No more overclocking on Nvidia mobile GPUs

interesting, because the people on baidu tieba did notice that some MXM cards came with a few capacitors less than the others, but that's just an OEM thing because they are the ones making MXM modules. I will link the posts later

 

I did. I didn't find anything in the INF files. I figured it was in that DLL the guy mentioned replacing on the forum but was hoping there would be an easier way in the INF files. I went through every single INF this morning comparing differences using winmerge and not only did I not find a fix, I found out that the desktop drivers and the laptop drivers are exactly the same for 347. Winmerge found no differences in the files.

Sent from my Nexus 6 using Tapatalk

 

Every press release can have potential securities ramifications. Hence, it is better to not make public statements and fix the issue than to say they messed up and have a lawsuit if the stock price is implicated. This is another reason companies wait until they can deliver good news with bad, to mitigate any stock price impact making securities lawsuits harder to sustain.

 

I've been really annoyed with how this story has been picked up. I seems it is mostly covered as Nvidia's corporate line and old wives tales about laptops in regards to overclocking.

So I made a new topic trying to connect the dots and give "the big picture".

I got called into work tonight so I have to go to sleep now, but my idea was to post directly to news outlets covering this and to their journalist directly if possible.

Toss this link to whomever you see fit.

http://forum.notebookreview.com/thr...rative-of-what-seems-to-have-happened.771440/

 

Well the infamous engineer strikes again

Did any of you geniuses create this by any chance?

 

Please please please, not all of us desktop users are like this. I am in full support of my notebook brethren in getting this overclocking fiasco turned around. I don't condone any infighting between notebook and desktop gamers. We are all PC gamers here in the end who love the flexiblity, mods, tweaking, etc. it brings. I love desktops AND notebooks and gaming on both of them.

Believe it or not, there are some desktop gamers who aren't totally dismissive of notebook gaming.

 

I don't want blocking on ANYTHING that I own. If I pay outright for my smartphone, I don't want it locked to a carrier or having its bootloader encrypted! If I pay outright for my gaming laptop, I don't want it limited in anyway.

I take full responsibility if I break anything while tweaking my valuable purchase. I am not some baby who needs to be protected by some scumbag corporation for my own good. These money grubbing pieces of garbage need to stop jerking their valuable customers around. > : (

 

Still no word from Nvidia?! They could, at the very least, issue a statement in response to all of this, but at this point, they're just pretending like the enthusiast community doesn't exist.

 

I think the bolded part is where the disconnect is. They do not value us as customers. We are end users that buy products from the OEMs that are their customers. They don't care what we want, and never have. If something they do is messed up, they blame the OEM. Until now, they pull this latest little god charade and pretend they have our best interests at heart.

 

He was talking about Desktop recording, which is disabled in Shadowplay on notebooks. Even SLI enabled, non-Optimus-using ones.

 

Yep I know now - 2 people have already told me! Thanks though.

 

This totally just gimped my 860m.... I got 200-250+ mhz out of the GPU after overclocking, and it never ran even remotely hot at all.. I usually gamed with the overclock when I had my cooling pad.

 

That would make me seriously mad at NVidia!

 

yeah it does, I ran my GPU at 790mhz (instead of 540) and my VRAM at 1,440 MHz (instead of 1,253)

It was like a new card, gave me 20-25% more FPS in games/benches

 

so as I went back to look at the 970m MXM comparison


the left 970m comes in the older p177sm model and the right one comes with p75x and p77x series.
as you can see the left one features two more MOSFETs on the top and one more choke(?) on the left, all marked by the red dots
so the VRMs on the newer cards have definitely shrunk.


in comparison to n=1's picture of 980m


the 980m also has one less MOSFET than the 970m on the p177sm and one choke missing on the left.
the people on tieba tried to test if there's any difference, but got mixed results all over the place.

Idk how this would affect OC in practice, but reduced VRM definitely kills power supply which is essential when OCing.
ALSO, this has ONLY been found on clevo cards so far...

speaking of OC damaging the cards, ironically, the only time I can remember when OC fks the GPU up is when NV FACTORY OC'D their 780m to be 880m. I have so far not heard of a case of 780m users getting screwed like that when they OC'd their cards to a similar level. so maybe nvidia did to...? I'll leave that up to you

 

Have a read of the following post on Tech Inferno, apparently the 980M uses a new 'type' of power delivery system which is more efficient so doesn't need so many of those chips lining the top of the cards - apparently it's a more capable power delivery system than that of the previous generation Kepler cards:
http://forum.techinferno.com/genera...us-nvidia-mxm-card-back-260-a.html#post123000

From what I gather, in terms of capability of the power delivery system, it looks as follows: 980M > (970M = previous generation Kepler). So, it looks like this is therefore not the reason for the ban on overclocking.

 

Wow a moderator closed the thread on anandtech

http://forums.anandtech.com/showthread.php?t=2420666&page=8

 

My IP is banned there even though I have never been even registered there and my rare yet multiple emails didn't change that for years so there is definitely something going on with moderators there.

 

I never posted there, was just refreshing the page every once and a while as I had heard that the moderators where a little idk power hungry or what not, if they finally found something they didn't agree with or something, bam! End of topic. Just rumors, but seeing that and hearing about your misfortune James D, makes me glad I don't get on there often.

 

I found help and now am able to use the latest drivers in myn GT72 again.



Best regards

phila

 

I found help and now am able to use the latest drivers in myn GT72 again.

Best regards
phila

P.S.: Sorry for the double Post IE hung up and did not show the post after returning to the site. So I wrote it again... Whoop threre it is...

 

What did u do to mod the drivers to OC?

 

he didnt, he flashed a modded vbios u can see that in nvidia inspector...

 

Hi! It was not my effort unfortunately ;-)

Svet a Moderator from the MSI Forums gave me a hand. And it works awesome.

Please feel free to check out his Mod Request thread: https://forum-en.msi.com/index.php?board=51.0

Best regards

phila

 

Yeah AnandTech mods make Guru3D mods look like saints. SMDH

OCN is pretty loosely moderated but there is a lot of ignorance among the user base

 

True that. Smartphones in my country have to be carrier unlocked by law - it's to discourage a monopoly over the market.

 

The 860M is supposed to run at 1098 MHz stock, not 540. Your card is being throttled or there's some power setting that is messed up in your notebook.

 

540 is base, not boost. Add that +250 MHz offset to boost clock too.

 

Interesting. I just read through the '860M is a Beast' thread, and noticed that some people managed to exceed 4500 in Fire Strike - mine's still around 4385, while my card runs at 1298 MHz and 2855 MHz (offsets in sig) according to Nvidia Inspector. I've overvolted by an offset of +275 mV. Temps don't exceed 70 degrees, and that's when propped up on a bed.

Is there something I'm doing wrong?

 

No what you thought was throttling was actually him reporting his base clock not boost clock. I'm sure his stock 860M runs over 1 GHz under load.

 

I don't follow. So is my card throttling?

 

No neither your card nor his is throttling. Some 860M cards get reported as 540 MHz when their actual boost clock is much higher. That's all I'm saying.

For example, look at this NotebookCheck review of the W370SS: http://www.notebookcheck.net/Review-Clevo-W370SS-Nexoc-G728II-Barebones-Notebook.114640.0.html

GPU-Z reports the base clock as 540 MHz. Boost clock is grayed out:



But in reality, it runs at 1097 MHz under load:

 

Can confirm, 540/790 are my base clocks. Boost is much higher. Makes a big difference honestly, and because nvidia is greedy/mad because we get all this free extra performance I'll probably have to mod my bios to continue to get that speed.

 

Thanks for the clarification. I thought Nvidia Inspector and GPU-Z were mis-reporting clocks.

 

There is, but I find they're not as bad as [H]ardOCP. Or maybe I just spend too much time on the GPU forums at [H]...

 

My experience OC the 970m 6gb:

I need to verify two things before I make any accusations on the card. First to check is a proper flash of the modded vbios. Second check is a proper modded driver (I have serious doubts about the driver I'm currently using because it was a first attempt and done hastily while I am studying for an exam. I will read every post and do my homework on this after said exam).

As I increase voltage, I receive lower scores in Valley. If I try going to within 50mV of 1.2, crash in valley within 2 minutes, even though I found it stable running 340mhz core at 1.125. I am going to play with it more after my exam next week and straighten out the inconsistent results.

As to benching performance, the scores in valley without changing VRam are linear up to 290:
Stock: 1540
+190 no voltage change: 1644
+290, voltage 1.1: 1690

This represents a 100%, 107%, 110% performance at these marks, rounding up to the nearest whole percent when above half a percent. If I jump to 1.2V and have the +290 core, then it will crash (same with 1.150). As I said, this may be due to a hastily done job modifying the driver, do not blame the hardware (yet). Also, at all times was the core temp below 70 degrees, so nowhere near the thermal throttle built into the modded vbios used.

If anyone has a correct modified driver for Win 7 x64, it would be greatly appreciated to compare as a control. If not, I will go through with more care at the end of next week and make sure my work is proper this time. BTW, thank you iata for the modded driver for Win 8. I have not tried it on Win 7 x64 yet, but appreciate your aid. This, apparently, will be something to learn for the future and the future is now.

Thanks to all that have posted resources throughout this thread and others.

 

Which modded driver you talking about?

 

I did a mod on Mr. Fox's real quick, but only changed one string in one .ini (hence why I'm sure I did not mod the driver correctly and this is my issue). I searched for the identifier mentioned in the thread in a specific ini mentioned, but only found the hardware id in reverse in the string at the bottom of the ini. Changed that and windows reads as 970m, but sure everything else is wrong due to my due diligence (laziness). Iata modded the 346 for me, but haven't yet used it. Just a matter of prepping for licensure at the moment and not having the time to devote my proper attention to this. As it works at stock for now, I may just wait until next friday and take the day to learn/write drivers.

 

Hmmm is there possibility of modding 347.xx to re-enable OC?

 

Which driver was this which you modded? if it's before 347xx, well you did nothing special..

 

Ay, take it easy. It's a learning process. Maybe ask yourself what have you done before making snide remarks about others.

 

The thread on geforce stated using the old .sys re-enabled OC. This means that someone needs to find a way to edit the .sys, compare the old and new one to narrow down the change, and copy the string of code from one to the other to enable OC on any driver 347.xx and beyond. Either that or hard flash the OC into the bios. I also will be looking into this over the next month. This used to be able to be performed in the system configuration tool in win 95 and 98. I will be checking if XP still has it baked in next week (virtual machine on my other dual boot). If so, then I will report back. If someone else knows of a similar tool and wants to track down the .sys change mentioned, all would be appreciative.

 

I also mentioned comparing changes in nvapi.dll and nvapi64.dll libraries because this:

 

Thank you both, and it's perfectly fine. I am not a driver person. Give me a month and I'll be up to snuff. I know I did nothing special, but I am brainstorming for an answer, both for pre-347 drivers and post-347 drivers. Someone may even drop a hint to NVIDIA that the best way to get us to quiet down, not admit wrong, and continue with what they are doing is to do something similar to what cyanogenmod did. Get a driver person to quietly get on this forum and provide a modded driver that would allow for OC without connecting that individual directly to nvidia. We get support, they get to decry OC, we still ***** about their actions, they still get to deny potential liability on burnt out OC cards, etc. But as I said, give it time, the solution will appear. I'm on 344 btw.
Much props to both of you!

 

Ahh ajc9988, please don't get offended or anything.. I'm genuninely wanting to know what you did... I actually think the 2 angles Octiceps mentioned are what need to be examined.. Now if I can figure out some way to open these damm files, I can try having a look.. Also Octiceps was mentioning about my earlier comment.. Please do not get offended..We are all in the same NVIDIA screw boat here..

 

AnandTech thread got re-opened. Mods there are weird. This is at least the second time they've locked the thread and then unlocked it shortly after.

 

What I did was nothing special, it was changing a string identifier. Here is where I think the solution is:
Open the pre-347 .sys file in sysedit or similar program (see https://support.microsoft.com/kb/232558/EN-US?wa=wsignin1.0 for details)
Open the 347 .sys file in sysedit or similar program
Compare and test all changes

 

Also, I usually use M$ notepad or similar for .ini editing and the like.

 

That +275mV is on the slider or actual recorded mV boost? The slider is just a slider and usually doesn't correlate to the amount of voltage increase. Usually it's 10-20mV at best improvement even at max slider OV.

 

Yup, Inspector voltage slider for my 650M doesn't start until +112.5 mV, which is the actual zero