6800le 'Black Screen' - The Complete Guide: Identification, modification and reason

Quote:

---

Originally Posted by markiemrboo

Well, tried benching 3dmark05 with 400MHz core and 900MHz memory (with the cap back on the back of the card) and it paused half way through, no black screen though! Lowered the memory to 850MHz and it completed the run.

Could be that my cap isn't good enough? My pausing before appears to be completely solved and was definatley heat related. I believe my pausing now is related to power, as the temps are good and it only pauses if I clock the memory too high..

With 1.3v to the core it's clocking at 430MHz on the core ( so far ;) )!! :D But if the memory is clocked too high i'll get pausing. Yet again confirmed, 430/900 = paused in 05. 430/870 = no pausing (running right now, nearly at the end!). I think I just need a really good quality cap on the memory? I mean, adding this cap back actually does seem to have made a difference. These kind of clocks weren't happening yesterday without that cap in place...

---

Thank you for voltage information.

Maybe the capacitor is now more optimally placed ? How long wires you have, and did you solder the leads ? These aspects DO have affect on performance and of course the capacitor quality (I wonder if it is 'low-ESR' -capacitor, not directly stated in the list you sent)

I think that flashing the memory voltage can be risky (due to possible stucking to wrong speed !!), and I would suggest to try clocking with different programs: coolbits, rivatuner...

Flashing memory voltage? No :) I adjusted memory voltage with a pot. But, I am flashing the BIOS to raise the core voltage from 1.2v to 1.3v.

I don't think it's a high quality cap :( It's just a general purpose 85C phillips thing. I am trying to get hold of a 105C low ESR one though.

If you like, I could probably remove the cap and see if any of the overclocks I had yesterday are still stable, or if they then experience pausing and black screens. Maybe that would be a good test?

I am overclocking with powerstrip ;) Coolbits is annoying, rivatuner's 'disable internal test' option doesn't work properly. I tried to tell them that when that power user option was enabled the core clock speed never changes from 325MHz, but they didn't want to listen to me :( powerstrip overclocks it just fine :)

Quote:

---

Originally Posted by SikaRippa

Good news for you :) .. thank you providing us a new capacitor model !!
Please post if you will get BS in future (it is not nice to be sooooooo sceptic BUT I do NOT rely these card too much anymore)

--

Figures would be nice, indeed :D :D :D ... having a 'proper' resolution i.e. one could see the component markings.

---

so the only camera i could get hold was usless [pics belows] so they show nothing, but i did a hand drawing instead :)

http://img19.echo.cx/img19/5418/untitled10cl.th.jpg

http://img19.echo.cx/img19/6889/untitled19br.th.jpg

me again :D

Gonna order some more caps today or tomorrow, giving in to the insanely high postage costs :(

I found some low ESR caps rated at 105C. I guess I should go for the 16v caps here? They have a higher ripple than 10v and 6.3v, a lower impedence and low "tan" :) Before I go spending my money, do these ones sound pretty good in general? I was looking at the 16v 3300uF cap on the spec sheet.

Also, at the bottom of the specs is:

Quote:

---

"* Impedance (max.) quoted at 20°C
**Ripple current (max.) quoted at 100kHz and 105°C
***Life expectancy quoted at 105°C"

---

Is it better that the ripple is measured at 100khz? Most of the others I see are rated at 120hz. Will it make much difference? 100khz is better?

Quote:

---

Originally Posted by sunn

so the only camera i could get hold was usless [pics belows] so they show nothing, but i did a hand drawing instead :)

http://img19.echo.cx/img19/5418/untitled10cl.th.jpg

http://img19.echo.cx/img19/6889/untitled19br.th.jpg

---

Awesome DRAWN figure :D !! Very nice indeed.

--

By the way, do you think your card is fixed/modified afterwards ? I mean misaligned capacitors (tiny light brown components) left side from switching regulator chip (see Fig. 10, version '1b').

Quote:

---

Originally Posted by markiemrboo

me again :D

Gonna order some more caps today or tomorrow, giving in to the insanely high postage costs :(

I found some low ESR caps rated at 105C. I guess I should go for the 16v caps here? They have a higher ripple than 10v and 6.3v, a lower impedence and low "tan" :) Before I go spending my money, do these ones sound pretty good in general? I was looking at the 16v 3300uF cap on the spec sheet.

Also, at the bottom of the specs is:



Is it better that the ripple is measured at 100khz? Most of the others I see are rated at 120hz. Will it make much difference? 100khz is better?

---

It is very hard to say what/how big difference will be if you would change the capacitor to better one. If the cost will be very high, I am sure you want to see better real performance increase...and it can not be guaranteed !

--

But, if you WANT better capacitors:

- It is BETTER that performance has been measured at 100kHz (vs 120Hz)
- 105 C is good
- 16 V caps give very nice ESR and ripple current values BUT are more bulky (maybe not a real problem, if capacitor is placed at bottom of the card)

Suggestion: 16 V, 3300 uF (from that list)

Thanks. I can't actually fit the caps on the bottom of the card anymore, as the Arctic Silencer I just fitted is massive :P My current one fits on the top just fine though, and that's a really big cap.. so size shouldn't be too much of an issue.

I am going to overclock the card tomorrow and try and keep a measurement on voltages until it freezes. I'm hoping it's due to fluctuation or something, but it could still be heat I guess. Not entirely sure where that black screen came from the other day, but it hasn't been back since I put the cap back on...

Quote:

---

Originally Posted by SikaRippa

Awesome DRAWN figure :D !! Very nice indeed.

--

By the way, do you think your card is fixed/modified afterwards ? I mean misaligned capacitors (tiny light brown components) left side from switching regulator chip (see Fig. 10, version '1b').

---

Thank You

I’m not too sure what you mean :confused:

But I did look closely at the top [back, whatever people call it] of the card, comparing it to version 1band, and there nothing like that, everything is on the straight and narrow, so I believe this card hasn’t been modified in the past.

if u have a load of money to spare, i think my old Gainward LE is being sold in the B-Grade on Overclockers

Ah ha! I think I might be able to reproduce a black screen on my card now!!!

I was playing with some overclocking again, literally 10 seconds ago. I did something different this time though, to try and eliminate the pausing issues I was getting with higher clocks. I enabled the power user setting in RivaTuner to allow adjusting the clocks of 'low power 3d' mode. I went ahead and set both 'performance 3d' and 'low power 3d' clocks the same, that being 395MHz on the core and 890MHz on the memory. It ran through AM3 fine, then I fired up 3dmark03. 10 seconds in to the "proxycon" test (number two), it paused.... and just sat there for a couple of minute, and finally black screen and monitor went in to standby.

So, it must be trying to switch to low power 3d mode when my pausing occurs, which defaults the same as 2d clocks apparently. I wonder what exactly is telling the drivers it has to switch to low power 3d mode?

Anyway, if you need me to measure anything while it's in the black screen state, or any other tests for that matter, I will gladly help.

I'm battling with Fast Writes too. After I try a benchmark with fast writes again (it will pause I am sure, but hey..) then I will disable fastwrites and just make sure I can reproduce that black screen by setting low power and performance clocks the same! I'm pretty sure it will though...

I think I might have to give in and just admit that I hate nVidia. I got another black screen just now, but low power 3d and performance 3d clocks were different. The difference was, fast writes was enabled that time!

A few curious things that may be worth noting:

When fast writes does it's pausing trick, the clock speed very very briefly clocks down to low power 3d mode, apparently. If I set low power 3d and performance 3d to the same clocks when the fast writes pausing happens, rivatuner shows no bump down to a lower clock speed. When the low power 3d mode is left default, it clocks back down.

When the black screen occurs, my computer isn't totally dead. It still responds to pings! Also, I am using 'ExperTool' from Gainward to adjust my GPU fan speed. It's set to 50% in 2D mode and 100% in 3D mode. When the black screen happened, I could clearly hear the fan slow down and go in to 2D mode!!! This could mean the black screen issue is actually some sort of driver bug, especially considering I could still ping my computer.

Strange eh? Why fast writes causes pausing is beyond me. Works fine with my nForce2 board, not with my nForce3 board though. Why it's trying to go in to low power 3d mode also has me puzzled, but the fact some people have success with adding a memory capacitor might suggest it's just bad caps. No idea why it isn't doing my card any good in that case :P Perhaps the cap I added is no better than the caps already on there to start with...

Quote:

---

Originally Posted by sunn

Thank You

I’m not too sure what you mean :confused:

But I did look closely at the top [back, whatever people call it] of the card, comparing it to version 1band, and there nothing like that, everything is on the straight and narrow, so I believe this card hasn’t been modified in the past.

if u have a load of money to spare, i think my old Gainward LE is being sold in the B-Grade on Overclockers

---

Sorry, about too bad explanation :( but I think you got the idea ! If the small capacitors are aligned properly the card is propably not fixed afterwards OR they have fixed it more accurately..

--

It might have a good idea to buy a BS-affected card for heavy-duty-testing-purposes :D . But I wonder that £145.00 (that one is yours ?) is too high cost for that purpose: removing capacitors etc. and measuring them i.e. most likely destroying the card.

Quote:

---

Originally Posted by markiemrboo

I think I might have to give in and just admit that I hate nVidia. I got another black screen just now, but low power 3d and performance 3d clocks were different. The difference was, fast writes was enabled that time!

A few curious things that may be worth noting:

When fast writes does it's pausing trick, the clock speed very very briefly clocks down to low power 3d mode, apparently. If I set low power 3d and performance 3d to the same clocks when the fast writes pausing happens, rivatuner shows no bump down to a lower clock speed. When the low power 3d mode is left default, it clocks back down.

When the black screen occurs, my computer isn't totally dead. It still responds to pings! Also, I am using 'ExperTool' from Gainward to adjust my GPU fan speed. It's set to 50% in 2D mode and 100% in 3D mode. When the black screen happened, I could clearly hear the fan slow down and go in to 2D mode!!! This could mean the black screen issue is actually some sort of driver bug, especially considering I could still ping my computer.

Strange eh? Why fast writes causes pausing is beyond me. Works fine with my nForce2 board, not with my nForce3 board though. Why it's trying to go in to low power 3d mode also has me puzzled, but the fact some people have success with adding a memory capacitor might suggest it's just bad caps. No idea why it isn't doing my card any good in that case :P Perhaps the cap I added is no better than the caps already on there to start with...

---

You have made quite nice survey :D Thank you for offering help to test, indeed I have something in my mind and I will ask later later :) Do you have a capacitance measuring capability on your multimeter ?

--

I wrote soooo long story that it take time to think..

I have tested little bit with 'FastWrite' and it does not solve the problem - but it propably have some impact to BS-feature.

I think that it is possible to ping (never tested) the computer after BS. As I know, BS kills only the graphic card and at least for me the CPU is still doing something. That could also explain the fan behaviour (?)

Settings 'performance 3d' and 'low power 3d' I have never tested, but the explanation you gave make sense: the driver forces the card 'togging' between those states during e.g. 3dmark2003. And toggling between different states can cause anything..

--

I was wondering about freezes you mentioned and found following thread:

http://www.nvnews.net/vbulletin/show...9&page=6&pp=15

.. there was told that main memery setting Dual Channel Mode vs. Single Channel Mode has been reported to cause freezes with 6800-cards. Do you have 'Dual Channel Mode'-memories ?

No worries. i defintly think it hasnt been fixed or they did a 'proper job'

My old card seems very expensive, especially if it would be your test rat. but if you wait a week or 2 and its still there overclockers normally reduce the price by quite abit. [example they was a MSI 6800 NU in the B-Grage for £165ish and it went down to £109]

yeah, that looks like the card. but im pretty sure its a 6800le. my system identifies it as a 6800le, but maybe thats just the drivers?

I dont know. its the NX6800-TD128 from MSI, and i dont see LE attached to it in any review or tech spec online, its just that the controller software, whatever u call it (the program that allows you to set your screen and graphics settings) calls it a 6800le.

Anyway, after sending that old card back and sending the company your great guide, asking them to pleeeeeease ensure that the replacement they send doesnt have one of the suspect capacitors, they appear to have not taken ANY notice of me and sent me back a card with TWO 136/143 "4 6 470 6E" capacitors.

im not happy, as this will be the 4th LE card ive had to send back if it doesnt work....

as for the gainward 2100 duel fan, i dont know. it didnt BS, it was just so unbearably noisy i sent it back after one day.

it was driving me insane.

Quote:

---

Originally Posted by SikaRippa

You have made quite nice survey :D Thank you for offering help to test, indeed I have something in my mind and I will ask later later :) Do you have a capacitance measuring capability on your multimeter ?

--

I wrote soooo long story that it take time to think..

I have tested little bit with 'FastWrite' and it does not solve the problem - but it propably have some impact to BS-feature.

I think that it is possible to ping (never tested) the computer after BS. As I know, BS kills only the graphic card and at least for me the CPU is still doing something. That could also explain the fan behaviour (?)

Settings 'performance 3d' and 'low power 3d' I have never tested, but the explanation you gave make sense: the driver forces the card 'togging' between those states during e.g. 3dmark2003. And toggling between different states can cause anything..

--

I was wondering about freezes you mentioned and found following thread:

http://www.nvnews.net/vbulletin/show...9&page=6&pp=15

.. there was told that main memery setting Dual Channel Mode vs. Single Channel Mode has been reported to cause freezes with 6800-cards. Do you have 'Dual Channel Mode'-memories ?

---

I only have capacitance measurements to something small, 22uF I think. Probably no good for what you were wondering?

I think that's what's going on. 835MHz + added cap started pausing again :( I lowered my memory clocks to 820MHz and I haven't had a pause or black screen yet. Why high memory clocks (for me atleast) throttles the card and then leads to a black screen is beyond me. There's no way to stop it either by the looks of it. Looks like I may be stuck with rather crappy memory overclocks, as the added cap doesn't seem to do anything for me :( I've yet to try a higher quality cap though..

I just had an idea. I am going to install VNC (remote desktop program) and try and force a black screen. If my laptop can still connect to my computer and see the desktop, well, I dunno what I'll do then. It'll be interesting to find out if it works though :woot: :woot: :woot:

I have a Winchester and 2x512MB sticks of RAM. So, yep, Dual Channel here! Don't think my black screen/pausing problem is related to that. I can see why running Single Channel may fix the nForce3 + FastWrites pausing/stutter issue though. I'll give that one a try next time I restart :)

Quote:

---

Originally Posted by sunn

No worries. i defintly think it hasnt been fixed or they did a 'proper job'

My old card seems very expensive, especially if it would be your test rat. but if you wait a week or 2 and its still there overclockers normally reduce the price by quite abit. [example they was a MSI 6800 NU in the B-Grage for £165ish and it went down to £109]

---

OK.. so we have to belive it is BRAND new !

--

I might buy a test rat in future, but this price is too high for that purpose.

Ok I just been playing DoD for like.. 3 hours straight with no pause :D Just by lowering my memory clock a bit. I get black screens if my memory clock is higher, like 900MHz rather than ~820MHz.

I take it most people here are seeing the same thing? Black screens go away when the memory clock is lowered? But all you people get it fixed by adding a cap? This is really annoying, as the memories on my card are even rated to 900MHz! There's not even any artifacts when running at 900MHz!!! The nvidia driver just seems to want to throttle the card and I have no idea why :(

Quote:

---

Originally Posted by dandaman

yeah, that looks like the card. but im pretty sure its a 6800le. my system identifies it as a 6800le, but maybe thats just the drivers?

I dont know. its the NX6800-TD128 from MSI, and i dont see LE attached to it in any review or tech spec online, its just that the controller software, whatever u call it (the program that allows you to set your screen and graphics settings) calls it a 6800le.

Anyway, after sending that old card back and sending the company your great guide, asking them to pleeeeeease ensure that the replacement they send doesnt have one of the suspect capacitors, they appear to have not taken ANY notice of me and sent me back a card with TWO 136/143 "4 6 470 6E" capacitors.

im not happy, as this will be the 4th LE card ive had to send back if it doesnt work....

as for the gainward 2100 duel fan, i dont know. it didnt BS, it was just so unbearably noisy i sent it back after one day.

it was driving me insane.

---

If I understood correctly you had exactly same card (in respect of those capacitors) before than now acquired from RMA ? How long it took when the first BS appeared with the your first card (days-months) ?

Have you checked with RivaTuner, how many active pipelines you have (6800 should have 12x1 and 6800le 8x1 ) ?

---

I should emphasize (again) that the logic 'certain label for capacitors c136' means BS-affected card is a THEORY. It has been seen that for some models e.g GW version 1 and 2 this theory has been worked quite well BUT e.g GW version 1b has been modified little bit AND this might have some effect (lower BS-propability or not, we will see..)

Quote:

---

Originally Posted by markiemrboo

Ok I just been playing DoD for like.. 3 hours straight with no pause :D Just by lowering my memory clock a bit. I get black screens if my memory clock is higher, like 900MHz rather than ~820MHz.

I take it most people here are seeing the same thing? Black screens go away when the memory clock is lowered? But all you people get it fixed by adding a cap? This is really annoying, as the memories on my card are even rated to 900MHz! There's not even any artifacts when running at 900MHz!!! The nvidia driver just seems to want to throttle the card and I have no idea why :(

---

Some quick comments:

- Measurements: even moderate low capacitance values can be OK (I will ask later what to measure)
- VNC is nice idea to test BS over net :D (it is pity we can not generate BS under Linux, since every UNIX-machine is SOOOO easy to debug over net)
- Did you test 'Dual/singel Channel' ? The point was that some people has reported it will help when card freezes, and you had also that kind of problems (?)
- Lower memory frequency helps allways.. have you EVER got problems at normal/stock speed ?
- I think that getting 900 MHz is a sum of many factors and the driver should not be the bottleneck. E.g cooling of the memories should be optimal when running at 900 Mhz, do have you original cooler or NV5 etc (maybe you said that before, but I can not remember everything)

Quote:

---

Originally Posted by SikaRippa

Some quick comments:

- Measurements: even moderate low capacitance values can be OK (I will ask later what to measure)
- VNC is nice idea to test BS over net :D (it is pity we can not generate BS under Linux, since every UNIX-machine is SOOOO easy to debug over net)
- Did you test 'Dual/singel Channel' ? The point was that some people has reported it will help when card freezes, and you had also that kind of problems (?)
- Lower memory frequency helps allways.. have you EVER got problems at normal/stock speed ?
- I think that getting 900 MHz is a sum of many factors and the driver should not be the bottleneck. E.g cooling of the memories should be optimal when running at 900 Mhz, do have you original cooler or NV5 etc (maybe you said that before, but I can not remember everything)

---

Memories are being cooled by the NV5, but the stock cooler (leadtek card) also cooled the memory aswell anyway :) Cooling shouldn't matter THAT much at 900MHz anyways, as the chips rated for 900MHz at 2.8v :P If they're anything like CPUs they should be able to run 'stock speeds' stable up till ~70C (althought not recommended)?

I don't think I got problems at stock speeds before, it's not quite as bad as some cards out there :) It does happen (alot) when I push the card though. The nvidia optimal test thing detects the memory overclock at ~830MHz, which it can't actually run without pausing.

I didn't test the dual/single channel yet. If that was supposed to help with 'fast writes' I will give it a try. But my pausing at the minute (fast writes off) is memory related :P Hmm.. I wonder if my fast writes on pausing/stuttering problem would go away if I lowered my memory overclock even more, maybe even stock speeds. I'll give that one a try too.

I've never measured a cap before :) Measure values with power off?

Actually, I did read that people get a lot of crashes and black screens with the Linux nvidia driver, but that's probably Linux driver bugs I suppose.

Guess I should get revising for my exams (in a couple of days) soon too :slobber:

the first card i had was an OEM Nvidia 6800le i got with the system. i dont know what capacitors it had. after working just fine for about 3 months it suddenly shorted my system. i took it to be looked at and the technician said it was the card at fault.

so I sent the card back. they sent me the Gainward dual fan as a replacement. I couldnt stand it.

So then they sent me the MSI card. It BS after less than 20 mins on doom and not much more on Far Cry. Plays Pro Evolution Soccer 4 just fine though.

And now..... the second MSi 6800 card they sent me, which i installed last night, BS after just 10 mins on Doom 3 and IMMEDIATELY (in seconds) on Far Cry.

I looked at the board and the four capacitors i took note of on the previous MSI board are EXACTLY the same as the four capacitors on the second replacement they sent me!!!

C 136 = 4 6 470 6E
C143 = 4 6 470 6E
C140 = 44 100 16A
C141 = 470 16 A 3EO



Could it be some other fault in my system do you think?

Im s fed up im thinking of just going for an X800.

Could you try something for me? Lower your memory overclock. Say, half of stock speeds to start with. Does it still black screen?

Obviously it shouldn't be black screening at stock speeds and you should probably either send it back or try the cap mod, but I want to know if lowering the memory clock solves the problem for you aswell :)

"Could you try something for me? Lower your memory overclock. Say, half of stock speeds to start with. Does it still black screen?

Obviously it shouldn't be black screening at stock speeds and you should probably either send it back or try the cap mod, but I want to know if lowering the memory clock solves the problem for you aswell"

U talking to me? Sorry, I dont know anything about changing clock speeds.

i justy ant to be able to play Doom 3!

why should I use a 1000uf capacitor?
Default the card has 470uf, don't you think it would be enough to solder a high Quality Sanyo Capititor there in place?

I get my card next week, whenn I have the BS, I will resolder ALL the big capicitors und replace them by Sanyo Capacitors.

12 Capacitors are on one card, right? (I don't think SMD Caps will make any problems)

5 very big caps (violet)
2 big caps (green)
5 samller ones (black)

Someone has an idea at which value caps DEcrease performance?
Is it only a problem at power on a system?
http://www.clubic.com/photo/00107274.jpg

Quote:

---

Originally Posted by dandaman

the first card i had was an OEM Nvidia 6800le i got with the system. i dont know what capacitors it had. after working just fine for about 3 months it suddenly shorted my system. i took it to be looked at and the technician said it was the card at fault.

so I sent the card back. they sent me the Gainward dual fan as a replacement. I couldnt stand it.

So then they sent me the MSI card. It BS after less than 20 mins on doom and not much more on Far Cry. Plays Pro Evolution Soccer 4 just fine though.

And now..... the second MSi 6800 card they sent me, which i installed last night, BS after just 10 mins on Doom 3 and IMMEDIATELY (in seconds) on Far Cry.

I looked at the board and the four capacitors i took note of on the previous MSI board are EXACTLY the same as the four capacitors on the second replacement they sent me!!!

C 136 = 4 6 470 6E
C143 = 4 6 470 6E
C140 = 44 100 16A
C141 = 470 16 A 3EO



Could it be some other fault in my system do you think?

Im s fed up im thinking of just going for an X800.

---

Thanks for complete explanation ;) .. you have been very unlucky with 6800le-cards :(

--

I think that nothing is bad with your system - except those cards, since all these models have been reported to some cases have been BS-affected (except that Gainward dual fan). Let's hope your new card will not make BS in future :)