PSU powerful enough? 5970 + i7-920. Artifacts? Other strange issues.

Jamrock

Distinguished
Nov 16, 2009
233
0
18,690
Hey all,

So I picked up Metro 2033 thinking that my rig would laugh at it same as it does with Crysis. Was I wrong, as were many ATI users.

Anyhow, I did some VERY marginal overclocking on my 5970. Granted, twice I upped the voltage to 1.1625v (which ATI Overvolt utility does automatically, no other options) and upped the core frequency to 800mhz and the memory to 1200mhz which would almost bring me to the stock 5870 clocks. The whole computer froze. Had to do a hard reboot. This was after I did the below.

The first time however, I took the OC slow. Running furmark 640x480 and adjusting in 25mhz increments with MSI Afterburner. As was recommended by the guru3D and of course the ATI Overvolt I adjusted the Core voltage to 1.1625V and got the Core/memory clocks to around 800/1200. Ran it for a few minutes in Furmark and the temps were staying pretty solid.

Started up Metro 2033, and noticed a pretty nice FPS increase! The game froze about 1 minute in and had to do a hard reboot. The GPU temps only reached ~55-60c so it wasn't a temp issue. Immediately, I thought it must be a PSU issue as it really was a small Overclock for the 5970. Which was made to be overclocked and has a "massive overclocking headroom."

After this, I ran into some strange and scary graphical issues which leaned me towards expletives that I had somehow fried my GPU *knock on wood*. Like it said, very small overclock and voltage increase that I did and the temps were PERFECT.

After booting up, my internet wasn't working. Odd. It said there was no network. I rebooted and all was fine. I preceded to load up Mass Effect 2 with my usual settings 8x AA edge-detect and 16x AF high quality. Eats the game for breakfast. No OC'ing applied.

After clicking the ME2 shortcut, Steam loaded up, and then BAM. A bunch of vertical lines and random anomalies showed up on the monitor with random colors. Odd thing is that the 5970 wasn't doing much if any work yet as it was still on the desktop. The lines went away as I moved my mouse cursor over them, most of them at least. Rebooted and haven't seen that again since played a few games without issue.

But damn I was worried after seeing that, certainly thought something had gotten fried. Could the freezing upon upping clocks/voltage (increasing the voltage quite a bit from the 300W 5970 stock) be the PSU being overloaded?

I don't understand why I had a (stable?) overclock in Furmark, which is burning my GPU to hell. And then loading up Metro 2033 with my GPU temps in practically the idle temp range and having the computer freeze on me. If it was a PSU issue wouldn't the PSU freeze as soon as it was overloaded?

I don't know, I'm not very informed on PSU's, this was the first computer that I built but I have taken care not to do anything too crazy. I have added a lot of stuff since getting the computer, maybe it is too much for the PSU. I ran my computer through the power supply calculator and it only came out to 623w with the OC cpu and both graphics cards. Here is all I have.

i7-920 @ 4.1 ghz
HD 5970
9500GT 1GB for PPU
Asus p6t Deluxe v2 mobo
6GB OCZ DDR3 (3 sticks)
Auzentech Prelude X-fi 7.1
Coolermaster Hyper 212
Antec 1200 case
DVD/CD RW drive
3 Samsung Spinpoint F3 1TB HDD's

And my PSU: 850W Corsair TX.

Is this my graphics card or PSU?

I noticed the above issues as explained after a couple reboots with no OC'ing which got me worried, especially the artifacts on the desktop. None of the OC I did should have done any damage to the 5970. Do I need a larger PSU?

Thank you.

-J
 

Jamrock

Distinguished
Nov 16, 2009
233
0
18,690
Well, that puts to rest some fears of a blown PSU. But, that also leaves me only to think that it's the GPU that is somehow damaged now.

The OC'ing I did shouldn't have done anything though to harm anything. These issues only showed up AFTER OC'ing so it only makes sense that it has something to do with that.

Even stranger though, is that just opening up Afterburner froze my computer once. Scared to go anywhere near that program now, lol.
 

Jamrock

Distinguished
Nov 16, 2009
233
0
18,690
ct,

Yeah I remember reading about that and had two defective 5970's before getting this one. Something to do with 2D/3D. Using 10.2 atm, will try 10.3.

Been running some games and everything seems stable at stock clocks. Furmark runs correctly too.

I don't get why any minor OC makes the 5970 have a fit though. I know bumping up the voltage on the 5970 gets it up near ~400W just for the card which lead me to think maybe it was a PSU issue.

Since I added the 9500GT pretty recently and added 2 new HDD's since building the computer.

I just am not sure what else would be the issue, as the temps on the card are just fine when these things happen.
 

LePhuronn

Distinguished
Apr 20, 2007
1,950
0
19,960
As englandr753 said, undo all overclocks on your system and do some stress testing at stock. See how stable things are from there.

If you're 100% stable at stock then it's just possible that the chips in your 5970 are as good as some others so you simply don't have any overclocking headroom. Which would be a shame, but that's just the way the cookie crumbles sometimes.
 

Jamrock

Distinguished
Nov 16, 2009
233
0
18,690
Ok, I think I found out the problem. I restarted a few times and made sure clocks were reset on the 5970. I got it overclocked easily to 1.2v, 850mhz core, and 1200mhz memory up the 5970 speeds and it worked great I played Metro 2033 for quite some time with no issues, as well as Furmark. Temps were great.

Restarted the computer, this is a couple days later, went into GPU-Z and it shows my 2nd gpu running at half memory clock speeds. Even when overclocked now, GPU-Z still shows the 2nd gpu running ~1/2 the clock speeds of GPU1. When memory was overclocked to 1025mhz, the memory clocks on gpu2 were only 555.8mhz.

When I overclocked prior to when the memory started DOWNclocking like above. I overclocked to 5870 speeds. 850Mhz/1200mhz Core/Memory and 1.2v. Looked at GPU-Z and I saw that the 1st GPU was indeed running at those speeds in GPU-Z. The 2nd gpu however, was running at 850mhz/2000mhz. I figured this was just a bug, as I set it to 1200mhz. I got a freeze shortly into the game, had to hard reboot, thought it must just be a bug with GPU-Z.

Now, the stock clocks (when in Powersaving mode, aka whenever your NOT playing a game,) are 157/300mhz (normal) on the 1st gpu, and 157/162mhz on gpu2 (NOT normal.)

I'm a little worried, I hope those GPU-Z reading weren't/aren't correct. Whenever I overclock the slightest amount now, the 2nd gpu memory clocks are always ~1/2 of what they should be.

Anyone please help, no way Afterburner just fucked the memory clock on my 2nd gpu.
 

Jamrock

Distinguished
Nov 16, 2009
233
0
18,690
The VDDC current going to GPU1 is 1.7A. The VDDC current going to GPU2 is only 0.9A. That has got to be the problem right there... I just checked all the cables and made sure they were connected and problem still persists.

Why is GPU2 not getting the same current, I know it should be.
 

AsAnAtheist

Distinguished
Sep 15, 2009
790
0
19,060


Could be a graphics card bios issue with the several stages (idle, mid-low, mid, full) of the 2nd GPU (the 9500 GT)... Look to see if there's an updated GPU bios for your card on the manufacturer's page.
 

Jamrock

Distinguished
Nov 16, 2009
233
0
18,690
Updated BIOS right after posting, and now all seems to working well! 1.7A on both and the correct idle clocks.

Unfortunately, the 9500GT 1GB that was in there seems to have gotten fried? Any way to fix this?

EVGA have the original BIOS for this card maybe? Which would change the BoardID back and the correct BIOS version?

It didn't literally FRY the card, ... did it?

Bunch of random blocks of colors and lines and such come up when booted from the 9500GT. Old card... but I hope there is a way to fix that?
 

AsAnAtheist

Distinguished
Sep 15, 2009
790
0
19,060


Wait did you update the bios on the 9500 GT, and then the card stopped working properly? If the card has been fried it is very unlikely that it would even display. Did you make sure to get the BIOS for the EXACT model 9500 GT you got? It has to be exactly for that model number.
 

Jamrock

Distinguished
Nov 16, 2009
233
0
18,690
Atheist,

I have contacted EVGA and they are extremely helpful, and seem to be either working on a BIOS update or finding one that will work to try and revive the card. I did update the 9500GT Bios and then the card started showing the odd artifacts and such. I was unable to see anything Windows related, just a bunch of random colors and blocks. Im crossing my fingers, but I am not giving my hopes up.

On another slightly related note, I am trying to update my Asus HD5970 BIOS. I have made a seperate post about it HERE.

If anyone could help me out there, it is really a simple question... it'd be much appreciated.

Thank you to everyone that has helped me out here.

-J