EVGA 7900GT KO issues

I just posted this info in two threads in the official EVGA forums along with others who are reporting problems, I thought some of you might be interested in what is going on. At least I am not alone:
http://www.evga.com/community/messageboard/topic.asp?TOPIC_ID=14973
http://www.evga.com/community/messageboard/topic.asp?TOPIC_ID=14843

Received cards (2) on 5/2 with other components and assembled my new system:

AMD 4400+
2 gigs Kingston HyperX memory
MSI N4 Diamond Plus
Audigy 2 Platinum
(2) EVGA 7900GT KO cards in SLI (obviously) - mine were 520/770
ThermalTake PurePower 500W (18A on each 12v rail)
2 SATA drives and a DVDRW
Dell 2405 - 1920x1200 resolution
Gigabyte Aurora case

Aside from initial system cooling problems (solved by installing 2x120mm fans in the side grill), everything loaded up fast and worked great for 8 days. I ran Prime95 for 24 hours with no errors, and did 3DMark 05 and 06 tests with excellent results. I have two installs of Windows XP SP2 going, one for fooling around and the other for gaming/benchmarking (with nothing else loading except drivers)

Latest Nvidia drivers all around. Oblivion and Civ4 worked great with all eye-candy on. Coolbits installed for giggles only. Nothing in my system was or is overclocked.

Temps are (idle/load):
CPU - 30/40
System - 44/50
GPUs - 48/65 (under 3dmark06 high stress)

Yesterday I came home and my screen saver was bugging out. Had to reboot to restore 3d-type video. Upon reboot, went into 3DMark06 and there is where my main issue lies, I cannot get through a loop of the Artic scene in 3DMark06 without heavy corruption by the end and then the blinking black desktop upon exiting. Sometimes it happens (while looping just that part) on run 2 and other times on run 3.

Downclocking the card (500/700) only allowed 1 or 2 more loops to complete succesfully before the terror began. Disabling SLI and the loop crashes to blue screen after flashing black. Trying each card individually was the same. Changed some settings in the BIOS (such as memory and board timings) and that had no affect at all.

Sorry about the length of this.

The only thing I haven't ruled out is the PSU, and to do that I am going to hook up a spare PSU, jumper the ATX connector (so it turns on), and try running one or both of the cards off of it, with the rest remaining on my Thermaltake supply. I want to rule out a heavy drain on the 12v rails as the problem.

Oh, and fooling with Coolbits to do clocks? Doesn't work at all for the 2D test...it fails at default or under/overclocks. For 3D tests, I get the flashing/resetting screen. I had to use Rivatuner to set the clocks without having to test before OK'ing out...and then run 3DMark06 or whatever.

I had bad luck with 2 Nvidia cards about 3 years ago, left them and went ATI, and only came back because of the possible great performance driving my widescreen with SLI. Now I fear I really made an expensive mistake.

I am also getting the "dots" in 2D mode - such as on the start menu, and window artifacts remaining as well in 2D, like from nTune's graphing. This happens usually after having anything 3D running, like the screen saver or a game.

I also tried holding a high-powered 80mm fan to the cards with my case open, and that did nothing (to rule out heat). I was mainly thinking the IC issue that people had stated, to do more to help that would be beyond what I would want to put up with for $320 cards.

Have you tried the cards individually to see if just one of the two cards is bad?

Don't forget to plug in the 4-pin molex connector right above the x16 slot into your power supply. Make sure the motherboard isn't auto-overclocking anything.

I read the post about the one guy getting 4 bad cards in a row. Just not likely... Most probable answer is that something in his system is either to blame or is causing the problem.

As Mod_JKOHL said in a reply: "I know it is aggravating Guys, to get a new card(With all the Excitement of Christmas) and then find out it does not work properly, and yes You do have the right to post here and complain about it. But please realize that though it may seem like a large number of the 7900 Series cards seem to be having problems, it is in reality only a small percentage of the 7900 cards being sold."

Some of the 7900 series cards have issues. Pop over to eVGA's forums and register, then talk to the moderators. They have a new BIOS for the 7900GT/GTX that seems to clear up (some) of the issues with the delta between the pixel clock and the memory clock.

I bought two Evga 7900GT KO and put one in so I could see what one card would do ,it booted into windows and I went to run 3Dmark 05 and then bam it just went black and I tried to reboot it ,but it had a direct short ,it kept throwing the breaker in my power supply.RMA it back to newegg and put my other evga 7900GT KO in and it ran perfect,I don't mess with the nvidia auto detector ,I had problems in the past where it would detect it to run at some super high frequency that I knew it would not run,But I install my other 7900GT KO and overclock it and I thought it was'nt reading the memory right I was up to 1900 on the memory and it was running perfect,I got it to 565/1950 with no kind of volt mod,and I got my other 7900GT KO back from newegg and the memory on it would'nt run as high but still got good overclocks 575/1860.

Yes, I have now tried each card individually, with similar bad results.

Nothing is overclocked, I have been through the BIOS many times. My memory timings are actually set low due to being set to Auto. I also have all 3 power connectors filled on the motherboard (ATX, 4pin, molex)

Have not tried my PSU test yet, will try that during the day tomorrow. I am not sure even my spare PSUs have enough amps on the 12v rails to rule anything out.

I saw that they have a beta bios out, and I have now sent an email to Mario there for his suggestions...and he may possibly send that bios to me. Not sure right now since I have not gotten a response yet.

After 5 days of working fine, I threw the boxes out...so returning to NewEgg is not an option. I have to go through the RMA crossship procedure with EVGA if it comes to that for these cards.

At high loads, the wattage viewer on my PSU at times bursts to 396...that's the highest I have seen it. Usually under load it's in the 200s...and idles at 170.

I've had the exact same problems as you described Astaroth with two different 7900gt cards in a row, though each was used as a single card.

I would install the card and run it at stock speeds on my system which also runs at all stock speeds, and everything would be great for a few days. But in both cases, after about a week I would start failing 3Dmark06 either during CPU test 1, or more often the Deep Freeze test. There would be tearing, artifacting, and blue and red circles flashinga nd moving about the screen and the desktop would start flashing and showing little dots.

I tried driver reinstalls, even fresh XP installs to no avail. GPU temps were always within norms and prior to the problems manifesting 3dmark would run just fine over and over.

I have a new Antec 550w PSU that seems to be working fine, asus probe tells me I'm putting out the appropriate amps on each line within +/1 0.02 or so, except for my 12v rail which seems to be putting out 11.9 which still seems well within norms.

Anyhow, I'm going for card number 3, the previous two were factory overclocks of 550core range with high mems too. 3rd is factory oc'd, but not as aggressively. We'll see if this one survives burnin, otherwise I'll just go back to ATI.

Well, so much for my testing.

I did hook up a 2nd power supply like I stated, and had it powering the 2nd card. At first, it looked like it was going to do the trick, but then around the 5th loop of Deep Freeze I got a few artifacts. It didn't seem as many as I was used to though. I then tried testing again, with the high-power fan blowing across the cards and the south bridge. Still got artifacts but not a total wipeout. So I was going to proceded further.

(I was going to try powering the molex connector on the motherboard from the 2nd power supply as well, just to pull more juice from the second).

Then my computer just up and shutdown. Come to find out now that one of my cards has completely died. With the other one, I tried the Deep Freeze 3DMark06 test again alone for the heck of it...and it artifacted so bad that I had to reboot.

So...in other words...both my cards are toast and I am proceding with the RMA. I may end up not even trying them and selling them off...and going in a different direction. Not sure.

unusual that you would get 2 cards with exact problem . unusual that a card would just die under default use. did you check card temps with the driver cp? these cards throttle at about 130C, which is very very hot, a lot hotter than a few loops of 3dmark can ever make them. look elsewhere, mainboard, psu probably.

The CP reported the same temps as NVMonitor, which was 40s idle (depending on the day) and high (during loop testing) of 63 (I had NVMonitor logging to a text file, so those readings were actually during the tests).

Now, I know part of me thought temp related at first, but other people report temps (reading for NVMonitor or wherever) up into the 70s and have no problems, so I don't think that is/was it.

I am just really scared now since there are a handful of posters on the main EVGA forums who have been through 2-3-4 RMA's of their cards.

No matter what happens, I still may nail down a PCP&C PSU 1kw model just to rule the power issue out, either if I stick with the (2) 7900s or go with a 1900xtx, for example.

I'm pretty sure its your power supply. Never trust Thermaltake for anything high powered.

Astaroth said:

The CP reported the same temps as NVMonitor, which was 40s idle (depending on the day) and high (during loop testing) of 63 (I had NVMonitor logging to a text file, so those readings were actually during the tests).

Now, I know part of me thought temp related at first, but other people report temps (reading for NVMonitor or wherever) up into the 70s and have no problems, so I don't think that is/was it.

I am just really scared now since there are a handful of posters on the main EVGA forums who have been through 2-3-4 RMA's of their cards.

No matter what happens, I still may nail down a PCP&C PSU 1kw model just to rule the power issue out, either if I stick with the (2) 7900s or go with a 1900xtx, for example.

Click to expand...

i ran a 7800gt sli rig with my psu, a $79 sunbeamtech 550w, and now run a 7900gt sli with same psu. it is nvidia sli certified. both rigs ran without problems of any kind. you dont need a 1kw psu. since these cards do go thru qa, i find it a little tough to take that some guys got 3 or 4 bad cards in a row. this would mean that almost every 7900gt card evga ships is broken. i think these guys may have helped break these cards just a little, ie volt mods, way overclocking, etc.

I just tried again, with the single card, changing some bios settings...most notably the freq/multipliers for the NB<->SB and SB(1)<->SB(2) connections. For these I dropped down to 400mhz/8x across the board, and also tried upping the voltage on the SB one notch (I believe it was 1.30 volts? MSI bios still reported that as being "safe"). No change, I still get artifacts in the first loop of Deep Freeze.

I set everything back to 1000mhz/16x and the voltage.

When I was running both cards, the TotalWattsViewer would report spikes to 399, mostly running in the 200s, and then 170 at idle. It is now sitting at 120 idle, and was in the 200s for testing just before. I know watts don't mean a hill of beans if your 12v is not up to the task.

I don't know...I am still thinking here. And waiting to RMA the cards at least.

these cards suck the amps out of the 12v rail ,and when overclocking it jumps on up there ,I had a new Antec 550 to start with but I kept getting lock ups and games just dropping out of the game and go back to windows and random reboots ,I thought my new two 7900GT KO's were just messed up ,but I order a Silverstone Zeus 560watt with 38amps on a single 12v rail and 45amps peak and I hook that up and took care of all my problems.

Predator_CITF said:

I'm pretty sure its your power supply. Never trust Thermaltake for anything high powered.

Click to expand...

This may be an overstatement. I own a PurePower 680W and it's serving me obediently with my 7900GTX. Of course, I don't have an SLI PC nor do I play BF2 or other games that stress the GPU like 200%, but I have ran '06 benchmark many times w/o problem, so PSU may not be the culprit.

Well, I have the RMAs for both 7900GT KOs now in progress. When I get them I may just end up selling them.

Come Monday morning, I am going to order a new power supply. As opposed to going PCP&C brand...how does this sound:
FSP Group (Fortron Source) FX700-GLN 700W Power Supply
http://www.newegg.com/Product/Product.asp?Item=N82E16817104015

Supposedly, going by EVGAs techs, you need 20A on the 12v rail for one 7900GT card and 34A for two cards. Think this new supply would handle that or maybe something beefer still (see below)???

Basically I am not going for overclocks or anything, stablility and performance (with no price limit - within reason) is the key for my system.

Now, while I wait for these two RMAs, I am going to try something at least temporarily on the video side. I am trying to decide between trying (2) 7900GTX cards or a X1900XTX card. I need to make sure the PSU I get can power that. I also need to sleep on who the manufacturer of these cards should be.

Besides welcoming any comments on the above, I also just made a stupid observation: why are there posts here, nvidia's and on the manufacturer's forums for evga, xfx and bfg cards, but the msi, asus, and similar forums have none? Same goes for the newegg reviews.

ITs not a power supply issue its the cards. I run a 510 sli and have a 3rd 7900gtx on the wavy. Nvidia is aware of the problem and working on the issue. My pwersupply can handle a hell of alot more than my rig.
.

dude.. that sucks man. sorry to hear it.. nothing like getting two bad cards.. gah! I hope they get you two good replacement, fast like! lettuce know what happens.

sounds like the overheating issue that many reported on other forums.

you should check under Event Viewer->System whether or not you have an error event from NV called temperature rising...

well, seems more like bad memory or memory controller issues.. or voltage regs overheating.. that nv heating thing is for the gpu core only I think.. none the less, would be worth a gander.

This is going to sound silly...but for those of you with borked 7900 cards, have you ever set your screen saver to be something 3D (like "3D Text"), and then come back to the computer after a while and seen it wigging out? That was the first sign I had that something was up 4 days ago...previously it was fine for over a week.

Oh..and Event Viewer is clean.

Ok...update and maybe some of you who have 7900GT and GTX cards can find this useful, especially owners of the overclocked ones.

I would say "SUCCESS!" but it isn't all that great. Reminding everyone that I am only running one card now...and I am still going to RMA it and the other one since I already started that process.

I underclocked the card to 400/600, I did underclock before but not that drastic. Ran 10 times straight through Deep Freeze, and for the first time had no issues (artifacts, black squares, flashing in 2D mode, black screen, etc).

I then thought I might as well try Nvidia's reference speeds of 450/650. Again, I ran 10 times through Deep Freeze with no problems. I then decided to really go for it all, and completely ran through 3DMark06 at defaults, and successfully completed it (for the first time since Thursday when this muck all started). My score was pathetic, but I feel now I have accomplished something.

So now that I have all this info from various testing the past few days, I think I have my solution. I am going to get a new power supply and 2 completely referenced-based (no stock overclock) GTX cards...and sell the corresponding pieces I have now. I want SLI, speed, and stability...and I think now I can achieve that with what I know.

I am completely sure about all of this? No. I am open to further ideas since I have never had this problem in 8 years of building/upgrading.

The question is now, why do these problems appear for people after 24 hours, several days, or even a month or two? If the memory just can't handle the speeds, and it is possibly because of the juice they are fed due to the VGA Bios, then why does it take so long to manifest itself?

Astaroth said:

The question is now, why do these problems appear for people after 24 hours, several days, or even a month or two? If the memory just can't handle the speeds, and it is possibly because of the juice they are fed due to the VGA Bios, then why does it take so long to manifest itself?

Click to expand...

Who knows for sure... only the board partners or perhaps only nVidia themselves. My guess would be that since they are running the memory out of specc'ed speeds (700mhz in the case of the Samsung 1.4ns BGA chips) it is reducing the lifetime of the chips at a drastic level. That's only a guess and I warn you that I could completely be blowing smoke outta my ass on that. But it sure seems like the vast majority of these issues are related to failure of the memory on the boards.

Astaroth said:

Ok...update and maybe some of you who have 7900GT and GTX cards can find this useful, especially owners of the overclocked ones.

I would say "SUCCESS!" but it isn't all that great. Reminding everyone that I am only running one card now...and I am still going to RMA it and the other one since I already started that process.

I underclocked the card to 400/600, I did underclock before but not that drastic. Ran 10 times straight through Deep Freeze, and for the first time had no issues (artifacts, black squares, flashing in 2D mode, black screen, etc).

I then thought I might as well try Nvidia's reference speeds of 450/650. Again, I ran 10 times through Deep Freeze with no problems. I then decided to really go for it all, and completely ran through 3DMark06 at defaults, and successfully completed it (for the first time since Thursday when this muck all started). My score was pathetic, but I feel now I have accomplished something.

So now that I have all this info from various testing the past few days, I think I have my solution. I am going to get a new power supply and 2 completely referenced-based (no stock overclock) GTX cards...and sell the corresponding pieces I have now. I want SLI, speed, and stability...and I think now I can achieve that with what I know.

I am completely sure about all of this? No. I am open to further ideas since I have never had this problem in 8 years of building/upgrading.

The question is now, why do these problems appear for people after 24 hours, several days, or even a month or two? If the memory just can't handle the speeds, and it is possibly because of the juice they are fed due to the VGA Bios, then why does it take so long to manifest itself?

Click to expand...

if you're starting all over and dropping all that cash, why not consider the ATi route?

I know, it would make sense...except that I upgraded alot of my system (MB) to go SLI and not Crossfire, and desire to drive 19x12 resolutions with this monitor. I am trying to do that with most eye-candy on.

For the 8 days that the 2 7900 GT KOs worked, it was great. I have most features in Oblivion cranked and was getting 20-40-60 fps (Gates, Outdoors, Indoors) with Vsync on at that resolution. The only other game I play right now is Civ4 and I also cranked everything in that (I know, big deal) and it was smooth and fast.

If my new route fails as well, then I will just go with a x1900xtx and be done with it.

I have two Evga 7900CO KO running in SLI altho when I first got them one was DOA ,but after I got one to replace it they have been working perfect I played Sin Episodes at 4x FSAA and 8x aniso at 1600x1200 all the way threw about 6to 8 hours and had no problems ,I hope it keeps working perfect.Its funny they don't mind running the 1.4ns memory at 2v out of spec but they just put the core at 1.2v and if they had ran the core at 1.4v they might not had all the problems with the Superclocks.