GeForce 7900 Inferno – Burn Baby Burn @ [H]

did any one else see this on nvnews

"NVIDIA wanted to pass along the following information in regards to "random freezing" issues on overclocked versions of the GeForce 7900 GTX:
In working with our board partners, we discovered the cause of the random slowdown and temporary lockup problems experienced by various users of certain over-clocked 7900 GTX graphics boards.

Essentially, it was a case where the core and/or memory clocks were driven a bit too high above NVIDIA specifications, and the over-clocking margins weren't available on those specific boards.

Users who experience problems should contact the graphics card vendor to fix the problem.

We have learned from our board partners that the situation should now be under control.

Also, in regards to 7900 GTX "Corruption upon exiting a 3D app (3DMark06)":
This is a new issue that has just been reported, and our software team is looking at this.

--Brian Burke, NVIDIA Corporation"
 
Blue Falcon said:
Of course their support forum had 7800GT problems listed. It's a support forum primarily designed for evga customers to consult others for technical problems with eVGA products. People don't go to a support forum to say 'Hey my video card works perfect!'. If you look there are just as many 'problem threads' about the 6800GT and other GPU chipsets as there are for the 7800GT. These are simply normal statistics at work here.

There is nothing wrong with the 7800GT, you must have just gotten a bad one. :)

Not trying to ruffle your feathers or anything, I just don't see the need to stir up a fire when things are pretty warm already.

Good point, but what I meant was that it looked like it was a "common" issue and that the issue wasn't just from a few erratic people. Not only that, but when the website says no cross shipping available, and someone from customer service posts that they will offer overnight cross shipping for the 7800, I assume that there were something wrong with the 7800's. I checked about a month ago.

I also wasn't trying to stir up anything more than there was. I just noted that their may be possible problems with the 7800's and thank the great service I got from the company I bought my card from.

I can understand that I was looking at the forums, and that that's what the forums where made for - problems. But I didn't expect the forums to be filled with that many problems similar to mine.
 
We've read that... problem is that many users using non-oc'd cards are having there cards die on them as well.

My Asus 7900GTX has been running fine... replaced an XFX XXX card that died. I'm still un-easy using it. Almost feel like I shouldn't use it.. which is ridiculous since I paid soo much for it.
 
GameDad said:
did any one else see this on nvnews

"NVIDIA wanted to pass along the following information in regards to "random freezing" issues on overclocked versions of the GeForce 7900 GTX:
In working with our board partners, we discovered the cause of the random slowdown and temporary lockup problems experienced by various users of certain over-clocked 7900 GTX graphics boards.

Essentially, it was a case where the core and/or memory clocks were driven a bit too high above NVIDIA specifications, and the over-clocking margins weren't available on those specific boards.

Users who experience problems should contact the graphics card vendor to fix the problem.

We have learned from our board partners that the situation should now be under control.

Also, in regards to 7900 GTX "Corruption upon exiting a 3D app (3DMark06)":
This is a new issue that has just been reported, and our software team is looking at this.

--Brian Burke, NVIDIA Corporation"

Oh wow, I experience random lock ups. The game would just freeze for like 20 - 30 seconds. But if I wait it out, the game would just resume like normal. This is rare though.
 
I've had terrible luck with the 7900 GTs, and thought I'd share some observations:

First, I ordered my card (initially) the day after the 7900s came out. I had two fully-functional power supplies, a Tagan 480-U22, and an Antec Neo HE 550 (A3 designation) which were rock-solid on a multimeter through any and all power draw. I had two different sets of memory, 2 different processors, but only one motherboard on which to test (Asus A8n-SLI premium). Any problems I indicate replicated with both power supplies, I tinkered with processors and ram, but didn't consistently test everything with both sets (ram was all tested with memtest, is stable, processors run forever on prime-95, appear stable).

First card I got was an XFX card, that had terrible artifacting out of the box, even when clocked BELOW stock clocks (350/600) was showing giant triangles, etcetera. After 3 minutes holding, the XFX tech had me check PSU voltage, fsb, etc etc, and asked me to put coolbits on and downclock the card just for curiosity's sake. When the card artifacted at this setting, XFX explained that they would not have the card in stock at their warehouse for at least a week and suggested that I RMA it through my retailer.

I did so, but they had no more XFX cards, so I got a BFG, as I've owned a handful of BFG cards over time. The BFG card I got was fine, worked great for about 6 days, then began artifacting, freezing, etcetera. I RMAed the card, after close to 1 hour on hold with Tech Support. The replacement card was DOA (with the motherboard reporting a VGA failure and refusing to post). I RMAed this card, and got a third replacement card. My third BFG card has MOSTLY worked, but has shown significant artifacting 4 times(Twice in Day of Defeat: Source, once in World of Warcraft, and once in 3dmark06), and I've had to shut down each of those times. I've not tried to overclock the card at all as I assume I'm on the razor's edge of failure all the time.

My computer is well-cooled, there's good airflow over the graphics card from a panaflo L1A 120 mm fan (at 7v) pointed directly across the card, and the coolbits reported temperature after 3-4 hours of RTHDRIBL is 54-55 C. Nonetheless, there are very small areas on the card that are hot to the touch.

I dunno, I'm pretty unhappy with nVidia atm.
 
Something odd is happening.

Normally when a site posts an article like this all the other tech sites have it as a news item. Seems all the other websites aren't talking about it.

It's in the Anandtech and Tom's hardware forums as well as Tweaktown.. but none of the sites have it as a news item. It's like it's being kept secret.

I mean when ATi's HDCP thing came out it was everywhere... and the issue wasn't nearly as big as this problem.

What's going on? Anyone know?
 
drakken said:
I had a card out of the first batch and it was utter trash BFG replaced the card several times but one of the issues was the peg link mode on a8n32,
needs to be disabled as it has issues with the 7900 cards, also asus has put out a new bios for the a8n32 boards to improve compatiblity with the 7900 cards:
A8N32-SLI Deluxe Bios version 1205
1- Enhance PCI / PCIE compatibility
2- Support 7900 Series VGA cards
3- Enhance SATA RAID compatibility
so the speed of your pci-e bus may be an issue too as it fixed my current problem at the time.

I completely agree. PEG link mode is a hazard. I wanted to test PEG again using my 7900GTX EGS after flashing the 1205 bios. When set to disabled, BF2 was artifact free and I looped Deep Freeze HDR test 50 times at 1600x1200 16AF with no problems.

The minute I set PEG to Auto, all hell broke loose. I fired up BF2 and within 10 seconds my screen turned to a multicolor snow.
 
killerbobjr said:
I'm reposting this message here that I posted over on Anandtech:

----------------------------------------------------------
I came across this post here which mentioned cleaning the areas around the pencil vmods. Well, I thought, what the heck let's give it a shot. Sure enough, my 7900GT stopped blanking on me. However, my memory overclocks were still limited to around 800 vs. the 880 I had gotten when I first overclocked the card.

I looked at the back of the board against the light and noticed that there was all sorts of residue left over from the assembly/soldering. Most of the time, the assembly factories do a quick wash to get rid of all the flux material on the circuit boards. This is usually good enough. However, when we start doing our mods, we're not always too careful and lay our sweaty, oily hands right across the components. This adds conductive paths across circuits and causes instabilities.

Here's what I did and what I recommend:

1. Get a bottle of 90% Isopropyl alcohol. Pure acetone (nail polish remover) should also work.
2. Get a roll of paper towels.
3. Take one paper towel, saturate a corner of it with alcohol.
4. Scrub vigorously the areas around the chips that control the memory and gpu voltages. Also scrub the components opposite each the ram chips. As your paper towels shreds, soak a fresh corner and continue. IMPORTANT: Don’t put your hands or fingers on the freshly scrubbed areas!
5. Get a clean paper towel and rub vigorously, sopping up all the alcohol. DO NOT BLOW ON IT TO DRY!
6. Hold up the board at an angle and look for any traces of residue. Use a magnifying glass if necessary. Repeat steps 4 and 5 if any residue is detectable. The board should look absolutely shiny.
7. Do your volt mods. DO NOT blow on the board to dry or to blow away graphite from a pencil mod. This will leave residue from the moisture on your breath. Use a can of compressed air instead (if you use an air compressor, make sure it has fresh oil and moisture filters).
8. Put your video card back into your system holding it only by the edges and bracket.

After you do this, your results hopefully will be similar to mine –

7900GT before cleaning:
GPU = 630MHz @ 1.4V and high fan speed (VF900), no artifacts, no blanking
Memory = 800MHz @ stock volts, no artifacts, no blanking
-- with pencil volt mod, 840MHz, no artifacts, but lots of blanking.

7900GT after cleaning:
GPU = 650MHz @ 1.4V and LOW fan speed (VF900), no artifacts, no blanking
Memory = 940MHz @ STOCK volts, no artifacts, no blanking!!

My 7900GT is now rock solid stable even at the low fan speeds.

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


There are two reasons why everyone has been getting artifacts/lockups on their overclocked 7900 cards. The first is dirty boards (literally!) from the factory and from our handling of the cards. This will not only cause blanking as described above, but also artifacting (and, to a lesser degree, lockups).

The second reason everyone is experiencing problems is overheating. My same 7900GT with a VF900 heatsink, will completely lock up after a while for about 15-60 seconds while playing F.E.A.R. (1600x1200 4x 16x, softshadows on) IF the fan speed is low OR the room is warm. If I crack open my case and turn the VF900 fan speed to maximum, I get ZERO lockups, even after hours of play.

There is absolutely nothing wrong with any of the manufacturer's overclocked 7900 cards. If you've grabbed the cards when inserting them and laid your sweaty, oily palms across any of the components or the PCB, you've created changes in resistance and capacitance that will quickly cause instabilities. If on top of this you are running things in an inadequately ventilated case, you will experience lockups, artifacts, and various other problems.

There is no need to RMA your card -- just make sure it is spotlessly clean as described in the guide above and get plenty of cool air to the heat sink.

- kbj


Your theory is flawed since it does not address the fact that I properly handle my cards and my first worked flawlessly for 60 days. I have exceptional cooling and a full tower with plenty of room to work. Your statement almost surely applies to your situation but not mine.
 
jacuzz1 said:
Your theory is flawed since it does not address the fact that I properly handle my cards and my first worked flawlessly for 60 days. I have exceptional cooling and a full tower with plenty of room to work. Your statement almost surely applies to your situation but not mine.

I've tried my hardest not to even acknowledge those posts, the whole idea of scrubbing down the backside of your card is just plain silly.

Hell, I think the post in question actually originated HERE and not at another forum. It's just been copied around a bit over the past few days. :)
 
I recieved the replacement for my 7900GT CO SuperClock just now via UPS next day air.

I havent had time to do an in depth comparison yet but just from first glance I can tell these differences between the replacement and the original card:

1. The cooler is much bigger, covering the memory too similar to the KO series cards. It's probably the same cooler used on the KO cards in fact.

2. The voltage regulation circuitry on this card seems to have more and different components than the original. Particularly in the 'voltage mod' area where everyone has been using conductive ink.

3. The card's part number is the same, but the serial number begins in 610565 instead of 608565. I believe this could indicate a different hardware revision, or it could be just a different 'batch'.

I'm going to pull the original 7900GT CO SC out of my machine now and take some comparison photos for you guys. Be back in a bit. :)
 
It's awesome that Kyle put this together and gave it front page exposure. I understand contacting both NVidia and the manufacturers to get their perspective on it, as well as get their responses 'from the horse's mouth'. Now, how about we do some [H]ard investigative work into what causes, or could cause this and help get us all back to gaming.

This, to me, seems similar to the old bad/leaky capacitor on the motherboard issue from several years back. There are too many different solutions to the problem and the success rate of each seems to vary. The best theory I have seen to explain the failures across the variety of the cards would be some supplemental circuitry that isn't up to the task of doing something that seems to be tickled by Oblivion and Deep Freeze.

This link on the eVGA forums suggests that maybe it's a memory voltage regulator going south after running too hot for too long. Seems like it'd be reasonable, as it seems that once the failure occurs, the hardware is, in fact, damaged and in many cases only seems to run in sub-stock configurations.

So, my question is this... Is there any way that [H] can bring its resources to bear (e.g. what's Brent doing ;) ) on examining the hardware aspects and similarities of the failures? Is the 7900 GT running significantly different supplemental circuitry from the GTXes? Are the three mfgs in question using different suppliers for said circuitry than Asus, etc.? Seems that this would be awesome information to gather and research. Anyone have access to multiple cards from multiple vendors and possibly some thermal imaging hardware :)

posterChild

p.s. is it just me, or is every workplace deader than a doornail on the Friday before the long weekend :)
 
I did a quick and supremely scientific statistical analysis of the EVGA thread where people report their bad cards. Overclocked GT cards seemed to suffer lots of problems (because they're so numerous?), but when comparing GTX cards I counted something like 5+ overclocked cards versus 30+ stock. This is, of course, incredibly accurate and fully scientific, but anyway it makes you wonder a bit. Like for example if the ratio is in fact real why would 5+ OC cards create a problem for manufacturers, but nVidia wouldn't have to care about 30+ cards that fail in stock configuration.
 
Blue Falcon said:
I recieved the replacement for my 7900GT CO SuperClock just now via UPS next day air.

I havent had time to do an in depth comparison yet but just from first glance I can tell these differences between the replacement and the original card:

1. The cooler is much bigger, covering the memory too similar to the KO series cards. It's probably the same cooler used on the KO cards in fact.

2. The voltage regulation circuitry on this card seems to have more and different components than the original. Particularly in the 'voltage mod' area where everyone has been using conductive ink.

3. The card's part number is the same, but the serial number begins in 610565 instead of 608565. I believe this could indicate a different hardware revision, or it could be just a different 'batch'.

I'm going to pull the original 7900GT CO SC out of my machine now and take some comparison photos for you guys. Be back in a bit. :)

Again...I think you're onto something there.

I dont think the original tiny heatsink was nearly enough for the overclocked 7900GT's. :(

Even if it wasnt just a heaksink problem...I think they've evolved the overall design of the original 7900GT's for a reason.
 
jacuzz1 said:
Your theory is flawed since it does not address the fact that I properly handle my cards and my first worked flawlessly for 60 days. I have exceptional cooling and a full tower with plenty of room to work. Your statement almost surely applies to your situation but not mine.

Do you live in a dust free environment? Does your humidity level never go above 5%? Even a small amount of dust lowers the efficiency of any heatsink by a disproportionate amount which greatly affects cooling ability. And if you have humidity in the environment, it gets slowly absorbed by any flux residue on the card eventually causing problems -- even if you never lay a hand across the circuit board.

Exceptional cooling means nothing if localized heat spots don't get cooled down. The voltage regulators put out quite a bit of excess heat that most folks don't adequately cool, even though the gpu and ram are being cooled.

- kbj
 
From what I've been reading here, eVGA forums and other places...the problem IS NOT only with the factory overclocked cards or in the case of the eVGA cards, the GT CO with the small cooler that doesn't cover the ram. I have a 7900GT CO KO (with the larger hsf) coming next week but I'm not totally convinced the problem isn't some inherent design flaw in the hardware that can't be fixed with new drivers or bios updates or more cooling. :( The question I ask is this: the 7600GT is a very popular card and many of them come with substantial factory overclocks yet we aren't hearing nearly the number of problems as the 7900 series...so why wouldn't it be something in the hardware design?
 
The original eVGA 7900GT CO SuperClocked. Ordered from Newegg at the beginning of May:



The replacement eVGA 7900GT CO SuperClocked, recieved this morning from eVGA. Shipped yesterday afternoon via UPS next day air saver:




As you can see there seems to be some added components in what I call the 'voltmod' area, and there are subtle differences elsewhere on the card.

The card has a sticker on the backside that says it has BIOS version 5.71.22.14.49 but in fact it has 5.71.22.14.45. (according to the nVidia control panel anyway)

eVGA said in an earlier email that they would be sending out 'slightly modified' and tested replacement cards so this looks like the end result.

Now as to wether they are really 7900GT CO SC cards that have been modified after the fact or not? Who knows. They look strikingly like the KO version of eVGA's 7900GT and I have heard from a few people that the KO versions of the 7900GT actually have these additional components in the voltmod area. So it may be that eVGA sent us all rebadged 7900GT KO SuperClocked cards. I have absolutely no complaints for sure, I'm just happy to have a (hopefully) working card. :)

I just want to again say "THANK YOU EVGA!!!". I am VERY happy overall with my RMA experience and with their customer service sofar. :D
 
Blue Falcon said:
I've tried my hardest not to even acknowledge those posts, the whole idea of scrubbing down the backside of your card is just plain silly.

Ahh, you must work in a mil-spec environment that requires conformance to ISO standards. You've probably never gotten any boards on your bench that were anything less than perfect. If you ever get a chance to work in radio or high voltage, you'll see that leftover flux residue is not so silly and can be a big problem. On these 7900 video boards with lots of surface mount caps and resistors, close traces, and high frequencies, it can also be a big problem.


Hell, I think the post in question actually originated HERE and not at another forum. It's just been copied around a bit over the past few days. :)

I actually posted it first at Anandtech, then reposted at OCWorkbench after reading this post after a rather desperate search in Google for a solution to my video blanking problem.

- kbj
 
killerbobjr said:
Do you live in a dust free environment? Does your humidity level never go above 5%? Even a small amount of dust lowers the efficiency of any heatsink by a disproportionate amount which greatly affects cooling ability. And if you have humidity in the environment, it gets slowly absorbed by any flux residue on the card eventually causing problems -- even if you never lay a hand across the circuit board.

Exceptional cooling means nothing if localized heat spots don't get cooled down. The voltage regulators put out quite a bit of excess heat that most folks don't adequately cool, even though the gpu and ram are being cooled.

- kbj

So people buying nVIDIA cards need to live in a dust free ultra clean bubble? Like those ones used during viral infection pandemics?

Not an excuse. There's something wrong and these companies need to fess up and tell us what it is and then fix it. If they don't... nVIDIA's sales will suffer.


Ahh, you must work in a mil-spec environment that requires conformance to ISO standards. You've probably never gotten any boards on your bench that were anything less than perfect. If you ever get a chance to work in radio or high voltage, you'll see that leftover flux residue is not so silly and can be a big problem. On these 7900 video boards with lots of surface mount caps and resistors, close traces, and high frequencies, it can also be a big problem.
Well that sounds fine and all.. but please explain why a more sopphisticated card with a larger cooler (which attracts more dust), more components on the PCB and which produces more heat doesn't exhibit any of these symptoms (I'm talking about the x1900XT and XTX of course).
 
The 7600GTs aren't coming with the substantial factory o/c's - at least not like the 7900s. The reference specs are 560c/1400m - so they're clocked pretty high out of the box..And out of 19 listed @ Newegg, 11 adhere to the reference speeds, 2 w/ cores @ 590, 4 @ 580, 1 @ 570, only 1 @ 600. Anyway, back to your point - perhaps the reason why there aren't nearly as many problems with them vs 7900s are due to the castrated pipes or something?

@ BF thanks for the pics - haven't clicked the thumbs yet, but can already seem some of the differences in the new revision..Keep us all updated on the cards progress.

[edit] Sure seems to be quite a few micro-resistors that (possibly) should have been there all along. Of course, if they're just rebadged KOs, that's kind of a moot point. I wonder if all of us will be getting the same cards, or at least similarly equpped (ie: new cooler, along w/ the new ICs)? Especially considering it looks like the GT CO/563 is no longer being sold (NewEgg and Monarch both dont have them - only 564/565/567 I think).

Also, any chance you could do a test with a DMM (if you have one) to check the vGPU and/or vMem levels? I'm curious to see if they actually upped the volts this time (vGPU at least), unlike with the N563 beta BIOS that had 1.35v programmed, but still only gave 1.2v.
 
kush said:
Also, any chance you could do a test with a DMM (if you have one) to check the vGPU and/or vMem levels? I'm curious to see if they actually upped the volts this time (vGPU at least), unlike with the N563 beta BIOS that had 1.35v programmed, but still only gave 1.2v.


Taken whilst running the Real Time High Dynamic Range demo:

V.GPU:
new7900gt_vgpu.jpg


vGPU measured at 1.2v while at the desktop, 1.45v when a 3D app is running.

V.MEM:
new7900gt_vmem.jpg
 
Cool, thanks...That answers that, no change in vMem (but thats expected as they're running @ rated max afaik), but 1.45v on vGPU now.

I guess the only question is if all the RMAs will have it [vGPU+] or only your model, ie: I'm going to be RMAing my 563/co 'vanilla'. Intersting though, now our lowly GTs have higher volts than standard GTX models (1.4v right?). No need really for a paint mod anymore (if it would even work at this point), unless you are someone on high end water/LN2/etc cooling and really want to try frying a card :p .
 
Nice.. I was thinking about adding another .05v to my gpu. :) looks like I should be able to do that with the comfort of knowing cards are shipping that way now. :) heh.

EDIT: I wonder if those new parts are doing anything other than just better voltage regulation?
 
Interesting....this got deleted from the Digg search engine. Seems as though somebody does not want it to be seen at all.
 
Too late...we all got secretly irradiated when Desmond shut down the device in the hatch on the LOST finale. ;)
 
XFX just asked me to overclock my memory by 5 mhz on my 7900 gt (520-1500) card. He asked me to play some games and come back with the results. I dont quite understand how this could help the situation, but ill give it a shot. My card would artifact in bf2 with blanking white blocks, and die in fear randomly.

-Steve
 
killerbobjr said:
Do you live in a dust free environment? Does your humidity level never go above 5%? Even a small amount of dust lowers the efficiency of any heatsink by a disproportionate amount which greatly affects cooling ability. And if you have humidity in the environment, it gets slowly absorbed by any flux residue on the card eventually causing problems -- even if you never lay a hand across the circuit board.

Exceptional cooling means nothing if localized heat spots don't get cooled down. The voltage regulators put out quite a bit of excess heat that most folks don't adequately cool, even though the gpu and ram are being cooled.

- kbj


Yes...........Roflmao
 
Here is the bottom line as I see it. Whatever the problem is , they have chosen a way to articulate the answer while keeping in mind the lowest common denominator. The only important thing too gleen is that they know the cause and are fixing it.

PS: EVGA rules, they are overnighting me a card to replace my non tested one and it will be retail boxed. I inferred from the tech support guy that the issue is solved, whatwever it is and its smooth sailing from here.

PS: I just ordered a Quad 1k for good measure lol
 
It's great that eVGA has decided to fix this issue on there end.. but this still leaves nVIDIA, BFG, XFX and God knows who else.

Also.. we don't know if eVGA has truly fixed the issue.. only time will tell.
 
ElMoIsEviL said:
It's great that eVGA has decided to fix this issue on there end.. but this still leaves nVIDIA, BFG, XFX and God knows who else.

Also.. we don't know if eVGA has truly fixed the issue.. only time will tell.

Good point, I can say that my "pre-tested card" if you will is humming along nicely. once I pass the 60 day threshold I will be totaly convinced. I might add that it would be a serious risk that they are taking shipping "fixed cards" at this point that end up failing as well and as such, my confidence level is high.
 
jacuzz1 said:
Good point, I can say that my "pre-tested card" if you will is humming along nicely. once I pass the 60 day threshold I will be totaly convinced. I might add that it would be a serious risk that they are taking shipping "fixed cards" at this point that end up failing as well and as such, my confidence level is high.

Good luck dude..:)

XFX did not give me nearly as much support.. the day after I posted in teh forums.. there forums got closed. I RMA'd through the store and got an Asus card instead.... hopefully my Asus card keeps on chugging.. :)
 
Blue Falcon said:
The original eVGA 7900GT CO SuperClocked. Ordered from Newegg at the beginning of May:



The replacement eVGA 7900GT CO SuperClocked, recieved this morning from eVGA. Shipped yesterday afternoon via UPS next day air saver:




As you can see there seems to be some added components in what I call the 'voltmod' area, and there are subtle differences elsewhere on the card.

The card has a sticker on the backside that says it has BIOS version 5.71.22.14.49 but in fact it has 5.71.22.14.45. (according to the nVidia control panel anyway)

eVGA said in an earlier email that they would be sending out 'slightly modified' and tested replacement cards so this looks like the end result.

Now as to wether they are really 7900GT CO SC cards that have been modified after the fact or not? Who knows. They look strikingly like the KO version of eVGA's 7900GT and I have heard from a few people that the KO versions of the 7900GT actually have these additional components in the voltmod area. So it may be that eVGA sent us all rebadged 7900GT KO SuperClocked cards. I have absolutely no complaints for sure, I'm just happy to have a (hopefully) working card. :)

I just want to again say "THANK YOU EVGA!!!". I am VERY happy overall with my RMA experience and with their customer service sofar. :D

Thanks for the post man. Ima give it a couple weeks to see how all this gets sorted out, and then start the RMA process for my two 7900GT's.
 
ElMoIsEviL said:
So people buying nVIDIA cards need to live in a dust free ultra clean bubble? Like those ones used during viral infection pandemics?

Not an excuse. There's something wrong and these companies need to fess up and tell us what it is and then fix it. If they don't... nVIDIA's sales will suffer.

Point taken. The stock heat sinks on the 7900GTs are inadequate for doing any sort of overclocking. If anything, the manufacturers of overclocked cards should be replacing the stock heatsinks with more robust ones (a la eVGA's KO version) before selling them. This doesn't mean however that there's something horribly wrong with the cards.


Well that sounds fine and all.. but please explain why a more sopphisticated card with a larger cooler (which attracts more dust), more components on the PCB and which produces more heat doesn't exhibit any of these symptoms (I'm talking about the x1900XT and XTX of course).

The X1900 cards have heat sinks with much a bigger surface area, quite a bit more CFM moving over them, and they ventillate outside the case. The 7900GTs with the stock heat sinks have a much smaller surface area, much lower CFM, and most importantly they don't ventillate outside the case. To me, that's just asking for trouble when overclocking.

- kbj
 
The X1900 cards have heat sinks with much a bigger surface area, quite a bit more CFM moving over them, and they ventillate outside the case. The 7900GTs with the stock heat sinks have a much smaller surface area, much lower CFM, and most importantly they don't ventillate outside the case. To me, that's just asking for trouble when overclocking.

Mmm kay what about use ATi x1900 users who've replaced the stock cooler with a Zalman VGA cooler or in my case a waterblock?

We're still not experiencing issues.
 
Just thought I'd add what ViperJohn said about my defective card when he evaluated it.

john said:
Looks like the card needs to be RMA'ed. I have the card on a ViperMagnum-II air cooler running the stock bios with the voltages turned down (although still a bit higher than stock and the card is locking up in 3DM05 FireFly Forest. It gets even worse in 3DM03 where the on-card power supplies completely shut down some where during the benchmark run every time.

Unmodded, these cards still aren't operating correctly and I second the notion that the issues are related to the power circuitry

I'm glad I hung onto my 7800 while my 7900 is getting sorted out
 
Well I called XFX last week out of curiousity and to see if they have any details on the problem. Over the past week they have been working with me to make sure my card is working fine (havent had any artifacts so far but I do get occasional pauses in games). They told me to overclock the memory by 10Mhz and see if it artifacts and it does. They said that I should try running the card at 450mhz and 660mhz (from 825mhz) for the memory over the weekend and see if I get any artifacts or pausing in games.

I asked them if it was a wide spread problem and they said no. He went on to say that they are selling thousands of these cards and the problems that people are having are enough for them to raise a red flag. He also said they are not sure if its an architectural problem with the cards or the overclocks they are doing on them or if its something to do with the motherboards people are using. I find the last one hard to believe as not many people are running an Abit AT8 with a 7900GT, he did say that most people are using the Abit AN8 SLI board though. He said that if over the weekend I get any pauses while running my games I am to use the "Detect optimal frequencies" in the coolbits options and see what it reports.

Sounds like they are just playing it safe right now and having people temporarily run the cards at nvidia specifications while they try and figure out what is wrong rather than have the customers playing the RMA game over and over.

Sounds reasonable to me.
 
Back
Top