WHO is right? (Final conclusion needed)

[dental]

First I put you in situation:

OS: Win XP SP.
PC:

Videocard: Sapphire HD 3870 X2
Processor: Q6600 @ 2.40GHz
Mainboard: GIGABYTE DQ6 X38 (Latest BIOS on Gigabyte website)
Memory: GeiL BlackDragon DDR2 2GB @ 800MHz
Soundcard: SB Live 5.1
PSU: Thermaltake Toughpower 750W Modular

Everything at default clocks.


When playing Crysis on very high (DX9) settings my PC shuts down at some point between 5 minutes and 10 (random). Other games that cause shutdowns are COD4 and the Racedriver GRID demo.

Things checked:

-I have the 6pin + the 8pin connected.

-I have the molex cable that supplies extra power to the PCI slots on the motherboard connected.

-I tried CAT 8.2, 8.3, 8.3+hotfix and 8.4 (using Driver Cleaner when deleting).

-Temp goes at 81ºC max when auto, but I tried with the fan blowing at 50% and 100% constantly with RivaTuner and it still shuts down.

I remember I played Crysis on very high (DX9) settings with my 8800gt for hours (rest of components where the same) and it never made my PC shut down just like if someone pulled off the power cable out.

Now the drama:

I asked at the Sappiretech forums and basically they told me that my PSU isn't enought for a 3870x2.

Read the history here:
SappireTech

At the same time, I asked at the guru3d forums and they said exactly the opposite.
Guru3D
As you can see, there was a confrontation and that G33ngecko guy banned a guru3d member that tried to dismantle their "you need a PSU with more than 20A on the 12V Rail" theory.

At this point, i've tried plenty of things, like downgrading the GPU, set a bit higher the CPU voltage, switch RAM slots... I always get a shutdown. An interesting thing is that it only aparently happens when Catalyst AI is set to Standard or Advanced (Crossfire is activated).

Already tried without the extra HD and DVD and case fans to leave all max power possible to the GPU. The thermal options in the BIOS are all OFF.

I got Everest to log everything while playing GRID with A.I set to Standard. It shut down after 1 minute 20 seconds and here are the stats until that point:

picture

Im totally left of ideas for things to try and I cannot swap components since I don't know anyone with a PC that has PCI-E (not gamers) and don't know a store where they will let you test stuff (at least without paying).

I don't have many options left besides:

RMA'ing card.
Buying another PSU.

But both things will cost me money and time, which means I need to know WHO is right (Sappire tech forum or guru3d forum).

Thanks all.

 

[Lorne]

Have you had the GPU heatsink off at any time?
If not pull it, Clean all the contact points and put new heatpaste, Make sure its making contact on all the spots its suppose to.

Could be heatsink warpage or the factory heat tape melted out, The more it over heats it gets worst, And a open case with external fan wont help at some point.

I had a gfx card with this, it would run fine most of the time but when it was worked hard I would get shut downs, It seemed that the heatsink was on spacers so the core couldnt be crushed but would warp at a certain temp and totaly pull away from the gpu and instant overheat and shut down, It was so instant that nothing recorded this.
I removed the spacers and planted that bastard hard on the GPU and never had problem after, It even OC's well now.

Just something to check.

 

[dental]

Originally posted by: Lorne
Have you had the GPU heatsink off at any time?
If not pull it, Clean all the contact points and put new heatpaste, Make sure its making contact on all the spots its suppose to.

Could be heatsink warpage or the factory heat tape melted out, The more it over heats it gets worst, And a open case with external fan wont help at some point.

I had a gfx card with this, it would run fine most of the time but when it was worked hard I would get shut downs, It seemed that the heatsink was on spacers so the core couldnt be crushed but would warp at a certain temp and totaly pull away from the gpu and instant overheat and shut down, It was so instant that nothing recorded this.
I removed the spacers and planted that bastard hard on the GPU and never had problem after, It even OC's well now.

Just something to check.

The thing is, the card is new (bought it like 2 months ago) and it happens since day 1.

 

[Jessica69]

OK...first, your 750W power supply should be more than plenty to run your setup without any problems.

I've got a very similar setup as yours, to wit:

Motherboard: Asus X38 (Maximus Formula)

CPU: Q6600 @ 3.6GHz @ 1.35V

Videocards: 2 x Asus HD3870 video cards in Crossfire (One is the "standard" version clocked at 775 core/2.25GHz memory and one is the TOP version clocked at 851 core/2.28GHz memory)

Memory: 4 x 2GB Mushkin Redline DDR2-1000

Also two hard drives, the obligatory two opticals, watercooling with a Laing DDC2 pump (D-Tek block on cpu, EK S-Max on NB, Thermochill 120.3 radiator)

And, surprising at it may seem to the idiots on Sapphire's forum (esp. Green Gekko or whatever his "Holiness" is named)....I've got the system running quite stabily with a Corsair HX620 power supply. And another interesting fact is that your power supply is certified by ATI/AMD to run a pair of 3870's in Crossfire or one 3870 x2 card.....Amazing, eh? So, since it can run two 3870's in Crossfire, it surely can run one 3870 x2 video card......ATI certified it to do just that. DUH!

This is my take on things......first, your problem is most likely one of your video cards.....and with them being Sapphire, I wouldn't be surprised at all. Sapphire, out of all the ATI partners/OEM's, seems to have the most problems of any of the OEM's out there. Add to that the idiocy that exists on Sapphire's forums from the Gekko-dude, who is there to help but ends up belittling anyone who points out "facts", like the one of your power supply being certified by ATI for Crossfire for those exact cards, or even had the audacity to disagree with him (not to mention his banning anyone who disagrees)....just leaves a sour taste in my mouth about Sapphire. Second, Sapphire branded cards are teh only ones I've personally had any failures from while using ATI-chipset cards, and I've been using ATI cards for a long time.

Secind, their assertion that while ATI lists your power supply as capable, not all mfgr's cards can do what ATI says is pure BS, unless Sapphire is playing fast and loose with the reference design.....but Sapphire insists it follows ATI's reference design, so anything ATI lists as being compatible with their cards are going to work with Sapphire's junk......

I did like the recommendation of an Enermax Infiniti 720 power supply as the "better" alternative to yours......which just goes to show there is NO clear understanding of anything related to power supplies over there.

I'd RMA the card.....and then sell the replacement and buy another brand.....like HIS or Asus or other......Sapphire is just much more problematic as a brand than any other out there....

 

[VirtualLarry]

Originally posted by: Lorne
Have you had the GPU heatsink off at any time?
If not pull it, Clean all the contact points and put new heatpaste, Make sure its making contact on all the spots its suppose to.

Could be heatsink warpage or the factory heat tape melted out, The more it over heats it gets worst, And a open case with external fan wont help at some point.

I had a gfx card with this, it would run fine most of the time but when it was worked hard I would get shut downs, It seemed that the heatsink was on spacers so the core couldnt be crushed but would warp at a certain temp and totaly pull away from the gpu and instant overheat and shut down, It was so instant that nothing recorded this.
I removed the spacers and planted that bastard hard on the GPU and never had problem after, It even OC's well now.

Just something to check.

That's interesting to hear. I have a friend with a P4 2.4Ghz mobile northwood chip on an 845PE mobo (200Mhz FSB), with an AGP GeForce 2 GTS card (a rather old one), and he gets random system shutdowns now. When he first got the system, this didn't happen. I've already swapped PSUs, and it still happens. So it shouldn't be the PSU. I looked at the mobo, and it's really hard to tell, but one cap might be a tiny bit buldging. Or it could just be the light.

I have a spare FX5900XT AGP card that I plan on swapping in and seeing if the shutdowns go away.

 

[dental]

I RMA'd the card, got it today, played Grid for 3 minutes and had a shutdown. I told them to test the card before sending me another one, so we can confirm now its not the card.

 

[superunknown98]

What voltage do you have the memory set to? I have seen ddr2 800 designed to run from 1.8 to 2.0v, that can cause some problems. Also what else do you have plugged in to the power supply that could possibly be drawing from the same rail as the video card. Maybe you should shuffle the molex connectors around. Other than that make sure your psu fans are spinning. Maybe even try the omega drivers?

 

[hooflung]

Sounds like you have heat problems possibly. I had to send my HD3850 and a 1900GT back to newegg. Protip: buy from newegg, they have a 1 year replacement warranty. You send the card into their RMA, they send you a new one no questions asked. Its good to live in the US of A with newegg in our borders.

 

[tcsenter]

Originally posted by: Jessica69
Add to that the idiocy that exists on Sapphire's forums from the Gekko-dude, who is there to help but ends up belittling anyone who points out "facts", like the one of your power supply being certified by ATI for Crossfire for those exact cards, or even had the audacity to disagree with him (not to mention his banning anyone who disagrees)....just leaves a sour taste in my mouth about Sapphire.

MSI Forums has a similar culture of mindlessly faulting the PSU based on specious notions of power requirements and brand bias. Although I don't think they've banned anyone over it, I did get a vacation once for likening the knee-jerk 'overselling' and 'overfaulting' of the PSU without merit to a creepy religious cult.

I once sifted through every topic on MSI Forums over a period of about 30 days, in which someone was caused to purchase a new expensive PSU on the advise or assertion that their PSU was not adequate. There were like five cases where the PSU did not solve the problem vs. one case where the PSU did solve the problem. And the one 'successful' case was a no-brainer that even I agreed was almost certainly the PSU.

I posted the results along with links to the topics and it was deleted. :roll:

 

[Sylvanas]

Things to try:

1.Raise mem voltage to 2.1v, run Memtest86+ for 4 hours at least.
2.Format+ Reinstall XP SP3 and try again- this time load only the mobo chipset drivers, sound and latest Catalyst 8.5's + Crysis (latest 1.2 patch) and see if problem persists (this eliminates the possibility of software conflicts).
3. Swap out HDD if possible and try again.
4. The most probable explaination is Heat try A)- take off the stock heatsink, reapply some quality TIM like MX-2 and reinstall. B) Add more ventilation to your case, 120mm fan on side pannels or even a 40mm fan on the mobo.

 

[lowrider69]

I didn't read through all of the threads on the other forums. Did you run memtest on the memory? How's your CPU temps?

 

[bryanW1995]

it's your memory most likely. I had a similar problem when I got my 3870. I installed the 3870 and, at the same time, installed a lot of software to test it (3dmark06, crysis demo, etc etc). I kept getting errors/lockups. I have an hx 620 so I was pretty confident that it wasn't the psu, and obviously the natural culprit was the 3870. however, I hadn't restarted my computer or even run a stressful program in several months, and my ram decided to give up the ghost. try different ram and you'll be amazed I'd wager.

 

[dental]

I ran memtest with only 1GB module to go faster for 24 hours. No errors after a shitload of passes. Then played only using this OK rated module and had a shutdown after 2 minutes. Definitely not RAM.

These are the min/max stats until GRID makes my PC shutdown (I keep Everest logging)

Click

Fan speed was set in auto at RivaTuner.

Again, it has something to do with the card, directly or indirectly, since the shutdowns started when I replaced my 8800gt (I was able to play any game for hours, including Crysis, without a shutdown). Now, after RMA'ing it and they telling me the card they sent me this time is 100% sure working, and knowing the shutdowns only happen when the CPU is stressed (high demanding games), and again, I was able to play any game with the 8800gt, it seems like must be something PSU retaled yes or yes. Now, If was able to play with the 8800gt, the PSU was working OK, and the guys at Shappiretech are right about that it needs more than 20A on a single 12V rail, or something else is going on.

Please HELP.

 

[zagood]

I don't think it's hardware. I think it's drivers, BIOS, registry, operating system, something like that. GRID shouldn't tax your machine that much even at max settings.

-z

 

[tcsenter]

You have enough spare $$ to change premium graphics cards like underwear in the pursuit of a few more FPS, but not enough for a new PSU?

You're a good American. i.e. trying to pursue a rich man's life on a poor man's income

This is the price of admission. Sure, its fun to ride for free while it lasts, but sooner or later, if you build one too many systems or upgrade one too many times, you are going to encounter an issue that will require changing out every last frigging component.

Pay up or get off.

 

[dental]

Originally posted by: tcsenter
You have enough spare $$ to change premium graphics cards like underwear in the pursuit of a few more FPS, but not enough for a new PSU?

You're a good American. i.e. trying to pursue a rich man's life on a poor man's income

This is the price of admission. Sure, its fun to ride for free while it lasts, but sooner or later, if you build one too many systems or upgrade one too many times, you are going to encounter an issue that will require changing out every last frigging component.

Pay up or get off.

Obviously your are completly wrong. First im not an American, second, you are an American who is talking without knowing about my situation.

I have this card because I had problems with my 8800gt. It totally sucked when playing older games (specially those using OpenGL). Apparently nVidia stoped giving support to these, and crashes happened all the time. Since the 3870x2 was released arround the time I wanted to RMA, and after alreadly RMA'ed the 8800gt one time, confirming its nvidia's fault, I exanged it for a 3870x2, which means I paid only 80? for the card (I bought the 8800gt when it came out). I build myself a whole new pc about every 5 freaking years, so your completely wrong on your statement.

 

[dental]

Originally posted by: zagood
I don't think it's hardware. I think it's drivers, BIOS, registry, operating system, something like that. GRID shouldn't tax your machine that much even at max settings.

-z

But how in hell. I already formated, reinstalled everything, tried with Vista, tried from 8.1 to 8.4 Catalyst, updated to latest BIOS on Gigabyte website... and still getting crashes. And a software issue seems to hardcore for complete shutdowns.

 

[tcsenter]

Hmm...story changing from hours of stability with 8800GT to lots of crashes and RMA of 8800GT before 3870X2 ever entered the picture.

 

[seemingly random]

Have you tried it without the sound card?

Tested both sticks of ram separately and together?

catalyst 8.5 is out...

Is it possible to try the gpu on another pc (probably a pain to install h/w and s/w though)?

Have you checked what the pc is drawing (spikes?) with a watt meter (kill-a-watt, etc)?

Does the problem persist if a powerful house fan (or equivalent) is directed into the open pc case while being used?

Don't see cpu/system temps mentioned - anything unusual?

 

[Davegod]

Definately try the omega drivers.

 

[dental]

Originally posted by: seemingly random
Have you tried it without the sound card?

Tested both sticks of ram separately and together?

catalyst 8.5 is out...

Is it possible to try the gpu on another pc (probably a pain to install h/w and s/w though)?

Have you checked what the pc is drawing (spikes?) with a watt meter (kill-a-watt, etc)?

Does the problem persist if a powerful house fan (or equivalent) is directed into the open pc case while being used?

Don't see cpu/system temps mentioned - anything unusual?

Yes, I tested the ram separately. The ram has no errors.

A couple of posts above there is a picture with all sort of temps.

 

[dental]

Originally posted by: tcsenter
Hmm...story changing from hours of stability with 8800GT to lots of crashes and RMA of 8800GT before 3870X2 ever entered the picture.

Dude, that was driver related, and ONLY with old games. I was able to play Crysis, COD4, and whetever you want for hours, without errors, crashes, shutdowns, whatever.

 

[dental]

Originally posted by: Davegod
Definately try the omega drivers.

Yes, I tried then. And whats the point anyway, if those are based on 7.12 they don't support Crossfire.

 

[seemingly random]

Originally posted by: dental
Yes, I tested the ram separately. The ram has no errors.

And together?

A couple of posts above there is a picture with all sort of temps.

The charting feature of speedfan can be useful.

I assume the other things mentioned have already been tested to no avail. What were the results with a big external fan? This will rule out a heat related problem. Any unusual events in the windows event log?

After trying another gpu (gtx, ultra?), I'd start to suspect other hardware - mb, etc. Does the same thing happen when everything is out of the case? This will implicitly cause the reseating of all connections.

 

[dental]

I dont have an external home fan atm, will try later. I've already bought another fan for the case (I have 4, but theres room for 5). That one goes just below the GPU so I hope it helps getting hot air out the case faster.