GTX 560 crash problems
  1 / 2    
Greetings, I got a Palit GTX 560 recently (822MhZ, 2GiB RAM), and have been encountering Windows blue screens and application crashes. I'm using driver 266.66.

Basically, when I run Team Fortress 2 or Crysis with the settings maxed out (1920x1080), I either get a BSOD (crashes in some random kernel code like dxgmms1.sys), or an application crash. This seems to be happening every 10 minutes.

I noticed if I enable vsync in Team Fortress 2, it reduces the crash rate. This might suggest I don't have enough power. I'm using this [url="http://www.amazon.com/Genuine-KDM-Energy-Efficient-Active/dp/B00421L0XO"]520W uATX power supply[/url] that supplies 30A total for 12V, which I think should be plenty since my previous GTX 260, core 216 never crashed and had a very similar wattage. I measured power draws before & after upgrade:
[code]
GTX 260 Core 216 GTX 560
idle 95 80
TF2 (maxed out) 200 +- 5 210 +- 5
Crysis (carrier level, maxed out) 255 238 +- 2
NVIDIA design garage (Optix ray tracer) 205 268 +- 2
[/code]

So it doesn't seem power is the problem. So is this a driver issue?
Greetings, I got a Palit GTX 560 recently (822MhZ, 2GiB RAM), and have been encountering Windows blue screens and application crashes. I'm using driver 266.66.



Basically, when I run Team Fortress 2 or Crysis with the settings maxed out (1920x1080), I either get a BSOD (crashes in some random kernel code like dxgmms1.sys), or an application crash. This seems to be happening every 10 minutes.



I noticed if I enable vsync in Team Fortress 2, it reduces the crash rate. This might suggest I don't have enough power. I'm using this 520W uATX power supply that supplies 30A total for 12V, which I think should be plenty since my previous GTX 260, core 216 never crashed and had a very similar wattage. I measured power draws before & after upgrade:



GTX 260 Core 216 GTX 560

idle 95 80

TF2 (maxed out) 200 +- 5 210 +- 5

Crysis (carrier level, maxed out) 255 238 +- 2

NVIDIA design garage (Optix ray tracer) 205 268 +- 2




So it doesn't seem power is the problem. So is this a driver issue?

#1
Posted 02/05/2011 12:29 AM   
[quote name='Uncle Joe' date='04 February 2011 - 04:29 PM' timestamp='1296865750' post='1189067']
Greetings, I got a Palit GTX 560 recently (822MhZ, 2GiB RAM), and have been encountering Windows blue screens and application crashes. I'm using driver 266.66.

Basically, when I run Team Fortress 2 or Crysis with the settings maxed out (1920x1080), I either get a BSOD (crashes in some random kernel code like dxgmms1.sys), or an application crash. This seems to be happening every 10 minutes.

I noticed if I enable vsync in Team Fortress 2, it reduces the crash rate. This might suggest I don't have enough power. I'm using this [url="http://www.amazon.com/Genuine-KDM-Energy-Efficient-Active/dp/B00421L0XO"]520W uATX power supply[/url] that supplies 30A total for 12V, which I think should be plenty since my previous GTX 260, core 216 never crashed and had a very similar wattage. I measured power draws before & after upgrade:
[code]
GTX 260 Core 216 GTX 560
idle 95 80
TF2 (maxed out) 200 +- 5 210 +- 5
Crysis (carrier level, maxed out) 255 238 +- 2
NVIDIA design garage (Optix ray tracer) 205 268 +- 2
[/code]

So it doesn't seem power is the problem. So is this a driver issue?
[/quote]
honestly you are running your system extremely close to its limits if not surpassing them. i wouldnt be surprised at all if it was your power supply. i would suggest putting your card into someone elses computer with a better PSU and test it out.
[quote name='Uncle Joe' date='04 February 2011 - 04:29 PM' timestamp='1296865750' post='1189067']

Greetings, I got a Palit GTX 560 recently (822MhZ, 2GiB RAM), and have been encountering Windows blue screens and application crashes. I'm using driver 266.66.



Basically, when I run Team Fortress 2 or Crysis with the settings maxed out (1920x1080), I either get a BSOD (crashes in some random kernel code like dxgmms1.sys), or an application crash. This seems to be happening every 10 minutes.



I noticed if I enable vsync in Team Fortress 2, it reduces the crash rate. This might suggest I don't have enough power. I'm using this 520W uATX power supply that supplies 30A total for 12V, which I think should be plenty since my previous GTX 260, core 216 never crashed and had a very similar wattage. I measured power draws before & after upgrade:



GTX 260 Core 216 GTX 560

idle 95 80

TF2 (maxed out) 200 +- 5 210 +- 5

Crysis (carrier level, maxed out) 255 238 +- 2

NVIDIA design garage (Optix ray tracer) 205 268 +- 2




So it doesn't seem power is the problem. So is this a driver issue?



honestly you are running your system extremely close to its limits if not surpassing them. i wouldnt be surprised at all if it was your power supply. i would suggest putting your card into someone elses computer with a better PSU and test it out.

#2
Posted 02/05/2011 12:38 AM   
Thanks EbbAndBlow, but I still think I have a small extra power margin. The Palit requirements are [quote]Minimum 500W or greater system power supply (with a minimum 12V current rating of 30A)[/quote]

I consider that power supply a some what higher end model (i.e. less ripple, more efficiency). I have a small form system and that was the biggest uATX PSU I could find. If anyone knows of a better one, please tell me.
Thanks EbbAndBlow, but I still think I have a small extra power margin. The Palit requirements are
Minimum 500W or greater system power supply (with a minimum 12V current rating of 30A)




I consider that power supply a some what higher end model (i.e. less ripple, more efficiency). I have a small form system and that was the biggest uATX PSU I could find. If anyone knows of a better one, please tell me.

#3
Posted 02/05/2011 12:49 AM   
[quote name='Uncle Joe' date='04 February 2011 - 04:49 PM' timestamp='1296866988' post='1189073']
Thanks EbbAndBlow, but I still think I have a small extra power margin. The Palit requirements are

I consider that power supply a some what higher end model (i.e. less ripple, more efficiency). I have a small form system and that was the biggest uATX PSU I could find. If anyone knows of a better one, please tell me.
[/quote]
eek, i see your dilemma sir. when you are running these games, are you also monitoring your GPU temps? this could be a factor if your fans are set to auto and being your case is a micro ATX so there could be less than adequate airflow.
[quote name='Uncle Joe' date='04 February 2011 - 04:49 PM' timestamp='1296866988' post='1189073']

Thanks EbbAndBlow, but I still think I have a small extra power margin. The Palit requirements are



I consider that power supply a some what higher end model (i.e. less ripple, more efficiency). I have a small form system and that was the biggest uATX PSU I could find. If anyone knows of a better one, please tell me.



eek, i see your dilemma sir. when you are running these games, are you also monitoring your GPU temps? this could be a factor if your fans are set to auto and being your case is a micro ATX so there could be less than adequate airflow.

#4
Posted 02/05/2011 12:55 AM   
GeForce GTX 560 Ti - 31A and a 500W PSU minimum

GeForce GTX 260 - 38A and a 500W PSU minimum


http://forum-en.msi.com/index.php?topic=104805.0;
power draw 170watts -15 amp draw alone

you have a 30 amp combined rail (a little under the recommended min)

and it's a 5th tier PSU (not even on that list...so unknown and not recommended)
http://www.xtremesystems.org/forums/showthread.php?t=108088

try card in a system with Adequate PSU to see if the card is good/bad

monitor temps to rule out overheat issues
GeForce GTX 560 Ti - 31A and a 500W PSU minimum



GeForce GTX 260 - 38A and a 500W PSU minimum





http://forum-en.msi.com/index.php?topic=104805.0;

power draw 170watts -15 amp draw alone



you have a 30 amp combined rail (a little under the recommended min)



and it's a 5th tier PSU (not even on that list...so unknown and not recommended)

http://www.xtremesystems.org/forums/showthread.php?t=108088



try card in a system with Adequate PSU to see if the card is good/bad



monitor temps to rule out overheat issues



ImageImage


Image

Join the NVidia Forum Team - Please Help Medical Research - Folding@home


#5
Posted 02/05/2011 12:56 AM   
OK, I've installed NVIDIA system monitor & logged the GPU temperatures:

[code]
Max temp
Crysis carrier level (intense, 2nd half) 78
Crysis (island fly over benchmark) 79
[/code]

These temps seemed a bit high, and I routinely get a crash within 5 minutes. So I opened the case. /shock.gif' class='bbc_emoticon' alt=':shock:' /> Wow, the side panel actually feels hot to touch even though the card is 1 cm away.

Now I run the benchmarks again:
[code]
Max temp
Crysis carrier level (intense, 2nd half) 71
Crysis (island fly over benchmark) 71
[/code]

I also went on Team Fortress 2 (maxed out, vsync off) for about an hour and didn't have any problems, except for 2 "hiccups" (locks up with stuttering audio - fixed by alt-tabbing).
To make sure everything's alright, I ran the Crysis island fly over benchmark again, this time with 15 loops instead of 4. This time, it BSODs on the 5th iteration /mad.gif' class='bbc_emoticon' alt=':mad:' /> ("page fault in non-paged area").

So, do you think the problem is still heat related or not enough power? I suppose it could be a combination of both since high temperature will increase leakage current.


[quote]and it's a 5th tier PSU (not even on that list...so unknown and not recommended)[/quote]
That list is 3 years old, so I don't think my PSU will be on it.

I think I might have found a [url="http://www.silverstonetek.com/products/p_spec.php?pno=ST45SF&area=usa"]better power supply[/url], even though it's 450W. It claims to supply 36 amps for 12V, which seems unusual, being very close to 450W itself. According to its GPU support list, it can even run a GTX 470, which needs 30W more than GTX 560 according to Wikipedia.
OK, I've installed NVIDIA system monitor & logged the GPU temperatures:





Max temp

Crysis carrier level (intense, 2nd half) 78

Crysis (island fly over benchmark) 79




These temps seemed a bit high, and I routinely get a crash within 5 minutes. So I opened the case. /shock.gif' class='bbc_emoticon' alt=':shock:' /> Wow, the side panel actually feels hot to touch even though the card is 1 cm away.



Now I run the benchmarks again:



Max temp

Crysis carrier level (intense, 2nd half) 71

Crysis (island fly over benchmark) 71




I also went on Team Fortress 2 (maxed out, vsync off) for about an hour and didn't have any problems, except for 2 "hiccups" (locks up with stuttering audio - fixed by alt-tabbing).

To make sure everything's alright, I ran the Crysis island fly over benchmark again, this time with 15 loops instead of 4. This time, it BSODs on the 5th iteration /mad.gif' class='bbc_emoticon' alt=':mad:' /> ("page fault in non-paged area").



So, do you think the problem is still heat related or not enough power? I suppose it could be a combination of both since high temperature will increase leakage current.





and it's a 5th tier PSU (not even on that list...so unknown and not recommended)


That list is 3 years old, so I don't think my PSU will be on it.



I think I might have found a better power supply, even though it's 450W. It claims to supply 36 amps for 12V, which seems unusual, being very close to 450W itself. According to its GPU support list, it can even run a GTX 470, which needs 30W more than GTX 560 according to Wikipedia.

#6
Posted 02/05/2011 06:31 AM   
download evga precision and setup a fan profile to keep the card cool
I don't let my 570 get above 55c (and my fan is around 75%) under load

seems your card works better when cooler

try testing it with a cooler profile...see how that goes (and see if you can improve airflow thru case)

poor power leads to the PSU compensating with increased voltage when power is low(which increases heat)

you could be Okay with The PSU...but without a new PSU chart I don't know how good the PSU is...you'll have to look at reviews
KDM been around for 20 years..supplying OEM manufacturers with PSU's (low cost)

the silverstone is a good brand...if you do opt for it..it's a good choice
download evga precision and setup a fan profile to keep the card cool

I don't let my 570 get above 55c (and my fan is around 75%) under load



seems your card works better when cooler



try testing it with a cooler profile...see how that goes (and see if you can improve airflow thru case)



poor power leads to the PSU compensating with increased voltage when power is low(which increases heat)



you could be Okay with The PSU...but without a new PSU chart I don't know how good the PSU is...you'll have to look at reviews

KDM been around for 20 years..supplying OEM manufacturers with PSU's (low cost)



the silverstone is a good brand...if you do opt for it..it's a good choice



ImageImage


Image

Join the NVidia Forum Team - Please Help Medical Research - Folding@home


#7
Posted 02/05/2011 06:48 AM   
Victory. 15 iterations of the Crysis island flyover benchmark finally pass! I left the case open and used the performance tab under NVIDIA control panel (need to install NVIDIA System Tools with ESA support) to set the fan speed to 70%. That greatly reduced the peak GPU temperature from 71 C to 55 C. Great solution Xavier.

The fan speed is *a lot* louder than the regular speed, which I never noticed increasing under load before. Is this a bug with the NVIDIA drivers?

I was considering abandoning my micro ATX tower case (see attached photo) for a similar sized case that can accommodate a regular ATX power supply, since powerful uATX/SFX PSUs are so rare. That's no longer necessary. I would like to reduce the need for such a high fan speed. Does anyone think it's worth to replace the thermal compound with IC diamond? [url="http://www.youtube.com/watch?v=rlO6Gx7Uf6g"]This video[/url] claims a 4C,11C idle/loaded decrease after applying IC diamond to a GTX 470.

Other questions:
1. Given that a GTX 470 runs fine at > 80C under load, is it correct to say that the reason for my problem is the power demand increases with temperature and my PSU was unable to meet it? Or is it because the GTX 560 is inherently unable to run at that high temperature?

2. The Palit cooler doesn't blow air out of the case, but rather all around. Is that better or worse than the closed shroud type cooler?
Victory. 15 iterations of the Crysis island flyover benchmark finally pass! I left the case open and used the performance tab under NVIDIA control panel (need to install NVIDIA System Tools with ESA support) to set the fan speed to 70%. That greatly reduced the peak GPU temperature from 71 C to 55 C. Great solution Xavier.



The fan speed is *a lot* louder than the regular speed, which I never noticed increasing under load before. Is this a bug with the NVIDIA drivers?



I was considering abandoning my micro ATX tower case (see attached photo) for a similar sized case that can accommodate a regular ATX power supply, since powerful uATX/SFX PSUs are so rare. That's no longer necessary. I would like to reduce the need for such a high fan speed. Does anyone think it's worth to replace the thermal compound with IC diamond? claims a 4C,11C idle/loaded decrease after applying IC diamond to a GTX 470.



Other questions:

1. Given that a GTX 470 runs fine at > 80C under load, is it correct to say that the reason for my problem is the power demand increases with temperature and my PSU was unable to meet it? Or is it because the GTX 560 is inherently unable to run at that high temperature?



2. The Palit cooler doesn't blow air out of the case, but rather all around. Is that better or worse than the closed shroud type cooler?
Attachments

desktop.jpeg

#8
Posted 02/05/2011 09:30 AM   
mine blows the hot air outside the case...that's better to keep it cool

the card has a built in profile in vbios to keep it cool (not as cool as some people like)
evga precision is just a program to manually control it thru software

if you have one that blows air all around the inside of the case... you need adequate ventalation
(i.e lots of fans for air movement thru the case..remember the cpu and the other stuff heats up too)

At least it's nice it working for you now..
did you set up a profile to control the temp or did you just push the auto button and manually increased the fan speed
the profile is better and you can tune it to gradually increase and then you only get noise when it gets hot
mine only makes a light whoosh sound like a normal house fan

check other temps like CPU...maybe it was overheating due to the heat you were dumping into the case from the card


if the cpu temps are fine then
You have a choice do you RMA the card back to say it overheats an 70's and crashes (it should run even at 95c)
or keep it

Id probably get in contact with Palit and see about an RMA
(or if bought at store in last 30 days see about refund or exchange for another one - like one that vents it outside the case)
mine blows the hot air outside the case...that's better to keep it cool



the card has a built in profile in vbios to keep it cool (not as cool as some people like)

evga precision is just a program to manually control it thru software



if you have one that blows air all around the inside of the case... you need adequate ventalation

(i.e lots of fans for air movement thru the case..remember the cpu and the other stuff heats up too)



At least it's nice it working for you now..

did you set up a profile to control the temp or did you just push the auto button and manually increased the fan speed

the profile is better and you can tune it to gradually increase and then you only get noise when it gets hot

mine only makes a light whoosh sound like a normal house fan



check other temps like CPU...maybe it was overheating due to the heat you were dumping into the case from the card





if the cpu temps are fine then

You have a choice do you RMA the card back to say it overheats an 70's and crashes (it should run even at 95c)

or keep it



Id probably get in contact with Palit and see about an RMA

(or if bought at store in last 30 days see about refund or exchange for another one - like one that vents it outside the case)



ImageImage


Image

Join the NVidia Forum Team - Please Help Medical Research - Folding@home


#9
Posted 02/05/2011 03:46 PM   
[quote]did you set up a profile to control the temp?[/quote]

NVIDIA SYSTEM TOOLS WITH ESA SUPPORT
No, I just changed the fan speed from custom to auto for testing. The performance tab in NVIDIA control panel lets you set up per app profiles, but I didn't bother to create one.


[quote]if the cpu temps are fine then
You have a choice do you RMA the card[/quote]

OK, I'll contact NewEgg about the problem. I don't like this because according to AnAndTech's GTX 560 benchmarks @ stock 822MhZ, the max temperature under load
on Crysis is exactly the same (71C) as I got when I leave the case open. If that's a normal operating temperature, then it shouldn't crash.
did you set up a profile to control the temp?




NVIDIA SYSTEM TOOLS WITH ESA SUPPORT

No, I just changed the fan speed from custom to auto for testing. The performance tab in NVIDIA control panel lets you set up per app profiles, but I didn't bother to create one.





if the cpu temps are fine then

You have a choice do you RMA the card




OK, I'll contact NewEgg about the problem. I don't like this because according to AnAndTech's GTX 560 benchmarks @ stock 822MhZ, the max temperature under load

on Crysis is exactly the same (71C) as I got when I leave the case open. If that's a normal operating temperature, then it shouldn't crash.

#10
Posted 02/05/2011 10:28 PM   
Update: I've tested the card in another desktop with a 850W power supply (36 amp 12V supply) and the Crysis benchmark runs without problem. The case was closed and the maximum temperature was 80C.

Rather than conclude the problem is the power supply (most likely it is), I'll leave it at the problem is something other than the GPU. The strange thing is I used a multimeter to measure the 12V supply and the supply even under load was 12.0V. This isn't outside the allowed +- 5% tolerance.
Update: I've tested the card in another desktop with a 850W power supply (36 amp 12V supply) and the Crysis benchmark runs without problem. The case was closed and the maximum temperature was 80C.



Rather than conclude the problem is the power supply (most likely it is), I'll leave it at the problem is something other than the GPU. The strange thing is I used a multimeter to measure the 12V supply and the supply even under load was 12.0V. This isn't outside the allowed +- 5% tolerance.

#11
Posted 02/08/2011 08:29 AM   
[quote name='Xavier Zepherios' date='05 February 2011 - 04:46 PM' timestamp='1296920805' post='1189308']
mine blows the hot air outside the case...that's better to keep it cool
[/quote]

No it's not. Almost all good aftermarket coolers exhaust hot air in the case, and yet they get lower temperatures then those with external exhaust.

Same like bottom placed PSU is not worse then top.




Neither PSU's, nor graphic card's purpose is to exhaust the hot air from the case.

That's what we have case fans for, and graphic card fan should ASAP and in the most efficient manner take the hot air of the graphic card itself.

Sorry for the /hijack.




[quote name='Xavier Zepherios' date='05 February 2011 - 04:46 PM' timestamp='1296920805' post='1189308']

mine blows the hot air outside the case...that's better to keep it cool





No it's not. Almost all good aftermarket coolers exhaust hot air in the case, and yet they get lower temperatures then those with external exhaust.



Same like bottom placed PSU is not worse then top.









Neither PSU's, nor graphic card's purpose is to exhaust the hot air from the case.



That's what we have case fans for, and graphic card fan should ASAP and in the most efficient manner take the hot air of the graphic card itself.



Sorry for the /hijack.








Image

#12
Posted 02/08/2011 06:32 PM   
That's provided you have good airflow (lots of fans)
he doesn't and may not be able to add case fans - best solution is a card that dumps the heat outside the case rather than inside in that case

if you have lots of air movement(fans) you can have the good non-stock coolers

I don't have a lot of case fans either. small tower (I gave my old full tower to my younger bro) - doing a new build with a new full tower this year
but seeing that my card gets under 55c at full load I don't have a temp issue - the stock cooler does a fine job
That's provided you have good airflow (lots of fans)

he doesn't and may not be able to add case fans - best solution is a card that dumps the heat outside the case rather than inside in that case



if you have lots of air movement(fans) you can have the good non-stock coolers



I don't have a lot of case fans either. small tower (I gave my old full tower to my younger bro) - doing a new build with a new full tower this year

but seeing that my card gets under 55c at full load I don't have a temp issue - the stock cooler does a fine job



ImageImage


Image

Join the NVidia Forum Team - Please Help Medical Research - Folding@home


#13
Posted 02/08/2011 07:10 PM   
Well, I've finally found the cause of my problem: over-heating/faulty memory.

I ran Prime 95 with 4 threads (I have Core I5 750) and it errors within a minute or 2. I then run it with 2 threads and still get an error. So the problem doesn't seem to be CPU heat related.

Then I take a look at the DDR3 RAM (1600 MhZ capable, but running @ 1333 MhZ) and notice there's 1 stick without a heat spreader (brandless EBay memory). I take it out, rerun Prime 95 (4 threads) and Crysis GPU benchmark and voila! no errors.

Note that I verified that the 3 remaining sticks (4GiB, 4GiB, 2GiB) are still capable of dual channel operation, to not skew the stress test.


This seems strange that memtest86 doesn't catch the error. I will ask the authors to make memtest86 multi-threaded so that it will catch overheating problems.
It looks like I will need to get a heat spreader. Can DDR3 memory be run without a heat spreader or is that a recipe for disaster?
Well, I've finally found the cause of my problem: over-heating/faulty memory.



I ran Prime 95 with 4 threads (I have Core I5 750) and it errors within a minute or 2. I then run it with 2 threads and still get an error. So the problem doesn't seem to be CPU heat related.



Then I take a look at the DDR3 RAM (1600 MhZ capable, but running @ 1333 MhZ) and notice there's 1 stick without a heat spreader (brandless EBay memory). I take it out, rerun Prime 95 (4 threads) and Crysis GPU benchmark and voila! no errors.



Note that I verified that the 3 remaining sticks (4GiB, 4GiB, 2GiB) are still capable of dual channel operation, to not skew the stress test.





This seems strange that memtest86 doesn't catch the error. I will ask the authors to make memtest86 multi-threaded so that it will catch overheating problems.

It looks like I will need to get a heat spreader. Can DDR3 memory be run without a heat spreader or is that a recipe for disaster?

#14
Posted 02/13/2011 01:24 PM   
WTF do u mean that you noticed "there's 1 stick without a heat spreader "?

Did it suddenly fall off? Was it always like that? You noticed just now WTF? Are you running different RAM models?

Unless some heavy OC + overvoltage is applied, RAM overheating is a non issue.

Mixing different RAMs or a vibrating case in which heat spreaders fall of by themselves OTOH...

WTF do u mean that you noticed "there's 1 stick without a heat spreader "?



Did it suddenly fall off? Was it always like that? You noticed just now WTF? Are you running different RAM models?



Unless some heavy OC + overvoltage is applied, RAM overheating is a non issue.



Mixing different RAMs or a vibrating case in which heat spreaders fall of by themselves OTOH...


Image

#15
Posted 02/14/2011 02:15 AM   
  1 / 2    
Scroll To Top