Problems with new gpus?

[npsgaming]

Okay. So not too long ago (maybe a week) I put a question on here about getting some new gpus. I ended up with 2x GTX670FTW+ cards from evga. They seem to be performing well. They installed quickly, run quietly, etc. tons and tons of vram (4gb per card), and the awesome feature of having the reference 680PCB means waterblocks won't be impossible to find later on. My problem however, is crashing in game. Lots of games. BF3, Red Orchestra 2:GOTY, Arkham City, Skyrim(doesn't crash but studders every 30seconds or so), etc.


So I am wondering if the cards are bad? I never had these issues before and I have tried a few different drivers to see if it was a driver problem. No luck.

GPU stress tests and synthetic benchmarks do fine so I am inclined to say its not the cards but just wanky drivers still. Anyone got an idea?

 

[npsgaming]

PSU: Corsair AX1200
I'm pretty sure they are all gone but I'll use DS to see. I've gone through all 3 recent driver revisions: 301.xx through 306.xx
I'm getting all kinds of errors whether it be:

1. green flickering in game
2. black screen followed by need for hard reset
3. sound loop followed by need for hard reset
4. straight up bsod

Temps are:
GPU1 Idle: 29c Load: 71c
GPU2 Idle: 28c Load: 64c
Motherboard:
Vreg: 88c
Chipset: 65c
CPU:
Core1: Idle: 32c Load: 56c
Core2: Idle: 33c Load: 53c
Core3: Idle: 36c Load: 55c
Core4: Idle: 31c Load: 55c

 

[manofchalk]

Those temperatures are fine. Though the voltage regulator is running a bit high, make sure you have good airflow over the motherboard.

Have you overclocked the cards at all, or just left them at their factory OC settings? Because that sounds like a bad overclock to me.

Try disabling SLI and running the games on a single card. If the issue clears up, try the other card. If both work just fine, then drivers are a possible cause.

If one of the cards does give problems, then it may not be able to maintain the clocks its set to. Try lowering the clocks or raising the voltage a bit and see if that fixes it.

 

[luciferano]

Your PSU, although unlikely, might be related to the problem. It's way overkill and not a particularly reliable brand.

 

[JackNaylorPE]

The AX1200 PSU (Flextronics is Corsair's OEM for this model) is one of the best on the market and garners a 10.0 performance rating from jonnyguru. It's also over sized so is running at peak efficiency and obviously running very cool.

http://www.jonnyguru.com/modules.php?name=NDReviews&op=Story5&reid=189

Of course that doesn't mean you didn't get a bad apple. I'd suggest that you run OCCT and monitor temps AND voltages, especially on the 12V rail.

http://www.ocbase.com/

The other thing I can't help but mention is I have gone thru this before myself w/ factory OC'd EVGA cards. 18 months, 20 support calls and 5 RMA's brought no results with Son No.2's box from EVGA. They blamed MoBo, RAM and everything else but had no answer to why the same box ran fine with two Asus cards (stolen from his brothers box) in it overclocked 30% while their factory OC'd cards from EVGA could not run over reference card speeds. They finally sent a "stock" reference card from the next generation which gave comparable performance, tho it also doesn't OC well.

 

[luciferano]

The AX1200 PSU (Flextronics is Corsair's OEM for this model) is one of the best on the market and garners a 10.0 performance rating from jonnyguru. It's also over sized so is running at peak efficiency and obviously running very cool.

http://www.jonnyguru.com/modules.php?name=NDReviews&op=Story5&reid=189

Of course that doesn't mean you didn't get a bad apple. I'd suggest that you run OCCT and monitor temps AND voltages, especially on the 12V rail.

http://www.ocbase.com/

The other thing I can't help but mention is I have gone thru this before myself w/ factory OC'd EVGA cards. 18 months, 20 support calls and 5 RMA's brought no results with Son No.2's box from EVGA. They blamed MoBo, RAM and everything else but had no answer to why the same box ran fine with two Asus cards (stolen from his brothers box) in it overclocked 30% while their factory OC'd cards from EVGA could not run over reference card speeds. They finally sent a "stock" reference card from the next generation which gave comparable performance, tho it also doesn't OC well.

It's overkill to the point that I am damn sure that it is not running at peak efficiency. The problem with going too high is that transformers get exponentially more like a Tesla coil as it gets linearly lower load by percentage. This causes damage over time, although it's usually not as bad as having an insufficient PSU.

 

[frombehind]

err, check with evga on your Gtx 670 cards... if by any chance they are the ones with the "shorter" circuit boards, there was a factory recall on those... long story short, they weren't built right and are plagued with problems... Evga apologized and offered free replacements to anyone with those cards, whether they had problems or not. Might want to look into it.

 

[dudewitbow]

err, check with evga on your Gtx 670 cards... if by any chance they are the ones with the "shorter" circuit boards, there was a factory recall on those... long story short, they weren't built right and are plagued with problems... Evga apologized and offered free replacements to anyone with those cards, whether they had problems or not. Might want to look into it.

the FTW models werent the recalled ones. it was the OC version. the FTW was the model given to people who had a faulty OC version in return.

 

[npsgaming]

Well the AX1200 has done me proud for about 3+ years now and I couldn't ask for a better unit. Mobo's airflow is a bit weak (using the 800D) but it does get some. The evga x58 classified mosfets/vregs always ran a bit warm so I'm not too freaked about that. It is either a problem with drivers or the 2nd card is dead. I removed the drivers again and replaced with current (306.xx) and disabled SLi and it seems to have fixed the issues. So I'm not sure what the problem is but a little more tinkering is required.

err, check with evga on your Gtx 670 cards... if by any chance they are the ones with the "shorter" circuit boards, there was a factory recall on those... long story short, they weren't built right and are plagued with problems... Evga apologized and offered free replacements to anyone with those cards, whether they had problems or not. Might want to look into it.

----- Good news on that front then. I bought the FTW+ cards so I could get the longer cards that use the 680 reference pcb so I could find waterblocks a bit easier (oh and 4Gb GDDR5 wasn't a bad call either :lol: )

 

[luciferano]

This statement shows you know absolutely nothing about power supplies.

http://www.jonnyguru.com/modules.php?name=NDReviews&op=Story3&reid=189

Sounds like it's just SLI that's giving you trouble ... if you try a different driver, does that help at all?

I misread the brand, my bad. It's still overkill and can be related to the problem, although again, it's unlikely.

 

[dudewitbow]

I misread the brand, my bad. It's still overkill and can be related to the problem, although again, it's unlikely.

in a deeper level. You should never judge psus by brand either. Each company usually has at least 1 good unit.

 

[luciferano]

Unless it's DiabloTek :lol:

Also @ luciferano - No, an overkill PSU can't be related to this problem. Obviously, any PC part can die, but that's different. Worst case scenario, he's losing like 2% efficiency. Did you even look at that review I linked? 88-89% efficiency at not even a 250W load.

Efficiency isn't the biggest issue at all. As PSU load is decreased roughly linearly by percentage, transformers exponentially act more like a Tesla coil. This damages it over time, especially when you don't stay within about 40-80% load (although it usually doesn't get severe until you get under 20%, possibly related to that being the minimum specification load specification for 80+ or better certification), especially in reducing efficiency over time. Do this for too long and it can screw up the voltage regulation in some circumstances.

It's a flaw with all computer power supplies that I've ever known and although it has been improving, it can still cause problems. I think that it can be greatly alleviated by adding additional low power components and letting a power supply switch between low power transformers and the high power components depending on the load, but I don't know of a PSU that does this.

 

[luciferano]

Obviously PSUs, like any component, degrade over time. Since you're the first person I've ever heard bringing this up as a problem, I'm going to treat it as

Even if true, there's nothing to substantiate the idea that his PSU might be the problem. It's just fear mongering.

To be fair, I was saying that it is unlikely that the PSU is the problem, only acknowledging the possibility of it.

 

[luciferano]

Proof:

If the output current becomes small enough, the output voltage of the power supply can reach the peak value of the secondary voltage of the transformer of the power supply. This occurs because with a very small output current, the inductor in the L-C low-pass filter does not drop much voltage (if any at all). The capacitor in the L-C low-pass filter therefore charges up to the peak voltage of the secondary of the transformer. This peak voltage is generally considerably higher than the average voltage of the secondary of the transformer. The higher voltage which occurs across the capacitor, and therefore also at the output of the power supply, can damage components within the power supply. The higher voltage can also damage any remaining electrical loads connected to the power supply.

SMPSs have an absolute limit on their minimum current output.[2] They are only able to output above a certain power level and cannot function below that point. In a no-load condition the frequency of the power slicing circuit increases to great speed, causing the isolated transformer to act as a Tesla coil, causing damage due to the resulting very high voltage power spikes. Low-load conditions increasingly case this phenomenon as the load gets lower on a logarithmic scale. Switched-mode supplies with protection circuits may briefly turn on but then shut down when no load has been detected. A very small low-power dummy load such as a ceramic power resistor or 10-watt light bulb can be attached to the supply to allow it to run with no primary load attached.

SMPS is Switched-Mode Power Supply, of which all computer power supplies fall under.

http://www.patentstorm.us/patents/5402059/description.html

 

[luciferano]

Fair enough. A big +1 for actually proving your claim! [:lutfij:3]

I still think you're way off at pointing to the PSU as a problem. The AX1200 is one of the highest quality PSUs on the market. It doesn't come with a 7 year warranty for nothing.

Like I said, I doubt that it's the problem. I just didn't want it completely knocked off the suspect list until the problem was solved. Even top-quality units can have issues occasionally.

 

[npsgaming]

Well I have an update: It seems the drivers were the problem? Not video but my Realtek HD Audio seemed to be messing with everything. I will keep testing in different applications to see if it proves true but so far with the 306.xx drivers and both cards active in SLi I've been able to run everything and there has been no studdering, green screens, or bsods. So here's to hoping!

On another note, interesting discussion on power supplies here. I must say though there is no chance this psu is being "underworked". On a regular day where I'm not having driver troubles this sucker is cranked. The ups battery backup has a power monitor built in. I have the thing log power usage to my pc and the logs are recording roughly 600w from the wall at idle. My next question is that too much draw for an idle rig?

 

[luciferano]

Well I have an update: It seems the drivers were the problem? Not video but my Realtek HD Audio seemed to be messing with everything. I will keep testing in different applications to see if it proves true but so far with the 306.xx drivers and both cards active in SLi I've been able to run everything and there has been no studdering, green screens, or bsods. So here's to hoping!

On another note, interesting discussion on power supplies here. I must say though there is no chance this psu is being "underworked". On a regular day where I'm not having driver troubles this sucker is cranked. The ups battery backup has a power monitor built in. I have the thing log power usage to my pc and the logs are recording roughly 600w from the wall at idle. My next question is that too much draw for an idle rig?

600W at idle seems extremely high to me, but I don't have personal experience with such a system as that. That might be considered further evidence to my claim. Very high power consumption like that at idle can imply that its running extremely inefficiently, a strong sign of under usage with such a power supply as that.

 

[manofchalk]

600W from the wall at idle, that is way too much. I would expect that under load.

Check some reviews on the efficiency of the PSU below 20% usage. 80+ certifications only cover efficiency between 20 and 100%, so the gold rating wont necessarily apply below that.
Though according to that Johny Guru review, its 84% efficient at 10% load (equivalent to 80+ Silver).
http://www.jonnyguru.com/modules.php?name=NDReviews&op=Story2&reid=189
So that evidently isnt the issue unless you drawing under 10% load. Which I dont think your machine can actually idle under.

Just to see, try running Prime 95 and Furmark while watching the wall meter. If the load goes down, then this may be the issue.

Its possible you have just got a bad unit.