New rig totally unstable in SLI, SLI-DR, FX-55, at my wit's end

[dobbs]

Here's my vital system specs, a machine I had prebuilt from Monarch Computer

 

 

Athlon FX-55 @ stock

DFI Lanparty NF4 SLI-DR (310p BIOS) - have also tried 414-3

OCZ ELDCGE-K DDR PC-3200 Dual Channel Gold (in orange slots)

Zalman Copper CNPS7000B CPU Fan

Enermax EG565P-FMA REV.2 FMA Series ATX 12V Ver.2.0 535W PSU

(2) BFG 6800GT OC @ stock (02.28 bios rev, then 03.10)

hp L2335 23" widescreen monitor (DVI)

 

In windows:

 

NForce 6.53 (SM bus driver only)

Forceware 71.89, have also tried 76.10, 76.44, 76.45, 76.50...

 

 

My memory timings and board voltages have been set according to specs here:

 

http://www.bleedinedge.com/forum/showthread.php?t=9942

 

All of my plugs on the board are in place, etc.

 

 

Let me begin by saying that I feel like I've tried every tip, trick, timing, etc. that I've found

in forums all over the net, and I can not for the life of me get this rig stable in any game in

SLI mode.

 

With no video drivers loaded and only 1 video card in the machine, the system has passed all Memtest,

SuperPI, StressCPU, and Prime95 runs that I've thrown at it. It seems like the underlying CPU/Memory

subsystems are doing fine, or at least appear to be.

 

Once I loaded the video drivers and began trying to test games, well...

 

Basically, the PC will crash any time SLI is enabled on the system and I go to play a game.

Half Life 2 is the main culprit, but I also get crashes in World of Warcraft. The system seems

to work fine with either of the cards operating individually in 1 slot. I did this just to make sure

that one of these 6800GTs being defective wasn't the problem. The system also works fine with both

cards in the machine with SLI disabled. And yet still, it also works fine with SLI enabled but with

the SLI bridge off ("software" SLI). It is only when SLI is enabled and the bridge is connecting

the two cards for full SLI operation that I get the issues. And yes, I have two SLI bridges and it

does it with both of them (and I've tried facing them both ways; I don't know whether this matters).

 

The main offending app, or the game that I have the most "success" with crashing the system is

Half Life 2. Usually it will crash inside of 10 minutes, but I've had it go for an hour or so

before crashing one time using the DNA Forceware 76.44 drivers. Of course, once I crashed with

these drivers, my system was no longer stable even inside of Windows, locking up about 10 seconds

within logon (with a noticeable high pitched squeal emanating from the motherboard). I quickly

went back to the 76.45 WHQLs and no longer encountered _this_ type of crash...

 

The crashes are particularly heinous as they knock my system to a state where the machine won't even

come back from a reset. I've had garbled, unreadable BSODs (nv4_disp.sys except totally unreadable),

nasty vertical banded test pattern type screens, and just 100% straight up black screen hard crashes

which trigger a reboot (even when I have the system NOT set to automatically restart on critical

errors). After I crash to a black screen, my monitor goes out of sync as if I were going to POST

... except it doesn't come back at all. I think this may have something to do with the cold/warm

boot problems on the SLI-DR board w/ a DVI monitor outlined in a thread on DFI-Street.com (don't have

link handy).

 

If I press reset, when I look at the 4 diagnostic LEDs on the NF4, all 4 light up and the system does

nothing except blink the 1st LED (which is CPU detect, I believe). It will not POST at all here.

I have to cold power it down and then power it back up in order to POST. All of this from a crash

in a game? Seems like a bit much.

 

The rare times when HL2 hasn't hard crashed my machine and when I've crashed in World of Warcraft,

I get a standard Windows application error with the "the referenced memory area can not be "read"

message. I think this traces back to some STUCK_THREAD Q article, not sure.

 

Here's what I've tried thus far:

 

Vid card

--------

 

1) Switching cards / SLI connector on/off

2) Flashed BFG 6800GT bios 02.28 to latest 03.10 (www.mvktech.com)

3) Different drivers, all of them crash equally

4) Underclocked the 370/1000 stock speeds on the 6800GTs to Nvidia spec 350/1000

5) Overlclocked the cards just to see if it'd crash the machine quicker. It did.

 

Memory timings / voltages / BIOS

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

 

1) Tried everything on AUTO

2) Set specs according to thread posted above

3) Updated to 414-3 bios, still same issues

4) Enabled / disabled all motherboard driven devices (IRDA, RAID, sound, etc.)

 

Software

--------

 

1) Tried Rivatuner Athlon "compatibility fix"

2) Multi-GPU / Single GPU rendering modes

 

A lot of times these were done on a fresh install of Windows. By my count I've reinstalled the system

10-15 times or so.

 

 

Conclusions and questions thus far

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

 

I'm getting ready to RMA this thing, but I don't feel like I should have to. I am sure there

are dozens of others that have an FX-55 with this board running 2 6800GTs in SLI mode just fine

without these problems. If you're one of them, please speak up and let me how it was accomplished.

 

* I have a 535W Enermax power supply with 18a each on 2 12V rails. This should be enough for SLI,

yes?

 

* The fact that the machine doesn't POST after a crash and requires a hard poweroff makes me

suspect there's some kind of issue with the motherboard. Am I right to assume this?

 

* The fact that I am getting these horrible test pattern crashes / garbled blue screens also makes

me wonder if either the cards are unstable at their gaming temps (71-75c) or if not enough power

is being supplied to them (which shouldn't be, right?) Yet the cards work fine one at a time...

 

 

Basically I guess I don't want to admit defeat. But I've already struggled with this thing

all week.

 

Any help anyone can provide would be welcomed. I'm crossposting this to a couple different

hardware forums so please excuse me if you see this twice.

[Whiskey]

Sorry about your problems, I salute you for sticking to your guns to determine the problem, you'll learn a lot more that way. I would guess that if you have tried both vid cards as singles, and not had problems with either one, that the next step would be to look at the psu. Good luck.

[dobbs]

You'd think though that this 535W PSU that claims SLI certification would be enough for this setup though, right? I mean I'm not saying it's not the problem, but it _should_ be enough power for this rig, yes?

[Sharp]

Hello,

 

The PSU seems to be ok, did you move the SLi jumpers so that both 16x PCie slot is running in 8x?

[jtworldwide2]

i would agree with the PSU to look at next. That setup is at the top of the heap in terms of power consumption, FX55 and 2x6800U is a lot. Add on top of that any hard drives and opticals that you have and a motherboard that is very fussy about power supplies and your talking about a ton. Not to say that ennermax makes a bad product, but from the symptoms you are describing i would say it to be a power issue IMO.

[h_Us3r]

Its definitely the PSU, that thing should be pushin *at least* 26A..

 

h

[dobbs]

Yeah I've moved the SLI jumpers on the mobo for SLI operation. Of course, the system was shipped to me in single VGA mode from Monarch... so I found out about that real quick.

 

jtworldwide -- actually its 2 6800GTs, not 6800 Ultras. Don't know if that makes a difference.

 

h_us3r -- In any event, the total amperage listed on the 12V rail is 36a, 18a apiece. You think this is not enough? For a company that advertises this as a SLI PSU to then not have enough amperage seems kooky.

[red930]

Just for clarity.

 

Have you qualified your RAM one stick at a time with your current BIOS settings using memtest?

 

Have you qualified your RAM in the OS by running prime95 or OCCT?

 

Until you get these two steps done you aren't sure the rig is stable.

 

Post back with your results.

[dobbs]

Exroadie -- per my post

 

---

With no video drivers loaded and only 1 video card in the machine, the system has passed all Memtest, SuperPI, StressCPU, and Prime95 runs that I've thrown at it. It seems like the underlying CPU/Memory subsystems are doing fine, or at least appear to be.

---

 

I haven't qualified the RAM one stick at a time; only with both in. But it did pass 3 full loops of memtest out of the OS without a hitch, it also did 8 hours of Prime95 blend and 18 hours of StressCPU without any issues. Not a super amount of hours, I realize, but enough for a quick stability check or so I was led to believe.

[Angry_Games]

the drivers I use for SLI are 76.41 beta drivers.

 

with 2x 6800GT in SLI + an FX-55...I would really suspect the power supply. No this isn't just me keeping blame from the board, as if i thought it was the board I'd be the first to say so (i've enough experience with these boards to have an idea when the board is the problem).

 

i would also try the 3/16/2005 bios or the 4/14-2 bios (based on 3/16) as it is the absolute best bios I've used.

 

but i keep coming back to maybe your psu is just not enough for the entire system with both SLI GT's in place.

 

try the 3/16 or later bios (anything with the -2 at the end) and let us know how it goes.

[dobbs]

I'll try the 414-2 bios -- what's different about it from the 414-3 that I already tried? I thought the dash was just to indicate the revision and didn't mean anything special.

 

Man, if this 535W powersupply with isn't enough, what should I get? OCZ 600?

[RGone]

Dobbs you need to put a full system spec up in your signature so all can take a look at what is in the case and better ascertain the problem. Something like this would be far more helpful than just cpu / video card / and power supply.

 

ILLUSTRATIVE SYSTEM SPECs

DFI XXXXX Motherboard

Bios AN81D107 01/07/04

Fortron 530WATT PowerSupply

3200XP+ Stock speed

2x256MB Corsair PC3200C2 Rev1.1

512MB Corsair PC3200C2 Rev5.1

120G Maxtor,160G Seagate

200G Western Digital

Sony DVD±R ±RW Burner

Pioneer 16X DVDrom

Promise Ultra 133TX2 IDE Contoller

BBI 9800 Pro 128MB

Audigy II

D-Link F5D5000 10/100 NIC

Windows XP Pro SP 2

Nvidia 5.03 chipset drivers + Omega 2.5.97 Radeon drivers.

 

See you have the problem and not us and we need every bit of information that will give clues. I checked your thread at bleedinedge and you seemed to say all was fine. Now you say it is not?

 

You don't mention really how fast you are trying to run the VX and not having any myself to test with I would bet 2.9V is not enough. It has been seen time and time again on the net that most of that stuff does not start to take off and do welll until about 3.0V or 3.1V. People who just hang on the bitter end of not upping voltages will forever see a problem is my guess.

 

One other thing I can now say for sure and that is that Memtest86 has little to do with how memory will work in windows. 240 passes of memtest86 on a setup I have now and will not run 9 seconds of Prime95. So the variables are many and change from setup to setup. FX is hotter and more of a power hog as I understand it. So you will have to loosen up the timings and give that memory some more voltage and it likely will quit all the crap. After that as test you can see if you need to do something else. But just hammering at the same old dead end of timings and voltages is going nowhere is my guess.

 

RGone...