750W enough?

[WojtekGee]

Hey guys,
I've been running into some issues as of late bsoding, and I'm trying to rule out possible issues.

I was considering that this may be a PSU issue, and I've tried to calculate myself and it seemed ok, but to make sure I thought i'd ask you guys.

It's a 750W Antec TP Blue (http://www.newegg.ca/Product/Product.aspx?Item=N82E16817371022)

Asus Crosshair IV extreme mobo, with 955 BE overclocked to 3.9, 1.5125V

8 Gigs corsair xms3 @ 1.5V

Two reference sapphire 5850s, overlocked slightly (i think +50/100 just in the CCC)

and a noctua NHD14 heatsink, with a few case fans and a noctua side panel fan.

One mechanical HDD
+ samsung 830 SSD
CD/DVD writer

CPU does NOT overheat, runs idle at 28C and under full load (stress tests) @ a steady 54-55.

GPUs do NOT overheat, run on full load (stress tests / graphic intensive games) @ a steady 60-65ish (I don't let them get higher with Afterburner)


Is this enough, would this setup cause BSOD issues because of power consumption issues?

BSOD error is atikmdag.sys... I'm aware this is a graphics card issue, googling brings up a lot of issues, lots that are random, and some indicate that it is related to power consumption as well.. so I thought I'd ask

 

[lehtv]

It should be enough to power your system. Atikmdag.sys looks like it's related to drivers

Have you tried reinstalling drivers or trying different drivers? Do you get BSOD with only a single GPU installed?

 

[WojtekGee]

I did not, I recently added a second GPU.

I'm afraid it may be faulty as well... However again trying to rule things out.

I have tried reinstalling drivers, I will try to the oldschool approach tonight with driver sweeper etc.


I also changed the board and from a 790GX to an 890 FX, did a sysprep and tried to give it a go.

I will be most likely formatting, but I'm waiting for another SSD.

Thanks for the response

 

[aaksheytalwar]

If after adding a second gpu it bsods then 100% psu related. Try with 1 gpu at a time, each gpu. If it is fine then you need a bigger psu.

 

[lehtv]

If after adding a second gpu it bsods then 100% psu related.

No, it could be crossfire/driver related. "BSOD error is atikmdag.sys."

Try with 1 gpu at a time, each gpu. If it is fine then you need a bigger psu.

If it is 100% PSU related, why bother testing?

 

[aaksheytalwar]

If it is 100% PSU related, why bother testing?

Read my complete reply. One word at a time.

And the chance of it being the psu is 10x times as much as it being a bsod issue because of drivers or cf.

 

[TemjinGold]

Read my complete reply. One word at a time.

And the chance of it being the psu is 10x times as much as it being a bsod issue because of drivers or cf.

Really? 10x? Exactly 10x? Are you sure it's not 9.75x maybe? I hope you know 5850 CF sucks about as much power as ONE GTX 480. And yes, 750 watts is plenty for that. The chance of it being a PSU problem I won't argue. But the chance of 750 watts being not enough is ZERO.

 

[lehtv]

Read my complete reply. One word at a time.

I did, and replied.

And the chance of it being the psu is 10x times as much as it being a bsod issue because of drivers or cf.

Thanks for sharing your source

 

[VirtualLarry]

GPUs do NOT overheat, run on full load (stress tests / graphic intensive games) @ a steady 60-65ish (I don't let them get higher with Afterburner)

Is this enough, would this setup cause BSOD issues because of power consumption issues?

BSOD error is atikmdag.sys... I'm aware this is a graphics card issue, googling brings up a lot of issues, lots that are random, and some indicate that it is related to power consumption as well.. so I thought I'd ask

I mostly remembered my older Cypress based GPUs having problems at 50-55C, so i wasn't sure if that was the culprit. mind you, these were 5830s.

http://forums.anandtech.com/showthread.php?t=2263620

From that, I would discern that your GPUs are indeed overheating.

 

[Sunny129]

http://forums.anandtech.com/showthread.php?t=2263620

From that, I would discern that your GPUs are indeed overheating.

i don't know...my Cypress (HD 5870 2GB) has been crunching 6 POEM@Home tasks in parallel for over 7 weeks straight now, and temps are in the mid 60's°C under a constant ~80% GPU load. it shares real estate w/ a Cayman (HD 6950 2GB) as well, which crunches Milkyway@Home and also reaches mid 60's°C under a constant 99% GPU load. granted i keep it pretty well ventilated, but even in that heat box, my Cypress has never given so much as a hiccup. before that, it was crunching Milkyway@Home 24/7 for a year or so...so it got pretty stressed.

OP, you mentioned your GPUs are slightly OC'ed via CCC. you also mentioned that you use MSI Afterburner to monitor your GPUs - you do know you can control the voltage to your GPUs using that software, right? you should try 2 things - first, put the GPUs back to stock clocks for a while and see of the BSODs go away. next, try OCing again, but this time bump up the voltage a bit in MSI Afterburner...and don't worry, your 750W PSU is more than enough to power everything, even after increasing the voltage to your GPUs.

 

[lakedude]

Kill-A-Watt, accept no substitute!

 

[WojtekGee]

i don't know...my Cypress (HD 5870 2GB) has been crunching 6 POEM@Home tasks in parallel for over 7 weeks straight now, and temps are in the mid 60's°C under a constant ~80% GPU load. it shares real estate w/ a Cayman (HD 6950 2GB) as well, which crunches Milkyway@Home and also reaches mid 60's°C under a constant 99% GPU load. granted i keep it pretty well ventilated, but even in that heat box, my Cypress has never given so much as a hiccup. before that, it was crunching Milkyway@Home 24/7 for a year or so...so it got pretty stressed.

OP, you mentioned your GPUs are slightly OC'ed via CCC. you also mentioned that you use MSI Afterburner to monitor your GPUs - you do know you can control the voltage to your GPUs using that software, right? you should try 2 things - first, put the GPUs back to stock clocks for a while and see of the BSODs go away. next, try OCing again, but this time bump up the voltage a bit in MSI Afterburner...and don't worry, your 750W PSU is more than enough to power everything, even after increasing the voltage to your GPUs.

Thanks for the reply/replies.

yeah at 775 / 1125.

I dropped them last night to stock (700 / 1000 IIRC) and it seemed to solve the problem, however I didn't play for a prolonged period of time (so not 100% sure on this).

When I was overclocking my BE I stress tested over a couple of nights succesfully for CPU / MEM, and when I was done with that I simply slid the CCC bars over to max allowable.

Can it be that the ATI cards aren't getting enough voltage for such a small increase ?

I'm going to avoid using CCC, and use MSI Afterburner to bump the voltages a bit.


Also I found an interesting issue/possible solution here?
http://support.amd.com/us/kbarticles/Pages/737-27116RadeonSeries-ATIKMDAGhasstoppedrespondingerrormessages.aspx
"The issue occurs when the Timeout Detection and Recovery (TDR) feature available in Windows Vista® and Windows 7 detects that the graphics card (GPU) has not responded within a predetermined period of time and reinitializes it with the Windows Display Driver Model (WDDM) to prevent the need to reboot. There is no single cause for a TDR event to be triggered."

* Also fair to say that furmark benched the gpus fine for an hour full load without crashes, where as a game usually doesn't last more than 20-40 mins.

 

[Sunny129]

Can it be that the ATI cards aren't getting enough voltage for such a small increase ?

i'm thinking its definitely a possibility. should most HD 5850s hit a 50/100mhz OC without a voltage bump?...probably. but one (or both) of your cards may just happen to have sensitive/lower-binned GPUs, which might result in needing a voltage bump for even the slightest of OCs. it might also mean that no amount of overvolting will solve the problem, and that a lower-binned GPU may not OC well regardless of how much juice you give it. when you get a chance, play for a prolonged period of time and determine if reverting back to stock clocks prevents any further BSODs. then, if you still feel the need to OC, just don't forget to overvolt as well when you attempt it, and see how that goes...

 

[WojtekGee]

Also what I've noticed, although I'm not sure if this should be an issue, is that both of the 5850's, although the same reference model card, have different BIOS versions.

Any comments on that?

 

[AnonymouseUser]

Probably not an issue with the GPU BIOS being different. The best thing to do to test the PSU is to temporarily disable all overclocking. If you still BSOD it's probably a driver issue.

 

[WojtekGee]

Probably not an issue with the GPU BIOS being different. The best thing to do to test the PSU is to temporarily disable all overclocking. If you still BSOD it's probably a driver issue.

Will do when I get some more time - I'll also try to get most up to date bios versions and flash both cards, uninstall/reinstall drivers at the same time.

Wouldn't hurt right?

 

[AnonymouseUser]

Yeah, that wouldn't hurt, but may be unnecessary. None of that really matters if the PSU is the problem.

 

[Sub Zero]

The PSU has 4 strong 12v rails. 25a on each on paper should work fine, if you have your video card wired with the right connectors.

If you have a single strong 12v rail, you can get by with less amps, like some Corsairs.

I have a number of UPS's with AVR and some can display / log the actual load at any given time. Can't get my Intel 980x with a 3gb GTX580 past 600w even with every hard drive doing copies in the background while I play high res video games. So capacity might not be the issue.

 

[WojtekGee]

The PSU has 4 strong 12v rails. 25a on each on paper should work fine, if you have your video card wired with the right connectors.

If you have a single strong 12v rail, you can get by with less amps, like some Corsairs.

I have a number of UPS's with AVR and some can display / log the actual load at any given time. Can't get my Intel 980x with a 3gb GTX580 past 600w even with every hard drive doing copies in the background while I play high res video games. So capacity might not be the issue.

Yeah I don't believe it was the issue.

I remember when playing a lower gpu taxing game (m&b crpg) it never gave me any issues, the load never neared anything close to 90%. So I went with the undervoltage theory, and even overclocked to 800/1200 to push that upping the mV to 1187. Safe limits apparently were to 1.25 V but the less the better right?


I put in about 2 hours of gameplay (usually would crash) last night, and it was fine.



I'm hoping that was the issue - Thanks for the feedback on the PSUs, I had a feeling my "old blue" wasn't letting me down.



And if anyone is still reading this... one gpu runs consistently hotter by around 5-8 C. I've cleaned both out with compressed air etc, again both exact same reference model sapphire 5850s. I have a feeling its the crossfire bridge that covers the exhaust (poor engineering imo).

I've read that you sometimes don't need the bridge. Is this true? I didn't want to experiment as I'm trying to narrow down the failure possibilities.

If it makes a difference I have an Asus Crosshair IV Extreme, and I wanna take that pos off if it isn't doing anything.

 

[Sub Zero]

I never OC a vid card. It just causes too much problems.

I find that in general, overclocking is more trouble than it's worth. Buy as much speed as you can, reset your BIOS to factory defaults so everything is all in sync and give it a whirl, tweaking 1333 memory to 1600 if it's rated for it, but leave as much default as you can, within reason.

Heck, maybe even try removing the side panel and see what happens.

Good luck.