Why does the 8250U require such a high voltage?

@Falkentyne, run it at the lowest possible voltage, you can probably do .940V or below.

 

That's interesting. When I try to run Prime95 + FurMark on my dGPU my CPU throttles down to 20W intermittently because of VRM throttling. I mean I'd prefer my VRM to throttle than to entirely shut off the laptop.

But it's better than Dell's VRMs

 

You can't increase it in XTU?


What about in RW-Everything

 

Boy you really want me to reinstall windows today huh?
You talking about 3.4 ghz?

I know the absolute limit for 4.2 ghz only. 4.2 ghz @ 1.08v will crash on FMA3 prime95 but pass small FFT's AVX disabled if I recall.
4.5 ghz @ 1.178v will pass AVX disabled small FFT prime95 forever, run for awhile on AVX enabled but eventually drop a core error (randomly, sometimes it can go 30+min) but this is an unrealistic load, so its rock stable 24/7.
4.4 ghz @ 1.178v will randomly drop a core in FMA3. At 1.182v, the EC will randomly decide it's had enough and shut the laptop off.
4.7 ghz @ 1.275v will run prime AVX disabled for hours but sometimes a WHEA correctable error will appear in windows event viewer.
If I enable AVX and press "Go", the EC INSTANTLY shuts off the laptop. Instantly.

 

I'm not going to touch XTU. Apparently it messes with the fan tables and with my current BIOS/EC the fans are completely silent at idle. I'm not going to throw that away! Here's a screenshot of TPL window.

I'll admit that I have no experience with RW-Everything. Do you have a link to a guide?

 

Changing TDP in RW Everything will cause system service exception instantly.
Changing Power Limit on HQ processor above 45W in RWE will be ignored.

He has to change IMON SLOPE in Bios to under 100. A value of 50 should let a 7700HQ use about 70W before it throttles.
But he needs to unlock Bios menus and hope it has this setting in CPU VR Config.

 

I'm not going to mod my BIOS. I have a three year warranty on my laptop and I'm not throwing that away. Aren't the tests with max power consumption and the Blend preset not enough?

 

I know, my reseller used Ek-TIM ectotherm (performs like MX-2, I'm told) and I have a tripod heat sink design. I've been meaning to repaste with Gelid GC extreme, but I haven't got around to doing that.

 

If you are happy bowing down to your manufacturer's potentially insufficient cooling/power components, sure.

Let me tell you this. It is unlikely that reputable software in Windows will permanently change your BIOS settings. Anything you do in XTU should be able to be resetted. I have never, ever heard of XTU modifying fan tables. I have absolutely no idea where you got that notion. As far as I know, fan speeds are controlled by the EC. on my laptop, howere, the fan table (The one that sets the temperature to fan speed) is in the ACPI SSDT.

 

Sorry @Falkentyne, don't want to corrupt your windoze. let me run run it.

 

@Vistar Shook
3.4 ghz @ 0.9v (900mv) prime95 small FFT.
Didn't go boom.

 

The power consumption at complete idle hovers between 0.7W - 1.5W, although it's usually under 1W.

I'm using the balanced preset with default settings.

My system is pretty slim (19mm) and there are virtually not options in the BIOS other than the usual boot order, Fast boot, virtualization and TPM settings. I seem to remember Prema saying that he didn't make a BIOS for my laptop model because there's very little headroom for tweaking.

 

if that is what you claim then you've got to first prove it, then let people know the settings and under what conditions so we can test the same for you. though i dont think thats happening at all, because what you showed is a peak minimum wattage. your average idle wattage may not be near as low as what you claimed.

 

@Che0063 created this thread to ask why his U CPU uses more voltage than a HQ. The reason is binning, where the U cpus are lowest bins. He did an amazing job bypassing the imposed limitations of his cpu, to have it peform unleashed, where he noticed the voltage and watts compared to an HQ. In terms of binning K>HK>HQ>U, but you want to determine if the HQ is more efficient at it's stock clocks and than a K running at those clocks. Not sure if it is the case and it would be interesting to see.... but really it doesn't matter, because binning is to select the chips that clock higher and not lower. Either way, I don't have a K, but a HK.
This is the blend test with AVX/FMA enabled on mine at .945V, so it runs at lower voltage, but my power package seems to be a few watts higher than your HQ.

 

@Vistar Shook
3.4 ghz FMA3 small FFT, 900mv. FMA3/AVX boosts voltage by 30mv according to some post on Overclock.net



Ok now that this thread is derailed, can someone please hack the EC to change the CPU Current trip flag from 1 to 0 without bricking it?
please.
I want to try 4.7 ghz AVX small FFT without the laptop SHUTTING OFF IN 0.01 SECONDS. i am not kidding.
I attached the EC. Someone in the world must be smart enough.

Or maybe someone can beg a MSI engineer to let the #1 7820HK in the world "Borrow" the unlocked EC ??
I attached the EC you guys.......

Go ahead and look in a hex editor.
you'll find it

CPU_CrtT = 1.CPU_ThtlT = 1.CPU_ThtlT = 0.SYS_CrtT = 1.SYS_ThtlT = 1.SYS_ThtlT = 0

And changing 1 to 0 with HxD will violate MULTIPLE CHECKSUMS on the encrypted image. Unless someone has a spare GT73VR system sitting around and a SPI programmer....you're going to have to decrypt it with Linux....

 

So my HK and Falkentyne's runs at lower voltages, but package power seem to be a few watts more. Now for binning comparison, perhaps CB15 runs at the HQ clocks?

 

Here you go:

The peaks in the graph are notifications I'm receiving. Y scale is from 0 - 2W

 

indeed i did win it because i paid for it its an 5.2 SL binned chip unfortunately, im on travels atm so cant sit down and properly test this. when i do get around to it, ill provide an update

Sent from my Xiaomi Mi Max 2 (Oxygen) using Tapatalk

 

Wait....what are we comparing now? Idling wattage? Oh...who cares. To compare voltage, watts under load is somewhat interesting, but at idle, which cpu uses less power, really? Let me find my son's tablet to compare.

Enviado de meu Pixel 2 usando Tapatalk

 

no wait, how about my phone forced to lowest clocks and voltages via kernel adiutor? xD

600 mV?! can the core go THAT low? dayum! tbh, ive never bothered to check, for obvious reasons *lol* but would be interesting to see what the max. stable clocks are that i could reach with the minimum voltage

its funny to see how this thread got derailed in like ten different ways but this race to the bottom is definitely something new haha!

Sent from my Xiaomi Mi Max 2 (Oxygen) using Tapatalk

 

It does matter if you care about battery life. I'll quote myself, from a previous post:

Comparing an ARM chip to a x86 chip is a bit of a fool's errand. Can your son's tablet run full Windows 10 with productivity apps like Word smoothly? I doubt it could even run Ubuntu compiled to AAarch64 smoothly. A cortex A75 core has far, far worse IPC than a Skylake or Ryzen core.

 

The only reason laptops with LGA don't have equal good battery life vs the BGA junk is due No Optimus.

 

Of course I was joking about the tablet.....just that if one wants to compare idling wattage and battery life then sure, go for U cpu or Y cpu and integrated graphics, optimus etc or even the new ARM windoze laptops. My RB gets amazing battery life compared to my EVGA turdbook. But on both, I am plugged in when gaming and such. There is a place for all types of cpus according to one's usage, although many in this forum might disagree, but in terms of performance there is no doubt where they are in the hierarchy.

Enviado de meu Pixel 2 usando Tapatalk

 

I don't recall if @RampantGorilla gave details about the laptop being used, but this is a well known issue with XTU and Clevo laptops - http://forum.notebookreview.com/thr...00k-and-sli-1080s.802786/page-2#post-10521192

There might be a work around by modifying one of the .ini files before installing XTU; otherwise, you're fan tables are all messed up... at least until a re-installation of the operating system.

 

I'm using a Clevo P950HP6, it's in my signature. I used Obsidian PC's Driver update utility to install the modded Control Centre without the XTU BS. Thank god for that piece of software, it's a marvel.

 

Whats the battery life on the P950 like?

 

theres something seriously wrong with screen shot u have taken. the readings might be from something else on the cpu/mobo itself though im not sure what it is,

your FID reading shows the frequency, 1.3w at 1.7ghz for 4cores 8 threads? intel is not god and even if they are the best chipmaker this isnt something they can accomplish yet. if we can do 4c at 1.7ghz for just 1.3w then we dont need intel setting those TDP value at 45w because 45w would mean we can easily do 5ghz 4 cores and still be under 45w. your machine is fked yo.

this also reminds me, my xpsm2010 with hwinfo showing chipset temp reading but in fact it is actually my GPU, the sensor in your machine is pretty messed up and hwinfo is picking it up.

 

How about 3.4GHz 4C/8T and 0.8W but 1.15v?

Clock speed is just about irrelevant to power consumption now. What determines it is the C state. A 'clean' system should remain >1% at C0, with over 95% of the cores in C7 and at least 80% of the package in C7/8/9.

A power efficient laptop doing absolutely nothing should draw no more than 5W. Laptops without optimus will use up to 10W. laptops with U CPUs should draw no more than 3.5W on the minimum brightenss setting. Turn off the display, a well engineered laptop should use no more than 2W. You have a lot of optimization to do @RampantGorilla

 

some of these readings are pretty wrong, even with optimus, or maybe because optimus is there. if cstate box is checked, while not underload and frequency not showing what its suppose to show, then that is not cstate which means readings are also off.

maybe @unclewebb can help with some answer how readings work.

 

Ole, you just ran into yet another cancer problem by ODM's and Micro$haft.
It's the C states as others mentioned.
On MANY laptops, even if you activate and unlock C states in the Bios, the chips won't go any lower than C3 state.
A CPU entering C9 or C10 state can consume that low watts at idle. But even Unclewebb has never seen a CPU enter C10 recently, at least not on windows 10.
And many can't go below C3.


BTW, i know I keep begging but since I *did* get #1 on HWbot for 7820HK, are there ANY EC modders or people able to disassemble EC firmwares on this forum? (Besides God Prema, who unfortunately cannot help with personal requests)?

 

yep, thats why those chip are very sluggish and requires speed shift, even with speed shift though the idea is nice, in reality it is just as cancerous. more feature is great if they actually worked and dont conflict with existing features.

its another reason why i wanted x299 in laptop with their turbo boost max 3.0, if it worked correctly.

 

Well I imagine it would be *ahem* as simple as dumping the BIOS, modifying a few values in the EC, and then reflashing the BIOS via a SOIC Programer+Clip.

I beg to differ about the CPUs not going under C3. Most laptops do. That's because they are still running stock Windows images from manufactures, which have included the appropriate ACHI controllers (or something along the lines of that.) Its not until many enthusiasts on this forum do a fresh install of windows, that they lose the manufacturer drivers which enables deeper C states. Nobody can blame Microsoft for doing this. At least they have drivers that actually work for the system. Remember Microsoft has to provide support for thousands and thousands (millions?) of system configurations. THey could just completely jettison support for older hardware (Which is what they are doing with Andromeda, Polaris (Core OS), because Win32 is not the way to head into the future) but then people would outrage and blah blah.

 

@Papusan My friend just ran Cinebench. He did it with a -125mV undervolt and no background tasks.



Look at that. My CPU is equal in performance.

But more interestingly is the powers and voltages. He sustains 3.4GHz at a mere 0.88v. Package Power is almost the same, which his averaging 33-34W and mine 30-31W.

So its clear that the 7700HQ uses more current...

But look at that cooling though.

 

can your friend undervolt it further?

 

I could hex edit the value right now. But without knowing the type of protection MSI uses on the encrypted or decrypted image, there is a 99% chance of a nice big instant brick.
If this were a GT75VR, I would have already bricked it, because the Bios chips and EC chips are right next to the video card fan. Easy to take a Pomona clip to it without disassembling anything.
But on the Gt73VR, the bios chip is under the MXM slot, blocked by the video card (yuck), and the EC chip (assuming the 128k flash is stored on that PLCC8 standard sized chip) is BEHIND the motherboard (assuming that is in fact the EC stored there and not a 'backup' Bios chip), which means a complete teardown would be required to access it.

That's why if someone, anyone, had a 'test bed' GT73VR and could test that, then I wouldn't have to rip apart my system. And I don't want to ruin my almost perfect heatsink sanding and LM repaste job. When you have a metal bar in your spine, any work like this is EXTREMELY painful and difficult. I don't think anyone here realizes how much I injure myself just servicing my own laptop....

 

5 hours with tweaking. If Clevo had put a 80/97wh battery instead the 55wh battery in my laptop, I wouldn't have to bring my charger with me.

 

@Che0063

What tweaks should I do? I've disabled my hard drive, turned off my screen and even disabled my GPU in device manager. None of these changes reduce power draw to under 10W, it usually hovers between 12-13W at idle. I'm not sure that optimus laptops use under 10W on average when idle, take a look at the gigabyte aero 15:
https://www.notebookcheck.net/Gigab...op-Review.215364.0.html#toc-energy-management

None of these laptops have average idle power consumption under 10W. Not even the Macbook Pro 15" 2017 has power consumption figures under 10W:
https://www.notebookcheck.net/Apple...op-Review.230096.0.html#toc-energy-management

 

Yeah my RB has 74WHr and I can get over 6 hours with light use, but it has optimus. My EVGA, with the same size battery, gets around 2-3 hrs but has G sync.

 

Stop. Notebook Check does not optimise their reviews to test their battery.


That may have been the machine under stock. But now at idle my laptop uses 4-5W. 3.5W if I turn the brightness down. I have no idea whether their brightness was set to max, but if it was there was no way the idle minimum could have been 3.7. A power efficient laptop (With U CPU) should use 5-8W on light browsing, 5W word processing, and >5W idle. Divide your battery capacity by these figures and you get the battery life you should be getting. I have no idea about macbooks, but all I know is that their 99Whr battery only gets about 10 hours. Well guess what, my 60Whr battery gets well over 10 hours as well, often peaking into 12 hours. That's why I charge my laptop every few days when I use it at school.

Many people here have the 7700HQ, since that seems to be the absolute baseline minimum CPU required for gaming (it really isn't) but once you optimize windows, you should get your package power consumption to IDLE at 0.3W. Considering many laptops spend a lot of their time idling, that can improve your battery life.

And no no no no do not under any circumstances disable your dedicated graphics card in Device Manager. Disabling your graphics card will disable all its features, including Optimus. Optimus usually keeps the dedicated GPU off when not in use. If you disable optimus you effectively turn the GPU on but leave it idle. I find in my 945M/920M previous GPUs they hogged an extra 3-4W on idle.

What program are you using to measure power consumption? I use BatteryInfoView since its very light on system resources. What people don't realise, is that with a laptop, every single milliwatt counts to improving battery life.

If you don't need that extra hard drive, remove it. If you do, set the APM to spin the drives down (But if you do this please please ensure you don't have any programs that run on the hard disk drive, they will spin up aggressively, not only putting strain on the hard drive motor but wasting A LOT of power.) FFS three of my friends had SSD+1TB HDD setups in their laptops and they never even realised they had another extra 1TB of space.

They know not very much about computing (I'm putting it nicely there ) but even with 20 minutes of tweaking their devices I got their laptop power consumption down to 6W idle. Their configurations are 6700HQ/7700HQ, 960M/1050 Ti/1060, 16GB Dual Channel DDR4 RAM, SSD+HDD, + 15.6" Screen., two have the newest Acer Aspire V Nitros.

Oh and look at that. Notebook check put the idle average at 10W for the Aspire V Nitro. Guess what? I got the exact configuration to idle at 6W. Check your GPU - is it off when you enable it? If it is off, you should see either 0.0C or -.-C or no temperature reading at all in OpenHardwareMonitor and ThrottleStop.

Maybe I should just create a guide on improving battery life.

 

I'm aware that disabling my graphics card will disable Optimus. I wanted to see if Windows would actually cut power to the GPU when disabled in Device Manager, it didn't. All laptops with dGPUs should come with a MUX switch in my opinion. As for my hard drive, I do use it frequently because my games and my movies are on it. I set Windows to shut down the hard drive if it has been idle for a 1 minute in the Windows power settings.

I'm 100% sure my GPU is off as it does not show up in HWInfo. I don't think that the Optimus cuts power to the GPU completely, because laptop with MUX switches tend to get much better battery life when only the iGPU is enabled in the BIOS as a pose to when optimus is used. I suggest that you check out B0B's review of the Ideapad 720s 15", which does have a (firmware) mux switch I used HWInfo and BatteryInfoView to monitor my battery discharge rate and they both report the same value.

I'm don't think that I can reduce the package power to anything below 700 mW. I did not win the silicon lottery; my i7 7700HQ can barely hold a -100 mV undervolt and runs at 990 mV @ 3.4 GHz under full load (prime95). Most people who have a i7 7700HQ manage -130 mV undervolts. I've currently set my speedshift EPP to 128, which provides a good balance between responsiveness and efficiency. I could set it higher, but then I'd have an expensive Intel Atom laptop on my hands. At idle, my CPU spends about 90% of its time in the C7 state (from HWInfo).

Are there any other tweaks that you think I should do? Oh and I definitely think you should make a guide about increasing laptop battery life, a lot of people here would find it useful.

 

Maybe you should! I'm sure it would be valuable to the community and highly appreciated!

 

Not so fast, my friend. You forgot one important thing in your statement... The power consumption rises with temperature (because of I2R loss). So if his 7700Hq run the same test but at higher Cpu temp, the power consumption will rise. From what I have read from your test, your processor run a lot cooler vs. him in the same test (65C vs. 84C degrees). And on top, none of you have a guarantee that the voltage reading is 100% correct. Mind you, two different noteboks and probably from two different brands.

 

Oops. totally forgot.
Well I guess we'll blame Acer for being unable to test this then

 

What did you use to do your logging and graphing? HWiNFO?

 

I used ThrottleStop and HWinfo.

@Arondel @RampantGorilla

Did it.
http://forum.notebookreview.com/thr...e-on-windows-enabling-deeper-c-states.815602/

 

.8w is low but the 8550u idles around .2-.3w actually so for a 7700hq that is pretty awesome actually. I tried doing that with a 3840QM and could only achieve 1.2w

 

Your 3840QM is Ivy Bridge. Idle power draw went down (and battery life went up) in a big way starting from Haswell.

 

C states - Cores going down to C7 states - normal. Package going to C8 - also normal. But then why is ThrottleStop reporting a limit of C9 for the package? Very rarely do CPUs ever go down so far because of lazy manufacturing. Is there any other software that can verify my C State limit? I'll try digging through HWInfo
Edit: Cache goes down to 6 on idle.

0.4W for the package is the best I can get down to. Any tips to get ever lower?

My experiences -

Haswell - i5-4210U 0.6W min power draw
Broardwell - ?
Skylake - i5-6200U and i7-6500U - 0.4W min power draw
Kaby-Lake-R (8th gen) still 0.4W min power

How is the 8550U drawing only 0.2-0.3W? I'm just curious, since my 8250U package draws 0.4Wat the minimum. My cache draws 0.3W, and according to Throttlestop the ratio is 8, which is as low as it goes


I mean I can definitely seet hat happening - the 7700HQs on my friends' computers idle at a constant 0.3W. Mine idles at 0.4W but with an average of 0.5W

 

I was just going to fire up TS and send you a screenshot and mine too is at .4w guess I lied or 1803 increased power consumption for me a little bit. What's funny is my voltage is at .53v and yours is much much higher at .69 yet our wattages are exact.

 

Voltages don't really matter on idle. How did you get your 8550U down to .53? For me, that would be a -400mV undervolt! I suppose 8550Us are just binned much better. For me, the 0.69v is with a 200mV undervolt. With a -200mV I have to disable Turbo to get the CPU stable. Anything above 1.8GHz requires the CPU to be dropped down to a -130mV undervolt.

I would agree that 1803 is not great for battery life. Microsoft has added many services. As a student who has multiple devices, I really need the future updates such as TimeLine/Sets, and the default Mail/Calendar apps are really useful for me. Otherwise, I'd downgrade straight to Windows 7.