Freeze/Crash & BSOD issues

[Kittra]

*snip*

Sorry about the snip, but quoting all of that would have put me over SE's stupid 1000 character limit.


My comments about the "danger zone" are all taken from personal experience and relate to Nvidia's cards between the 400-500 series.
My comments on throttling are also taken from personal experience when related to the 600-700 series cards.

You are right about the temperatures for the 600 and 700 series cards being higher (70C and 80C respectively), I was simply trying to recall from memory when I made the post. I'll edit it to reflect the correct information.

My personal observations include a GTX 670 and GTX 780 dropping well below their "base clock" when over 80C with the lowest clock at 732mhz where it no longer dropped.

As far as the 500 series goes... I have gone through about 4 different GTX 570s which all started to artifact at above 80C and one's transistors even melted at a sustained 93C (don't buy MSI...).
The biggest problem I have with the 600 and 700 series cards IS the GPU Boost 1.0/2.0 as it causes some "frame rating issues" (smooth frame rate vs. number of frames per second) when it kicks the Boost in. To combat this, I simply overclock it so that it never needs to Boost, but I also needed to install additional cooling so that it would never throttle the speed down above a certain temperature.
The listed factors are complete bull, I'll always trust someone with personal experience over someone who just reads off of a tech article.

[Pandoro]

I'm not sure about all the technical stuff, I will just write about my experience.

I saw a lot of the crashes. Most of the times I wasn't getting a BSOD a lot, but it would freeze for a second and then unfreeze again when lucky. Otherwise I would need to hard reset my pc, or sometimes, but rather rarely get a BSOD.

What I noticed regarding drivers, and I tried all between 314 and the latest beta. The 314 runs more stable for me. The latest beta gave me LOT more crashes and more frequent BSODs.

Then I tried the TdrDelay "fix" which is proposed in many youtube videos. This didn't help. Neither it helped to set all FFXIV settings to the lowest possible value. Eventually I checked my temperatures and my GPU was around 80 degrees most of the time. This didn't seem to be a big problem though and I didn't really notice a big correlation between crashes and temperature spikes. I also set my nvidia settings to "performance always" or something like this, which also didn't help, although I had the idea that crashes were less frequent after all of this together with the 314 drivers.

As I am playing on my Dell laptop, holding a nvidia 540m with optimus technology, I wasn't able to do a lot in the bios. What "solved" the problem for me, was to download some nvidia system config tool and underclock my GPU to 630mhz. Since then I haven't had a single crash, apart from the moments where it would unload the profile and return to the normal 672mhz. So at least for me, there seems to be a problem with the general way my GPU works. A lot of people have pointed out that this kind of stuff happens when your GPU doesn't get enough power. This kind of makes sense as you would figure that it makes a system run unstable if it has a too low voltage. So since I couldn't crank up the voltage, I just underclocked my GPU.

I'm not saying this is the fix, or the way to go, I just hope that it helps some desperate soul to get to play in a normal way.

[Squa11_Leonhart]

I'm not sure about all the technical stuff, I will just write about my experience.

I saw a lot of the crashes. Most of the times I wasn't getting a BSOD a lot, but it would freeze for a second and then unfreeze again when lucky. Otherwise I would need to hard reset my pc, or sometimes, but rather rarely get a BSOD.

What I noticed regarding drivers, and I tried all between 314 and the latest beta. The 314 runs more stable for me. The latest beta gave me LOT more crashes and more frequent BSODs.

Then I tried the TdrDelay "fix" which is proposed in many youtube videos. This didn't help. Neither it helped to set all FFXIV settings to the lowest possible value. Eventually I checked my temperatures and my GPU was around 80 degrees most of the time. This didn't seem to be a big problem though and I didn't really notice a big correlation between crashes and temperature spikes. I also set my nvidia settings to "performance always" or something like this, which also didn't help, although I had the idea that crashes were less frequent after all of this together with the 314 drivers.

As I am playing on my Dell laptop, holding a nvidia 540m with optimus technology, I wasn't able to do a lot in the bios. What "solved" the problem for me, was to download some nvidia system config tool and underclock my GPU to 630mhz. Since then I haven't had a single crash, apart from the moments where it would unload the profile and return to the normal 672mhz. So at least for me, there seems to be a problem with the general way my GPU works. A lot of people have pointed out that this kind of stuff happens when your GPU doesn't get enough power. This kind of makes sense as you would figure that it makes a system run unstable if it has a too low voltage. So since I couldn't crank up the voltage, I just underclocked my GPU.

I'm not saying this is the fix, or the way to go, I just hope that it helps some desperate soul to get to play in a normal way.

There are some issues with certain implementations of nvidia gpu's that don't work too well with the newest drivers.

[Squa11_Leonhart]

Sorry about the snip, but quoting all of that would have put me over SE's stupid 1000 character limit.


My comments about the "danger zone" are all taken from personal experience and relate to Nvidia's cards between the 400-500 series.

400/500 series cards that are overclocked by the manufacturer are likely the blame there, msi/gigabyte sold a heap of 400 cards that leaked tdp like crazy and artifacted at high temps because of it.

My comments on throttling are also taken from personal experience when related to the 600-700 series cards.
You are right about the temperatures for the 600 and 700 series cards being higher (70C and 80C respectively), I was simply trying to recall from memory when I made the post. I'll edit it to reflect the correct information.

My personal observations include a GTX 670 and GTX 780 dropping well below their "base clock" when over 80C with the lowest clock at 732mhz where it no longer dropped.

I've had several 600 gpu's at 85c in tombraider/folding and not had them drop back to 732, the lowest they got was to the base clock.
there were driver issues that were causing it to get stuck in low power 3D,
A cpu throttling due to heat will reduce gpu load, and thus result in the clock rate falling back as well.

As far as the 500 series goes... I have gone through about 4 different GTX 570s which all started to artifact at above 80C and one's transistors even melted at a sustained 93C (don't buy MSI...).

MSI (and others) were caught using way to much voltage on those gpus, which increased heat, increased tdp leakage and thus increased instability of the transistor when under load. they were again caught doing so on 600 series cards

The biggest problem I have with the 600 and 700 series cards IS the GPU Boost 1.0/2.0 as it causes some "frame rating issues" (smooth frame rate vs. number of frames per second) when it kicks the Boost in. To combat this, I simply overclock it so that it never needs to Boost, but I also needed to install additional cooling so that it would never throttle the speed down above a certain temperature.

I see reports of this, but have never experienced it myself - even though I'm rather sensitive to input, audio and visual lag.