Computer shutting down with SLI 7800GTX 512MB

[Ezekrael]

Hello all,

This is my first post in the forum. I don't usually post on any forums but I'm absolutely stuck with a problem.

After having installed a 2nd XFX GeForce 7800 GTX 512MB into my system and enabling SLI, the computer now shuts down at the exact same point in both 3DMark 05 and in FEAR.

It shuts down straight after Game Test 1 on 3DMark 05 and about 2 seconds into the system test built into FEAR. I don't think it's an overheating issue as I have plenty of cooling, and monitoring the temps didn't show any excessive heat. I also doubt it's a power supply issue as I have a 600w Silverstone ST60F power supply which should be more than enough. Besides, if it was either of these two things, would it shut down at the exact same point every time?

The system ran fine with just the one card installed, the problem has only come up with the 2nd card installed. I've tried nVidia driver versions 82.12 and 81.98, reinstalling software and drivers (incl Direct X, 3DMark 05 and FEAR). I'm hoping that I won't need to go as far as reformat / reinstall to troubleshoot this.

Any ideas / help would be greatly appreciated!

 

[fliguy84]

try uninstalling the forceware drivers, then run driver cleaner and reinstall forceware drivers.

 

[RussianSensation]

Is the cpu, ram, and videocards all running stock? I know the PSU should be sufficient but still download Motherboard Monitor 5.3.7.0 or similar version and check the rails on load. Make sure the 12V rail doesn't dip below 11.5 and 3.3 below 3.23. If they do you can have instability. Also it might be good idea to run Memtest86 or Prime95 if your other system components are overclocked. Although the 2 cards themselves might not be causing issues in such a scenario, the extra load could potentially make voltage fluctuations more sensitive for cpu and ram components.

 

[Ezekrael]

Hi fliguy84,

Thanks for the suggestion. I downloaded and reinstalled using Driver Cleaner as per your suggestion but still no luck...

 

[Ezekrael]

Hi RussianSensation,

CPU, RAM and vid cards are all at stock speeds. I did download Motherboard Monitor and the 12v rail is fine, the 3.3 seems a little low at between 3.12 - 3 .14. Do you think this could be enough of a problem to make the system shut down in this manner?

 

[Elfear]

Does that Silverstone have separate 12V rails or one massive one? I don't know if that matters a ton but it seems like I've seen people have the most luck running a single fat 12V. Might be worth it to you to borrow another PSU from a buddy or something. A guy over at NvidiaNews had the same problem as you and after switching PSU's it went away.

 

[allies]

Ezekrael - I'll take one of those 512 GTX's off your hands, and your crashing problem will be fixed as well.

 

[Matthias99]

In my experience, a system that shuts down randomly (as opposed to locking up or getting a BSOD) is usually the result of electrical trouble of some sort. Either your PSU is defective or inadequate (can you post the specs on that model?), or the PSU is overheating (saw a system behave like this, and it turned out the PSU's fan was shot), or something is shorting out and causing the system to shut down.

 

[SickBeast]

AFAIK most motherboards will shut down when the temperature reaches a certain point. It's either your PSU or too much heat. Touch the side of your case. How warm is it?

 

[nitromullet]

Originally posted by: Elfear
Does that Silverstone have separate 12V rails or one massive one? I don't know if that matters a ton but it seems like I've seen people have the most luck running a single fat 12V. Might be worth it to you to borrow another PSU from a buddy or something. A guy over at NvidiaNews had the same problem as you and after switching PSU's it went away.

The Silverstone has 4 rails IIRC...

 

[RussianSensation]

Originally posted by: Ezekrael
Hi RussianSensation,

CPU, RAM and vid cards are all at stock speeds. I did download Motherboard Monitor and the 12v rail is fine, the 3.3 seems a little low at between 3.12 - 3 .14. Do you think this could be enough of a problem to make the system shut down in this manner?

That it right there. Many people do not realize that 3.3 is VERY important. This is the primary cause of the shutdown here. Everyone always talks about 12V rail which is very important for high overclocks. But it wont make your comp shutdown necessarily but 3.3 certainly will. It also happens that shutdown occurs are around 3.12 area. Try with 1 videocard and see if the 3.3 goes higher. Make sure the PSU 24 pin is plugged in all the way and that i doesnt interfere with an IDE connector or the floppy connector.

Also looking at specs:
+3.3V@33A (this should be sufficient actually), +5V@24A, +12V1@13A, +12V2@18A, +12V3@16A, make sure that you power the dual 512 cards with 12V2 and V3 cables.

 

[cronic]

3.12-3.14v on 3.3v line is definitly an issue

 

[Ezekrael]

Thanks for the replies everybody.

Allies - interesting solution, the only one so far guaranteed to work... probably the only one so far I'm guaranteed not to try... hehehe

Matthias99 - PSU is brand new, actually, the whole PC is brand new so I'm hoping it's not down to defective hardware. Specs on the PSU can be found here http://www.silverstonetek.com/products-60f.htm.

SickBeast - Side of the case is cool. You can obviously feel warmth on the PSU but it's not excessive.

RussianSensation - I'll try it with one card and check the volatages again... if it reports increased voltages with just one card, does that necessarily mean the low voltage on the rail is the culprit for my problems?

 

[Matthias99]

Originally posted by: Ezekrael
Matthias99 - PSU is brand new, actually, the whole PC is brand new so I'm hoping it's not down to defective hardware. Specs on the PSU can be found here http://www.silverstonetek.com/products-60f.htm.

While you can't rule out a defective PSU, it's certainly not a case of an inadequate one.

A low 3.3V rail *can* cause problems -- but you also can't really rely on software monitoring for that. Get a voltmeter and check it (you can check 3.3V on some of the pins of an unused RAM socket, IIRC; google for details). Although if you just want to see if it goes up with only one card installed, you *can* probably tell that reliably.

Another thing that could be problematic would be accidentally plugging too much stuff into one of the +12V rails on the PSU. Also, from the specs, +12V1 and +12V4 are considerably lower-power than the other two, so don't plug your video cards into those!

 

[moonboy403]

i had that exact same problem with my antec neopower 480w with 7800 gt sli

the problem was that my case ventilation wasn't good enough....thus i guess the psu's overheating

when the computer turns off...does the psu gets really hot???

1 solution i suggest is that...try running your rig with the case open
it worked for me

 

[Ronin]

Another vote for the PSU, and I'll tell you why. I have the same exact PSU (not in use currently, and I'll tell you why about that, too).

Box is as follows:

1 80GB Raptor
2 250GB Hitachi T7K250's
2 160GB WD 8MB Cache
1 Pioneer A09
1 Sony DVD-RW (Don't recall the model # off the top of my head)
1 FX57
2x512MB G.Skill TCCD
2 120MM Fans (intake/exhaust), and a third on top of the XP-120
2x256MB 7800 GTX
1 Audigy2 (for testing purposes)

This was too much for the Seasonic. I had to transition to a 680W TT (I started with a 500W Antec SmartPower, went to the Seasonic, and now I'm on the TT, which I had laying around). The TT is the ONLY one that hasn't locked up on me in games with the setup. It's possible you're not providing enough power. If I recall correctly, 2 7800 512MB under load push almost 400W of power on their own (someone feel free to correct me, as I'm going on memory from a month or two ago). For what you've got, get a PCP&C 850 and remove all doubt.

 

[Matthias99]

If I recall correctly, 2 7800 512MB under load push almost 400W of power on their own (someone feel free to correct me, as I'm going on memory from a month or two ago).

Sounds too high; I think they're supposed to be around 110W each at most.

GamePC (not the best, I know, but I would hope they can plug in a wattmeter correctly) showed a jump from 284W to 389W under load going from one 7800GTX512 to the SLI config. That would be 105W of extra power, and that may be measured at the wall (which means it's not taking the PSU's efficiency into account, so card usage would be more like 75-80W at most).

ComputerPowerUser found:

While we were evaluating Nvidia?s GeForce 7800 GTX 512MB cards in single-card and dual-card SLI configurations, we used a Seasonic power meter to monitor the total power consumption of our entire test system, excluding the monitor. If you were using only a single GeForce 7800 GTX 512MB card, you?d be fine, as our test system never consumed more than 330W. Power consumption, however, increases when you install two cards. When our test system was sitting idle at the Windows Desktop with a pair of 512MB GTX cards installed, it consumed almost 240W of power. And when we threw a heavy 3D workload at the pair of cards running in SLI mode, power consumption shot all the way up to over 410W.

About 80W increase from installing a second 7800GTX512MB. Again, that's measured at the wall. Both systems were drawing around 400W at the wall, which means actual PSU load was more like 250-350W.

The PSU should be MORE THAN ENOUGH. Ronin, maybe you had a flaky Seasonic sample or something (can't vouch for the Antec), but a good 450+W PSU should be able to drive a system like yours (or the OP's) without any problem. Although, from your sig, you are overclocking quite a bit... that CAN up your power requirements considerably.

 

[Ezekrael]

Hi guys,

I've tried removing all devices like sound card, floppy drive, optical drive etc. This leaves 4 FLC0 Raptors running in RAID 0 and the 2 7800GTX 512's. Still shutting down...

Some of you are saying the PSU should be more than sufficient, some are saying it may not be enough, some are saying it may be defective... The PCP&C 850w PSU is a lot of money, do you really think I need to go as far as that to power the system?

Here are the full specs:
Athlon FX-57 water cooled with Aquagate Mini R120
2 x 512MB OCZ Platinum Elite
2 x XFX 7800GTX 512MB
MSI K8N Diamond
4 x 74GB WD Raptors in RAID 0
1 FDD
1 Pioneer 16x Dual-Layer DVD-R / 8x RW
X-Fi Fatal1ty
3 x 120 case / system fans
Wi-Fi / Bluetooth adapter

After reading all that and the specs on the Silverstone PSU, does it seem insufficient? If anybody thinks it may be which 12v rail I've plugged things into, there are 2 PCI-E outputs specifically designated for the task of powering the gfx cards, so I've used those and presumed they would give the most power to those outputs. I can't plug the GFX cards into any others due the cables provided and modular nature of the PSU. I've tried plugging each of the Raptors into their own individual IDE output, whereas they all shared the same output before. I've tried using a combination of the SATA and IDE outputs - still no good.

I just find it hard to believe that 4 Raptors and 2 7800GTX's with nothing else installed are enough to shutdown a well-rated and solid 600w PSU.

Sorry for the big rant by the way, I just wanted to convey as much information as possible as you guys have been a great help.

 

[Polish3d]

Have you tried running the second GTX alone through the same tests?

 

[Ezekrael]

Yes, I tried each card alone to discount either card being faulty. Individually, they work perfectly...

 

[Ezekrael]

Any further thoughts guys? Maybe recommend a power supply given the specs? I don't think a PCP&C 850 will physically fit in my system... Any others I should consider?

 

[Elfear]

Originally posted by: Ezekrael
Any further thoughts guys? Maybe recommend a power supply given the specs? I don't think a PCP&C 850 will physically fit in my system... Any others I should consider?

Definetely sounds like a power draw issue to me if one card works great. Especially after seeing a few threads from people with the same problem. Check this thread out. Some good info there about PSU selection with SLI'd 512MB 7800GTX's.

Also head over to Xtremesystems.org to the 3D section. You'll find some good info there about what those guys' are using for their crazy SLI'd 7800GTX rigs.

 

[Cuular]

I would suggest finding another hard drive and pull out the current 4 that you have raided.

I setup a SLI 7800GTX(256MB) system, using the previous silverstone model, and with over 2 hard drives in the system, had the same problem as you. I pulled out 4 of my 6 HD's and it made everything stable. So try removing 2 or more HD's and give it a try.

Running in raid mode is going to increase your total usage of power since anytime you access a single file, it's going to be spread out across all the drives in the raid.

So try with a single HD if you have an extra laying around you can test with.

 

[Ronin]

Originally posted by: Ezekrael
Any further thoughts guys? Maybe recommend a power supply given the specs? I don't think a PCP&C 850 will physically fit in my system... Any others I should consider?

As I said, you're seeing exactly what I did, until I found a PSU that would supply enough juice. You don't *have* to do the PCP&C. You could try the TT 680W that I'm using (which is a fair amount cheaper), and give it a go.

 

[brownzer]

I have a very similar problem to the one you are describing.

Here are my symptioms:

- Works fine out of SLI.

- Runs for a while 5-20 mins with games like quake 4 1280x1024
4 AA, ultraquality, in SLI, then sudden powerdown.

- Take side of the case off, runs for a little longer ( 30-40 mins), then powers down.

- Add 120mm blower, directed into the gap between cards =>
system is stable on 1280x1024 lcd for q4, cod etc.

- Try 1920x1200 24 inch fancy dell monitor, with call of duty everything
maxxed out, problem is back.

Factors at play affecting the heat:
1. Drivers: The latest nvidia ( non-beta) makes things a bit better.
2. Amount of extra processing (anti aliasing)
3. Resolution you play at
4. The game
5. Quality of the air flow.

Things to try:
0 Stick your hand into the air space between the cards and onto the card heat sink, see how hot they are.
1 Take the side off and run your test again.
2 Blow air _between_ the cards, this is were a heat trap seems to be.
3 If 2 works, mount a 120mm fan to blow between the cards.

My next experiment is to try reversing the air flow on the 120mm and
see if suck works better then blow.

My extra 120mm is blowing down, from a position opposite the mobo, into the gap
between the two SLI cards.

Rig:
AMD Athlon 64 X2 4800+ 2.4GHz Socket 939,
2MB, BOXED w/fan

Corsair TWINX2048-3200C2 DDR-DIMM 2048MB
Kit w/two matched CMX1024-3200C2 DIMMs

Asus A8N-SLI Premium, nForce4 SLI,Socket
-939, Super Cooling pipe, PCI-Ex16

Western Digital Raptor 74GB SATA
8MB 10000RPM

Thermaltake Silent Purepower 680W
Powersupply, 20/24pin

Antec P160 Miditower, Silver,
Side window, (Without PSU)

LG DVD±RW burner, GSA-4167RBB, 16x,
Dual, Black OEM

2 XFX GeForce 7800GTX 512MB GDDR3 XXX,
PCI-Express, Dual-DVI/HDTV/ViVo, Retail