NVIDIA Kepler VBIOS mods - Overclocking Editions, modified clocks, voltage tweaks

[Kross]

message after flashed official bios for modded bios !

bios :

before "80.04.29.00.01" (official)

after "Clevo_680m_-_80.04.33.00.10_'OCedition'_revised_01"

[gokica]

Only svl7 can answer the question about "Invalid" message with the modded bios.

I am using the official bios because no modded bios is available for my GTX 670MX and the Nvidia inspector is partially functional (in 1.91 sliders work and later versions of this utility only allow command prompt orders for increasing the GPU and the memory frequency above the bios limits +135/+1000). The voltage is locked and can not be tweaked with Nvidia Inspector.

[svl7]

Regarding the 'invalid' - that will only happen with a modified vbios and doesn't matter (at least not so far). The current Nvidia vbios generation is digitally signed, if the file gets modified the signature consequently becomes invalid. Nvidia inspector seems to have implemented the check for the vbios integrity, I assume it's done per driver call via nvapi. It has no effect on the overclocking features of Nvidia inspector or any other tools. Things might change if Nvidia ever decides to implement the vbios signature check in their driver and if invalid refuse to clock to 3d clocks or similar... If this ever happens I'm definitely going to say goodbye to team green.

I contacted Nvidia Support direct. Really good bunch of people and you can open live text chats with a support representative. I was told that anything up to 85C is considered safe and healthy, 85C - 105C is considered to be critical, and quoting from memory, 105C and above you can expect to see smoke coming from your GPU!! Nice!.

If Nvidia considers 85°C safe and healthy then you have to ask yourself why so many of their cards start to throttle at like 78°C... While 85°C isn't really a problem the current Kepler cards should all maintain temps below 75°C at stock 3d clocks, at least as long as the system is properly designed... Under load the 680m hits 90°C in most P170EM systems, while temps hardly ever go above 75°C in a AW system... even in my almost ancient M15x, same card. That's just poor design and unnecessary. Cooler temps are preferable, the card degrades faster at higher temps and the fan causes more noise.

[isaacucb]

Thank you!

This card is seriously underclocked... lol, glad to see it working like this. Temps are totally fine, great actually. I don't see any issue with a constant +420MHz overclock on the core as long as it is stable, it's stock voltage after all, but I would recommend to keep it easy on the memory overclock for daily use. Maybe +200 - +250MHz on the vram.

Thanks for the advice! It seems I get a stable 87-89 C and 30 FPS running Kombustor Tessy Furry test at 1080p, no AA, fullscreen with these clocks. Past that, more memory overclocking didn't seem to yield too much in terms of performance, and more overclocking on my cores just made it crash so I'll stick to this. However, I'm a bit confused. What's bad about running the memory higher than the threshold you stated?

[jamesst20]

Wow seems amazing to push our Alienware NVidia 660M to their limits! Looking toward this right now! Thank you sir!

[Kross]

Regarding the 'invalid' - that will only happen with a modified vbios and doesn't matter (at least not so far). The current Nvidia vbios generation is digitally signed, if the file gets modified the signature consequently becomes invalid. Nvidia inspector seems to have implemented the check for the vbios integrity, I assume it's done per driver call via nvapi. It has no effect on the overclocking features of Nvidia inspector or any other tools. Things might change if Nvidia ever decides to implement the vbios signature check in their driver and if invalid refuse to clock to 3d clocks or similar... If this ever happens I'm definitely going to say goodbye to team green.

If Nvidia considers 85°C safe and healthy then you have to ask yourself why so many of their cards start to throttle at like 78°C... While 85°C isn't really a problem the current Kepler cards should all maintain temps below 75°C at stock 3d clocks, at least as long as the system is properly designed... Under load the 680m hits 90°C in most P170EM systems, while temps hardly ever go above 75°C in a AW system... even in my almost ancient M15x, same card. That's just poor design and unnecessary. Cooler temps are preferable, the card degrades faster at higher temps and the fan causes more noise.

Thanks for information

I want a regular overclock with nvidia inspector, it is in performance level [2] - (P0) must be modified based Clock and memory clock offset? and a stable clock for 680MGTX 4GB please ?

thank you very much

[rmmrolf]

Thanks for this information! I want to OC this card so bad.

[dasnorsk]

Well, the fact mine is marked INVALID means that MSI / Medion modified the bios without signing it? i hope it means that you cannot modify our bios anymore, slv7 ^^

I tested Nvidia Inspector 1.91, but no changes can applied.They always get back to stock.

[frostie]

@gokica.

I now understand why you were asking about power draw and loading. That's a nice laptop you have, just been reading the review on Notebookcheck.net. Towards the end of the review they state the peak load at 138 Watts, so your PSU has plenty of headroom to OC. I previously had an Alienware M15x with a GTX670M and a 45watt i7. I used to measure the power usage with a current meter and found that under full load I was averaging 133 Watts. I upgraded to an extreme i7 so a 55watt chip that could be OCd, and with a 40% OC on the i7 and a 10% OC on the 670M I averaged 154 - 168 watts. The only way to really know your power draw is to meter it, but I don't think you need to with a 180 Watt supply.

Your GPU load temps are excellent, on the M15 with the old Fermi technology the 670 averaged 75C for me and rose to 80C with a 10% OC. I wish you all the best with flashing the vbios, you are brave with it being an Asus custom fitted GPU board. Have you removed the bottom cover to see if you can locate a vbios chip, just incase anything should go wrong and it needs reprogramming?

[gokica]

svl7, thanks for the information on the GPU safe temps. Now I feel less enthusiastic about this venture. Ha, ha.

Kross and everyone trying to use Nvidia Inspector see whether you could use the following guide:

nVIDIA mobile Kepler 6xx MASTER overclocking guide. Workaround for max clocks/OC stability.

[GUIDE] Nvidia Inspector GTX670/680 - disable boost / fixed clock speed / undervolting

For example I use Total Commander utility (trial) instead of the CMD prompt which seems more complicated. Within Total Commander I just navigate to the folder where Nvidia Inspector is located and than at the bottom of the Total Commander window I give the command parameters.

The command does the job very well.

Regarding the sliders (I am at work right now and not infront of the computer) but I know that for raising the memory clock above the limits on my GTX 670MX I have to equally move the memory and shader sliders to a same frequency. Moving only the memory overclock slider does not work and sets the slider at the beginning position. First unlock the max of course.

And yes, for overclocking you need to play with the Performance state.

If you need more detailed instructions please let me know and once I am back from work I will provide screenshots from the laptop and how I do it.

frostie, thank you for the information. For the moment I can only measure the power draw on my desktop with Zalman MFC but I do not have appropriate device for the laptop so your data is very helpful. I did not check whether the bios chip can be reprogrammed since the card is completely covered with a copper/aluminum plate including memory chips, mosfets, chokes etc. If I take it off I will probably need to replace the paste and the cooling pads touching those components so I guess I will have to take my chances once a modified bios comes up.

Thank you.

[roo7y82]

hello,

Thanks you for your help gokica, i have medion x7819 with 670mx and like all pepoles here i cant OC my cart more than +135 core clock, so if any one know how to break this limite it's will be so nice ! sorry for my bad english.

[gokica]

A brief tutorial on Nvidia Inspector's command line parameters (with Total Commander utility):

1. Download Total Commander.

2. Install and run (press key 1, 2 or 3 at startup as required by the shareware version of the software)

The program will run and you will see a window which looks like this:

Basically you see two windows for file management. In one of the windows above (left in the example) navigate with your keyboard or mouse to the folder where you have Nvidia Inspector installed. Once nvidiainspector.exe file is visible in the Total Commander window add the following parameters in the bottom line (marked with red square in the picture above):

For 750MHz overclock on the GPU of GTX 670MX:

nvidiaInspector.exe -setVoltageOffset:0,0,0 -setGpuClock:0,2,750 -setMemoryClock:0,2,1400 -forcepstate:0,0

and press enter.

For 760MHz overclock on the GPU of GTX 670MX:

nvidiaInspector.exe -setVoltageOffset:0,0,0 -setGpuClock:0,2,760 -setMemoryClock:0,2,1400 -forcepstate:0,0

press enter.

For going back to default state of your graphic card:

nvidiaInspector.exe -setVoltageOffset:0,0,0 -setGpuClock:0,2,135 -setMemoryClock:0,2,1400 -forcepstate:0,16

press enter.

During the process open EVGA precision in a separate window on your desktop and in it follow the frequency and the temperature. Please note that the voltage will be fixed and maximum while you are overclocking.

Play with the parameters as you wish but do not exceed the numbers.

Please note that the above values are valid for GTX 670MX. If you have different card in your computer modify the frequency values upon need. Do not touch the voltage in the parameters. Leave the value at 0.

So you do not need to run the Nvidia inspector by double clicking. Just enter the command parameters while Total commander has nvidiainspector.exe folder open.

If you have more questions please let me know.

For changing the memory frequency in the nvidia inspector just move both sliders equally:

If you want to increase the memory limit just press "unlock max" in the upper right corner. Than move shader clock and memory clock sliders both to 3000 (or any other value) for example. After both are pointing at 3000 (i.e. both same value) press "Apply Clocks & Voltage" at the bottom right corner. (I am sorry for the Asian language example).

As you can see the option for raising the GPU frequency in Nvidia inspector is dimmed in my case. I can only play with the GPU frequency via the command line parameters or via EVGA precision. I do not know why is like this.

These options work with my GTX 670MX.

In case you see a message popping up like "divider not 0" or similar once you run Nvidia inspector try and use version 1.91 instead of the latest build.

Do not forget to monitor the changes in frequency and temperature in EVGA precision or MSI afterburner.

I hope this helps.

[Kross]

Thank Gokica !

nvidia inspector 1.9.6.9 I have here and the table

base clock and memory clock has changed is that?

http://img.techpowerup.org/130208/nvidia_20130208_161552.png

EDIT : something else too please :

this little thing that bothers me I try to run any game on battery and each time the PC turns off suddenly, something I sure did other laptop before.

So I thought at first was a battery problem not powerful enough, or too powerful GPU, so I was in the Nvidia config panel, switch on the Clevo integrated GPU (HD 4000) and restart the game on battery no worries there battery like no problem but when I go to Nvidia as favorite graphics and the PC shuts down to reboot.

I said that when plugged into sector there are no worries on that side of the PC does its job!

So where is the problem material ? because then I dry if it is normal for you I would learn something: p

Thank you to you

[gokica]

I am sorry but I do not have experience with Clevo laptops and I do not have Optimus technology enabled so that I can switch back and fourth between the dedicated and the onboard GPU. My guess is that you are trying to run a game which is not demanding and can be played on the HD 4000 equally well as with the GTX 680m. Therefore HD 4000 uses much less power and enables you to play the game longer.

What I am sure of is that my laptop has the ability to detect plugged and non plugged (in A/C power outlet) mode and can adjust the performance automatically. I get lower FPS in games if I am running the laptop just on battery and I can not change this.

Hope this helps and maybe Clevo user can help you more or you could address the issue to the Clevo forums and support.

[Kross]

i test it thanks

[gokica]

Since it is 3D and not 2D game title obviously GTX 680m is activated while you are playing.

Is the battery empty when the laptop shuts off? Maybe there is a safety feature that stops the computer running on battery if the power draw is unusually big (with overclock). If you have battery power remaining and no overclock is applied while the computer shuts off than maybe it is best if you contact your vendor or try and consult the manual first for example.

Once again I have no experience with Clevo so maybe someone else will help. I hope that your issue (if there is one) will be resolved.

[Kross]

when the pc stops I always full battery.

RESULT: just changed the vbios> for BACK original BIOS 680MGTX of my clevo

When i run a game, the VBIOS Directly launch the nvidia GPU under the battery in place of the HD 4000 chipset => CRASH REBOOT!

The original bios Directly launch the HD 4000 chipset and not the nvidia GPU even if I force Nvidia GPU in panel config under battery => Security?

reference bios problem?

my original bios 80.04.29.00.01

my vbios : 80.04.33.00.10 rev 1

for vbios I have to take the same reference, or there is no effect?

in DOS :

I had an error on the front of pci flash but I force (nvflash -6 files.rom)

[Khenglish]

when the pc stops I always full battery.

RESULT: just changed the vbios> for BACK original BIOS 680MGTX of my clevo

When i run a game, the VBIOS Directly launch the nvidia GPU under the battery in place of the HD 4000 chipset => CRASH REBOOT!

The original bios Directly launch the HD 4000 chipset and not the nvidia GPU even if I force Nvidia GPU in panel config under battery => Security?

reference bios problem?

my original bios 80.04.29.00.01

my vbios : 80.04.33.00.10 rev 1

for vbios I have to take the same reference, or there is no effect?

in DOS :

I had an error on the front of pci flash but I force (nvflash -6 files.rom)

I think the battery can't put out enough power to run the 680m at full 3d clocks, so clevo sets it to be disabled on battery. The crash with svl7's version is likely the system BIOS detecting too low of a battery voltage and shutting the system down to prevent battery and voltage regulator damage. Since slv7's version lets you try using the 680m on battery, you could likely use Nvidia Inspector to disable the top 1-2 P-states to keep the power draw down. A 400 MHz 680m is still a hell of a lot better than a HD4000. Your battery wouldn't last much more than an hour though.

Running the 680m at full power would likely require an extra 8 battery cells jammed into the laptop chassis.

[dasnorsk]

@gokica : Thanks to your explanations, i made some testing on my side, the gpu works nicely up to 796mhz, so +181mhz. Not that bad. I didn't push it any further. Concerning vram, +1450mhz works pretty well, pushing it to 2850mhz (1400mhz stock).

With roo7y82, we manage to get nearly P5K on 3DM11 with that. We only have nearly P3800 without o/c. (P4932 for me ; Gpu @ 796mhz and VRAM @ 2850mhz). The temps never exceeds 69° on my side, for the gpu.

[gokica]

@gokica : Thanks to your explanations, i made some testing on my side, the gpu works nicely up to 796mhz, so +181mhz. Not that bad. I didn't push it any further. Concerning vram, +1450mhz works pretty well, pushing it to 2850mhz (1400mhz stock).

With roo7y82, we manage to get nearly P5K on 3DM11 with that. We only have nearly P3800 without o/c. (P4932 for me ; Gpu @ 796mhz and VRAM @ 2850mhz). The temps never exceeds 69° on my side, for the gpu.

That is a very good OC. My GTX 670MX has 3GB of video memory and will not run above +1200 and the GPU will not run smoothly above 750MHz with the latest official drivers. So I need more voltage.

BTW ASIC quality of my card's GPU is 84.4%. You can read it in GPU-Z by left clicking the upper tab and selecting ASIC. What is your GPU quality?

EDIT: Thanks to your reply and results I tried running the card at higher frequency once again. I did a clean install of the driver the other day and it seems like 800/2400 is stabile in Unigine Heaven. The driver was holding me back.

More testing for me now. Thanks.

[dasnorsk]

Well, it tells me 72.1%. I don't know if its good or not, but i will not get over 775mhz for gpu for daily use anyways.

My card is only 1.5GB VRAM, this may explane why i can get higher frequencies than yours.

[gokica]

Well, it tells me 72.1%. I don't know if its good or not, but i will not get over 775mhz for gpu for daily use anyways.

My card is only 1.5GB VRAM, this may explane why i can get higher frequencies than yours.

Yes the ASIC quality is sufficient for a nice OC and yes the memory size as well as the quality of the memory modules on board can affect the OC results. Thanks.

[dasnorsk]

Okay! Well i learned something new once again thanks to you! I hope we can get a nicely modded vbios one day for ours cards!

[roo7y82]

Thanks @gokica , it's work for me i can up it to 825+ core and i think its the max limite with no OV !

NVIDIA GeForce GTX 670MX video card benchmark result - Intel Core i7-3630QM Processor,MEDION X781X/X782X score: P4894 3DMarks

now we will wait for VBIOS with OV

[gokica]

I am glad you did it guys. What is your card's voltage when running 825Mhz?