How well does the 7700HQ undervolt?

I know these CPUs haven't been out long but has anyone seen how well they undervolt compared the 6700HQ?

Most 6700HQs will drop about 100-150mV with around 5-10C, I'm hoping the 7700HQ is similar.

My google-fu isnt strong enough to find anything, or its not there yet.

 

Undervolt amount is irrelevant. Resulting voltage is all that matters. My 4800MQ gets under 1v with a simple -50mV adjustment. Others have done -100mV and still been above 1v.

And, nobody knows.

 

Well, those that have them can report their results, and then taking all of those into account we can offer a suggested range to try.

The Skylakes had a wide range of undervolt at stock settings, from -80mV to more than -200mV - it's nice to know in this case that great than -100mV has been attained.

Most vendors are shipping with the CPU voltage default in BIOS set too high, so undervolting isn't really optional, it should be done.

The temperature drops have been high enough to avoid re-pasting - which for most people is beyond their comfort zone.

It's much easier to tune in XTU or TS the undervolt, than it is to disassemble a laptop and re-paste.

So asking what the range of undervolt for Kabylake is a valid question, we just haven't had enough reports yet to suggest any change from Skylake experiences.

As always start at stock settings and test there fully with a wide range of benchmarks and games/apps, then once you have a good measure of temperature under various loads, pick a method for finding your best undervolt, and proceed with the process.

I recommend getting comfortable with the idea of testing to Blue Screen on your laptop; it's really the only way you will know the limits.

Early Skylake owners were stopping their undervolt testing around -100mV because they thought that was the limit - Haswell was around there - but it turns out Skylake undervolts on some CPU's to well over -200mV at stock settings. If they had tested to Blue Screen during their testing they would have found the upper limit months earlier.

Just remember to do that testing to failure outside of your normal use - don't experiment on a device doing mission critical work.

Save your work, load the test profile(s), and if you do crash then be sure and do a boot-time Disk Check on your boot drive and any drive with open files (pagefile too) at the time of the crash.

I'd start Kabylake testing at stock CPU settings (Intel's numbers, not OC'd laptop stock numbers) with -100mV, then -150mV, then -190mV, etc by -10mV increments from there.

If we find early on a lower limit, I will try to remember to update here

 

Yes, but nobody really has them yet. We'll find out when people do. I'll be doing another thread like I did for Skylake asking for OC results and such. In another month or so we'll be good to start gathering information =D.

Yes, Skylake had EXTREMELY high default CPU voltage for most chips, though in general far lower than Haswell did. Kaby Lake is designed to use a lot less voltage; we don't know what stock voltages look like or how they will handle undervolting since they're supposed to be voltage-tuned.

Vendors aren't actually doing anything. Stock voltage is different for each CPU. It's reliant on the CPU. If you could switch the CPU out, you'd find each one has a different default voltage. That's all. It is high, and that's mostly Intel's fault, but we'll see how it turns out for Kaby Lake.

I don't understand what you mean by "OC'd laptop stock numbers"? 6700HQ couldn't change its multipliers at all; only turbo boost on or off. 7700HQ is more than likely the same thing. There would be no OCing.

 

So until then why not answer with the Skylake undervolt range of results found?

The undervolt range was much higher for Skylake than for Haswell.

Each CPU can have a different maximum absolute undervolt value, that's why we experiment to find that value for each laptop

But, by keeping up with user reports of their maximum undervolt we can give a range of results seen that can help guide new owners on their path to finding the unique setting for their CPU / Laptop.

For now I am going by the assumption that the BIOS settings - all base numbers recommended by Intel - including voltage - are set up by the vendor in the BIOS they provide.

There are some laptops like the Asus G701VI that come out with BIOS settings for multiplier (and other supporting settings?) that are higher than stock, which can reduce the range of undervolting normally seen.

Here is such an example just posted recently, where the expected range of undervolting isn't available:

Asus ROG G701VI - Air-cooled single GTX 1080 laptop
http://forum.notebookreview.com/thr...-gtx-1080-laptop.796850/page-11#post-10441328

 

Because since Kaby Lake is tuned for less voltage than Skylake, I can't say whether it would undervolt as far. If Skylake's average voltage on a 6700HQ was 1.21v let's say, and Kaby Lake's average on 7700HQ is 1.1v, it stands to reason that -200mV is a long shot (0.9v for 3.4GHz). We don't know until we see some chips.

The range is something that can be averaged within a certain architecture. Skylake undervolted more, because they used too much voltage out of the box. Haswell also used too much voltage, and undervolted by a lesser amount, but haswell's "too much voltage" was on average more than Skylake's. I saw a lot of 1.3v+ haswell chips, even one person with 1.42v at stock. I also saw a lot of 1.15v or so haswell chips. But in general, the people who got the most out of undervolting had very default voltage. Skylake didn't launch much 1.3v chips; most of them were pretty much under 1.2v as far as I remember. Skylake's "too much voltage" isn't as high as Haswell's "too much voltage".

Yes... because each CPU's voltage is different. Every single CPU's default voltage is different, and their lowest voltage tolerance for any speed is different. My 4800MQ's default voltage is 1.0474v or so. -50mV at 3.5GHz (stock turbo for 4 cores) is 0.9976v on this existing chip. At 3.8GHz however, I can't even use a -20mV offset. I have to use stock voltage, and my CPU suddenly requires 1.09v now. This also shows default voltage scales with speed of the processor. This is why each CPU's undervolt tolerances are different. Some CPUs have a low default voltage, and cannot undervolt far. Some CPUs have a higher default voltage, and can undervolt really far. Some have a higher default voltage and can't undervolt really far. It's all wild card. This is why I always say to pay attention to the effective voltage after an undervolt. If I get a 1.1v 6700HQ and use a -100mV offset and it sits at 1v flat, and you get a 1.2v 6700HQ and your limit is -180mV, you're STILL using more voltage than my 6700HQ even though your undervolt is "more". But your effective voltage, 1.02v, is very close to my 1v, and thus we can look at a lot of laptops and say "oh, 6700HQs average around 1.02v stable with undervolting" and then a user can aim for that, and if stable, then push from there.

It's 100% the chip. Intel probably set an overly safe voltage for most CPUs, but it is seriously dependent on the CPU.

As for laptops that can overclock in the BIOS with an auto-OC, that's not stock. That's a 6820HK in your provided link; overclocked to 4GHz. Those chips are known to have extremely bad binning, and require a lot of voltage at 4GHz+. The average for 4GHz is about 1.2v or so, with 1.25v to 1.3v needed for 4.1GHz or higher (you can even see in the post under the one you linked that the user was at 4.1GHz with over 1.3v). If the auto-OC left voltage near that, and you then try to undervolt a lot, it'll break with extreme ease. If you're at 1.2v then you should try aiming low, like -10mV steps; you're probably already at your limit. But most importantly, that isn't stock, and 6700HQs (and likely 7700HQs, which is the primary question of this thread) cannot be overclocked in such a manner. It doesn't matter if you take a 6700HQ out of one machine with one of the units to install a BGA chip and put it in a laptop that previously had a 6820HK at 4.1GHz via the BIOS. That chip is GOING to sit at 3.1GHz 3-core and 4-core load, 3.3GHz 2-core load, and 3.5GHz 1-core load. It'll never change. You can only enable or disable turbo.

 

Where are you getting that the default voltage is now .1v lower for Kabylake? I have only heard that the Skylake / Kabylake operating environment is the same, not different.

If it were true that Kabylake CPU's were being "undervolted" -100mV in the BIOS as opposed to Skylake, that's would be big news, and it would change the user undervolt range by -100mV.

Links? References?

I realize you are fantasizing a scenario, but there's no need to do so, unless you have solid info as to global changes to the operating environment for Kabylake vs Skylake to back that up.

You are simply introducing Hyperbole to avoid admitting that the first statements made are correct and reasonable when estimating how Kaby lake will perform compared to Skylake.

No need to go off in the weeds man, you can stay on the well worn intellectual path for estimations in answer to valid questions.

 

I was making an example, to prove that you cannot simply translate undervolt amounts from Skylake to Kaby Lake. The entire point of the Kaby Lake architecture is that it's meant to use "less voltage per clock" over Skylake, and has a stronger IMC (integrated memory controller) allowing for better/faster RAM. It is not any "faster" clock for clock. However, since it is faster (by 300MHz), and is designed to use less voltage than Skylake for all clocks, it is possible that we will see very low voltage Kaby Lakes out of the box.

Kaby Lakes are not being undervolted in the BIOS. Kaby Lake CPUs are designed to use less voltage. That is all. I am "expecting" to see Kaby Lake CPUs using less voltage out of the box than Skylake ones.

 

Dude, you didn't prove anything

We already know that Intel CPU's are variable in the amount of undervolt they provide, for many years we've known this - that you need to tune each CPU in place.

A range of values can be established for owners to use to focus their tuning, which for many provides a reference point to ease their tension - making it easier for them to proceed given a range to try.

There's no need to spin it all into nonsense, give them a range, it's not tough, then they feed back what they find and adjust the range to fit the new line of CPU characteristics.

Why confuse this "for a month or more", give them something useful to work with now.

 

Mostly because the people who have got them just havent looked yet, or thought that others may want to compare. Some people do already have 7700HQ chips and if interest is shown its more likely that people will look at it and give details.

In the various reviews floating around for 7700HQs I havent come across any which specficailly look at CPU voltages. I did find what which did happen to show some CPU related voltage but its not clear which one it is. (An ASUS GL502VS showing 0.965V in BIOS and 0.639 in CPU-Z, both quite low even for at idle?)

 

MSI and Clevo owners on this forum are generally undervolting nuts. ASUS to a lesser extent. Most of them will be looking to undervolt. You just need to wait for the active/helpful ones to get some. If you find any of them who will give you their readouts, then great. But like the number of responders in this thread, the number of helpful/savvy people who already have one is low.

Reviewers will almmost never undervolt. Unless HTWingnut (or myself) is doing the reviewing, at least, for the most part. But 0.965v for 3.4GHz is extremely little... I don't imagine it would have so much room. Idling at such a low voltage, however, is normal. I just watched my CPU idle in Throttlestop and it hit 0.6312v at a low point, so... Kaby Lake's idle voltage isn't impressive. When a CPU is clocked low, voltage almost always remains really low (from intel anyway).

 

mine 7700HQ is working fine with an undervolt of -150mV on CPU cache and core with speedshift activated. In Prime95 the temperature now does stay in the low 80's under full loads. You can find my idle and under load (prime95) Voltage attached .

i did not check any lower settings but i will do this if i find some time for testing. I only had tried to apply -200 instead of going little steps but it freezes immediateley. ;-)

 

Prime 95 did fail on one of the cores at -170mV after 24 minutes. Also graphical anomalies while outputting to 4k TV despite not downvolting the iGPU yet and running just on the dGPU. I output to my 4k 40in TV as my main monitor when I'm not using it as a laptop so this was a weird bug but is unacceptable.

Prime 95 failed at 17 mins on -160mV on one of the cores. Same graphical anomalies above, when i move windows it leaves a shadow of itself in effect. Opening other apps in full screen and what not to try to clear it doesn't fix it.

At -150mV while prime 95 doesn't fail or show errors for 44 minutes during the test, but it froze my computer when i stopped prime 95 so i could go to sleep. There was also a graphical glitch while clicking around in TS but prime 95 didnt show errors either. Because of that. I don't believe -150mV is stable.

-130mV CPU Core and Cache and was stable for 3 hours 3 minutes. Another test showed it stable for 2 hours on prime 95. Temps are reaching 85C max but usually hover around mid to high 70s. I would like to note that no game can push the CPU temps past 51C including Witcher 3, For Honor Beta, and GTA 5.

Its exhausting to also test the effects of disabling turbo boost and clamping the Package Power Limits to a certain watt independently, and seeing the effects on battery life. So i'm just going to use the laptop in normal operation with turbo off and will keep lowering the package power until i see noticeable lag before I do a battery drain test. Will take a week or two & will update this post.

Updates: I got graphical anomalies bring back my computer from sleep while undervolted at -130. This is with discrete graphics mode only, and connected via HDMI to a 4k 40in TV/monitor. Thought it was something else but turning TS off completely solves the problem. Lowering voltage to -110mV and will see.

Update: Computer automatically restarts when coming out from sleep after not using computer 1-2 hours at -110mV. Reducing it to -100mV and will see if that's stable. If i put my computer into sleep manually and then wake it up, its fine. If the computer falls asleep for a short period of time it also wakes up fine. But if it falls asleep for more than 1.5 to 2 hours, it automatically restarts. Very troublesome for critical work. Anyone else having issues coming out of a long sleep?

Update: -100mV to CPU and Cache seems to be stable coming out from sleep. That plus disabling turbo, -40mV to iGPU, and clamping the Package Power Limits to 15 watts allowed me to get a battery life of 4 hours before hitting 7%! Before the laptop only got 3 hours 41 minutes before hitting 7% battery on iGPU and 2 hr 50 mins if on dGPU. Despite clamping the Package Power Limits so low, I did not notice any slowdowns and the laptop still feels a lot faster compared to my Dell XPS 9350 with a Intel i5-6200U. I will continue to try lowering the iGPU voltage and clamping the Package Power Limits to see if i can get further battery life improvements.

 

Currently testing -150mv on CPU core and cache. Using throttle stop and hitting 3.4 GHz on all 4 cores. 60 min stable on AIDA64. Integrated graphics at -125mv and 30 min stable in Unigine Heaven.

Voltage:
Vid max of .957v (3.5GHz on all 4 Cores)
VID min of .558v (897 MHz on all 4 Cores)

On Desktop with no programs open idles between 1.3 and 1.4 GHz at .6v

Testing done on an XPS 15 9560 (FHD with 97wh battery)

Will do a 6+ hour Prime95 run tonight.

Update: Stable after 12 hours with a max package temp of 84C.

 

You are right about 150mV. Actually 150mV isn't much when you run full throttle - this is less than 15% decrease of the normal processor voltage at that speed. But look at lowest energy modes when processor itself run as little as ~650mV and you undervolt it even more (up to 500mV) - that's relatively tooooo large decrease (more than 20%) and may cause transistors just not work as expected and triggers does not trigger .... (also 500mV itself on the edge when logical units still works ...)
So assuming:
1) doing undervolt - take into account lowest possible power mode as well as highest.
2) you can increase undervolt by some number if you limit processor power saving capabilities (for example disable FID less than 20) - you can run colder at high speed but not able to benefit from low power modes

BTW:
Throttle Stop 8.4 on my XPS 15 9560 i7-7700hq
CPU/CACHE: -135mV
GPU: -60mV

 

You missed one very critical stability test:
Test your CPU at MAX FID by launching just one prime95 threads (you can launch 2 or even 3 if you configure processor to run at 3.8GHz)
As you mentioned - you tested it at 3.5 GHz which is relatively low to 3.8 GHz

 

I've back off to -120mV on the CPU and its been running well. I might need to raise the voltage some more to increase stability. Last week, I had a BSOD after quiting csgo directly out of a deathmatch. I might just do -100mV on the CPU and GPU to make it rock solid. Temperatures aren't a huge problem since I also re pasted and padded.

 

Good to hear you've found stable undervolt for you..
I stopped at -125... no BSODs for ~1 month.
Also enabled SST for better responsiveness - PC is fast and stable at the moment and able to run over nine hours on battery (I have 4k display ... if I would have FHD I'd probably get even better battery life). Thanks to ThrottleStop developer (I use this piece of software since i7-820QM released) and those people who made research on how to permanently enable SST.

 

I've only heard good things about speed step. Is throttle stop the best method to activate it permanently? Also how does turbo @3.8Ghz across all cores affect battery life? I was on a long road trip and got 12.5 hours of battery life watching movies using VLC player.

 

I have undervolted my dell xps 15 with i7 7700hq 150millivolt and it seems stable till now. Now max voltage is 0.9V and temperatures have dropped 12 C from 79 to 67 degrees. Total power consumption under stress has dropped from 33 to 23Watts. I will have to wait and do further tests to tell you if this undervolt is stable, 155 millivolts resulted in sistem freezing immediatly but no blue skreens.

 

I would step a bit for 24/7 undervolt if just 5mv more crashes your system. What program are you testing with because that V-core and those temps are very low.

 

That's a really nice undervolt if it works for you, but you have to be a bit careful: as far as I can tell, there is no reliable way to determine that an undervolt is stable except for simply using the laptop. My XPS 15 is currently at -100mV CPU and -75mV iGPU. I used it at -110mV for some time and while this is rock solid in benchmarks and doesn't cause any problems when coming back from sleep, it turns out to very rarely cause crashes in games (about 1 crash every 5 hours vs. 0 crashes in 30+ hours at -100mV).

 

If you undervolt enough to avoid TDP limit errors in Prime95, showing up in Throttlestop (=45W), then small FFT test will work in prime95 to test the undervolt. But even with a small undervolt, you will exceed TDP....

 

I used TS and started the long test a few times in immediate succession to test temperatures and did some 4k gaming to test stability. I also undervolted gpu 100mV probably that helps with temperatures too. Since I don't use my pc to the ful potential often I think stability will be ok in case I experience crashes I will reduce undervolt. I also noticed that the clock dosen't goes above 3400GH even if it is set to 3800, probably that helps with stability and temps.

 

My Asus GL502VMK (K for Kaby-Lake), with an Intel i7-7700HQ, u ndervolts -110mV.
The core voltage - which is the main focus - is around 1V, staying between 0.942V and 1.043V.

This laptop suffers from Power throttling, which basically is the CPU demanding more power than it is defined as the limit power (in Watts). So, the CPU will throttle to 0.8GHz, raising slowing to the normal frequency by steps: 0.8; 1.2; 1.8; 2.7; 3.8GHz.
The undervolt solution solve this issue.

I tested using Intel XTU (Extreme Tuning Utility) and maintained that Voltage with Throttle Stop running at Windows logon.

Despite all my efforts, I noticed I am not able to reach the full 3,80GHz, staying shy of 3,73GHz.
This is due to reference clock not staying at 100%, being lowered to ~99,7%.
I am not sure what is causing this, but I will keep digging...

I also undervolted my GPU, using MSI Afterburner.
For more details, check the ROG forum here.

 

While I didn't have the power trottling issues my i7 7700hq does not reach ful clock either. I get aproximatly 3.6 ghz single core and exactly 3.4 ghz whith all cores. Its wery annoing because temps are never realy hight so I would like some more performance. Does sombody knows a fix to this? I alredy tried with trottlestop but klock speed is apparently locked. I have a dell xps15 9960

 

Probably tasks in the background who load more than one core. Disable unnecessary processes - software and test again. Use Autoruns.Exe

 

The fix is to buy an unlocked processor. Locked processors are bad. There were a few locked processors which allowed you to increase the turbo bins a few steps, like x3 higher, after a few tweaks (e.g. 4700HQ), but there's no fix for any newer processors. The only processors worth buying are the HK processors, or desktop processors (in barebone laptops). It's also very difficult or impossible to force only one thread usage because windows will still use other threads by itself. See if your Bios allows you to disable 3 cores or if there is an unlocked Bios available to let you set the # of active cores (don't be surprised on HQ processors if the most you can do is disable hyperthreading on an unlocked Bios).

There may be INI tweaks you can do to windows to only force it to run with 1 core or 2 cores. There were for windows XP. No idea if there are any for windows 10 or 7.

 

On my AW 15 R3 I undervolt my 7700HQ to -140mv and stable for gaming.

 

hello everyone, I own a Asus GL502VMK. which i have repasted with Thermal Grizzly Conductonaut, along with a few more things like new high quality thermal pads, tapped the gap between the heatsink and fan AND doubled the height of the fan intake spacers so the when the case cover is put back, the fans have more area to pull air from. i have been using Intel xtu to undervolt as i just found it easier and i settled with a core&cache undervolt of -0.108mV. for the past 8 months i have been running this setup. but i read somewhere on the net that because the core and cache slider is sync'd and cannot move independently of each other. people have got consistantly better undervolts with throttlestop. ( although the ui is not so straightforward ). so after trying it out it turns out to be true, in intel xtu , i could not manage more than the -0.108mV on the core or cache. but in throttle stop, i managed to undervolt the cache to -0.125mV and the core to -0.150mV.

AND ITS COMPLETELY STABLE.

i used occt/aida 64/ prime 95

and irl games like bf1/4 , gta 5 online ( and offline ) , and lighter games like cs go as well

the computer is completely stable and has not locked up or crashed for the last 2 months.

what i don't understand is why when i asked intel xtu to do a -0.110mV on either the cache or core, it outright ust failed and locked the system. so that's worth noting

room temp 35c humidity give/take 80%

without either undervolt, in aida 64, the cpu would consume 44w and maintained [email protected] ([email protected]) 0.00mV

with the undervolt from intel xtu, in aida 64, the cpu consumed 36w and maintained [email protected] ([email protected]) -0.108mV (both cpu/cache)

with the latest undervolt using throttlestop, in aida 64, the cpu consumed 32w and maintained [email protected] ([email protected]) -0.125mV for cache and -0.150mV for core

thats just my research into cpu testing and stability.

alongside that, i have the GTX1060 6GB running at [email protected].

temps in gta 5 online after about 6 hours ( remember laptop is on default feet, not raised in anyway)

cpu peak of 90c but averages 86c

gpu peak of 87c but averages 85c

if the laptop is lifted or placed on a cooling pad, take away 5-6c from the avg and max temps.

have any more questions? feel free to ask.

cheers everyone!

PS: sorry if my english is a bit weird, i did type this in a hurry .

take care!

 

Intel writes the XTU software, and so it does follow the capabilities of the Intel CPU's closely, whereas ThrottleStop is an independent development - but I think @unclewebb gets input under NDA for details he uses, so it should also follow Intel specs and capabilities.

I am answering because recently @Buffalord was testing his 8th gen 8750H and found ThrottleStop would also allow independent cache and core settings for undervolt, but it seems that it doesn't really work - he had to double and triple check several tests to realize it was picking one setting and going with the - as it turned out the smaller of the two undervolts -180mV set as cache vs -330mV set as core.

@unclewebb can you maybe help us out to understand what's going on with ThrottleStop and the independent cache / core undervolt settings on CPU's we didn't think would support that feature? Thank you!

 

That is the problem with ThrottleStop. I do not have a NDA agreement with Intel which means I have zero access to any insider information. None of the information about CPU voltages is publicly documented by Intel so individual programmers like myself have to do lots of testing, some guessing and ultimately, I depend on user feedback to keep this project going forward. ThrottleStop seems to work OK on 8th Gen hardware. The only laptop I own has a 4th Gen Haswell CPU. This means there has not been any thorough testing done on any recent mobile CPUs during the last 5 years.

On all Intel CPUs, 4th Gen through the 8th Gen, the CPU Core and CPU Cache voltages are written to two different registers within the CPU. Intel XTU writes the same information to both registers at the same time. ThrottleStop allows users to write different voltage information to each register. Think of this as the voltage request registers. If you write something to the CPU that it does not understand, your request will be ignored or the CPU will use the lower value.

With 4th Gen Haswell, the CPU Core and CPU Cache voltage adjustments were completely separate. I think the 5th Gen Broadwell were still the same. On most 6th Gen Skylake and later CPUs, these two voltages seem to be locked together. When I found this out, I wrote a version of ThrottleStop that locked these two voltages together and there were immediately complaints. It turned out that on some CPUs, it seemed that you could still adjust the CPU Cache and CPU Core voltages independently. As long as the CPU Cache offset voltage was set lower than the CPU Core voltage, this still seemed to work but this was not possible to do in Intel XTU.

-100 mV Core
-150 mV Cache

Users told me that the above seemed to work OK and gave better results compared to having both set to the same -100 mV. After hearing that, I went back and decided to leave these two voltages independent. This may only apply to the 6th Gen but I am really not sure.

When the Core offset was set higher, this did not seem to work.

-150 mV Core
-100 mV Cache

The CPU would default to the lower number so you would end up with -100 mV on both.

Without access to proper documentation or a wide variety of hardware for testing purposes, it has always been my opinion that things should be left unlocked so users can experiment.

While recently testing a 7th Gen desktop CPU, I discovered that even though changing the offset was not changing the actual voltage going to the CPU, it was able to change the VID voltage significantly. Power consumption in Intel CPUs is not measured. Instead, Intel uses a formula to approximate power consumption based on VID voltage. Tricking the CPU this way can reduce reported power consumption which can eliminate throttling based on power consumption. I have not had enough time or hardware to see if this trick can be harnessed on laptops but I will post a few pics that show what I am talking about. Leaving those two voltages unlocked and independent allows tricks like this to be pursued.

 

My 6600K is horribly binned, requiring 1.47 V just to get to 4.7ghz, and my 4700MQ undervolts by -80mv using Throttlestop, doing day to day tasks.

 

Neither of those are 7700HQ's

Every CPU is different, if that's what you are saying, so the very title of this thread isn't very helpful, and I agree.

The average won't apply to every CPU, maybe not even 50%...

The range of undervolt has been growing by each new generation, new process, and the 7700HQ average was around -120mV, while the new 8750H is hitting -170mV -> -180mV and more.

Thanks for sharing, it's helpful to get perspective from the long term. I remember my 4720HQ that would only do -15mV out of the box, and 18 months later with lots of coaxing finally went to someone else undervolted at -45mV

 

@Buffalord @Rannon Anthony

Wow, that's amazing, great work then getting by without information from Intel under NDA. Someone a while back mentioned NDA with Intel being a problem handing off the ThrottleStop development - I guess they were just guessing

Hows that going? Are you going to hang in there, or end development sometime soon as I heard? Or, is someone going to pick up development?

Ok, well IDK how to characterize it based on your examples, like why a certain value is better than another - while keeping cache undervolt to less than core undervolt.

@Buffalord set cache @ -180mV and core @ -330mV, and he said he thinks the -180mV was the value used - but the -330mV on core did make some kind of change in effect vs setting both to -180mV.

I don't know how to recommend owners use TS like this to tune for best effect in this situation.

Initially setting both undervolt core/cache values to the same value to find the lowest undervolt, then try to vary the CPU core undervolt higher - keeping the cache undervolt at the previous maximum...trying to tickle the VID trick.

It's time consuming enough to find the stable locked cache / core undervolt stable at load / idle, to then add chasing "ghosts" along a side-effect trick path for a little improvement seems like something I wouldn't bring up when trying to help someone get through this the first time.

Maybe mentioning at the end after everything is tuned and done, 'oh btw, you can trying dropping the cache undervolt a bit at a time until you notice a drop in power draw and temperature, even though its not supported to unlock them by the CPU...

Thank you for taking the time to explain, I think that helped "lock it down" as to what is going on.

 

@Buffalord @Rannon Anthony
If you aren't already, you'll want to read up on @unclewebb 's thread, where this is recent info about keeping undervolt for iGPU, locking power limit's, and Turbo Boost vs Windows power settings:

The ThrottleStop Guide

http://forum.notebookreview.com/threads/the-throttlestop-guide.531329/page-812#post-10758751

http://forum.notebookreview.com/threads/the-throttlestop-guide.531329/page-810#post-10757575

http://forum.notebookreview.com/threads/the-throttlestop-guide.531329/page-810#post-10757517

And, post #1 / #2...
http://forum.notebookreview.com/threads/the-throttlestop-guide.531329/

 

Wrong thread!

 

I have a 6700hq uved by 130mv fully stable (before i unlocked my dgpu Vbios and oced), now i run it at - 110mv.

I also been working on a dell inspiron 7577 with the 7700hq, and i managed - 125mv np, evej stress tested.

So yea, i think those chips can be uved by at least 100mv across the board.

 

That was the consensus when they were new, -100mV was a safe starting point, some unstable there, most were stable higher.

And, it's gotten better each with each new "process generation", now the 8th gen are -150mV to -180mV plus or minus.

If you screw with the BIOS settings, adjust the line load calibration or voltage curve the recommendations can change wildly, causing the need for +100mV or more to balance things out. Causes more of mess than helps in actuality.

End of the day, undervolt is enough correction, don't confuse things pulling both ends of the cat.

Same goes for OC'ing, after you find the undervolt balance point, solid and stable, overclocking upwards reduces undervolt - CPU needs more voltage as you raise clocks - and overlocking downwards (underclocking) allows for more undervolt - CPU needs less voltage as you drop clocks.

Seems like small changes, but when you stop thermal throttling, reduce power throttling, and get better results overall, it's well worth the effort.

 

I'm well aware of all thay, all i'm saying is that skylake and kabylake locked mobile chips do undervolt very well.

I have an AW 17 r3 with the 6700hq and gtx 965m, i uved the cpu by 130mv initially (needed to reduce the uv to 110 when i oced the gpu), and oced the gpu by 100mv (allowing me to raise the clock to 1.3ghz, up from 950mhz) so yeah, i'm always tweaking this baby

Also, stock TIM job