920Hq. This technology is intended more for business and not so much for private users. Do not forget That Intel makes much quirky things nowadays. Nothing surprised me anymore.
What a joke... Dynamic overclocking. This is what Intel make now. A name for this; Garbage. It is not only Intel that make garbage nowadays.

 

What?

It's not new. I knew about that the whole time. Why do you think I was talking about the pitiful 3.1GHz 6700HQ vs the 3.3GHz 471xxQ and 3.4GHz 4720HQ? The 3.2GHz 6820HQ vs the 3.5GHz/3.6GHz 4800MQ/4810MQ is even MORE of a joke. Far less the 3.4GHz 6920HQ (WITHOUT Iris Pro graphics) vs the 3.6GHz/3.7GHz 4900MQ/4910MQ or the 3.8GHz 4980HQ. It's going BACKWARDS in available power, especially since as Prema pointed out, all the HQ chips cannot overclock at all. They are flat out ALL inferior to previous generation things.

I don't understand what you're talking about here.

Yes, you did. You couldn't bump speed beyond +600MHz from the default 4-core turbo, as far as I know. But you had TDP/thermal locks so it was pointless to overclock.

Prema just said above that only the HK chip can overclock. A 6700HQ is stuck at the numbers you see on that table. Same for the 6820HQ and 6920HQ. The 6820HK can OC but we're not sure if it has a TDP limit properly. Seems someone said 78W is the limit, which is very low if a user is using a 240W PSU and a 100W GPU and wants to clock it pretty high with some cooling mods.

 

I'm running Level 3 with no problems. It definitely improves performance if evaluated in a real world way. A lot of the benchmarks are not very real world; still, even with TSBench, I saw a good 15-20% performance improvement.

I think the R3 is extremely quiet so I don't know if our laptops are different or we just have different definitions of loud but I can say without doubt, it's the quietest laptop I've ever had. Unless, I'm playing games, the fans completely silent. They definitely come on during games but if you have the sound on at all, you don't hear them much at all.

I did notice all the drivers disappeared off the support site. Weird. I also noticed they shipped the laptops with bad free fall sensor drivers, making them take 1 1/2 minutes to start. wtf Dell! After fixing that, 15 second start times. My first Alienware . . . very mixed feelings so far.

 

Could you run a wprime 1024 (8 treads) stress test with level 3 Oc. Put win power plan to high performance and not balanced performance. Monitor with Hwinfo/xtu and post pict of the results in the tread. http://downloads.putme.net/wPrime155.zip
Wprime is maybe the bench test nearest real world use.

 

wPrime 1024, 8 threads, level 2 OC. Level 3 hasn't been completely stable, so I'm using Level2

 

Sorry, it's Level 2 I"m using not 3. I was thinking there were 4 levels but there are only 3. Level 3 wasn't completely stable.

 

wPrime 1024, 8 threads with NO overlclocking. 20% improvement with Level 2 Overlcocking. Not bad.

 

Level 3 in bios. What is the problem ? Unstable only in bench or also in other use? Have you tried with lower core voltage? 1.253v is quite a lot for Oc at 4GHz. Went multi under 4 ghz during the Wprime test? What is voltage in Oc Level 3 during the bench?

 

No, I was just getting random occasional blue screens. Completely gone at level 2.

 

LOL. Yeees. Dell is not good with this. It is not the first time Dell's Oc profiles are crippled with incorrect power setings.

 

Unfortunately @climbbike I can't really use those results, you need to check if there's any throttling going on during the test, and you should be using Throttlestop and the updated Limit Reasons too.

What it looks like to me is that you topped out around 57W with the 4GHz OC, but your minimum processor state is under 4GHz. I can't tell if it just downclocked like that for the sake of being idle or not.

And the 3.6GHz has you topping at 42W, but at the same time, you're showing to be around 3.2GHz and not 3.6GHz in the screenshot. So I can't tell. Basically you need to let most of the tests run, then near the end, with Throttlestop and limit reasons and HWMonitor up, take a screenshot with all of them, so we can see if things are consistent.

 

Hope this is still needed. My laptop came in last week Wednesday (after a week more delay), and I haven't had time to rumble with it because of work.
So, i've been up all night trying to OC the CPU and little luck. I am very confused and couldn't find any reference online to base on my settings/results.

My stock BIOS doesn't have CPU modifications settings (I thought it would as it's OC-able) so I used XTU to overclock it. The settings is attached below.
Basically, it's 4Ghz and default core voltage. Turbo boost power caps are all maxed.
...

 

Following your post below:

Throttlestop ran fully with no throttling at all
XTU Stress test had ~6% current throttling
XTU Benchmark had massive current throttling running up and down between 6% to 100%

 

OCCT and prime95 tests: 100% current throttling at all times. Didn't bother running them for long
wprime test: throttled 1% in the beginning, then stopped. This guy wanted to burn my CPU

Edit: prime95 crashed the system, would have to dial down

 

From all indicators, it definitely means the CPU can exceed intel's TDP limit and stay stable.
I believe my current OC setup is not optimal (as I don't even know what am doing) so if you can direct me to a guide to OC it and rerun any tests you would like.

 

From what I see, the only thing you may need to do is bump voltage some more. In this case, you need to find out what your average voltage is (TSBench and OCCT are good for this) and then change voltage from "default" to a bit above what you have.

It seems you have an average of 1.19v but with XTU stress and bench you hit 1.23v. I'd suggest setting voltage to 1.195v and see how that handles stress tests, since the prime95 screenshot you showed was only 1.14v.

If it works for XTU bench/stress as well as OCCT/TSBench, then that's awesome. If it does not, watch if the voltage is going up or down. If it's remaining below 1.195v in any test and then crashing you, I suggest making a profile for heavy load where voltage is static and not adaptive, and then trying it. If it crashes only in tests that require higher voltages, then I would suggest 1.23v as your voltage amount, and trying again. If it's stable all round, keep it until you do crash and are certain it's the CPU. Then you might want to try 1.25v. That's my basic advice for now.

The thing is current limit throttle needs to be bypassed. I'd need you to get a Prema system BIOS mod so it is unlocked. Once current limit throttle is working, we can run the Prime95, XTU Bench and OCCT tests again to make sure both that current limit works and that TDP is kept high. But you've provided some valuable information, and thank you very much!

 

I bumped the voltage up by 0.09v, averaging 1.193v on stress testing and max 1.265v on idle. Multiplier is 38
Prime95, XTU benchmark, OCCT: throttles current 100% and limit reasons Ring displays EDP (red).
XTU stress test: throttles current a few times at 1% and limit reasons Ring displays EDP a few times (yellow).
TSBench: no throttling at all on both 38 and 40 multipliers

I can't seem to change voltage to static as the system crashes whenever I do so. All tests that throttles are run around 5mins. As it never hit the set core speed, I don't see the need running them further because it will never crash the system (i suppose).
Which brings us down to prema BIOS. Which one should I use? There's no list for my clevo (P650RG) on the website.
Also, can I revert to my stock BIOS anytime?

 

You can backup your stock sBIOS and revert, yes, but you'll probably need to contact Prema and ask for the sBIOS for your machine. See, his mods aren't available to the public just yet. It's a benefit from purchasing from his partner shops, like Mythlogic and Eurocom. See if you can get him to give you a sBIOS for testing here? I don't know what else to say. We need to fix your current limit throttling (or try to).

 

I see. Well am not even in the USA so it wouldn't have been possible. Hope he makes an exception with me ;D
@Prema

 

He has partners elsewhere, but you're right, there is by no means worldwide coverage.

 

Hi there....... ive just signed up and would be interested in maybe running my ''AW15 (2015) with i7 6820hk'' for some testing in relation to overclocking...

now, i wont lie laptop OCing is very new to me (seems strange not to spend most of my time in the BIOS).. but my laptop does throttle back under continued load like everyone elses, but ive just scratched the surface of XTU and im sure theres a way to hold 4.1ghz without throttling..????

it'll hold 4.1ghz for day to day use ...but would be nice to stop the throttle..

 

calling out to @D2 Ultima !!! uve got urself a willing AW tester

 

Thanks for the alert =D

Hi. You might want to check the post on the first page for some tests to run, and then report back with your findings? There's detailed instructions there on what you should be capable of doing.

 

I showed him the way He claimed that his BGA **** overclocked excellant

 

@D2 Ultima
Here you go! Prime95 blend torture test. I had another screenshot about a minute prior showing a steady ~77W package power, but I forgot to tab into HWI. Then suddenly the power shot up to like 91W and it hit 96C so I got scared and turned it off. D: Methinks I need a repaste. I recall hearing that increased temps cause increased power consumption, so maybe it was some sort of runaway positive feedback effect? In any case, the window on the XTU graph is 5 minutes, so you're seeing at least 4 solid minutes at 77W.

Spoiler

Edit:
Round 2! Taken shortly before my CPU temp explodes (and you can see it's already started to), but it shows the package power well over the TDP after 3+ minutes.

Spoiler

 

interesting!

btw, the shooting up of temps is due to the pulsating nature of xtu stress test and prime95. they cycle through different kinds of loads on the cpu, thus ull see alternating higher and lower temps / tdps.

Sent from my Nexus 5 using Tapatalk