MAJOR discovery! CPU throttling to low speed seems NORMAL when hybrid power unavailable (normally when battery under 30%): https://forum-en.msi.com/index.php?topic=268588.0
JUST found that. So the hybrid power not working might be the only problem we have!!!
@Pedro69, @streetunder, @Falkentyne
-
Falkentyne Notebook Prophet
Ok then I need people with GT72VR's (or GT62VR's) from North America to please post whether their laptops drain battery while gaming at extremely high loads.
BONUS POINTS would be given if someone posted the screenshot from the EC (Embedded Controller) tab in RW Everything on a north american 7700HQ GT72VR.
I know where the NOS (battery boost @Papusan) flag is on the GT73VR. I know how to disable the NOS flag, but this causes a 160W AC power limit. Bypassing 160W (which would normally trigger NOS battery boost) throttles the CPU via POWER LIMIT 1/2, not via BD PROCHOT, however if you bypass 160W by a HUGE amount, the CPU gets throttled to 1000 mhz, even though the power draw is reporting 25W, it's drawing WAY less than that. I assume this "system" is different than the GT72VR since this is an unlocked CPU that has a built in 45W TDP to begin with, so PECI power limit restrictions are not required to override "BIOS" power limit restrictions.
Fun fact, if anyone cares.
On the HM175 "GT73EVR" system, if you unlock the Bios menus and set PSYS_PMAX to 2700 (337.5W), and put ANY load on the GPU, the CPU gets POWER CUT to 800.0 mhz.
WITHOUT BD PROCHOT BEING TRIGGERED OR ANY POWER LIMIT WHATSOEVER. NO power limit flag, NOTHING. Just the CPU running at 800 mhz with NO flags being shown!!! It just says.....800 mhz. I am NOT kidding.
If PSYS_PMAX is set to 2400 (300W), then it doesn't happen.
ON THE CM238 CHIPSET GT73VR, or if using EC 17A1EMS1.108 or 17A1EMS1.109 even on teh HM175 chipset version, you can set PSYS_PMAX=2700 and it will NOT throttle at all.
Why?
Easy.
The CM238 version EC has to allow for SLI configurations, which can draw 230W * 2 = 460W.
The HM175 GT73EVR (which 17A1EMS1.112 EC is based on) does NOT have a SLI configuration available, so the absolute max limit is 330W.
I have not seen ANY restriction like this on the CM238 EC. Even setting PSYS_PMAX to 8191 (1023W), does not cause CPU throttling at load, even though HWinfo64 reports the power draw is over 600W LOL.Last edited: Mar 8, 2018 -
Falkentyne Notebook Prophet
See my post i just wrote just now when you were writing yours.
For GT73VR and GT75VR, it triggers 45W TDP power limit override when 160W of system power is exceeded, overruling Bios power limits (PECI has highest priority. Priority level is 1) PECI, 2) MMIO, 3) MSR (Bios).)
If 45W is not enough, it then cuts it to 25W. if 25W is not enough, the EC "Cuts" CPU cycles/multiplier levels downwards towards 800 mhz.
EC RAM register C6 changes from C0/C1, to 40/41 when NOS is disabled.
Changing the value of EC RAM Register C5 (normally a 9E on all MSI laptops) to between 00-80 (hexadecimal values), DISABLES NOS and forces EC RAM register C6 to 40/41 again. Setting a value in register C5 from 81-EF, ENABLES NOS and EC Ram register C6 gets set to C0/C1, meaning full power is allowed. Setting a value in register C5 from F0-FF causes NOS to enable and disable itself each second, with C6 changing repeatedly from C0 to 40 and back and forth, which seems totally useless.
I can not find anything significant about the value "9E". Maybe i didn't test enough, but I saw no difference at all between using 81, 9E or EF. Maybe I didn't do enough work. Maybe I'm just lazy. -
@Falkentyne, I also have unlocked cpu (6820HK) in my gt72vr 6re. Rare model ^^ Thermally totally capable of handling it even overclocked. Interestingly there is no separate support page for it or EC firmware. I will buy the kill-a-watt thing (if they sell something like that here) before picking the laptop up and see if it still throttles, at which point it is. If at 160W, then we know it's because of hybrid power disabled for some reason >.< I know my system should be able to pull over 200W easily when fully loaded which is way beyond the 160 (based on benchmarks I have seen on notebookcheck). I totally agree that it affects other laptops differently. I suppose due to my cpu being different I can also trigger the problem easier.
I will also download the rw everything and check if hybrid power is truly not being triggered.Pedro69 likes this. -
Falkentyne Notebook Prophet
If Hybrid Power is being used "past" 160W on your system, you should see EC RAM Registers 46 and 47 have values in them that are not "00, 00". 46 and 47 are the battery drain or charging rate registers.
You don't need a kill-a-watt.
You can just use available tools directly. Wall power will always be overreported because PSU inefficiency it will be drawing more watts than the system is using, and releasing extra watts as heat.
1) GPU will use 115W. You can use GPU-Z to show this.
2) Throttlestop 8.50 + "Limit Reasons"-this will show CPU power draw. Your CPU's base TDP is 45W, but unlocked SKU allows overriding by the BIOS--IF THE EC ALLOWS IT.
Notice that 115W + 45W is 160W.
Monitor backlight+LCD uses about 20W at maxiumum brightness.
Then there is mainboard, DRAM and peripherals too.
You should post your EC RAM page when you get the laptop back (RW Everything--EC) so I can look at it.
EC controls PECI power limits. On GT73VR and GT75VR, if the "power ID" is unknown (in EC register E3), the CPU will be throttled to 45W.
It seems strange but at least I found a pattern:
EC register E3: FULL LOAD (prime95) testing:
Values 00-0F = 6820Hk and 7820HK will be throttled via Power limit 2 to 62W, then after 28 seconds, turbo will be disabled and will be throttled to Power Limit 1 flag, 45W, or max non turbo limit (2900 mhz or 2600 mhz, whichever is HIGHER in mhz...e.g. if 2900 mhz is using 55W, it will not reduce below 55W).
Value 10=default for SKYLAKE 230W ID (GTX 1070).
value 11=default for SKYLAKE 330W ID (GTX 1080).
Values 12-1F "seem" to use the last set "master" power ID value that was initialized (230W or 330W).
Values 20-2F do not make sense. The CPU will draw "Unlimited" Power for 28 seconds, and then will suddenly get throttled instantly to POWER LIMIT 1 (2900 mhz) or 45W, whichever is a higher speed.
Maybe it's trying to access the MMIO power limits, which are "unused" (empty) on the GT73VR, and then uses a default value. I do not have enough experience with using MMIO, either way, pretty pointless.
Values 30-8F seem to be the same and useless, afaik, triggering PL2 then PL1.
90=Default for KABY LAKE 230W (GTX 1070).
91=default for kaby lake 330W (GTX 1080).
92-9F function as 12-1F.
A0-AF function as 20-2F
B0-FF all seem to be useless and trigger PL2 and PL1. -
So, some old dragon center have the power consumption section but I still can not get a relationship between bd prochot and consumption ... but I have already been able to visualize values of 240W.
Bad thing is that i need play the game in Windows Mode.
-
@Pedro69 that does not really make sense considering the gpu is reporting only 71W of power at that time >.< There is no way you would reach over 230 with that, that just isn't right.
-
@heliada
But the values of power consumption should be right even if they do not match in the game. Also im using Dragon Center from MSI GT60 20D...but for some reason i dont see the icon of hybrid power...Last edited: Mar 9, 2018 -
Falkentyne and Pedro69 like this.
-
We might be onto something. How is hybrid power regulated... via windows? Dragon center drivers? SCM? Or is it hardware-based? If the system that regulates hybrid power thinks it is not needed (let's say 150W use), but the motherboard/EC think that over 240W is being drawn... would that not send the mb/EC into panic mode and possibly trigger bd prochot...? At least seeing how there is discrepancy between obvious power usage and one that dragon center reports in this case. But then again, could just be due to old/incompatible dragon center version.
-
Another pic with all usb on fire and game in Windows Mode, max gpu power in HWInfo64 was 143 W
-
Totally not right. Unless the power consumption would be trying to take power adapter inefficiency into account and calculate draw at the wall in some weird ways. I think even then it still reports way too much. This is your specific laptop: https://www.notebookcheck.net/MSI-G...ok-Review.190406.0.html#toc-energy-management and it does not get even close to even 200W under max load. No way it would go over 240. But it indeed seems that the older gen CPU (like my laptop) consumes up to 20W more during load, which again explains why I can trigger the problem so easy. No idea where these wattages are measured tho, if at wall or elsewhere.
-
Falkentyne Notebook Prophet
I thought I explained this already.
at least ON THE GT73VR AND GT75VR:
registers C8 and C9 are power consumption (does not seem to be in watts or at least in something not directly convertible except "higher=more"
@Papusan
You guys are making mighty Papusan happy. Giving Turdbook a new name: DeadVomitBook.
Register c9 will go to FF (highest value), then at overflow, register C8 will go to "01" and C9 will start at 00 again and keep going up. Then up to FF again, then C8 will change to "02", etc....
Register c6 says whether NOS is enabled or disabled. 40/41=disabled, C0/C1=enabled. 40/41 means no battery boost, and system will throttle at the "point" where battery boost would start being used (about 160W on a 230W system).
Register C5 seems to be some sort of master flag for NOS. 00-80 forces register C6 to 40 and disables NOS and forces 160W power limit. 81-D (something) enables NOS and you can draw full power.
But something funny happened again today. I set it to DF to mess around this morning at 4.5 ghz and no problems. So I rebooted to 4.7 ghz @ 1.272v and started running prime95 small FFT, no AVX, no FMA3. After a few minutes, the CPU started going from 4700 mhz to 4692 mhz then 4700 mhz then 4693 mhz repeatedly. I checked "Limit Reasons" and PL2 flag was being turned on in yellow briefly each second. So I opened EC RAM, and EC register C6 was changing from C1 to 41 every second again. Uh ok whatever. So I changed EC register C5 from DF to "DE", and then C6 went back to a value of C1 and stayed at C1, and the CPU remained at 4700 mhz during the next 2 prime95 tests without that toggle or stutter or PL2 thing trying to come on. Note that the default value of EC register C5 is 9E on all MSI jokebooks. I still don't know the "point" of the value 9E and what the hell it's for. Because 81 still has NOS enabled, but set it to 80, and register C6 changes to 40 (or 41), and NOS gets disabled again. I have found it IMPOSSIBLE to have full power without AC power limit throttling if register C6 is set to 40. completely impossible. Because that's internal code. However other things can force EC register C6 to 40 even if C5 is sets to 9E. Battery level below 30% forces C6 to 40. Unplugging battery cable also forces C6 to 40. It's possible to force C6 back to C0 (C1) with the battery unplugged, by setting EC RAM register 31 to "09" and EC RAM register 42 to "64".
The NOS "Cutoff" point of 160W seems to be completely hard coded in code. Can't access this and can't change it. It also seems to be set during POWER on time. GTX 1080 systems (330W AC limit) seem to have this at 235W or 240W instead of 160W, however changing the power ID to the GTX 1080 power ID, while it DOES change the maximum power allowed to 330W (no more CPU power limit 2 throttling if 230W is exceeded), it does **NOT** change the original NOS cutoff point of 160W, which is what I've been trying to change. Seems to NOT be in the EC RAM. Probably done in internal code, meaning:
-
Falkentyne Notebook Prophet
I was able to draw 220W from the wall with it. FAR past 160W. System probably using 180W and maybe 20W from the battery boost.
115W: GTX 1070 (PWR LIMIT)
30W: Cancer CPU 7700HQ
20W: LCD panel
5W: USB keyboard or mouse
Unknown W: Mainboard, disk drives, DRAM. (YES these use power too!)
And max speed fans use at LEAST 5W. And there's more than 1 fan too. Maybe 10W...yes fans use power....
180W system AC power usage is being a bit conservative.
And I still remember battery boost draining almost 10% battery in 1 hour when I was sitting there watching those dudes fight in Witcher 2 during that tutorial loop. I was not happy...Last edited: Mar 9, 2018 -
Maximum values in the third column -
-
So I just picked the laptop up. Could not test in store as they closed at 6 and it was 5:40 when it arrived. Well... I missed my laptop for 5 weeks for nothing and will send it back monday. Not only is there filth on the keyboard and about 5 warranty stickers stacked on each other, the original cardboard box is damaged on all corners, but the ONLY thing that was done was windows reinstall and .8A ****ing vbios installed. DO THEY think I accept that as a solution when I said I DID NOT WANT a limiting vbios installed? WTF. So pissed I have no words to describe it.
-
But Dragon Gaming crap can trigger things, so either its reading wrong registers/wrong scalings applied to the readings that would explain a lot of the problems, I still stay on my narrative, nuke windows install, reset NVRAM, load defaults on BIOS, clean install of OS, NO software from MSI.
I already said this X times, my clocks where all kinds of screwed due to Gaming Center and CSM or whatever its called today where both fighting to set power schemes...
The fact that my GPU would only go to full clock if I enabled the Cooler Boost implies that there is something wrong being done on either the software, the firmware, or both sides.. -
Falkentyne Notebook Prophet
I went so far past 160W it kicked my CPU in the nuts (Prime95 blend is running and 150W TDP GTX 1070).
@Papusan dont look. This is NSFW porn. You'll be hating LGAbooks after seeing this!
-
Falkentyne likes this. -
First it was with the damaged graphics and now this ... MSI really wants to lower is reputation.
Girl, I wish you all luck in your case. -
I am sick of this.... So disappointed. If I install the normal vbios back it will be back to bd prochot. More than month of my time wasted. I want my money back :'( I am actually literally crying. Waiting for it to charge fully to test how the ec looks >.>
@Pedro69 thank you. I am not buying from msi again. I will just build a nice pc with damn asus motherboard and gpu. **** this. And get myself a big suite case to transport the pc for when I travel xDPedro69 likes this. -
-
Falkentyne Notebook Prophet
@heliada please show me a picture of your EC RAM.
I asked people in GT62VR and GT72VR thread to test for battery boost and post their EC RAM But no one replied. I think no one cares -
Yeah it says 10 mins to full charge now, then I will test it. SS it and post it too ^^
-
-
All is inside the info file. Bios is .118 and EC .107. Vbios ends with .8A.
I am a bit concerned tho as it has been on 99% battery for the last 30 minutes. I mean, just my luck right. Maybe my battery is dying hahahaha. Anyway, here is the help desk file of MSI reset system (cause Norton is awesome).Attached Files:
Pedro69 likes this. -
-
Oh my god idk what msi did to the laptop but it takes like 2 minutes for it to shut down now >.> Seriously sigh.
-
-
Falkentyne Notebook Prophet
@heliada
I need a screenshot of the EC window, not the system dump.
Download RW Everything (it's very small), install it, go to the Embedded Controller tab, open it, then screenshot the window.
99% battery being charged slow to 100% is normal. Battery charge rate drops extremely low for "some" reason (can be seen in Batteryinfoview.exe) once it gets past 98%. This may have something to do with wear level being "negative" (e.g. fully charged capacity being HIGHER than designed capacity). Can't help with that, not a battery guy. -
I noticed. It is pathetic that the laptop was gone for more than 5 weeks for this. I am so upset. I don't even want it around now >.> I will just pack it in a box and drop it off at the shop where I bought it on monday. Cause for some reason msi does not work through the weekend and thus cannot call them to inquire about this poo situation (wondering if that comes through as stars).
-
Falkentyne Notebook Prophet
Just relax a bit and please take a screenshot of the EC window?
Please?
-
@Falkentyne, I don't have negative wear level, in fact the laptop never reached design capacity. It seems kinda laggy, not sure if due to the bloatware. Anyway here is the screenshot, not sure how it will help since the battery is being charged now. I will post another one with gaming later on to see if battery drains while gaming.
Attached Files:
-
-
-
Falkentyne Notebook Prophet
Now this is confusing.
You have the same notebook as Pedro, right? Except you have a 6820HK while he has a 7700HQ?
Your E3 register has a value of 20 in it. Yet his has a "00" there.
That actually explains a LOT. "20" and "A0" seem to be duplicated in the GT73VR, if 20 is for 6820Hk and A0 is for 7820HK. But that's for a different product ID.
When I set 20 in my GT73VR, my CPU was able to draw "full power" for 28 seconds (without Power Limit 2), then PL1 throttle came on afterwards. Same thing for "A0".
When I set any other value (not in range 20-2F or A0-AF), example, 00, 80, etc, instant PL2 TDP throttle for 28 seconds (62W), then PL1 throttle afterwards.
I'm guessing I would have to mess with the MMIO register that Unclewebb mentioned in the TS thread, because one time I set a 25W power limit in that MMIO space (00000FED19500 i think it was) and it 'overrode' the bios set power limits (PL1 and PL2) of 200W. (I'm talking about CPU TDP power limits not AC/system power limits btw).
Only with 10, 11, 90 and 91 (which are direct matches for this SKU) was I able to draw full power unlimited.
Now this doesn't help -your- situation but the same pattern is there:
EC RAM registers C5 and C6 have nothing in them.
Someone with a USA shipped model needs to post their EC....because the one *I* had for 3 weeks had battery boost..., but I never even knew about RW Everything back then, I only used throttlestop and saw the battery vomit draining at full load.....
@heliada
do me a favor please.
Can you set a value of 21 in EC RAM register E3 and tell me if you get PROCHOT?
This will only work if your SKU has a GTX 1080 configuration, and I don't think it does...
or if we're lucky, MSI copied code.... -
@Falkentyne, values 46 and 47 stay on 00 the entire time when playing assasins creed. I have insane sound distortions now too lol probably the drivers all screwed. Anyway, no battery drain at all. My old dragon center also shows the power usage tho and it seems weird. It would spike to numbers higher than 160 randomly (even over 200W) but it then returns below REALLY fast (talk less than second, impossible to ss). Also the whole EC windows flickered couple times, I actually say places 46 and 47 have "90" in it for less than a second, as well as plenty other registers had 90 at the same time. It is all weird >.<
-
Falkentyne Notebook Prophet
Helida i know.
Please set a value of "21" in EC ram register E3. Your default value is 20.
I want to see if your EC has a possible "unused" GTX 1080 configuration (e.g. reserved).
If it does, that would change your power limit to 330W, however if NOS is DEACTIVATED, it would still not remove the 160W cutoff at all.
Take a look at this picture.
This is from a REALLY OLD laptop, GT60. The first laptop to use NOS (Battery Boost @Papusan )
You can see that "9E" is there, and 40 is there, and 40 means he's being throttled by something...unless the code was different BACK THEN and 40 meant full power...but 9E is the main NOS flag and 40 is the status flag...
Assuming MSI is using similar battery code as current laptops, it looks like his battery is not even installed...I don't even see the current battery voltage registers (the ones after registers 46 and 47 are the battery voltage and other bizarre registers). That would explain why EC RAM register C6 has a 40 there...Last edited: Mar 9, 2018 -
Pic from my GT72VR to see if help in case of @heliada
In my case, registers 46 and 47 only change when charging the battery.Last edited: Mar 9, 2018 -
-
@Falkentyne I changed to 21 and it did nothing. I attached a screenshot in reply to the message you sent me. All is deleted now and laptop packed back into the box where it belongs.
-
Falkentyne Notebook Prophet
That's good. Get rid of that cancer and just live your life. Consider yourself lucky you still have what many people don't have. You have a good boyfriend and a decent life there. Don't worry about BGA Cancerbooks. it's not worth it.
I'm already upset that MSI simply recycles old code from old laptops into their new laptops. They just use reserved registers or cycle stuff around. The entire basic EC structure is still the same as old laptops. -
@Falkentyne thanks for your help man, no amount of troubleshooting will fix this. I hope to get my money back now ^^ Looking at this really cool setup:
Attached Files:
-
-
Anyways tell me, you will deal directly with MSI with RMA or reseller? -
Falkentyne Notebook Prophet
Amazed you can get a GTX 1070 without paying 200% MSRP because of the mining craze. Can't get anything in USA right now without paying double the price.
BTW check your PM. I gave you "Confidential" MSI information. I cant discuss this more here, but it looks like your region has "NOS" completely locked out. This can't be done through Bios or a "different" EC Because the EC Is the same as others. Something on the mainboard must be telling the EC to disable NOS. like how the GT73VR system power limits are determined by if a GTX 1070 or GTX 1080 is INSTALLED in the system and set at POST power on, then the AC limit is set to 230W or 330W.
If it's a hardware component on your mainboard or *MISSING* from your mainboard because of region, that is causing NOS to be disabled, that's impossible to fix through EC RAM, at least not without me having the laptop here. You would have to compare a USA GT62VR (with a 6820HK) that has NOS enabled, with your EC, side by side, then look for differences in the register values that are NOT read only values, and then hope changing one that is different on your system, to the same value as the USA system, would disable the lockout. (That's basically how I unlocked the 330W power limit on my GT73VR. But I still can NOT remove the 160W NOS limit from my GTX 1070 being installed, which SHOULD be 240W on GTX 1080 systems); only way is to disconnect the battery completely and change EC registers 31 and 42........ -
-
@Falkentyne thank you I read the message ^^ it's a mess they would remove parts of mainboard or alter them. No wonder they cannot sort their own stuff out. I am done with it big time, if they cannot make it run as promised, I want money back ^^ I don't plan to troubleshoot any further in my own time. If Pedro69 still wants some help, feel free to help him out but I am going passive on the forums atm. It's just making me upset while I have to focus on university. I am so far behind (frightened smiley face) it's not even fun. Had to troubleshoot the old pc few times already too but seeing it's at least 8 years old and a total mess it runs 1000% more reliably than the msi laptop. Just look at this beauty! Such green LED's, cute mini cpu cooler, tiny gpu single fan, 50% of connectors either hanging around or broken... isn't she beautiful? It even runs final fantasy xxiv on 30 fps on low laptop settings!
Attached Files:
Falkentyne and Pedro69 like this. -
-
Falkentyne Notebook Prophet
Is that an Asus motherboard from the P5WDH era? Oh I remember those...those ran stable as a rock. Like the old P5Q Pro....
Only problem I had was in my P5WDH, I put a QX9650 CPU in there and the power draw was too much for the 4 pin connector and it burnt the connector. It still worked though. I had to use the second connector that was on the PSU and unburnt, and I turned the overclock down. Eventually I upgraded to a 2600K and a Seasonic 1000W PSU, which is still in my desktop right now. -
@Falkentyne yeah it is an asus mb. My old bf's pc tho ^^ he bought it premade for him like 8 years ago (so he was not the one to cause the mess of a cable management but never bothered fixing it up). I believe it is this one: https://www.asus.com/Motherboards/M5A78L-M-PLUS-USB3/HelpDesk_CPU/
** 1070 laptop: "bd prochot" causing cpu throttling to 800MHz and stuttering **
Discussion in 'MSI' started by streetunder, Feb 13, 2018.