AGP Aperture

[TimBlake84]

There may be another thread on this but I can't find it. Anyways, I am just curious. I heard that the AGP aperture setting in my BIOS can adversely affect the performance of my card. Is this true? What should it be set to and why does this happen. If it is not true, could someone please explain to me what this setting is for and what exactly does it change? Thanks a bunch... :attn:

[Slimmy]

I am not sure on this one. I had and ati card for the longest time and changing the agp aperture never really did much for me. I believe on the box or manual it says to set it to the amount of ram on your card.

Off topic. Have you ran 3dmark2k3 on this card yet? I have the same card and have unlocked everything to 16x1,6vp like yours. Stock i got 9100, No unlocked pipes and oc to 420/820 i scored 10600. With pipes unlocked stock speed i got 9900. Then i tried ulock all pipes and same oc and got 10600. Wierd i have no artifacts on any ot these.

[TimBlake84]

Thats quite strange... when I did my OC on this card, I benched it stock, benched it OCed, then benched it unlocked and OCed and that third bench resulted in my highest score. However, I have heard, that with it unlocked, there is a point in the OC at which you actually lose performance due to heat. How did you OC it to 420/820 anyways? Thats REALLY high on the core. I only have mine at 381/822. But hey, your score rocks. I only got 10,098 with everything modded. But I am planning on working on the OC and seeing if I can play with my BIOS settings or voltmod it maybe. Just a suggestion... in my adventures, I found that the chrome piece on the HSF severely restricts airflow. I would suggest taking it off. and while you are at it, apply some AS5 to the core and RAM. BFG uses thermal pads. I got about an 8 degree drop in temp after I did those two things.

[TimBlake84]

Got a few ideas for ya' here. with the unlocked and OCed settings, try upping the voltage to the card about .2 volts just to give it some more juice. I can't imagine that card running completely stable at 420mhz. Also, if that doesn't work, I have spoken to some people who said that their BFG card had a bad (masked) pixel pipeline. Therefore they only opened the VP and recieved a better score. Would you mind PM'ing me if you try these? I would like to hear how you do because I would really like to squeeze more out of mine as well.

[Slimmy]

Ya i thought of just opening just one or the other. Haven't tried yet. You only got 10098. :-/ With your system you should be killing mine.

Yea i benched several times to watch it for artifacts and none what so ever. I used the coolbits auto detect and that is what it gave me for the core. :attn:

Oh yea i checked the temp after benches and games and haven't seen it go above 50c. Yet it is cold in this room and i have great airflow. Damn thing sounds like a hovercraft! :p

[veryhumid]

woohoo!

http://www.ocfaq.com/article.php/overclocking/vidcard/43

[TimBlake84]

Veryhumid - perfect man... exactly what I needed. Thanks

Slimmy, I think your system has the upper hand with that RAM you have... Also, you have your cpu OC'ed and mine is stock. If I OCed my CPU I could probably break 11K but I just got a new mobo and am still fiddling with it. I am thinking of opening up the windows in my room, turning my room into a walk in freezer, OCing the hell out of everything I can, and just push for one really good bench. (ahhh, the joys of living in NH and gaming on your PC with a snowsuit on....lol) hey, if coolbits gave a return for your core like that, maybe that is the way I should go... you think it will give me the same numbers?

[Slimmy]

hey, if coolbits gave a return for your core like that, maybe that is the way I should go... you think it will give me the same numbers?

Coolbits "detect optimal frequency" would only work if it was at 12x1,5vp. Would let it oc if the other pipes were unlocked.

[gigabit]

Yeah i use it now since i dont need RT for my 16/6. Coolbits keeps me at a steady 371/800 no problem.I forget what it detects my optimal frequency to but usally what ever it detects for me is to high. i dont know why but it over estimates my oc by a lot.You will also get better scores runnig in 2D when OCed.I think thats why i have trouble because it detects for 3D.

[Slimmy]

I see you both have your core at 381 or lower. I have tested mine at 420 core and no artifact and no crashes. Is this too high or did i just get lucky w/ this card.

[gigabit]

Your lucky.I cant get my core past 415 with the screen freezing up.Then i have to reboot but anything over 400 on the core makes my scores and frames rates seem to diminish.If i oc my ram past 850 my core wont do 405 its weird and i tried upping the voltages but it dont help.Plus im not big on pushing mine to its limit.I want it to run fast but without stressing it to much..My best score to date is 11484 in 3D03 and 4119 in 05 if i push my cpu to 3.7 or better my scores would jump even better ,I get 65.000 in aquamark But only 19000 in 3D01 thats weird it should be much higher my old BFG fx5900 flashed with 5950 bios got 18,000 and some change.

[Slimmy]

My best score to date is 11484 in 3D03


Just curious, what setting did you use on the video card for this?

[TimBlake84]

okay guys... here is the link to my best score. http://service.futuremark.com/compare?2k3=3401057

It's weird, I have been comparing this score to many others and my Pixel Shading is blowing everyone away... double their score! What would cause this? And my audio sucks... oh well. Just remember, this is all on a stock cpu... wait until I get my hands on that OC....=)

[gigabit]

:rolleyes:
Score: 11486
Date: 2004-11-30
CPU: Intel Pentium 4 3409 MHz
GPU: NVIDIA GeForce 6800
394 MHz / 803 MHz
OS: Microsoft Windows XP
Res: 1024x768@32 bit Active

Published
Delete
Dynamic Compare URL
Compare URL: http://service.futuremark.com/compare?2k3=3412209

Score: 11404 Date: 2004-11-29
CPU: Intel Pentium 4 3409 MHz
GPU: NVIDIA GeForce 6800
385 MHz / 803 MHz
OS: Microsoft Windows XP
Res: 1024x768@32 bit
Active

Score: 11196
Date: 2004-11-06
CPU: Intel Pentium 4 3718 MHz
GPU: NVIDIA GeForce 6800
371 MHz / 776 MHz
OS: Microsoft Windows XP
Res: 1024x768@32 bit
Active

Score: 11180
Date: 2004-11-05
CPU: Intel Pentium 4 3718 MHz
GPU: NVIDIA GeForce 6800
371 MHz / 776 MHz
OS: Microsoft Windows XP
Res: 1024x768@32 bit
Active

Score: 11150
Date: 2004-12-02
CPU: Intel Pentium 4 3434 MHz
GPU: NVIDIA GeForce 6800
371 MHz / 780 MHz
OS: Microsoft Windows XP
Res: 1024x768@32 bit

:burn:
Here are my latest scores so you can compare them to yours

[gigabit]

Its your driver's they are probably optiomized more to the pixel shaders.Thats only thing i can think of.And they are old why are you runing 6 month old drivers try the new 67,42 or even better the67.02 they seem to be the most balanced drivers yet but they are beta
http://www.nzone.com/object/nzone_downloads_winxp_2k_67.02.html
Also dont be afraid of blowing that mobo up give it some juice gigabyte boards tend to under volt.

Candyman.
I have used it and i might give another go now that i dont need RT

[TimBlake84]

gig- great work... but compare your pixel shaders to mine... that is the only thing I blow you away in. Anyways, we have the same audio, but yet you kill me with audio. How does that work? Could it be the different chipsets?

[CandymanCan]

Yea idont see how ur getting like 90 points in PS 2.0 over mine ....... im betting it was just bugged. Btw guys do you use Coolbits to Auto overclock youre cards also for 24/7 use or do you set it @ certain speed you like and leave it thier.

Here is mine you can see my ps is much lower then his that doesnt make sense

http://service.futuremark.com/compare?2k3=3379791



{EDIT} damn 3dmark03 likes P4's more then A64's doesnt make any sense why a 3.4gig p4 beats my 2.5ghz A64. Its like this for everyone on P4's and A64's its wierd.


Timblacke re-run the test youre PS 2.0 is higher then anyone on the ORB from what ican see its just a buggy run

[TimBlake84]

I use RT to keep mine clocked 24/7.

EDIT: Okay, this is really weird guys. I did a multicompare of my three benches, and the bench where I overclocked it I got a dramatic drop in CPU score. The other two (the stock bench and the unlocked bench) the score stays the same but when I OC it, I get a drop... hmmmm

[CandymanCan]

I personlly use Powerstrip, just wondering if anyone actually uses the Auto overclock feature on Coolbits.

[gigabit]

Yeah most likely it has something to do with the chipsets and the drivers.Your drivers a very old you would do much better with a newer set since most of the newer driver are tweaked to optimize the 6800 gpu.Your pixel shading thats crazy maybe you have a freak card with an awesome pixel shader`or some how you have an extra one...lol

[TimBlake84]

I don't know if this will work, but here is what I am talking about it my previous post. These are my three benches. The third bench is where I get a drop in CPU score.
http://service.futuremark.com/servlet/Index?pageid=/orb/multicompare_view&p0=3398707&n0=Project+03&p1=3398778&n1=Project+03&p2=3401057&n2=Project+01&t=MultiCompare&k=10&s=1&l=1&c=-1265563316

[CandymanCan]

That is wierd, try rerunning the test oced again. I never used that multi compare before that pretty cool, lol.

[Slimmy]

Ok this is what i got so far...

3dmark2k3

420/835 12x1,5vp-10531

420/835 12x1,6vp-10635

420/835 16x1,6vp-10411

400/835 16x1,6vp-10418

380/835 16x1,6vp-10407

[TimBlake84]

dude.... thats amazing me. I still don't get how the hell you got your core that high... and then with 12 pipes??? Im so CONFUSED!!!!! maybe you just got a card that has been doing steroids or something cause that is not normal.

[TimBlake84]

Yeah, that multicompare is cool. it gives you a great idea of just how much of a performance gain each little tweak gives you.

[gigabit]

Slimmy
You pipes are bad but Vp is good thats why your score is the best with 12/6 rather than 16/6 or 12/5.

[CandymanCan]

To bad you need to have the pro version to do it i think.

[Slimmy]

Something weird here a 40mhz difference would make the score change alittle. Maybe 3dmarks isn't taking the oc from rivtuner.

I am going to test at stock speed w/ 16x1,6vp , 12x1,6vp , 16x1,5vp and 12x1,5vp. This should show me if i have a bad pipe or something. ;)

[TimBlake84]

You might be right candyman... I did pay about 20 bucks for it I think. Anyways, gigabit, wouldn't bad pipes result in a far greater difference in score? Also, artifacts and crashing typically occur with bad pipes... I don't know if this is Slimmy's case.

[TimBlake84]

slimmy, I have disabled the Speed Test in rivatuner so when I get home late tonight, I shall try benching it at the speeds you have posted. If your card can handle that speed, mine should at least be able to get close to it.

[Slimmy]

Anyways, gigabit, wouldn't bad pipes result in a far greater difference in score? Also, artifacts and crashing typically occur with bad pipes... I don't know if this is Slimmy's case.

That is what i thought. I have watched this thing a dozen time or so looking for artifacts or anything, but nothing still looks awsome. Dang i need to purchase this cause it takes forever to run through every single test.

[TimBlake84]

I would definately recommend buying 3dmark03... there are great tools that come with it such as the multicompare and dynamic-compare tools. But yeah, bad pipes is kind of like OCing too much, or at least thats what I was under the impression of. I seriously don't know what is going on with your card. Try running it at my speeds (381/822) and let me know what score you return on it.

[gigabit]

opps

[gigabit]

woohoo!

http://www.ocfaq.com/article.php/overclocking/vidcard/43
Whats up VH havent seen you in a while

[Slimmy]

ok here is me latest

3dmark2k3

420/835 12x1,5vp-10531

420/835 12x1,6vp-10635

420/835 16x1,6vp-10411

400/835 16x1,6vp-10418

380/835 16x1,6vp-10407

350/700 16x1,6vp-9924

350/700 12x1,6vp-9223

350/700 16x1,5vp-9801

350/700 12x1,5vp-9135 <--STOCK

[gigabit]

you never ran it 16/5 OCed to 429/835 and i would because you got a nice jump from12/5 to 16/5 which loos to be the best combo but youdidnt run it oced like tha

[TimBlake84]

that is really weird that your score is that unpredictable... I don't even know what else to say... (::shakes head in disbelief::)

[Slimmy]

got me baffled too....

[Slimmy]

ok i just ran it at 16x1,5vp and oc at 420/834 and only got 10240. :(

These aren't very consistent. I don't know what the heck.

Any suggestions.

[Sevin7]

I changed my agp aperture size from 64 to 128 and my frame rate in the source vid stress test went from 88 to 94 (max settings including AA and AF, 1280x1024)

[gigabit]

Turn off vsync and and fastwrite.Well tried all kinds of thing and just got a 200 increase but anymore and i get corruption big time.I wish i had taken a screen shot at 399/820.
http://service.futuremark.com/compare?2k3=3435376