New PC Build With Graphics Dropouts

[phsyraxion]

I'm not new to PC builds but components are a LOT more powerful than 5 years ago on my last big build and I typically can sort out all my issues as I go but I need some assistance with this one since I have not had hardware failures before.

New build and all hardware components in my profile. Issue I am having is when I tested the game GTAIV after playing many other games (Borderlands, NFS, Gas Guzzlers etc) I started to have lockups of the video card.

I have 3 monitors connected to this cars (only one for gaming obviously) with the main 42" monitor connected via HDMI and the 2 smaller monitors via DVI. I have CPU and all components running at stock speeds with no overclocking.

When I play GTA I can play for about 10 minutes before aall 3 displays will turn to rainbow colours and then all go blank. Sometimes the PC will reboot and sometimes it will be "frozen" but the game audio is still there so this is where I believe I have a video card issue and probably power or heat related.

I have a system monitor running for all CPU and GPU temps while I play and I note that the GPU reaches about 70c when the cutout occurs and the CPU is at about 42c. Looking at power consumption of the P/S it seems the system is only demanding 400W max from the supply so there is loads of headroom.

Where I believe I may have something wrong with my installation is the card has 2 PCIe power connectors on it. My card has 2 x 8pin connectors but Gigabyte specs say that a 6+8pin connector are fitted. I have a SINGLE 8 pin power feed from my PSU which has a TWIN 8 pin at the end and I have this single power feed supplying the card. I have not worked with a monster card like this and I am thinking I should be using 2 separate feeds from the PSU? I am mainly asking for ideas while at work to try and get a plan for when I start troubleshooting.

 

[flmatter]

First I would try gaming on just one monitor try one of the 22" you have and if that is fine then move up to the 42". If that is fine connect both 22's and try again. Sounds to me like your video card is running out of memory pushing 3 monitors, especially that 42", and locking up. As long as you have the 2 8 pin power connected you should be fine with that single rail, you have plenty of power with that 760W ps. You can try using 2 separate rails but I believe your issue is from trying to push 3 monitors. Downsize to the 22's and see if you have the same problem or just single monitor either 22 or 42.

 

[Solaris17]

STUFF

Where I believe I may have something wrong with my installation is the card has 2 PCIe power connectors on it. My card has 2 x 8pin connectors but Gigabyte specs say that a 6+8pin connector are fitted. I have a SINGLE 8 pin power feed from my PSU which has a TWIN 8 pin at the end and I have this single power feed supplying the card. I have not worked with a monster card like this and I am thinking I should be using 2 separate feeds from the PSU? I am mainly asking for ideas while at work to try and get a plan for when I start troubleshooting.

BINGO!

is it TWIN 8 pins? (16) or twin 4 pins? (8) because 2x 4pin blocks go to the CPU

Also you should always run 2 diffirent rails.

though to be honest this sounds like a memory problem with your GPU.

 

[phsyraxion]

Thanks for the reply.

I considered that it could be because of the multi-monitor setup but really hope not. I spcifically chose the 4GB card because of the claims it can run 4 full HD monitors and I need the 3 monitors for my production and video editing work. My previous i7 920 CPU / GT 220 rig ran 2 x 22" LCD for close to 5 years without an issue.

While monitoring the temp and GPU memory the memory usage with all monitors on would only be about 60% usage according to the bar graph readout so it isn't as if it is running out of memory.

One thing I forgot to mention was the cards cooling didn't appear to be working hard enough. Even when the GPU reported 70c the fans still only worked at about 20% speed (so quiet I couldn't hear them). I then used my Corsair utility to adjust the temp curve to push the fans up to 80% when the GPU reached 55c and I ran out of time to test but it appeared I was ble to run for a good 30 minutes without a crash instead of 10.

I also updated my MOBO BIOS from the stock supplied bios to the latest so not sure if that could make a difference here.

I will run more tests tonight but I mainly wanted to get advice on the power rail connector if you needed 2 separate rails or if 1 was sufficient. I did test the card on a different rail with the same result.

 

[phsyraxion]

BINGO!

is it TWIN 8 pins? (16) or twin 4 pins? (8) because 2x 4pin blocks go to the CPU

Also you should always run 2 diffirent rails.

though to be honest this sounds like a memory problem with your GPU.

It's twin 8 pins (2 sets of 8 pin).

This is where I am trying to find what the issue isnt so I can focus where the issue IS. The first 2 posts though contradict each other with one saying a single power rail should be fine and one saying you should always use 2 rails. Since I am only using a single rail I will try and add another rail and see if it helps the situation as a start.

EDIT: This is the card I have: http://www.gigabyte.com.au/products/product-page.aspx?pid=4647#ov

As far as I can tell it should allow 3 monitors to display fine while gaming.

 

[Solaris17]

He may be right its not necessarily a contradiction. if one works it works but to rule it out use 2. you couldn't possibly over stressing the RAIL (not the psu) if you split it. I am actually leaning toward a ram issue. VRAM not RAM the next step might be to see if you can crank furmark or something absolutely maxxed out to try and stress things.

you might be in for a benchmark session but not for numbers you will be attempting to count artifacts.

 

[phsyraxion]

Cool

I will certainly run a 2 rail supply just to test. It's very rare I have issues like this which is why I am needing to turn to more experienced minds. It also appears to only be an issue with GTAIV on maxed settings which would make sense on the VRAM since that would require a huge amount of video processing and memory which apparently my card has but I am not even sure if it is all accessible.

I will run a test with Furmark since I only stress tested my CPU and RAM when I did my overclocking (which I have removed while troubleshooting this issue). In the past I never needed to stress test GPU but that should tell me some more.

Would that also test the VRAM for errors too?

 

[flmatter]

I said single rail because that is what I have from my PS, a single rail with a 6 pin and a 6+2 pin on it. If your PS has 2 dedicated rails for pci-e power use them.

There is a contradiction that you can use that card for 3 monitors. Some say yes and others say no unless you sli them, depends on which google site you go to. The issue at hand is and is opposite of everyone else's set up, is your driving a 42" and 2 22" monitors while everything I have seen are multi 27" and smaller. Try your 22" first then the 42" if both test fine singly then move on to dual 22's. If they duals test fine and benchmark out ok, go back to the triple set up, benchmark it/test it and if problem returns now you where problem is. I know my 760 2gb can push my 42" Toshiba in 1080p but it is not a happy card gaming and games will sometime lag a bit and so I went back to my 23" Acer, maybe get a 24" LED one of these days. Could be your card runs out of Vram quickly pushing all 3 monitors.

Yes the furmark will test/stress your vram. But start testing easy, one monitor at a time. 22" then 42" then dual 22" then one 22" and the 42".

 

[phsyraxion]

Interesting.

So it requires more power/memory to display an image on a 42" over a 22" even when set at the same resolution?

 

[flmatter]

the more power/ram is for displaying on multiple monitors. 3 monitors, 3 times the amount of pixels the 770 has to produce.

as for the difference between 42 vs 23(or 22 in your case) I spoke from my personal gtx760 2gb experience, I could have had my game settings way to high for my lack of performance BUT same settings on my 23" monitor and I have no issues. I just know my card/system was not happy with my 42 Toshiba. 1080 should be 1080 no matter what the real estate it is on.

So you can try benchmarking the 42" first then the 42 with a 22 and see where you stand, if you get a stable system with1 or 2 monitors and not 3(then you will need a second 770 to sli and run the 3rd monitor), you know where the problem lie's. If you benchmark single or dual monitors and have same problems then the issues is the video card and rma it.

 

[phsyraxion]

OK so I have run some tests and found pretty consistent results. All tests have failed but under controlled and repeatable conditions which should make this a bit easier.

Firstly I connected 2 separate power rails to the card.
I then disabled the other monitors using just a single monitor (the 42") on HDMI

I ran a test using the 1080 15min Burn In test and after about 30 seconds the temp on the GPU climbed from 38c to 68c and the video went rainbow and failed. It did dump me back to the desktop though with the attached error.

I then used the corsair tool which controls all chassis fans to increase the fan speed on the card to 100% from start (the card does not seem to be increasing the fans even with all fan control software unloaded and disabled) and ran the test again. This time it lasted a bit longer till 1 minute and again the temp had reached about 67 and the card failed and this time the machine rebooted. I did the above test 3 times with the exact same result each time.

I then tested the card at 720 15min test and it lasted about 1:30 before the same issue occured. I tried twice with the same result both times.

To me it sounds like an issue with the card if it cannot even display 1 monitor but the temp also seems to climb just a bit too quick. I havent overclocked the card but I believe this is a factory overclocked GPU. Would lowering the overclock solve the issue or should I look at taking the card back.

 

[Shou Miko]

i like how to flipped in ur i7-4770k into a AsRock X79 Extreme9 board, got a tutorial for that?

use the newest Nvidia Driver even Beta and try set PhysX to GPU (GTX 770) in Nvidia Control panel and see where that leads u ^^

 

[phsyraxion]

i like how to flipped in ur i7-4770k into a AsRock X79 Extreme9 board, got a tutorial for that?

use the newest Nvidia Driver even Beta and try set PhysX to GPU (GTX 770) in Nvidia Control panel and see where that leads u ^^

Well picked... Was going off what I thought I ordered at work. It's an i7 4820K

I have the absolute latest driver and there is no Beta that can be found. I will give your suggestion a try.

 

[phsyraxion]

I changed the PhysX setting from Auto to the GPU and I still had the same crash but the test at 1080 lasted about 1 minute longer this time.

EDIT: I just ran the FurMark on another machine I have here which has a GTX660 with a 4:3 LCD at 1280x1024 and that ran a full screen test for 10 minutes and peaked temp at 71. Knowing that card seems good I might rip that out, uninstall these drivers and see if it passes in this new build. That will have to wait for tomorrow. Grrr, all I want to do is play a couple of games

 

[Shou Miko]

could it be a unstable card u have gotten?

bcs i know from the 5x0 series Gigabyte had some cards that didn't receive correct voltage of their overclocked cards so they crashed or was in basic just unstable.

 

[phsyraxion]

could it be a unstable card u have gotten?

bcs i know from the 5x0 series Gigabyte had some cards that didn't receive correct voltage of their overclocked cards so they crashed or was in basic just unstable.

Certainly could be which is what I am trying to figure out. My PSY should be able to deliver upwards of 63 Amps on the 12v rail and has no complaints doing a full system stress test on ram+CPU. Additionally when there have been failures of the GPU on occasion the PC still seems to be "operating" as there is still audio being processed. It all points to the PSU not being the issue but I really need to be sure to avoid paying for any tests they will do on the card when I take it back.

Tonight I will take the card out and put it in a rig running a GTX660 and do the same stress test. The 660 passes just fine so if the 770 fails in another rig I will take it back for replacement. As someone suggested earlier it is starting to look like a memory issue.

 

[Solaris17]

Looking at your results it looks like you have a dude card friend.

 

[phsyraxion]

Certainly looks that way. All warranty covered and I am hosting a big LAN weekend in 2 weeks so i'm glad it's been discovered now. Will post back when I determine what the outcomes are. Was hoping to start setting up the system with SSDs this weekend but that looks blown to taking back hardware and they aint going to swap it on the spot. Could take weeks. Ugh..

 

[phsyraxion]

So I ripped out the card and tested in a different rig running a 680w psu and was running a gtx660 and same result. Fails exactly the same so looks like a dud card

 

[Shou Miko]

So I ripped out the card and tested in a different rig running a 680w psu and was running a gtx660 and same result. Fails exactly the same so looks like a dud card

if you still got any warranty left on ur card, ship it back to get a replacement or try adding more voltage with MSI AB or another OC program to see if that help.