Poor 6800GS performance

[Unf]

Intel D845PESV mobo
Intel(R) Pentium(R) 4 CPU 2.80GHz
768MB RAM
previously: RADEON 9550 (250/400 OCed to: 425/552) now: Geforce 6800GS (non-OC)
40GB Maxtor HDD (60E40L0)
160GB Seagate HDD (ST3160023A)
FORTRON FSP400-60PFN 400W PSU

I got my 6800GS yesterday and proceeded to test it. I was hoping for some good results, even though I'm rather short on RAM. I've read this article in which some guy got a really nice boost out of this upgrade (and scores more 3DMarks than me with a worse setup, but I'll get to this later)

First game: Oblivion. Not the highest FPS of course, but definitely playable and it looked gorgeous with HDR enabled and stuff set near High (Oblivion set graphics to High Quality anyway via autodetect).


Second game: CS:S. FPS a bit higher. HDR was playable and without stuttering until the scene was full of models, that's when FPS dropped to 14. Still no HDR for me, then.


Third game: BF2 (that's what I mostly play and that's where I wanted to see the difference most). I played it on my old card at 1024x768, medium settings, except for low lightning, low dynamic lightning and dynamic shadows turned off, 90% draw distance.

I thought that High wouldn't work too well with 768MB RAM, so I decided to put everything on medium (I've played on these settings before but there was a slight performance hit and maps loaded longer, so I set lightning stuff to low). I was expecting that loading lag and some stutter won't disappear, but at least I would get higher FPS while there is no stuttering. Wrong. Fps were a bit higher in some places, but the same as on my previous card (or even lower!) in others. One time I was looking one way and FPS were ok, I turned around to look at 3 humvees and a few people behind me and they dramatically dropped. Didn't experience that on my old GPU.


3DMark05 score:

Radeon 9550 (let's count it as 9600Pro since it was OCed) - 1745
GeForce 6800GS - 2229 (Forceware 91.31) - 2836 (Forceware 84.12)


I think it's way too low, looking at the article I linked to and considering the fact that my friend having the exact same card an Athlon 2500+ and 1GB RAM scored over 4000. I browsed the net and people were getting over 4k points with this card.


What's wrong?

 

[Pain]

Did you update your MB drivers, and remove the old ati drivers?

BTW, I play games, not benchmarks. Do your tests in the games you like and forget the benchmark results.

 

[Unf]

Yep I did, removed drivers with the uninstall app and with driver cleaner.

It looks that there is not much of an improvement in games and yes, I don't play benchmarks but they can give a good idea of how the GPU performs. Mine doesn't perform as it should.

 

[Unf]

Remember my 2836 score? Scores on the same drivers:

One after another (PC has been working for some time, I was browsing the net etc.)
1979
2171
2210

Right after booting:
2064

What's with the randomness... :?

 

[Scarchunk]

Well with that much variation in your 3dMark score, I would say that you have some problems. A small variation is normal(+-15), but your having almost 1000 point leaps in performance(1979-2836 not normal) My first suspect would be power supply. I went from a 380W to a 485W when I bought my 6800gs. I'm assuming that you didn't have this kind of score variation when using the 9550. Next I would look into hard drive. Run the Seagate diagnostic off there website and defrag with your pagefile disabled. Having a 512MB and a 256MB stick of RAM isn't helping since you can't run dual-channel. As for BF2, it likes 2GB of RAM for higher settings, especially playing online. Wish I could help more, but your score should be much higher. I'm assuming your using the AGP 6800gs, which tends to perform slightly worse, but my pci-express 6800gs scores 5955 on 3dMark 05 with rig below. It's OC'ed but even at stock I was in the 5700's.

 

[Unf]

The Fortron PSU is a new one, previously I had a 350W one with which I used the 9550, then switched to Fortron and still used the card with it for a few days.

I though insufficient power would mean crashes or freezes, not overall bad performance and score deviation.

Is there any way to check the stability of the power supply ( a software way preferably)?

Could it be the GPU itself? Maybe I was unlucky and bought a damaged card and should exchange it?

UPDATE: 2194 points on a fresh Windows XP after formating, so it doesn't look like it's a software problem if anyone still wondered.

 

[Shinobi13]

I have my 6800GS AGP OC`d to slightly above ultra settings (unlocked with 16/5) and my games have never looked and played better. Have you tried to unlock it using riva tuner?

I have a XP 3200+ rig with 1gb of corsair value select ram and a 450w PSU and I get mid 5k scores with 3d mark 05. Quake 4 (ultra settings and playable), GRAW, high settings and playable. Keep messin with it and you`ll find ur sweet spot. GL.

 

[Scarchunk]

Well even after reinstall I think you're scoring still well below where you should be. Are you still getting the wild score flucuations? I would still run the Seagate diagnostics and memtest to eliminate possible hardware issues. Intel has a utility that you can download off their website that will allow you to monitor your temps as well as your voltages, I think its called Intel control panel or something. I would also open up the nvidia control panel and leave the temp tab open while using 3dMark 05 to get an idea of temps while at load. In between tests you can check the temp when it flashes back to desktop for a second. If none of this turns up anything suspect then, yes, I would RMA the card, since you aren't getting the performance you should expect.

 

[1Tanker]

Intel D845PESV mobo
Intel(R) Pentium(R) 4 CPU 2.80GHz
768MB RAM
previously: RADEON 9550 (250/400 OCed to: 425/552) now: Geforce 6800GS (non-OC)
40GB Maxtor HDD (60E40L0)
160GB Seagate HDD (ST3160023A)
FORTRON FSP400-60PFN 400W PSU

I got my 6800GS yesterday and proceeded to test it. I was hoping for some good results, even though I'm rather short on RAM. I've read this article in which some guy got a really nice boost out of this upgrade (and scores more 3DMarks than me with a worse setup, but I'll get to this later)

First game: Oblivion. Not the highest FPS of course, but definitely playable and it looked gorgeous with HDR enabled and stuff set near High (Oblivion set graphics to High Quality anyway via autodetect).


Second game: CS:S. FPS a bit higher. HDR was playable and without stuttering until the scene was full of models, that's when FPS dropped to 14. Still no HDR for me, then.


Third game: BF2 (that's what I mostly play and that's where I wanted to see the difference most). I played it on my old card at 1024x768, medium settings, except for low lightning, low dynamic lightning and dynamic shadows turned off, 90% draw distance.

I thought that High wouldn't work too well with 768MB RAM, so I decided to put everything on medium (I've played on these settings before but there was a slight performance hit and maps loaded longer, so I set lightning stuff to low). I was expecting that loading lag and some stutter won't disappear, but at least I would get higher FPS while there is no stuttering. Wrong. Fps were a bit higher in some places, but the same as on my previous card (or even lower!) in others. One time I was looking one way and FPS were ok, I turned around to look at 3 humvees and a few people behind me and they dramatically dropped. Didn't experience that on my old GPU.


3DMark05 score:

Radeon 9550 (let's count it as 9600Pro since it was OCed) - 1745
GeForce 6800GS - 2229 (Forceware 91.31) - 2836 (Forceware 84.12)


I think it's way too low, looking at the article I linked to and considering the fact that my friend having the exact same card an Athlon 2500+ and 1GB RAM scored over 4000. I browsed the net and people were getting over 4k points with this card.


What's wrong?

This might sound silly, but you did plug the Molex connector into the end of your card, right? I highly doubt you would get far into a game, but you would definetely be limited without the extra power. You aren't even getting 1/2 of what you should be, as stated by others.

 

[rcs2749]

Also did you check your AGP settings in the BIOS? Are you are 4x or 8x?

 

[1Tanker]

Also did you check your AGP settings in the BIOS? Are you are 4x or 8x?

4x won't make that much difference, if any.

 

[Unf]

I performed the seagate diagnostics and memtest which gave me no error. Voltages according to Intel Active monitor look fine as you can see here.

I've got both molex connectors plugged (there is a 6-pin in the card but since my PSU doesn't have such connector, I have to use an adaptor that came with the card with 2 molex connectors).

I did an experiment, though, and the results are surprising. I unplugged one molex connector. You would think the performance would go downhill but no - I scored one of my best scores - 2858. Then I plugged in ONLY the other molex - result - 3077. I thought that maybe the whole adaptor or the port in the GPU don't work. I checked how it would act with both connectors unplugged and that's when I got a notification that the card switched itself into some low performance mode due to lack of power. I checked in 3dmark and the first scene didn't go above 5 fps (I didn't finish the test, didn't want to risk any damage). Both molexes back in - slightly above 2000.

I hope that will shed some light on what the cause might be.

 

[1Tanker]

I performed the seagate diagnostics and memtest which gave me no error. Voltages according to Intel Active monitor look fine as you can see here.

I've got both molex connectors plugged (there is a 6-pin in the card but since my PSU doesn't have such connector, I have to use an adaptor that came with the card with 2 molex connectors).

I did an experiment, though, and the results are surprising. I unplugged one molex connector. You would think the performance would go downhill but no - I scored one of my best scores - 2858. Then I plugged in ONLY the other molex - result - 3077. I thought that maybe the whole adaptor or the port in the GPU don't work. I checked how it would act with both connectors unplugged and that's when I got a notification that the card switched itself into some low performance mode due to lack of power. I checked in 3dmark and the first scene didn't go above 5 fps (I didn't finish the test, didn't want to risk any damage). Both molexes back in - slightly above 2000.

I hope that will shed some light on what the cause might be.

Sounds like it's PSU related then, as messing with the molex connectors increased your score. Either a bad/failing PSU or possibly underpowered(yet i doubt it), or bad molex connector. Try another molex if you have spare(s) available. GL

 

[Scarchunk]

That makes it interesting. I had said something about psu earlier because I had heard of graphics cards behaving strangely when underpowered(not shutting off, but underperforming) so that might be the case. I would definetly try switching molex connectors as 1Tanker has suggested. I ran my 6800 on an adaptor like you're doing for about a week, but never came across any problems. Try switching out as many different molex connectors as possible to see if it affects your score. If you continue to get bad scores with different connectors, it would seem to me that the problem is with the card, as it unlikely that you would have that many bad connectors and your psu should be able to handle the card. Maybe the problem is with the power connector on the card(RMA) and not the psu. What temp does your card idle at? In any case, that is an interesting experiment running the card without molex plugged in.

 

[Unf]

I tried 2 molexes from one rail, 2 from the other rail and one from each of the 2 rails I have (though I couldn't check all mixed combinations as the cables are too short and I couldn't supply power to my HDD).

What makes me wonder is indeed this experiment. When I plugged in one molex and got no score penalty (even a bonus!) I thought that maybe the other one is not working. When I plugged in the other one and the situation repeated, I thought that maybe the whole port or adaptor is not working and it doesn't make any difference if they are plugged or not. As it turned out, it does make a difference and the PC is aware when there is no additional power. So I've no idea how it works.

I could try and see how the card works with my previous 350W PSU but I saw a bad review stating there's a high risk of damaging the PC with that PSU. Then again it worked fine for nearly 4 years, causing no problems when I added 1 HDD and 1 DVD drive to the 1 HDD and 1 DVD I already had. If I plugged in the GPU and one HDD only it should need even less power than for 2 HDD and 2 DVD I think. So maybe it won't be that risky after all...?

And BTW - i'm planning to notify the shop I bought the card from and I'm wondering whether I can admit that I've been messing with the molexes, plugging only one at a time or none at all. That's the only way to give them an idea what's wrong but maybe I'm voiding my warranty there.

 

[1Tanker]

I tried 2 molexes from one rail, 2 from the other rail and one from each of the 2 rails I have (though I couldn't check all mixed combinations as the cables are too short and I couldn't supply power to my HDD).

What makes me wonder is indeed this experiment. When I plugged in one molex and got no score penalty (even a bonus!) I thought that maybe the other one is not working. When I plugged in the other one and the situation repeated, I thought that maybe the whole port or adaptor is not working and it doesn't make any difference if they are plugged or not. As it turned out, it does make a difference and the PC is aware when there is no additional power. So I've no idea how it works.

I could try and see how the card works with my previous 350W PSU but I saw a bad review stating there's a high risk of damaging the PC with that PSU. Then again it worked fine for nearly 4 years, causing no problems when I added 1 HDD and 1 DVD drive to the 1 HDD and 1 DVD I already had. If I plugged in the GPU and one HDD only it should need even less power than for 2 HDD and 2 DVD I think. So maybe it won't be that risky after all...?

And BTW - i'm planning to notify the shop I bought the card from and I'm wondering whether I can admit that I've been messing with the molexes, plugging only one at a time or none at all. That's the only way to give them an idea what's wrong but maybe I'm voiding my warranty there.

Trying the old PSU sounds like a good idea. Unplug all unnecesary stuff-floppy, any and all optical drives, spare HD, any extra(non-crucial) fans, UV lights, etc to lighten the load on the PSU. If that doesn't work, then it's gotta be the video card. :? GL

 

[Unf]

Forgot to say: idle temp is 47 at the moment. After the 1st 3dmark05 test it was 65 but dropped to fifties quickly, before the next test loaded.

 

[buckiller]

try bumping the stuff up to high and see what it does.

 

[Scarchunk]

I could try and see how the card works with my previous 350W PSU but I saw a bad review stating there's a high risk of damaging the PC with that PSU. Then again it worked fine for nearly 4 years, causing no problems when I added 1 HDD and 1 DVD drive to the 1 HDD and 1 DVD I already had. If I plugged in the GPU and one HDD only it should need even less power than for 2 HDD and 2 DVD I think. So maybe it won't be that risky after all...?

And BTW - i'm planning to notify the shop I bought the card from and I'm wondering whether I can admit that I've been messing with the molexes, plugging only one at a time or none at all. That's the only way to give them an idea what's wrong but maybe I'm voiding my warranty there.

Well I don't think that the whole story about how you messed with the molexes is really relavent here as you had the problem with the under-performance before and continue to have it now. I would approach it as you are certain that the card is under-performing and you need to have the defective card replaced. If you seem uncertain or ask their opinion, they will probably start to suggest that you might have done something to cause it. Tell them that you have tested your other components and your voltages and that you plugged it into other molex connectors, with all the same results. They will probably dock you some sort of restocking fee or offer store credit. Anyhow, trying out another psu would be a good idea to further narrow down your problem. Diagnosing problem equipment can be a pain, especially when it's half-way performing, like in your case.

 

[Unf]

I've e-mailed the shop, didn't get a response yet.

I also plugged in my old PSU, but got the same low performance.

 

[Unf]

UPDATE: It's definitely the GPU. My friend has got the same GPU model and I brought mine to compare on one his PC. He achieved 4455 points with his card, then we switched the card to mine and got slightly above 2600.

We think something might be wrong with the part(s) responsible for processing shaders. My card was mostly struggling with the 2nd test (Firefly Forest).

The only difference between our cards that we noticed, was an additional sticker on my card. While that shouldn't make a difference, my friend suggested that maybe the parts covered by it are overheating making the card underperform. Doesn't sound likely, though I'm no expert. It's marked here. I didn't want to take it off, as I don't want to void my warranty or anything, so I'll just e-mail the shop the additional info (they have yet to respond to my previous one).

 

[Scarchunk]

RMA the card and DON"T remove the sticker. It won't help and it might seem suspicious to the shop where you bought it. Glad you found you're problem.

 

[Unf]

Thanks for the help everyone