Overclock.net banner

1 - 19 of 19 Posts

·
Registered
Joined
·
10 Posts
Discussion Starter #1 (Edited)
Hello guys,

I got my hands on a fairly good(and a bit old) gaming PC recently. Everything seems to work fine, except for one thing - the Asus GTX 770(a Direct CU II model, 2GB DDR5) that is inside it. The card is surely not dead yet, as it boots up normally and there is zero artifacts/screen glitches. The problem with it is that it is locked at idle clock speeds, which I believe is caused by the 430% TDP shown by GPU-Z and EVGA Precission. The core clock is at 135 MHz and the memory clock is at 162 MHz and they won't change under any load. I guess it is the card's "defense mechanism".

I have tried a few simple things, as at first I thought it was a bad clock/driver issue:

- I sweeped all the drivers and installed the latest 391.35
- I reseated the GPU
- I reset the CMOS
- I played around with the power management settings both in Windows and in NVIDIA's settings

Sadly, none of those worked(as I expected). The drivers doesn't seem to be a problem. It also seems that this is not caused by a bad overclock, as both the clock speed and the voltage are at stock.

I have read a few threads here and there about this issue and I saw somebody mentioning a fix with power management tweaking with Kepler BIOS Tweaker. This, however, is the breaking point for me as I will likely kill the GPU completely, as I have no idea on what and how to edit the VBIOS.

So, to sum up, it seems it is more like some faulty hardware part(maybe a power controller?). My questions here are as follows:

Have you guys had similar issues?
Is there any way to reduce the insane 430% TDP, so the clocks are not locked?
Do you guys have any idea what harware fault may cause this and can it be fixed/changed?
Is there any way to "somewhat" fix this by tweaking the VBIOS?

I am looking forward to any answers, so keep them coming! I really hope there is some fix or workaround to this! Also, thanks for you time!

Best regards!

P.S. I can't post any pictures of GPU-Z as I am a free/new user.

P.S.S. I now added the pictures from GPU-Z.
 

Attachments

·
Registered
Joined
·
293 Posts
Hello guys,

I got my hands on a fairly good(and a bit old) gaming PC recently. Everything seems to work fine, except for one thing - the Asus GTX 770(a Direct CU II model, 2GB DDR5) that is inside it. The card is surely not dead yet, as it boots up normally and there is zero artifacts/screen glitches. The problem with it is that it is locked at idle clock speeds, which I believe is caused by the 430% TDP shown by GPU-Z and EVGA Precission. The core clock is at 135 MHz and the memory clock is at 162 MHz and they won't change under any load. I guess it is the card's "defense mechanism".

I have tried a few simple things, as at first I thought it was a bad clock/driver issue:

- I sweeped all the drivers and installed the latest 391.35
- I reseated the GPU
- I reset the CMOS
- I played around with the power management settings both in Windows and in NVIDIA's settings

Sadly, none of those worked(as I expected). The drivers doesn't seem to be a problem. It also seems that this is not caused by a bad overclock, as both the clock speed and the voltage are at stock.

I have read a few threads here and there about this issue and I saw somebody mentioning a fix with power management tweaking with Kepler BIOS Tweaker. This, however, is the breaking point for me as I will likely kill the GPU completely, as I have no idea on what and how to edit the VBIOS.

So, to sum up, it seems it is more like some faulty hardware part(maybe a power controller?). My questions here are as follows:

Have you guys had similar issues?
Is there any way to reduce the insane 430% TDP, so the clocks are not locked?
Do you guys have any idea what harware fault may cause this and can it be fixed/changed?
Is there any way to "somewhat" fix this by tweaking the VBIOS?

I am looking forward to any answers, so keep them coming! I really hope there is some fix or workaround to this! Also, thanks for you time!

Best regards!

P.S. I can't post any pictures of GPU-Z as I am a free/new user.
If GPU-Z is showing 430% TDP and the card locks to P8 clocks, it has to be a bad sensor on the card. Reducing the reading of 430% should be possible by raising the TDP in the bios to the maximum but the sensor could l still read above 100% even with 999W theoretical TDP. You would need to test it.
 

·
Registered
Joined
·
10 Posts
Discussion Starter #3
If GPU-Z is showing 430% TDP and the card locks to P8 clocks, it has to be a bad sensor on the card. Reducing the reading of 430% should be possible by raising the TDP in the bios to the maximum but the sensor could l still read above 100% even with 999W theoretical TDP. You would need to test it.
Hmm, sounds viable. But is there any chance to mess things up by raising the TDP so high? I am kinda inexperienced and I don't want to brake it completely, yet.
 

·
Registered
Joined
·
10 Posts
Discussion Starter #4
Hello guys!

Could I get some more suggestions or maybe more details on how exactly to play around with the TDP in the VBIOS?
 

·
Registered
Joined
·
293 Posts
Hello guys!

Could I get some more suggestions or maybe more details on how exactly to play around with the TDP in the VBIOS?
Rasing TDP so high wont harm your card at all. It wont pull more power than usual.
You can just post yopur bios and I'll do it for you. Hard to explain to someone new to bios modding.
 

·
Registered
Joined
·
10 Posts
Discussion Starter #6
Rasing TDP so high wont harm your card at all. It wont pull more power than usual.
You can just post yopur bios and I'll do it for you. Hard to explain to someone new to bios modding.
There it is. Also, thanks for offering your help!
 

Attachments

·
Registered
Joined
·
293 Posts
MSI AB allows it to go up to 110%.
OK. So your stock 100% TDP is 228W.Since GPU-Z is showing 430% the card thinks its pulling 980W. To start I raised the TDP to 999W. After flashing the bios, GPU-Z should show around 95-99% TDP. If so, the card could also be at a bit higher clocks and not lock to P8. If one or even both of these things happen, we can try to completely fix the card by raising the TDP even more until the card doesnt go over 100% and doesnt throttle.
 

Attachments

·
Registered
Joined
·
10 Posts
Discussion Starter #10 (Edited)
OK. So your stock 100% TDP is 228W.Since GPU-Z is showing 430% the card thinks its pulling 980W. To start I raised the TDP to 999W. After flashing the bios, GPU-Z should show around 95-99% TDP. If so, the card could also be at a bit higher clocks and not lock to P8. If one or even both of these things happen, we can try to completely fix the card by raising the TDP even more until the card doesnt go over 100% and doesnt throttle.
Thank you very much! I flashed it and so far it works fine. GPU-Z now shows around 98%. What I noticed so far is that when I try to stress the card a bit, the PC freezes. This all so far. :)
 

·
Registered
Joined
·
293 Posts
Thank you! Could you also help me with the command line for nvflash? I really don't wanna mess it up relying only on my own knowledge, hehe.
I`m just gonna give you my way of flashing a bios and prepare everything for you in a folder so you wont have to do anything other than execute a file.
Make sure you unzip the zip file. In the folder is the bios I did for you, nvflash.exe, nvflash.sys and a file called "execute". Everything you have to do then is to execute the file called "execute".
Then the nvflash command prompt will open, after a few seconds you will have to press "y". You'll notice when since it will ask you if you are sure and want to proceed.
Before flashing turn off your display drivers.After the flash restart your computer and activate your drivers again.


nvm
 

Attachments

·
Registered
Joined
·
10 Posts
Discussion Starter #12
I`m just gonna give you my way of flashing a bios and prepare everything for you in a folder so you wont have to do anything other than execute a file.
Make sure you unzip the zip file. In the folder is the bios I did for you, nvflash.exe, nvflash.sys and a file called "execute". Everything you have to do then is to execute the file called "execute".
Then the nvflash command prompt will open, after a few seconds you will have to press "y". You'll notice when since it will ask you if you are sure and want to proceed.
Before flashing turn off your display drivers.After the flash restart your computer and activate your drivers again.


nvm
Sorry for the bad edit. I did the flash, but it seems pretty unstable now. Tried running a game twice to see what actually changed and first time the whole system froze, the second time the screen started flickering and went off.
 

·
Registered
Joined
·
293 Posts
Sorry for the bad edit. I did the flash, but it seems pretty unstable now. Tried running a game twice to see what actually changed and first time the whole system froze, the second time the screen started flickering and went off.
Yeah since the card is at its 98% TDP at idle and you hit it with load, it hits the TDP limit of 100% instantly and just crashes.
To fix this we have to set the TDP even higher. Its now set to 4000W. At idle GPU-Z should now show ~25% . Hit it with load and see what happens (maybe use the GPU-Z render test and see what happens to clocks and the TDP).
 

Attachments

·
Registered
Joined
·
10 Posts
Discussion Starter #14
Yeah since the card is at its 98% TDP at idle and you hit it with load, it hits the TDP limit of 100% instantly and just crashes.
To fix this we have to set the TDP even higher. Its now set to 4000W. At idle GPU-Z should now show ~25% . Hit it with load and see what happens (maybe use the GPU-Z render test and see what happens to clocks and the TDP).
Okay, I did the GPU-Z Render test, but the clocks did not change at all, nor did the TDP(it stays at around 25%). I ran it for a few mins without any changes, then I ran CS:GO to see what happens and after a minute or so, the screen went gray and I had to reset the system.
 

·
Registered
Joined
·
293 Posts
Okay, I did the GPU-Z Render test, but the clocks did not change at all, nor did the TDP(it stays at around 25%). I ran it for a few mins without any changes, then I ran CS:GO to see what happens and after a minute or so, the screen went gray and I had to reset the system.
Hmm, nothing happening to the TDP is understandable because the GPU-Z test is such a low load. But nothing happening to the clocks is weird. Do you know if the clocks changed in CSGO ?
 

·
Registered
Joined
·
10 Posts
Discussion Starter #16
Hmm, nothing happening to the TDP is understandable because the GPU-Z test is such a low load. But nothing happening to the clocks is weird. Do you know if the clocks changed in CSGO ?
Clocks and TDP stay the same in CSGO as well. Seems like it won't change from P8.
 

·
Registered
Joined
·
293 Posts
Clocks and TDP stay the same in CSGO as well. Seems like it won't change from P8.
Its possible to force the card to higher clocks (and eventually the cards normal 3D clocks) in P8 state but I'm afraid the card wont boot anymore if we set it to P0 clocks in P8 state because it hits TDP during the boot process.
We could raise the TDP even more and slowly increase the cards P0 clocks and see what happens to the TDP everytime we increase it a little bit. That way we can fully avoid what I described one line above.
It's on you. Do you want to try it ?

EDIT: The crashing could still be because of the TDP. Since the typical idle power draw is ~20W (even lower I thing maybe 10-15W) and your card shows 430% (980W) at idle, every 980W we increase the power limit results in ~20W more power allowed in the real world. So even with 4000W in the bios, the cards is still only able to hit 80W in the real world before it crashes.
 

·
Registered
Joined
·
10 Posts
Discussion Starter #18
Its possible to force the card to higher clocks (and eventually the cards normal 3D clocks) in P8 state but I'm afraid the card wont boot anymore if we set it to P0 clocks in P8 state because it hits TDP during the boot process.
We could raise the TDP even more and slowly increase the cards P0 clocks and see what happens to the TDP everytime we increase it a little bit. That way we can fully avoid what I described one line above.
It's on you. Do you want to try it ?

EDIT: The crashing could still be because of the TDP. Since the typical idle power draw is ~20W (even lower I thing maybe 10-15W) and your card shows 430% (980W) at idle, every 980W we increase the power limit results in ~20W more power allowed in the real world. So even with 4000W in the bios, the cards is still only able to hit 80W in the real world before it crashes.
Currently I am more prone to think that it might actually be a faulty power controller than a sensor. I might flash the original bios back and take it to some hardware repairs place to have it checked, cuz it is surely not dead. I read a thread on the NVIDIA forums talking about the exact same GPU with the exact same issue getting fixed by changing a fried controller, which is rather cheap.

Anyway, for now I will stick to having the hardware checked. Thank you very much for your help though, I really appreciate it and I've learnt quite a few things! Thanks! :)

I will update the thread once I have it checked(which might not be soon, but I will do my best).
 

·
hkgamer
Joined
·
12 Posts
Currently I am more prone to think that it might actually be a faulty power controller than a sensor. I might flash the original bios back and take it to some hardware repairs place to have it checked, cuz it is surely not dead. I read a thread on the NVIDIA forums talking about the exact same GPU with the exact same issue getting fixed by changing a fried controller, which is rather cheap.

Anyway, for now I will stick to having the hardware checked. Thank you very much for your help though, I really appreciate it and I've learnt quite a few things! Thanks! :)

I will update the thread once I have it checked(which might not be soon, but I will do my best).
Hello! I'd like to ask you what kind of hardware your GPU ended up with, such as what chip broke, and I have the same question as you
 
1 - 19 of 19 Posts
Top