GTX 970 Dying - Need Replacement ASAP

T

the.ronin

Limp Gawd
Joined
Mar 25, 2008
Messages
508
I'm pretty sure my EVGA GTX 970 is on its last legs. I do a lot of gaming for very long stretches namely in a game called Ark Survival Evolved. Only while playing, I started to get blue screens and memory allocation errors. Since some recent driver upgrades, things started to stutter which I determined was not server lag. The fans have also been running much higher than normal. At one point after a crash, my computer would be stuck in a pre-post loop which was fixed only after a cmos reset. I'm running an i7 920 which I ratcheted down from 3.9 to 3.7 to be safe. I've switched all my game settings from what was generally high / epic to medium and the crashing and stuttering has only gotten worse.

I've run Windows memory test and everything was fine. I checked my SSDs with CrystalDisk and all were in good health. I am using a Corsair HX750 power supply which I think is very reliable but have not replaced it or tested to see if there were problems with it. I tend to think since the issues only occur while I'm playing a game, it is likely video card related.

At this point in my life, I can't afford an upgrade. Keeping things under $200 for a replacement 970 GTX would be great otherwise I'd be willing to consider under $300 for an AMD or Nvidia replacement you might suggest. New or used makes no difference to me.

Thank you for any advice! I need to get back to the Ark before the EU raiders from Ragnarok arrive!! LOL @ the things we do.
 
They way you describe the issues I think it point more towards CPU/Mobo than GPU.. and the fact that you reduced graphical settings (Thus increasing CPU burden) increase the crashing issues and stuttering point to that even more, pre-post boot loop? also point to CPU/Mobo.. so I honestly think the issue is not related to GPU more than getting Dusted (explain the increased fan speed to keep temps below limits)
 
Araxie said:
They way you describe the issues I think it point more towards CPU/Mobo than GPU.. and the fact that you reduced graphical settings (Thus increasing CPU burden) increase the crashing issues and stuttering point to that even more, pre-post boot loop? also point to CPU/Mobo.. so I honestly think the issue is not related to GPU more than getting Dusted (explain the increased fan speed to keep temps below limits)
Click to expand...

Agreed. This doesn't sound GPU. This sounds memory to me. Windows Memory Test is not a solid test. Memtest86 or HCI MEmtest is what I'd recommend.
 
Shoot ok thanks guys. The GPU is relatively new compared to all the CPU, mobo and RAM. I did switch out the ram sticks to see if that would help.

I'll run Memtest86 and see if that says anything. Any recommendations to test the CPU and mobo? Infos of my current build are in my sig.
 
the.ronin said:
Shoot ok thanks guys. The GPU is relatively new compared to all the CPU, mobo and RAM. I did switch out the ram sticks to see if that would help.

I'll run Memtest86 and see if that says anything. Any recommendations to test the CPU and mobo? Infos of my current build are in my sig.
Click to expand...

Prime95
 
Are you using any hardware monitoring software so you can watch temps, usage, and such?
 
Take your 970 out, put it in another machine. Play the same game, reply with your results.
 
Alright so I ran one pass with Memtest86 and 0 errors on default. I plan to run it overnight to see if anything comes up with more passes.

I also ran the Intel Burn Test. It again passed on default settings 10 passes.

Any recommendation for GPU tester? I'm looking into FurMark and others.

And yes, I do use HWMonitor now and then - I'll get a log going. Unfortunately, I don't have a separate gaming computer lying around to test.
 
Sorry for the back-to-back post ... but I ran Furmark and the GPU peaked at 88/89C and remained stable from there. I only ran it for about 3 minutes since it stuck at 88/89 after 2 minutes into the burn. The fan was going at 100%.

Now I'm not sure at all what the problem is??
 
the.ronin said:
Alright so I ran one pass with Memtest86 and 0 errors on default. I plan to run it overnight to see if anything comes up with more passes.

I also ran the Intel Burn Test. It again passed on default settings 10 passes.

Any recommendation for GPU tester? I'm looking into FurMark and others.

And yes, I do use HWMonitor now and then - I'll get a log going. Unfortunately, I don't have a separate gaming computer lying around to test.
Click to expand...

memtest should be ran for at least 12 hours to have a reliable result, I do prefer at least 24h.. but 12h should be OK.

Furmark, it can point to a PSU Failure and/or GPU failure if you left running for at least 30 minutes, and well I would recommend to run a realworld application to test the system, such Asus RealBench in Benchmarking loops.. it test both system, CPU and GPU together and separately so it's a good indicator on what's going on with the system.. Also, what model exactly of the GTX 970 you have? 88-89 it's really very hot for a low power GPU like that, specially at 100% fan speed.

Also you can try Aida64 system stability test first CPU+FPU+Cache for at least 2 hours, then CPU+FPU+Cache+RAM for at least 2 hours and then add the GPU to the mix to have a complete system test. I know it's very long time but it's also good at pointing what's really wrong..
 
Those temps are high but with the blower cooler not incredibly surprising. What does it get to in ark after you play for 10mins or so?
 
I think you'd be better taking that blower off & zip tying a few high quality PWM fans to the heatsink instead. If its out of warranty though, you should tear it down and clean it & apply new thermal paste to the GPU.
 
You said you are on an i7 920, right? That's Nehalem with triple-channel memory. I had similar problems on an Alienware Aurora R1 with the same chip, and it turned out that over time, the CPU "shifts" in the socket just a hair and it stops seeing one of the RAM channels (reports seeing the correct amount of RAM, but only 2/3 of is listed as useable on the Windows System Info screen). Reading around, this was a common problem for these systems as they aged, and the big symptoms were random blue screens and weird slow downs, just as you said. The solution (for me) was to reseat the CPU and cooler. Some people have had to replace the mainboard to fix it, though :(
 
Last edited:
Yeah I would agree sounds more like Memory and CPU than GPU, would also agree with the GPU temp is abnormally high for a 970. Good ol system tear down and cleaning might be in the works.
 
Hey guys thanks again so much for all of this feedback.

I actually just put in a good hour or so of gaming. But first with Day of Infamy which is arguably far less resource intensive than Ark. No problems, stutters, or any issues at all on high settings. GPU temps maxed at 80C. Then I threw on Ark and immediately experienced stuttering and a GPU crash (unable to allocate memory). I went back and set everything to Low settings and while I still get some slight stutters, I have not crashed. Temps after this were 82C max.

I did reseat the RAM and the GPU but did not touch the CPU - I'll check to see if I have any arctic silver left and will try that.

mvmiller12, in the Windows System Info screen, I'm seeing:

Installed 12.0GB
Total Physical Memory 1.99GB
Available Physical Memory 392MB
Total Virtual Memory 7.98GB
Available Virtual Memory 4.71GB

I have 6x 2GB DDR3 ram installed - so it seems to be seeing the correct installed size but is that Total Physical Memory being so low right?
 
No, that is very, very wrong. Your total physical memory should match your installed.
 
Thank you guys for clarifying that!

Ok so I'll inspect the pins on the CPU make sure nothing looks off and reseat it. Of course, can't find the arctic silver so may need to wait a bit.

Could it still be an issue with the RAM itself rather than the CPU not recognizing? Would that single pass on memtest provide some confidence at least tthat it is the CPU and not the RAM?

Thanks again everyone very much.
 
I just noticed something odd - in CPU-Z, under the SPD tabs and cycling through the 6 slots:

* slots 1, 3 and 6 have the same part number
* slots 2, 4, and 5 have the same part number

Is this supposed to be like this?
 
the.ronin said:
Thank you guys for clarifying that!

Ok so I'll inspect the pins on the CPU make sure nothing looks off and reseat it. Of course, can't find the arctic silver so may need to wait a bit.

Could it still be an issue with the RAM itself rather than the CPU not recognizing? Would that single pass on memtest provide some confidence at least tthat it is the CPU and not the RAM?

Thanks again everyone very much.
Click to expand...

My experience was that the RAM itself was fine, and would test fine in MEMTest, but it would only test the RAM it saw as physically attached. YMMV. The Nehalem CPUs don't have pins, the chips are LGA, but it would not be a bad idea to look at the pins in the CPU socket to make sure they are OK looking. It might not be a bad idea to also clean the CPU with Isopropyl Alcohol as well (90%). I had to reseat the CPU in that Alienware twice before it settled down correctly. That system is currently owned by my cousin and he hasn't had any trouble with it since the reseatings.

Good luck, man. I hope you're able to get it sorted. While you're at it, you should probably clean out the blower on your 970. The temps you were getting on it were a bit high, indicating either poor airflow around the card, or a dirty fan/heatsink on it.
 
the.ronin said:
I just noticed something odd - in CPU-Z, under the SPD tabs and cycling through the 6 slots:

* slots 1, 3 and 6 have the same part number
* slots 2, 4, and 5 have the same part number

Is this supposed to be like this?
Click to expand...

I can't speak to this - I do not recall CPU-Z reports on that machine, and it is not readily available to me to check.
 
Yes i have an x58 myself.....it does sound like your mb/cpu/ram is having SERIOUS issues....Also try running IBT on max settings If/when possible to stress all your ram. My guess is your MB is starting to go out or needs and or needs so voltage adjustments. Where exactly are your checking Windows System Info screen with the exact info? Does cpu-z agree with it?
 
the.ronin said:
Installed 12.0GB
Total Physical Memory 1.99GB
Available Physical Memory 392MB
Total Virtual Memory 7.98GB
Available Virtual Memory 4.71GB
Click to expand...

I got some thermal paste and did some cleaning. Now I'm at 5.99GB of Total Physical Memory and 4.10GB of Available Physical Memory. So a step in the right direction? LOL.

I'm on Windows 10 Pro x64 so it should be recognizing all 12GB right? Anyway I'm going to play some Ark see if this made that big of a difference.
 
Quick update ... been playing for just about hallf an hour and so far a huge difference. Been able to crank up the graphics again but the smoothness is a huge difference. I dont play with FPS indicators but this is a big difference. Temps have been 82C max.
 
the.ronin said:
I got some thermal paste and did some cleaning. Now I'm at 5.99GB of Total Physical Memory and 4.10GB of Available Physical Memory. So a step in the right direction? LOL.

I'm on Windows 10 Pro x64 so it should be recognizing all 12GB right? Anyway I'm going to play some Ark see if this made that big of a difference.
Click to expand...

Yes you should see all 12GB with Win7/8/10 x64.

Have you tried using a rubber eraser to clean the RAM contacts? This has solved memory related issues for me on occasion with C2D and earlier machines.
 
Yes, you should see all 12GB of RAM under Total Physical Memory. I'd try the reseat again - if you can't get all 12GB to show, your mainboard may be defective. Again, Google search reveals that this seems to happen a lot with Intel X58-based systems. :(

As it is, you aren't getting the full advantage of your RAM.
 
Thanks guys. I'll try the rubber eraser. I cleaned the CPU contacts but only blew air into the RAM slots. Don't think it matters but I also matched the part numbers to the two color coded channels. As it stands though, I've never seen this game run so smoothly ... lol right??
 
I used to run an i7 950 and it was the absolute pits to get it to recognize all 6 slots at once. It would say it had 12gb installed but only 6gb available. I ended up changing out my 6 2gb sticks for 3 8gb sticks. Much MUCH happier system.
 
Jorona said:
I used to run an i7 950 and it was the absolute pits to get it to recognize all 6 slots at once. It would say it had 12gb installed but only 6gb available. I ended up changing out my 6 2gb sticks for 3 8gb sticks. Much MUCH happier system.
Click to expand...
I may do that as a stopgap until I do an upgrade. Mind if I ask which ones you went with?

Also, how can I be sure that it's an entire channel that's bad rather than single slots from both channels?
 
the.ronin said:
I may do that as a stopgap until I do an upgrade. Mind if I ask which ones you went with?

Also, how can I be sure that it's an entire channel that's bad rather than single slots from both channels?
Click to expand...
Pretty much EVERY board has an easier time running 3 sticks of ram vs 6......mine doesn't seem to mind but thats been that way since the dark ages....Why cause it eliminates the possibility of issues by half. Weather its failing ram or Mb slots is a coin toss. In general overclocking the ram is easier with fewer sticks as well.
 
Intel has a diagnostic. You can run all or some of the tests in each iteration. It is NOT the Intel Burn Test. I forget where to get it. Try intel's website.

I found an error on the first run. Each run took about 5 minutes, doing all the tests. The next time that error occurred was run 242. I got lucky finding it right away. Very lucky.

I'd run it on all tests for 24 hours. It'll test every friggin' transistor in your cpu and every electron in your RAM. (Okay, I -may- be exaggerating.) It found my error when nothing else showed anything wrong.

For you? Obviously something's up with your memory. It could be the RAM, the mobo, or the CPU. Test 'em all...
 
the.ronin said:
I'm pretty sure my EVGA GTX 970 is on its last legs. I do a lot of gaming for very long stretches namely in a game called Ark Survival Evolved. Only while playing, I started to get blue screens and memory allocation errors. Since some recent driver upgrades, things started to stutter which I determined was not server lag. The fans have also been running much higher than normal. At one point after a crash, my computer would be stuck in a pre-post loop which was fixed only after a cmos reset. I'm running an i7 920 which I ratcheted down from 3.9 to 3.7 to be safe. I've switched all my game settings from what was generally high / epic to medium and the crashing and stuttering has only gotten worse.

I've run Windows memory test and everything was fine. I checked my SSDs with CrystalDisk and all were in good health. I am using a Corsair HX750 power supply which I think is very reliable but have not replaced it or tested to see if there were problems with it. I tend to think since the issues only occur while I'm playing a game, it is likely video card related.

At this point in my life, I can't afford an upgrade. Keeping things under $200 for a replacement 970 GTX would be great otherwise I'd be willing to consider under $300 for an AMD or Nvidia replacement you might suggest. New or used makes no difference to me.

Thank you for any advice! I need to get back to the Ark before the EU raiders from Ragnarok arrive!! LOL @ the things we do.
Click to expand...

If the card is under warranty send it in for a replacement. If you waited too long to send it in and the warranty expired then there's nothing one can say except it sounds like another 970 will work for you and is definitely the less expensive solution
 
Maybe Side-grade to a GTX 1060 3GB? Here's the best price for one I've seen today. Bear in mind, you'll lose 500mb of RAM, but performance is the same or better.
 
Hey guys thanks again for all the great advice - without you I would've likely continued to think it was the GPU.

So I've tried to clean out the ram and also re-arranged and did more general mobo cleaning mostly with an air can - still getting half of the 12GB recognized. But like I mentioned, my games run very well so at least it's a good temporary fix.

Given that Cyber Monday is coming up though, I'm thinking of just getting a CPU/mobo/ram upgrade likely to a Rysen 5 (maybe 7 if the deals are good enough). I think even with the 5, it will still be a decent enough upgrade from my existing i7 920 setup thats clearly on its last legs. Leaning from this experience, I also just plan to go with x2 8GB ram rather than x4 4GB.

Thanks again guys, really, SUCH a huge help.
 
the.ronin said:
Hey guys thanks again for all the great advice - without you I would've likely continued to think it was the GPU.

So I've tried to clean out the ram and also re-arranged and did more general mobo cleaning mostly with an air can - still getting half of the 12GB recognized. But like I mentioned, my games run very well so at least it's a good temporary fix.

Given that Cyber Monday is coming up though, I'm thinking of just getting a CPU/mobo/ram upgrade likely to a Rysen 5 (maybe 7 if the deals are good enough). I think even with the 5, it will still be a decent enough upgrade from my existing i7 920 setup thats clearly on its last legs. Leaning from this experience, I also just plan to go with x2 8GB ram rather than x4 4GB.

Thanks again guys, really, SUCH a huge help.
Click to expand...
Cant blame you there as our boards weren't designed to go on forever.....I got way more cpu power that your's an even i am looking to the "next" build.....im either going AMD x399 (STR4) or Intel x299 (LGA2066)
 
You must log in or register to reply here.
Back
Top