NF3 + 6800 + 7x.xx drivers = game freezing issue

[Nazo]

Hmm, I saw a refurbished X800 XT 256MB on Newegg for $218. A bit over. Still, if you're curious: http://www.newegg.com/Product/Product.asp?Item=N82E16814102492R

Anyway, I'm inclined to say go with the 6800GS. Frankly, ATI doesn't care much about mid-range or low end (especially not mid-range) and you want mid-range. Talking current gen that is. I'm just guessing that the 6800GS could have as much risk of this occuring as a 6600GT based on the fact that we're talking a rather different card here. The impression I have is that it has more to do with performance though. The 6600GT is a bit weak, but, it's kind of close to a 6800, especially with a good overclock, so that might explain why it gets a lower risk and occasional occurance of the issue versus a high risk on the 6800 series. Still, the 6800GS is different enough that I think you have a fighting chance. Also, if you make sure to get it from a good source, you should be able to send it back if it's not satisfactory (and I, for one, would consider buggy drivers causing jerking unpleasant playing to be unsatisfactory.) This isn't like where you knowingly void a warantee like by upgrading cooling, if a product doesn't do what it's supposed to do, you can legally AND morally send it back. Anyway, from what you said, I felt like you maybe thought the 6600GT has a high risk of the bug, but, it actually has a much lower risk than the 6800 series and only a few people have complained here and there about 6600GTs.

In this pricerange, nVidia now currently offers better choices IMO, not to mention cooler and quieter with the stock cooler (don't know what ATI was thinking with the cooling on the X800s and X850s, but, it sure gets loud and lets the card run hotter than what I at least am comfortable with since I want my card to last years. Not to mention that the temperature doesn't just shoot up in a hurry, but, it goes down in a hurry and rapid temperature changes costs lifetime like heat does.) The 6800GS almost guarantees an overclock so that it becomes at least equal to a good 6800 Ultra (the lowest 6800GS on Mentok's list has a 5.2K 05 score while the highest 6800GS has 6.7, versus the lowest 6800 Ultra getting 4 and the highest getting 6.9.)

 

[Ka0s]

Hmm i do see less problems with the 6600GT overall. But I am just wondering for sure if its going to be a guarenteed problem if i get the 6800GS on my gigabyte K8NS nforce3.

I sent my X1600 back and lost around 30$ because of restocking fee ;/. So i still lost some money on just sending a card back . Would you say getting the 6800GS is worth risking all the problems...or is it safer to go with a 6600GT for now with this situation arising since AGP cards are slowly dying it seems?

Thanks for your input thus far. I just want to make a smarter shopping decision come next week when the refund money is back in my account

Also i have no real intention of overclocking the GS unless i have 0 problems whatsoever..why risk it?

The previous card of the X1600 i had was rated in the high end section on the oblivion forums yet the 6600GT is ranked in the midrange..although from benchmarks ive seen it outperforms the X1600? so i wonder why its ranked a midrange card?

 

[Nazo]

Heck no. Spend the extra $18 to get the X800 XT refurb before you even think about a 6600GT. X800 was made to be high end when they made it, 6600GT was made to be midrange when they made it (last gen, maybe two generations back if you consider all the new types of 7800s and such.) The problem with the 6600GT is that on paper it sounds great with it's gDDR3 running at such high speeds and a relatively fast core, but, what they kind of try to downplay in their marketing hype is that the GPU core is rather crippled versus a 6800, and the bus is half of what it really should be basically defeating the advantage of the memory running so fast. In other words, in a few raw performance tests, the 6600GT may actually do surprisingly well (especially if you test OpenGL instead of Direct3D) but, when you throw a real game at it where it's going to have to do more than execute a predefined path you find it's limitations. In truth, a 6800 nu with it's full memory bus actually comes out ahead of the 6600GT typically, despite the memory being as slow as I think even 700MHz stock in a few rare cases. Heck, a 6800le, which is even more stripped down might have a fighting chance against the 6600GT from what I've heard, though there aren't any comparisons that I've seen since 6800les are almost impossible to find. If you're thinking why, the answer is simple: 6600GT is easy and cheap to make, 6800 is not (except maybe 6800GS.)

Even though it is a small risk, I do feel that you have a decent fighting chance with the 6800GS if you're willing to give it a shot. In the end, it's completely up to you, and you should get great performance out of the X800 XT if you choose it as well, I just feel that it should be slightly better overall thanks to running cooler and quieter and supporting Shader 3.0. Otherwise a 6800GS is actually roughly equal to a X800 XT at least judging by benchmarks (which, admitedly, aren't perfect for judging and in fact implies the 6800GS would do even better in OpenGL games than a X800 XT. Oblivion will surely have to be Direct3D though, so it doesn't matter if you're like me and looking towards that more than anything else.)

As for overclocking, well, why not? As long as you keep it within reason (eg no artifacts and acceptable temperatures) it really doesn't matter. Still, whether you're overclocking or not, temperature matters. A stock clocked video card running too hot can still produce artifacts.

 

[Ka0s]

ill probably give the GS a try although im quite hesitant.....just based on the fact that ive already had to send a card back for this game already...and i dont feel like spending 219$ on a XT just because i see this as the last AGP card id probably be getting

 

[Ka0s]

what about this?

SAPPHIRE Radeon X800GTO

165$ better than the x1600..a little bit below the GS but i wonder if it has any problems

 

[Ka0s]

bleh after reading about the X800 GTO it has alot of problems too with nforce3 boards..so might as wel just go with the GS ;/

 

[d2thez28]

It has been ages and I don't have this combo or problem any more, but I may have stumbled across some info and solution. Looking through the release notes of the 91.45 drivers I found this:

AGP and PCI-E Programs May Hang With AMD K7 and
K8 Processors
• Issue
Microsoft® Windows® 2000 and Windows XP systems using AMD K7 and
K8 processors can hang when an AGP or PCI-E program is used.
• Root Cause
There is a known problem with Microsoft® Windows® 2000 and Windows
XP systems using AMD K7 and K8 CPUs that results in the Microsoft
operating system allocating overlapping 4M cached pages with 4k writecombined
pages. This condition results in undefined behavior and data
corruption, and is explicitly disallowed by the AMD CPU manual.
This problem can affect any device driver in the system that allocates writecombined
system memory, but is usually most easily reproduced with
graphics drivers since graphics drivers generally make heavy use of writecombined
system memory for performance reasons.
• Resolution
Microsoft has a knowledge base article on the issue, the text of which is
unfortunately quite outdated. While the article only mentions Windows
2000, AGP, and K7, both the root cause and resolution also apply to
Windows 2000 or WindowsXP, AGP or PCI-E, and AMD K7 or K8. The
article can be found at http://support.microsoft.com/?id=270715.
The issue is resolved by applying an operating system registry key as
described in the referenced article that instructs the Microsoft operating
system to not use the 4M pages, thus avoiding the conflict.
The registry key is automatically applied by installation of the latest NVIDIA
nForce platform driver package (including 4.57 SMBUS or later). It is
imperative for the package to be installed or for the registry key to be applied
before the NVIDIA graphics driver or any other device drivers are installed.
The registry key takes effect only after an operating system reboot.

I think it is worth a try to all of you still experiencing this combo and problem!!

 

[oldpablo]

It has been ages and I don't have this combo or problem any more, but I may have stumbled across some info and solution. Looking through the release notes of the 91.45 drivers I found this:

AGP and PCI-E Programs May Hang With AMD K7 and
K8 Processors
• Issue
Microsoft® Windows® 2000 and Windows XP systems using AMD K7 and
K8 processors can hang when an AGP or PCI-E program is used.
• Root Cause
There is a known problem with Microsoft® Windows® 2000 and Windows
XP systems using AMD K7 and K8 CPUs that results in the Microsoft
operating system allocating overlapping 4M cached pages with 4k writecombined
pages. This condition results in undefined behavior and data
corruption, and is explicitly disallowed by the AMD CPU manual.
This problem can affect any device driver in the system that allocates writecombined
system memory, but is usually most easily reproduced with
graphics drivers since graphics drivers generally make heavy use of writecombined
system memory for performance reasons.
• Resolution
Microsoft has a knowledge base article on the issue, the text of which is
unfortunately quite outdated. While the article only mentions Windows
2000, AGP, and K7, both the root cause and resolution also apply to
Windows 2000 or WindowsXP, AGP or PCI-E, and AMD K7 or K8. The
article can be found at http://support.microsoft.com/?id=270715.
The issue is resolved by applying an operating system registry key as
described in the referenced article that instructs the Microsoft operating
system to not use the 4M pages, thus avoiding the conflict.
The registry key is automatically applied by installation of the latest NVIDIA
nForce platform driver package (including 4.57 SMBUS or later). It is
imperative for the package to be installed or for the registry key to be applied
before the NVIDIA graphics driver or any other device drivers are installed.
The registry key takes effect only after an operating system reboot.

I think it is worth a try to all of you still experiencing this combo and problem!!

Nice find, it would have been nice to have this in the earlier driver release notes.

 

[d2thez28]

After having battled this issue for so long myself, I am extremely interested in seeing if this actually fixes the problem. Any testers?????

 

[morry71]

After having battled this issue for so long myself, I am extremely interested in seeing if this actually fixes the problem. Any testers?????

hey guys i gave it a go with the ms reg and i am using nf2 mb, 6800GS 128, forceware 91.31
and was having the stuttering due to dcp and certain temp. but since i tried your suggestion i can not cause the problem to re-occur. it has only been 1 day so i'll keep testing and send back the results.
a big premature thankyou for your efforts
cheers
maurice

 

[morry71]

sorry people
when i started up again this morning all went well for about 20 mins than it started stuttering again in 2D. i heard a clunk/thunk sound as if it was changing voltages on a relay switch but i had rivatuner running and there was nothing noticable happening
bugga
bugga
bugga

 

[morry71]

i have now loaded 91.33 beta and have not had the problem .......yet for 6hrs tryin to make it happen.. bloody crazy ha.. can't believe i won't give up on this and just send it back
once upon a time i would use nvidia and asus but i hate being burnt 3 times

 

[morry71]

i have got it back again after 9hrs .. wtf i'm going to look for the receipt
not happy jan

 

[d2thez28]

To be honest Morry, it sounds like you may have more issues that the aformentioned bug I posted about. The 6800 series has had it share of stuttering issues related to several causes. Consider these:

-is that a native AGP card or a PCI-E bridge to AGP card?
-the post I made previously said it was imperative to install the registry fix/SMBUS driver 4.57 and then reboot prior to installing any other driver
-Is the fan working correctly on the card?
-What are your power supply specs?
-Are you overclocking at all?
-Are the card's 2D and 3D clocks the same?

 

[morry71]

Thez i assume i have an agp card
i followed the install procedure
the fan is fine
the psu is a Antec smartblue 350, 16A on the 12v and for the first couple of months it worked fine. then stuttering appeared and the occational hang
no overclocking and the card's 2d/3d clocks are at the default which means they are not the same

edit i have now put the split power adaptor to the card on the same rail and pushed the graphics and seem to be more stable.. i have a strange feeling i may have chaged that when i was replacing the floppy drive. i think that the psu was makin the thunk sound just before i would have problems about 20 -30mins warmup
i also have the a strange feeling it is a pci-e bridge card (NV41) 6800 GS/XT

edit 2: took the 2D/3D clocks question as a hint and reprogrammed the bios so that all was the same .. we'll see how that goes
cheers

 

[d2thez28]

It is great that you changed the clocks to match each other. That should definately help. But I have always thought it sounded power related. By today's standards 16a on the 12v rail is a bit small. What else you do you running in your system?

 

[morry71]

d2thez28 i to am comming to that conclusion. i think with 3 hd's and a DVD + extra fan the 6800 just sucks up too much. just weird how i can get it to be stable for 5+ hrs before it goes back to it's old ways. although now after 9hrs it went into stutter then out back in and only in 3D. 2D is fine and also there is no thunk warning . i just want to be sure that it's the psu and not the mb combination cause i don't want to waste $$'s
thanks for the advice it does help to let it out with someone listening
cheers
edit: what advice do you guys recommend on a psu up to the task
i've looked into a few but there are so many. what about Hiper Type-R 580 (which i can get for A$149 here in oz ) or stretch the buget to a TT ToughPower 600 .. or something else?

 

[d2thez28]

Well I certainly can't speak for the Hiper brand as I have never used them, however, a reputable brand with at least 20a on the 12volt rail should be good. Enermax makes nice units, and I use 2 Antec (a Trupower 550watt and a SmartPower 500watt) and they both have dual 12volt lines with plenty of amps between them. Both of those units can be had for under $100. To make sure your problem isn't power related, I would try to play a few days with every bit of excess hardware unplugged to try to save some power!

 

[morry71]

again great advice, will unplug as much as i can. i won't be able to get anything until next week anyway. wish i was able to get those sorts of prices in western australia. your about 30-40 US dollars cheaper.
thankyou

 

[d2thez28]

I guess Western Australia would throw a wrench into the pricing mix. And thanks for the compliment regarding my advice, but I am simply trying to help.

Unplug and optical drives, harddrives, floppies, fans, lights and anything you can spare electricity from to test and see if it still happens. Another idea would be to borrow one from somebody or otherwise get your hands on a stronger power supply and use it to see if it still happens.

It is ideal to have whatever power cord running to your videocard to not be shared with any other device. I.e. make sure it is dedicated to just the videocard.

I used to run a 6800gt and 6800Ultra in the Abit NF7-S 2.0 with no problems what-so-ever, so I don't think it is compatability issues.

 

[morry71]

ended up getting the hiper type-m 580(also has a bit of future resistance) and d2thez28 everything so far has smoothed out
my wife will be sooo happy(might even get lucky). thanks so much d2 for sending me up the right path
now i want one of these psu's for mine except i've just spotted a new Hiper Type R 730W Modular PSU
a big cheers
maurice

 

[d2thez28]

Well glad to hear things are looking on the upside for you Morry! And best wishes on "getting lucky."

Hey man, just remember, lots of watts doesn't necessarily equal good. You really want plenty of amps on all of the rails, especially the 12v rails. That is where it counts the most!

 

[fuelvolts]

Turning off AGP fastwrites works for me. I have an AGP dedication of 256MB and a 6800nu w/ Gigabyte NF3 250gb mobo. Fixed the HL2:E1 issues just fine.

Installing 91.47 Forceware drivers now, so hopefully this doesn't hose my rivatuner settings.

Will let you all know if it does anything bad

 

[d2thez28]

Well leaving Fastwrites on if my other proposed fix works would certainly be the ideal solution. Fastwrites has its benefits, though they may not be readily apparent. Why overcompensate to fix something which may work fine after one simple conclusive fix??

 

[GSDragoon]

So ya, I got an NF3 board and a 6800nu. I've actually been through 2 motherboards and they both had stuttering and/or freezes. Fresh windows install, tons of different drivers, that fix on MS's site, messing with agp busses, voltages, underclocking, ect, ect.

Well, just now I finally fixed the problem. What I think the problem is, would be a conflict between the chipset and drivers/software with fast writes. I had fast writes dissabled in the BIOS. After trying almost everything out theres, I noticed the "AGP Settings" in the classic nVidia control pannel showed some weird info. Yeah, removing/not installing GART drives fixed the problem too, but that's not really a "fix" if you have to run in PCI mode. Anyways, that AGP settings menu showed something weird. They list settings for GPU, chipset, OS and Optimal. I noticed fastwrites was a "Yes" for each category still and the Fast writes box what checked in "other settings". I also noticed it was still enabled in RivaTuner. So, I unchecked it in the NV control pannel and in RivaTuner. Restarted and OMG, it was FINALLY gone. So my guess would have to be something with fast writes and maybe not being dissabled. Like, the drivers were thinking it's enabled but the BIOS having it off. So there would be a conflict of the drivers/programs trying to use fast writes and the option being dissabled in the BIOS causing problems. But that's just my guess. Hopefully this can help others out there.


On a side note: the AGP settings shows 8x for max AGP rate for GPU, chipset and Optimal, but 2x OS. Shows that for any driver I use and I have 8x enabled in the BIOS.

 

[morry71]

i thought everything was fine so i put back the original bios with original fan speeds
and bugga me it was back. so re applied the bios where there is only one level plus no checking of temp and it still came back makin the cpu work at 40 - 50%. i then adjusted the fans down to 25% and now it never reaches 48C, alway above, and i have not seen any stutters or unusual cpu usage for the last 3 days. i can only conclude there is some wierd temp issue and conflict with nforce as i don't have these problem on a via chipset