PDA

View Full Version : BFG 6800GT OC Overclocking issues.


mongoose
12-04-04, 03:50 PM
With stock cooling, I could hit 435/1150 using Detect Auto Frequencies (using Nokos modded BIOS.) I swapped out the stock cooler for a Koolance Exos using the GPU 180 watreblock and ramsinks.

My core temp dropped 13C according to coolbits but now auto detect gives me only 395/1050. Turning it up to 400/1100 and playing UT2004 for a bit cause the screen to display several colored "sine waves" then a black screen and sound loop, and I lose signal to the monitor. Reboot to stock clocks and the card is fine.

I thought the block wasn't making good contact with the core but it is, and there are no leaks. The card is relatively cool to the touch.

I tried running on a clean line with no other devices attached and I could hit 415/1090 on auto detect, but still crashed in UT2004 and Doom3 even at 400/1050.

With the stock BIOS I now detect higher with watercooling but get immediate freezes in 3DMark @400/1100 like I always have.

I'm suspecting my power supply can't handle the load when I overclock the GPU since I installed the Koolance as it draws a bit off the 12v rail. I tested the 12v rails with a multimeter and they seemed solid 11.94 but I suspect I don't have enough current. Note: One time after a crash in UT2004, the scrolling bar on the Windows XP logo upon bootup slowed to a crawl and took forever to boot.

I'd appreciate some input on this.

Here are my system and PSU specs.

Antec Truepower 550 rated 24A peak on 12V, fans maxed out@3000RPM. 530 Watts max at 25C ambient.
DFI Lanparty UT NF3 250GB
AMD 64 Clawhammer 3200+ OC'D to 3700+ (10X240FSB)
Lite-On CDRW-DVD combo drive.
Memorex 8X DVD burner.
2X 74GB Western Digital Raptors in Raid 1
BFG 256MB 6800GT OC
1 512MB Mushkin PC3200 Level II Black
Two 92MM Case fans, 3000RPM
Thermaltake Hardcano.
Koolance Exos EXT-A04 Watercooling, fans running at 100%

Thanks in advance.

NemesisChild
12-04-04, 04:21 PM
Well, your PSU shouldn't be the issue, BFG recommends a min. of 300w. Do you have the BFG reference card or the newer one with twin cooling fans? I suggest running Riva Tuner hardware monitoring in the background while playing a graphic intense game to determine what your GPU and Ambient temps are. I had heat related issues OC'ing at Ultra speed (400/1100). I upgraded up my cooling fans in my rig and this solved my heat related, lock-up problems. Btw, don't use auto detect to determine your OC frequencies. You seem to be having overheating related issues, IMO.

Good luck......

noko
12-04-04, 04:53 PM
Try just OCing your core and not the ram and see what is your limit. Then just the ram with the core at stock. Try to isolate which one if any is the limiting factor.

Now you did use a thermal compound right? Possible air pocket on core somewhere. Some thermal compound founds its way to the chip base? Reseat your card, check the card terminals for any junk on them, try a different power supply connector (Just some suggestions). You should be able to OC higher not lower. . . Ponder ponder . . .

aldamon
12-04-04, 05:01 PM
Do you have air running over the card? I had the same problem with my Swiftech MCW50. The stock cooler blows air across a number of components and some cards can't handle the lack of air. I had to reinstall the original RAM cooler plate and then point an 80mm fan at the card. Otherwise the card wigged out even at stock speeds. To accomplish this, I put some velcro on top of my Mag 3 pump and pointed it directly at the card.

noko
12-04-04, 05:05 PM
ahhh, that makes sense. Hopefully he figures it out and its something simple.

mongoose
12-05-04, 01:34 AM
I'm on water so the stock cooler and fans have been removed. And yes, I used thermal compund. I have the Newer BFG with the twin fans, the heatsink with the 1/8" thick sponges to make contact with the ram which likely acted as an insulator. The GPU 180 block by Koolance is stock on the Ultra OC Watercooled so I'm certain it's good enough for a GT.

I even faced my computer toward the window with a fan blowing in cool Iowa air over the card (40 degrees) and it made no difference.

Nvidia recommends 350W for the GT series and I have an OC model that I'm trying to hit Ultra Speeds on. I have a PSU rated at 550 max (at 25C ambient in the PSU. Simply put, I'll never see those temps inside the PSU.) However I have a heavily overclocked CPU and am running a watercooler, both on the 12v rail. Yes, the BFG alone wouldn't max the PSU, but I'm curious about the BFG and everything else together.

I'll try Noko's suggestion and monitor with Rivatuner then report back.

bkswaney
12-05-04, 01:53 AM
I've had 3 GT's now.
2 would not run 400/1100 without locks or artifact.
The one I have now will run 420/1150.
But all GT's will run 400/1100.
Even with a huge PSU and over volting some just
will not run at ultra speeds.

My guess is you have one that will do 390/1100 Max.
Be happy with it. Even at those speeds u only loose
a very few FPS.

mongoose
12-05-04, 01:57 AM
My overclock is Prime95 stable for 11 hours (when I self terminated the test.) But it may be a UT issue so I'll keep that thought in mind.

I don't call anything a successful overclock without extensive, stressful testing.

bkswaney:

You're running a Koolance as well. Are you watercooling the GPU?

Ran UT2004 at 400/1100. Core never rose above 53C (about 13C cooler than the stock cooler) but the ambient was 53C.

I do appreciate everybody telling me to look for the simple first. I'm not quite ready to shell out 200 for a PCP&C yet.

mongoose
12-05-04, 02:19 AM
This time I ran UT2004@400/1100 until it crashed. Unfortunately Rivatuner didn't save the log. This time it simply went to a black screen, lost the monitor signal. I heard a two sound through the speakers like scratches then three rapid beeps and I had to power cycle.

I'll try Noko's suggestion now.

mongoose
12-05-04, 06:24 AM
Tried 370/1100 and 400/1000, both resulted in crashes like before. GPU core temp never went above 53, but the ambient temp approached 56.

rflair
12-05-04, 09:29 AM
I was just wondering if it could be a thorttling issuse. Does anyone know where the heat sensors on these 6800's are? Is the GPU sensor under the GPU or built in and where is the ambient temperature sensor?


rflair

p.s. ButterBandit 6 hours of prime95 is nothing, if your cpu can't do 48 hours of Prime95 its not stable.
________
Honda RC114 (http://www.cyclechaos.com/wiki/Honda_RC114)

mongoose
12-05-04, 02:15 PM
UT2004 won't find the CPU instabilities Prime95 does.

However, running UT2004 with my CPU at stock speeds still results in a crash when I overclock the 6800GT.

I can run UT2004 for hours without a crash with the CPU overclocked.

We can say with certainty the issue is isolated to the card when it's overclocked.

mongoose
12-05-04, 10:40 PM
"That being said, Prime95 won't find the same instabilities that UT2004 may find. No single benchmark/stress test is the find all/end all."

I can agree with that. I've found UT2004 to be a good test for system RAM. Some assault maps need a 512MB video card and if you don't have it, it will load textures into whatever available system ram. I've troubleshooted many a ram issue with it and discovered problems memtest86 didn't.

I also loop 3dmark.

I prefer Prime 95 for CPU and sometimes RAM issues. But 48 hours is overkill, IMHO. Eight hours has been good to me.

Back to the issue: I'll try reseating the card and redoing the wiring. I've pretty much eliminated the PSU since I still crashed after resetting the CPU to stock and disabling a few devices. I am considering the possibility that the card still might need some consistent airflow over it. My test last night may not have blown the air where it was needed.

mongoose
12-06-04, 03:09 PM
I flashed back to the orignal BIOS and tried running 3DMark with Rivatuner monitoring.

It seems the card is throttling. Since the symptoms are the same (5 second freezes in 3dmark) I can assume it happened with my stock cooler @ 400/1100.

What puzzles me is the BFG is much cooler with water cooling by 20C under load, however ambient temps are roughly the same. Where is the sensor for ambient temps?

I get no artifacts whatsoever when overclocked. No snow or tearing at all. Just the monitor shutdown.

I checked temps around other areas of the card and found that around the power connector, it is hot to the point it hurts a bit to touch.

mongoose
12-07-04, 12:22 AM
I've heard BFG doesn't void the warranty for installing watercooling but I'll check anyway.

In the meantime, I'll experiment with Rivatuner and see if there's a correlation between the ambient temps and the throttling.

mongoose
12-07-04, 03:53 AM
Gotta love accidents.

By accident, I found that the molex wasn't very secure in the socket, which is a very loose fit in the BFG compared to a Radeon. I tapped it ever so slightly and POP, CRASH.

I have it snug now (I bumped it a quite a few times and no crashes) and noticed my ambient temps dropped seven degrees at idle, wierd.

Haven't tried overclocking yet, but I can finally loop 3DMark at stock speeds without crashes (it had become that bad.)

Ludic
12-07-04, 09:14 PM
BFG sells a card with a waterblock on it, so they have to honor a warranty on that particular watercooling.

mongoose
12-07-04, 10:11 PM
If I recall correctly, BFG uses the same cooler I am, the GPU 180 L06 from Koolance.

Right now, the card is far more stable after I fixed the molex issue. I flashed back to the original BIOS and the card doesn't throttle as much as it did with the stock cooler and BIOS when I'm at 400/1100. Noko's BIOS replaced the throttling with crashes. Since my GPU and ambients temps are fairly low compared to air cooled cards, something else is likely factored into the algorithm for GPU throttling.