Is one of my cards broken?

[-=HaXoMaTiC=-]

Is one of my cards broken? (Updated! Need a bit more help!)

I'm running this rig currently.

Asus A8N32-SLI Deluxe
FX-55 AMD
2x Seagate 250gb HDD
Onboard AC97 Audio
Sony DVD-DL +/- Drive
PNY 9600GTx2 SLI-Setup PCI-E 2.0 Compat Note: I'm using a 2.0 16x card in a 1.0 x16 slot but I read that they are backwards compatible
OCZ gameXstream 700w Power Supply

Now everything runs smooth and I never crash outside of games and even then it occurs rarely if that. However, I am quick to paranoia and I am getting something that worries me quite a bit.

It has only crashed twice in 1 month but both times it crashed like this.

------------------------------------------

Instance 1: I was playing oblivion I had overclocked my pci-e bus to 2650 (6 mhz), raised my mhz on my core and mem by 50 mhz each, and overclocked my cpu a mere 8 mhz. I was jumping down a mountain heading towards a city and my moniter shut off and went into power saving mode. It was as if my cards stopped sending a signal completely and my monitor just responded by shutting down. I had to force power-down the computer to rememdy the situation but it started up fine and dandy and ran flawlessly (and the overclocking reset).

------------------------------------------

Instance 2: I was playing Synergy a Halflife2 modification which allows you to play HL2+ games in 2 player online. I was hosting a server and only 1 person was connected as we played. My game instantly locked about 30 minutes into gameplay and the sound looped at around a 0.1 second interval (very fast short sound repetitions). My monitor shut down just like in instance 1 however the sound was going nuts. After a hard power-down it ran fine.

------------------------------------------

Now here is the odd thing, I've played TF2 (Team Fortress 2) for up to 8 hours with this exact setup minus the overclocking. In fact I've even ran it for at least 4-6 with 50/50 increase in core/mem clock speed and it ran roughly around 50-65 Cel. I've played oblivion on 1440x900 with fairly high settings for 2 days straight in fact and it never seemed to crash one time or show any signs of instability. The only thing I've changed recently is the video cards, previously I had 1 single video card (7900 GS KO) now they are 2x PNY 9600 GTs .

Overclocking wise the cards crash well before they even reach a temp remotely around a dangerous temp (It's never peaked 70 deg and thats with gpu fans on LOW (25-30%). They don't seem to like getting above a clock of 725/975 in fact the core clock stays near 714 if i clock it that high.

Of course all in all I've left my computer at stock speeds only instance 1 was a time I overclocked and it was for a very short time at which I monitored the temps of my system very closely.

Specifics: I have the most recent drivers for my chipset, graphics and sound. I used driver sweeper to remove the graphics drivers before installing the most recent as well. My settings in my drivers are fairly stock except for texture filtering being high performance and single display performance mode.

So here is my question after this wall of text... Is SLI still a bit unstable at times causing hardware-level lockups and I experienced an oddity or is one of my cards faulty? I never crash anywhere else and as I said it has only happened twice in 3-4 weeks under very certain circumstances.

Any help would be greatly appreciated as I have little information left to go off of. Also, if you need anymore information let me know and I will post it.

I talked to a PNY rep who claimed my 700w power supply was insufficient and the black screens and hardware locks are because the cards require 22 and the rails only show 18 (forgot the measurement he said of voltage)

Any feedback on this would be greatly appreciated.

 

[Zehnsucht]

In instance 2 did you overclock the graphic cards? If not, it could be the nvlddmkm.sys acting up. I had similar experiences with Bioshock and nvlddmkm.sys stopped responding following a looping of the sound and black screen. Fortunately, vista can ignore the error and restart the driver (thus crashing all 3d applications though).

In my experience, it isn't the temperatures that limits how much you can overclock the GPU (if not voltmodded). Some cards clock well, others simply don't.

 

[-=HaXoMaTiC=-]

Yes, the second instance I was not overclocking. Instance one I was just slightly overclocking everything including the ever useless PCI-E bus.

 

[Zehnsucht]

I don't know if it's different with AMD, but with Intel it is not recommended to overclock the PCie bus, most people lock it at 100 Mhz and call it a day. The SATA controller is also connected to that bus, overclocking it can lead to data corruption. Again, I don't know if this applies to AMD.

In general, when overclocking, only one item at a time is overclocked, then checked for stability. Overclocking everything at the same times makes it hard to know what caused the error.
When I search for 24/7 clocks I often start with the CPU, if it runs great for a couple of days with varied types of loads, I start overclocking the GPU. But then again I'm quite conservative, aiming for stability first, speed second.

 

[-=HaXoMaTiC=-]

Ironically my cpu only overclocks 8 mhz before it crashes... 10 makes it unstable. My friend was a master overclocker and melted many a computer to perfect his skill, however, I was unable to pick his brain too much before we fell out of contact.

Do you have to loosen your memory timings when you overclock because mine is running like this: (Dual Channel 2 sticks, 2 gigs)

CAS# Latency (tCL) 2.5 clocks
RAS# to CAS# (tRCD) 3 clocks
RAS# Precharge (tRP) 2 clocks
Cycle Time (tRAS) 5 clocks
Bank Cycle Time (tRC) 9 clocks
DRAM Idle Timer 16 clocks
Command Rate (CR) 2T

My cpu only runs 38 Cel (100F) and I only use a venus 12 fan cooler.

Anyways, not meaning to deviate from the subject, thanks so far for the input. Knowing it may in fact be a driver issue sets my mind at ease. I cannot afford to replace these cards so I was worried I would have to revert to a old card.

Edit: On a funny sidenote my friend ran nTune and it overclocked his pcibus to 2900 mhz from the stock 2500 without crashing. It detected it in the red yet it kept pushing it up without faulting...I worried and told him to set it back lol.

Second Edit: One last question, do you think my computer would run faster with Vista-64bit then my current windows xp 32bit? I have a 64bit cpu and DX10 capable cards but ... I was thinking the bulk and excess of Vista would be too much for my simple rig.

 

[Zehnsucht]

That is some seriously tight timings. What sticks are you running?

2.5-3-2-5

 

[-=HaXoMaTiC=-]

I bought these sticks to replace my Platinum Rev.2 ram 2 512 Sticks.

http://www.newegg.com/Product/Product.aspx?Item=N82E16820227210

The timings are actually 2-3-2-5 but the stupid bios detected it as something different. I flashed my bios to a beta 1405 (runs fine so far) and it reset my timings. Anyways now its the OCZ Platinum 2GB (2 x 1GB) 184-Pin DDR SDRAM DDR 400 (PC 3200) Dual Channel Kit Desktop Memory

 

[-=HaXoMaTiC=-]

3rd crash now playing farcry on max settings only overclocking 50/50 on the card and thats it. I called PNY and told them about my problem and they identified at as a power issue. He claimed that black moniter and system lockup is caused when the powersupply falters due to insufficient power.

My powersupply is a OCZ GameXstream 700w, I was mixing up what ps i had in my original post.. he said the card requires 22 but the psupply only carries 18 in each rail...I forgot what measurement he said along with the numbers but ... is this true? Is a 700w Power Supply insufficient for 2 9600GT's? I'm about to facepalm myself hardcore for not thinking this one through if it is in fact true...

Any feedback would be insanely helpful since I'm at a loss. If its true I have a friend who has a 1,000w psupply and could use them and I may be able to salvage my blunder partially.

 

[Zehnsucht]

I highly doubt it's a power issue.

You could borrow your friends though to easy your mind.

The guys at PNY are plain wrong though. The 9600GT uses one PCIe 6-pin connector. 22 amps recommendation? Yeah right. As long as you don't connect the cards to the same rail you are ok.

Here's an interesting find:
http://www.anandtech.com/casecoolingpsus/showdoc.aspx?i=3413
107W at full load during FUR. Which translates 9A.

Worth mention is that the second graphics card in an SLI/CrossFire setup never actually uses 100% of the theoretical maximum power consumption.

No, it's something else. Have you tried increasing voltages throughout the motherboard? Particulary the chipsets (northbridge/southbridge).