GTX 570 Woes

Dan221081 · 12 Dec 2011 at 00:32

Until recently I my EVGA GTX570 SC has been reliable, however over the last couple of months its started to crash. The crashes resemble hard resets, and while this initially started out as occasional, they have grown increasingly regular.

The other day it crashed which it would not recover from and I had to leave the machine for a couple of hours before I could start it up again. The system would start fine but nothing would get sent to the monitor (tried all the usual stuff connections etc).

My motherboard now shows a red light occasionally when I experience one of these crash / hard resets and go to start the system again. The strange thing is this started happening after nothing had really changed. I hadn't move any hardware around or made any significant software changes. It just started to get worse.

I have now tried just about everything I can think of :-

- Tried latest Beta Nvidia Drivers - and Reference Drivers ( No Difference)
- Flashed the BIOS of the graphics card. ( No Difference )
- Flashed my motherboards BIOS. ( No Difference )
- Tried a different graphics card in my system (which worked fine without any crashing even though it generated just as much heat when running)
- Tried underclocking the graphics card ( Slightly lower temps but still crashed and hard reset the system )
- Tried resetting the CMOS on the motherboard. ( No Difference )

Upon doing the above I have come to the conclusion that the issue must be a hardware fault that has developed over time with the graphics card and its time to send it back to OCUK for them to have a look at it.

Essentially it sometimes boots fine but then when you are using it, it will either about 1 - 3 hours of constant use lock up with a blue screen of death or hard reset your system and then either wont boot for a while and will fail any POST test the MB throws at it.

I don't have another system to try it in, however I am confident the results will be the same if I placed the card in another system and pushed it for a while.

Anything else you think I could try before sending it back ?

Cleeecooo · 12 Dec 2011 at 00:35

You seem to have tried everything!

You could try budging the voltage on the graphics card up a little?

Also, you may want to RMA straight to EVGA as it'll save you avfew days. Downside is you have to pay postage to Germany

Dan221081 · 12 Dec 2011 at 01:11

Yeah I was afraid that was the case

I will give them a ring tomorrow, I didn't want to jump into voltage changes in case that would void the warranty.

I will look to get another card while this one is being processed if it will take a while to do.

Shame BFG don't do 570's my BFG 260 (which I am having to use now) is built like a tank and has been rock solid. As it stands looking at the selection available from OCUK it will either be a

http://www.overclockers.co.uk/showproduct.php?prodid=GX-129-MS&groupid=701&catid=1914&subcat=1010

or possibly a

http://www.overclockers.co.uk/showproduct.php?prodid=GX-116-MS

I just hope this next one is more reliable. At least those two cards have decent fans. Just wish they did non-over clocked versions with the same fans !!

Jakus · 12 Dec 2011 at 02:18

And why do You think it's the GFX Card ?

RichardH754 · 12 Dec 2011 at 06:50

Could be PSU failing.

Dan221081 · 12 Dec 2011 at 17:51

Yeah could be the PSU but its new and it certainly kicks out enough juice for what I am running so in this case I think its unlikely.

I rang OCUK this afternoon and explained what I have tried and they agreed that it definitely sounds like the graphics card.

I have got an RMA code and will be sending it off tomorrow.

Thanks for you your help

ALXAndy · 12 Dec 2011 at 17:57

What temps are you hitting when it switches off?

My friend's 460 2win was shutting down his system all the time. Pooey NZXT phantom, cards were reaching 110c thermal shutdown.

Thing is, it was only certain games using certain engines that cause it to happen.

Example - Need for speed : The runs. It made my 295s hit 104c during a cut scene and switched off the PC. TRINE 2 makes my 6970L hit 75c and 100% fans, and so on.

Some games are super hard on the GPU, causing it to reach max temps very quickly. I would install Afterburner and monitor the temps to a file. If it happens? reboot and look at what temps you were hitting..

I had the NFS the runs issue with the latest Nvidia beta drivers. They stopped my card's fans hitting 100% and were stopping at 75%. Not enough when the cards are parked next to eachother and overheating. Nothing else made that happen, though.

Dan221081 · 12 Dec 2011 at 18:21

Yeah I have Afterburner installed and I also had EVGA Precision installed just in case that gave me something different

Here is a clip from the logs :-
02, 11-12-2011 01:26:13, GPU temperature ,GPU usage ,Fan speed ,Fan tachometer ,Core clock ,Shader clock ,Memory clock ,Memory usage
80, 11-12-2011 11:20:12, 88.000 ,99.000 ,73.000 ,3030.000 ,735.000 ,1470.000 ,1803.000 ,951.770
80, 11-12-2011 11:20:13, 89.000 ,99.000 ,73.000 ,3030.000 ,735.000 ,1470.000 ,1803.000 ,951.770
80, 11-12-2011 11:20:14, 89.000 ,99.000 ,73.000 ,3060.000 ,735.000 ,1470.000 ,1803.000 ,951.395
80, 11-12-2011 11:20:15, 88.000 ,99.000 ,73.000 ,3030.000 ,735.000 ,1470.000 ,1803.000 ,951.695
80, 11-12-2011 11:20:16, 88.000 ,99.000 ,73.000 ,3030.000 ,735.000 ,1470.000 ,1803.000 ,952.063
80, 11-12-2011 11:20:17, 88.000 ,99.000 ,73.000 ,3030.000 ,735.000 ,1470.000 ,1803.000 ,952.438
80, 11-12-2011 11:20:18, 89.000 ,99.000 ,73.000 ,3060.000 ,735.000 ,1470.000 ,1803.000 ,953.098
80, 11-12-2011 11:20:19, 88.000 ,99.000 ,73.000 ,3000.000 ,735.000 ,1470.000 ,1803.000 ,953.281
80, 11-12-2011 11:20:20, 88.000 ,99.000 ,73.000 ,3030.000 ,735.000 ,1470.000 ,1803.000 ,953.281
80, 11-12-2011 11:20:21, 88.000

Pasted the headings in so ignore the very first time stamp

Cleeecooo · 12 Dec 2011 at 18:22

all of that looks normal to me ?!

Dan221081 · 12 Dec 2011 at 18:31

90 Degrees is very hot, but apparently that can be normal, I would put the card in start up a game, play it for about 20 minutes it would crash and the card would literally be too hot to touch I would have to wait for it to cool down before I could take it out. Perhaps the temp sensor is faulty ? I am out of ideas....if I take that card out and put another one in that also gets pretty damn hot I have no crashes and the system is rock solid.

The chap at OCUK said he had encountered this before, but its a new one to me !

I have had plenty of graphics cards start producing artefacts (usually green) and then start to crash regularly which usually means the memory has gone, and I have had cards completely die on me but not this before... I am going to make sure I get a cooler card next time...

Cleeecooo · 12 Dec 2011 at 18:49

Dan221081 said:
90 Degrees is very hot, but apparently that can be normal, I would put the card in start up a game, play it for about 20 minutes it would crash and the card would literally be too hot to touch I would have to wait for it to cool down before I could take it out. Perhaps the temp sensor is faulty ? I am out of ideas....if I take that card out and put another one in that also gets pretty damn hot I have no crashes and the system is rock solid.

The chap at OCUK said he had encountered this before, but its a new one to me ! I have had plenty of graphics cards start producing artefacts (usually green) and then start to crash regularly which usually means the memory has gone, and I have had cards completely die on me but not this before... I am going to make sure I get a cooler card next time...

90's hot but not too hot.

My cards overheat a lot aswell, the cooler is pretty crappy TBH

Also, your sig's a line too big

ALXAndy · 12 Dec 2011 at 18:51

Dan221081 said:
90 Degrees is very hot, but apparently that can be normal

OK well according to whom? Nvidia?

For the past three years any electrical equipment needs to be made using ROHC lead free solder. Old solder had lead in, with tin (a alloy) and then flux resin (glue that goes hard) to keep it together.

As the card gets hot the PCB which is made of glass fibre will flex. Expand and contract. As time passes with non lead solder (which is harder and made from tin and doesn't flex as much) the joints will fail and crack.

Thus, after extensive research it was concluded that a GPU really shouldn't be ran over 80c, without seriously shortening the lifespan of the solder (hence why the oven bake it trick is a temporary fix, 'cos it reflows the solder).

Oddly my friend's 460 2win is also EVGA, and also shuts down at 90c. Yeah it's gone higher, but only for a split second, then wham, idle in L4D2, have to wait for him to reboot his PC.

The solution was 100% fans all the time....

I just find it a little odd that you also have an EVGA and EVGA give a nice big long warranty. 10 years isn't it?

Dan221081 · 15 Dec 2011 at 23:25

Well I ordered a MSI GTX570 Twin Fozr III which came today. Put it in and fired the machine up got the red light again, but continued. Fired up BF3 played it for about 20 minutes and then bam it did the same thing...

So I'm in new territory now, I have never had a power supply degrade on me before... Looking at the back of the power supply at the switch (its lit up) the light does appear to be quite faint and flicker slightly. I have no idea what that means ? Does that mean its on its way out ??

So I need to send back the graphics card tomorrow and then ask them to send me a PSU out asap. If that doesn't work then its the motherboard. If it is I am going to huuuuuugely annoyed. As it stands I will have to take the cooler off the CPU to get the PSU out and disassemble everything...

Can anyone recommend a decent power supply make ? The one I have in there at the moment is rated at 800W (although it' likely it isnt kicking that out anymore ....) If I am going to get one and spend £100+ on one I might as well get a decent one .

Any suggestions ?

Jakus · 16 Dec 2011 at 01:55

Yeah !

Start listening to the advice people give You !

I already said..... "And why do You think it's the GFX Card ? "

I suppose eventually If You waste more time and money replacing everything you will sort the problem, but why not start getting more analytical ?

I'll give you another clue, mains neons flicker all the time, it tells you nothing more than the psu has mains power connected !

When a system works perfectly but falls over after 20-30 mins at full load the fault is generally a heat related stability problem.
Electronics become less stable the hotter they get, component life is also considerably shortened when run at elevated temps.

welshdragon · 16 Dec 2011 at 16:37

Dan221081 said:
Can anyone recommend a decent power supply make ? The one I have in there at the moment is rated at 800W (although it' likely it isnt kicking that out anymore ....) If I am going to get one and spend £100+ on one I might as well get a decent one .

Any suggestions ?

Seasonic; Antec; Enermax. Any of those also rebadged ie XFX

Just having a quick look this would appear suitable

chaparral · 16 Dec 2011 at 16:51

Dan221081 said:
I have now tried just about everything I can think of :-

- Tried latest Beta Nvidia Drivers - and Reference Drivers ( No Difference)
- Flashed the BIOS of the graphics card. ( No Difference )
- Flashed my motherboards BIOS. ( No Difference )
- Tried a different graphics card in my system (which worked fine without any crashing even though it generated just as much heat when running)
- Tried underclocking the graphics card ( Slightly lower temps but still crashed and hard reset the system )
- Tried resetting the CMOS on the motherboard. ( No Difference )

Essentially it sometimes boots fine but then when you are using it, it will either about 1 - 3 hours of constant use lock up with a blue screen of death or hard reset your system and then either wont boot for a while and will fail any POST test the MB throws at it.

Anything else you think I could try before sending it back ?

It sounds like the problems you get when the CPU is overclocked to far

Have you tryed with your cpu at it stock speed ??............(As this would been the first thing i would have tested )
Also What is your CPU temp under full load ??

Dan221081 · 16 Dec 2011 at 21:09

Temps for the CPU appears to be stable at around 34-40 idle and maxing out around 55 which seems fine when over clocked.

I have been doing some testing with the CPU running at stock speeds. Which appears to make no difference did a reset to defaults to make sure in the BIOS but still the same effect.

Powering up the machine and going into the BIOS and watching the temps on IDLE and it looks like the MB temps are higher than the CPU temps most of the time. CPU around 37 MB at around 39.

The only thing I have noticed is when I have a 570 in there the chipset gets really hot. On the P8P67Pro it has a blue heat sink on top of the chipset which the graphics card practically sits on top of which you don't appear to be able to take off (ruling out the possibility of the chipset heatsink loosing connectivity with the actual chipset).

I have an 900W Quad Rail Antec PSU (Thanks welshdragon) coming tomorrow to replace my el -cheapo PC World one thats in there at the moment which I was intending on replacing anyway. Will have to see if that makes any difference.

Fun and Games !!

Thanks for all the suggestions guys

Jakus · 16 Dec 2011 at 21:22

Why don't You try it with the side off and position a big desk fan blowing onto the motherboard ?

It should eliminate memory and chipset overheating

Dan221081 · 16 Dec 2011 at 22:12

Yeah I don't have a desk fan but I have had the case open for all my testing I will try and get hold of one tomorrow.

The following show details of my system running in stock :-

Idle :

Full Load

CPU-Z

Silly thing is
prior to that full load test screenshot being taken I had Photoshop open about 20 webpages , MSN , and BF3 running for about 40 minutes and it was rock solid and that must have been pushing the system pretty hard (inc the chipset all of this was done using the GTX260) I was alt tabbing and jumping about the place and it was all responding well..

I am keeping my fingers crossed for the PSU tomorrow :S

Jakus · 17 Dec 2011 at 10:16

If you have an Asus mobo You can run probe and log (i think) the voltages, In 10 mins you will know if the power supply is not holding up !

I'm really trying to help you here but I don't understand why You want to keep replacing parts before doing some basic testing :confused: