6700hq all cores at 3.5ghz possible using throttlestop?

I recently saw someone say that they used throttlestop to increase their 4 core clock speed to 3.5ghz from 3.1ghz using the turbo ratio limits and setting them all to 35x. I have a Alienware 17R4. I tried this and ran OCCT and my clock speed never went above 3.1ghz according to HWMonitor. https://i.imgur.com/xcnrSDD.png <-- this is what the settings were originally. Was this guy lying and the laptop is locked in at what it is regardless of what throttlestop says?

Also, Ive not seen it ever run at 3.5ghz despite having cpu performance mode on. I use OCCT and set the threads to 1 and it stays at 3.1. Am I doing this wrong?

 

From what I know, the 6700HQ is hard-locked at 3.1 for 3 and 4 cores, 3.3 for 2 cores and 3.5 for 1 core - I do not know any way to bypass that using traditional means such as throttlestop.

 

Some people do not understand ThrottleStop or their CPU. You can go into the ThrottleStop FIVR window and you might be able to set all of your turbo limits to 35 but this is only a request. A retail 6700HQ is going to ignore that request and will only give you the 31 multiplier when 3 or 4 cores are active.

http://www.cpu-world.com/CPUs/Core_i7/Intel-Core i7-6700HQ Mobile processor.html

To make sure the 1 core active, 35 multiplier is working correctly, try running a 1 thread test of the built in TS Bench test. If ThrottleStop is not reporting a multiplier higher than 31 then that usually means that you have disabled your C States. On a locked CPU, you need to have at least C3 enabled so your CPU can use the maximum multiplier. Not all monitoring software can report the highest multiplier correctly but ThrottleStop should. You also have to realize that while the TS Bench test is running, Windows background processes will be randomly waking up additional core or cores which immediately reduces the maximum multiplier. ThrottleStop uses high performance timers to calculate and report an accurate average multiplier using a method recommended by Intel. The less background processes you have running on your computer, the closer you will be able to get to the 35 multiplier.

 

i just wanna say, it's wonderful that you guys are being active here.

Idk how i lived without TS for so long.

 

@unclewebb: I have a doubt that bugged me even till this day, Intel states i7 6700hq doesn't have TSX but when I ran CPUz TSX shows up. Why is that?
One feature request: Can you link CPU core and Cache offset to change at the same time during undervolt. I mean those who used TS will find it, but newbies doesn't know what they do and end up with XTU which feels heavy by sucking up huge resources for the work it does. I feel TS is superior and ALL OEMS must use your software.

 

Guess what? I added that feature to ThrottleStop a while ago and I was expecting that everyone would be happy. Then I found out that there are some Skylake and Kaby Lake CPUs where these two voltages can be adjusted independently. The CPU Cache voltage can be adjusted lower than the CPU Core voltage and users proved to me that this works. Core temps went down so the CPU Cache must have been getting less voltage. XTU does not let you do this so when I removed this feature from ThrottleStop, users asked me why and the only explanation I had was because XTU does that so it must be right. After seeing the testing, I back tracked.

The other thing I recently noticed is that the latest version of XTU that I am using only allows voltage adjustment in steps of 0.005 V. This isn't accurate either. Voltage has always been adjusted and continues to be adjusted within the CPU in steps of 1/1024. Intel decided to simplify things and got rid of access to the intermediate voltages between the steps of 0.005 V. No more following the competition for me.

Intel documentation is not perfect so maybe their ark site is not accurate.

http://ark.intel.com/products/88967/Intel-Core-i7-6700HQ-Processor-6M-Cache-up-to-3_50-GHz?q=6700hq

CPU World shows that the 6700HQ supports TSX.

http://www.cpu-world.com/CPUs/Core_i7/Intel-Core i7-6700HQ Mobile processor.html

CPU-Z is usually correct. When you run the CPUID instruction, the CPU reports back to software all of the various features that it supports. Unless there is a bug in CPU-Z, it is likely that the CPU is telling CPU-Z that it supports TSX. Send me a message next week if you would like me to look into this further. It takes a while to hunt through the Intel documentation mountain. After that I will try to write a simple program that you can run to see if it agrees that TSX is available.

I don't mind if the newbies start out with XTU. People should try both programs and use whatever one gives them the best performance.
Some people like colorful graphs. I prefer maximum performance.

 

I got some Intel Docs, I dunno I can execute the same code using C/C# on Visual Studio, maybe you know better than me by looking at the assembly code mixed with some C codes https://software.intel.com/en-us/ar...he-4th-generation-intel-core-processor-family https://software.intel.com/en-us/ar...he-4th-generation-intel-core-processor-family

 

Intel never makes life easy. Software should be able to check a single bit but instead that code looks like a mess.

The interesting thing I read about TSX is that Intel decided to disable this in some of their processors due to bugs. A microcode update was released which can disable TSX. These updates can be part of a bios update or they can be released along with Windows updates.

 

@unclewebb:
So, they should have initialized the values and computed directly instead of taking variables and complicating stuffs at the end.
I check TSX is enabled or disabled, after every BIOS update from Dell. Till now, they haven't disabled it. Good thing!
It is disabled on Linux after microcode update though, that's what I got after running lscpu.
After all, TSX is helpful in Servers that does parallel R/W to database at the same time, ther it might be helpful.

 

I`m very satisfied with TS. It got my cputemps in wither 3 down from 87 til 76 There are a few things I wonder about: Is there a point of undervolting analog I\O and system agent? I havent tried undervolting cpu cache more than core so will try that Also I noticed that CPUID often reports 3500 MHz on all cores while gaming, must be an error since you say it isent possible. CPUID reports -4 mv lower voltage than TS.

 

I recommend you stick with same values for CPU,cache, S/m agent and Analog IO.

 

Tbh, I never bothered undervolting analog and system agent. According to people it makes almost no difference.


Sent from my iPhone using Tapatalk

 

Lower voltages mean less heat.

 

Best way to reduce the heat is with a garden hose /s

 

If he ain't on high oc, I would put lower voltage on Cashe vs. Core and see how it works out.

 

No OC possible with 6700HQ

 

6820/7820hk ain't so much better But still, you can or should undervolt so you can hold Boost a longer time.

 

That's not on all cores. That's because the load was low on one core, so the other cores went up. Plus if it's not updating in real time, it probably showed all the cores at 3.5 ghz, which is impossible. use Throttlestop to see what your real speeds are).

 

Okay. That makes sense

 

After tinkering some more get core stable at -169 and cache at -198. That was quite a difference! In witcher 3 I usually get 86C on CPU after an hour of playing with no UV ad 78C on GPU. With -169 on both I get 75C on CPU and 76C on GPU. With -169 core and -198 cache I get max 73C on CPU and 75C on GPU I have not tinkered With system agent and I\O, but will try, maybe another degree or 2?

 

Interesting and quite odd. During my first experiments with TS beginning of this year, the processor seemed to ignore my settings, if these two voltages were different. Now it works ...

True

Every CS student should see this design & code as a warning example

Why ?

Wow, this is nice.

Intel seems to set SA to the same as the graphics, but I also didn't notice any differences here ...
(I'd guess it'd be better, if everything is undervolted in a similar fashion, to avoid higher difference voltage between the parts, but as long Intel allows separate settings, it should be irrelevant)

My CPU has a maximum of -179mV but only for CPU load (e.g. prime95).
If I start a 3D application, it crashes / BSOD.
Only -145mV is safe.
(Graphics & SA are at -125mV)

So I'm using 2 profiles, ("3d unstable") maximum as the default,
and I need to remember to switch to graphics_safe if I want to use the nvidia GPU

Did you experience something similar ? Can you undervolt more in prime95 load only ?

 

My UV og -168 core, -196 cache is stable in everything. I tried UV I\O and then system crashed in games when at -140, with -110 it works perfectly. Will try systemagent NeXT, got it at -100 now. I saw when running prime95 the system used a maximum of 35W total. Without UV it uses about 10W more. After 30 min of stresstest With Furmark and Prime95 CPU got to 83C and GPU 76C. Without UV CPU went to 94C and GPU 80C so there is a big difference!

EDIT: CPU maintained cores with UV, without it throttled down a few hundred MHz. Also GPU was stable at 1240 to 1320 so slight throttle there.

 

Don't use Furmark and P95 for extended periods because it ruins your cpu and gpu. Use OCCT, AIDA64 to test CPU and GPU.

 

I really recommend experimenting with this, it makes a huge difference!

 

Yeah, I`m aware, the only time I used without UV was a year ago to test temps. The temps I got now seemed safe by a big margin. Witcher 3 gives same GPU temp and higher CPU-temp With no UV. I will probably never use any of them again, but thanks for advice

 

Well known firmware/vBios modders have posted warnings about use of Furmark!! Bad enough on desktop, but laptops is worse. FYI... Your graphics card haven't temp sensors for the other components outside the graphics cores!!

 

Ah nice results !
Yeah, first prime95 tests were at almost 60W !!
Now it's nice with about 38W. (@ -179mV VPU)
I'll test with a lower cache value as well.

Totally agree here !
Funny thing: The German c't magazine wrote an article, that benefits from undervolting (for modern processors) are a myth:
https://www.heise.de/newsticker/mel...ren-durch-Prozessor-Undervolting-3697442.html
There was ofc a backlash in the comments section. (The auther also refined it ofc, b/c many users use static voltages, that's ofc stupid. And his main point was, that modern AMD cpus have some sort of their own integrated UV logic for each unit. I don't know the details about this, so I can't make a statement about these CPUs ...)

But I'd say 40W instead of 60W and 10-15° lower temperatures on my i7-6700 are quite real !

Now to the part, which caught my attention:

Ehm ... why ? (Or how exactly ruins it your cpu and gpu)
And how do you define extended periods ?
(Is this also true for UV CPUs like mine, which is even during P95 lower than it's TDP now ?)

I'm just really curious.
(I used these tools maybe 10 mins for P95 and 20mins for furmark (in 30-60 sec chunks to test max power and correct computation, so I probably/hopefully didn't burn my NB )

Could you write more details about this / some links or for what to search exactly ?
For the most part, I just find threads from 2010-2013 with the last in 2015 http://www.tomshardware.co.uk/answers/id-2637189/running-furmark-harm-gpu.html
Seems AMD and nvidia both implemented protections in the drivers, after problems with the 5 & 6 gen of their cards ...
So if you use stock settings with furmark and the drivers, is it still bad ?

 

Regarding overide voltage... From an known oc'ers viewpoint.

Statement from one of the best firmware/vBios modders we have today, @Prema
You find more if you search in this forum or on the web. @Phoenix

 

Looks like your CPU isn't limited by power and firmware. I tried my max to make the cpu reach 45W but it stays below 40w even without undervolt.
Protection in drivers of nvidia is disabled by default for furmark and kombustor.

 

One thing I noticed in Furmark was that clockspeed of GPU was hovering around 1240-1340 MHz even when at 60 C. In Witcher 3 GPU often reaches 74C after a couple of hours and Turbo stays around 1670-1750. It seems like there is a Furmarkprotecton somewhere in the system. Furmark never exceeded 76C on GPU, thats very Close to W3, but the clockspeed is sh*t in FM.

 

Ok, got it:
It was in the TS TPL settings: Package Power short is set to 55W (and that's actually, what Aida shows ...
So it fits perfectly.

That means, undervolt is really awesome, and basically helps to protect the cpu in these crazy scenarios.

I always thought, this warning message at start of furmark was for people with bad cooling b/c then devices could overheat. And I thought, the CPUs / GPUs had internal protection against overheating, most of them actually have:
I could even remove the cooling fan of my old Core 2 Duo E6700, and the system would just become super slow and the CPU throttled.

My current 960m wont pass 82° even during Furmark and GPU-Z shows thermal throttling.

I'm really puzzled, why the manufacturers don't put out clear statements, not to use these tools. (<tinfoil hat mode>Ok, maybe these tools wont kill the hardware instantly, but just reduce the lifetime after the warranty, and people have to buy new hardware sooner * conspiracy theory* </tinfoil hat mode>)
But for the moment, the info is quite convincing, and I will avoid using these tools.
Thank you all for pointing it out !

 

Yeah, for the core(throttling algorithm). But the graphics cards PCB contain more components than the graphics core!! People forget this

 

From where do you have this quote ? I couldn't find it with google ...

So anyway, I think there are 2 different points here:

Sure, when *overclocking* other rules apply, and you'll just need to find & set the minimum stable voltage for the maximum OC you want. (At least as far as I know).
The Auto Voltage is not optimal / will get far too high during OC. I saw this for myself while OC of my i7-920 back then, the auto voltage for idle was even higher than my (constant) OC voltage (in the light profile) ...
And I think, this stuff is more interesting for desktop/workstation computers in general (as it becomes more relevant the more power consumption happens during load phases).

But in the c't article the author was writing about *undervolting* explicitly (and modern (!) CPUs). Sure, the task is basically the same, to find & set the minimum stable voltage.
But I'd say, your main goal here is to reduce the overall power consumption, this means the idle voltages have to be as low as possible as well. I think this is most relevant for notebooks, and especially in the cases idle, writing/office stuff & browsing the web.

Maybe gaming notebooks are a bit different here (as they are somewhere in between)
I just tested one Asus with i7-7700, and it crashed at the following offsets:
idle@-160mV
and
load@-220mV

So sure, as expected, this stuff is not constant over the whole clock / load range. (And I guess the idle -160mV get probably too close the physical limit of the transistors ...)

And the load case (during gaming, etc) is more similar to the first (workstation) case.
So such hardware & these use cases will probably require 2 profiles:
A "mobile" profile, with maximum possible idle UV and (auto voltage settings)
vs.
a "gaming/performance" profile, which sets a constant voltage (minimum stable V for maximum turbo clock).

 

Skylake – Overclocking – Power Consumption and Voltage Scaling