PDA

View Full Version : AGP Aperture size


AndyH8
01-09-02, 08:08 PM
Read somewhere that if you increase the mobo Agp aperture size beyond the size of the card it speeds things up?
Bull or what?
Thanks.
Andy.


AMD XP 1600 512meg PC133
HERCULES GF3 Ti 200 with extra fans cooling. 200/460. rock solid.
3D 2001 - 5900.
Is this good , average or could i get more safely?

Jon
01-09-02, 08:20 PM
I generally use half my system memory. Twice the onboard video memory if it's not a PC that I care for video performance on.

Magnus
01-09-02, 09:37 PM
I have a gf2 mx200 and a7a266 mb

when i set mine to 256 in the bios i got about a 10 fps raise
in quake 3 @ 1024x768 ... not much but for such a small task
it was great.... also have never had a prob from it

Kent
01-09-02, 11:37 PM
Some say half system ram. Some say 64, 128 tops! Just play around and see what's best for your system. I have 512mg PC-133 and mine is set to 64. Tried 128, but it made no difference. 256 gave me all kinds of trouble.

TruckChase!
01-10-02, 12:25 AM
Do some benchmarking... most people seem to have different opinions as to what's the best.

Best performance/stability for me is 64 meg. 32 was actually a bit quicker, but sometimes 3dmark2k1 wouldn't completely finish. As I went up from 64 it got slower until I got to 256, which was totally unstable.

o770
01-10-02, 10:56 AM
its not an opinion matter. there r technical rules to set the agp aperture size and it will depend on your system, mostly system memory and video memory sizes. rojakpot had a good explanation about how to set the agp aperture size but theres nothing wrong on testing the system using some benchmark utility. do that and find out the best setting for u.
that Magnus said is very uncommon, differences use to be small with new video cards.

Despotes
01-10-02, 11:56 AM
Originally posted by oTTO
its not an opinion matter. there r technical rules to set the agp aperture size and it will depend on your system, mostly system memory and video memory sizes. rojakpot had a good explanation about how to set the agp aperture size but theres nothing wrong on testing the system using some benchmark utility. do that and find out the best setting for u.
that Magnus said is very uncommon, differences use to be small with new video cards.

Rojakpot's site is not available:( Has it gone the way of Guru3d??

oTTO, have you done any visual quality and speed testing with S3tc activated and disabled??
Thanks:)

rUfUnKy
01-10-02, 12:05 PM
Originally posted by AndyH8
Read somewhere that if you increase the mobo Agp aperture size beyond the size of the card it speeds things up?
Bull or what?
Thanks.
Andy.


AMD XP 1600 512meg PC133
HERCULES GF3 Ti 200 with extra fans cooling. 200/460. rock solid.
3D 2001 - 5900.
Is this good , average or could i get more safely? The lower you can keep the aperture the faster your card will be (usually not by much) The higher aperture you use the more of the mother board memory you will use which makes for a higher latency time...Also if it is to low during intense graphics it can bog down waiting for more available memory...

o770
01-12-02, 10:50 AM
Originally posted by Despotes


Rojakpot's site is not available:( Has it gone the way of Guru3d??

oTTO, have you done any visual quality and speed testing with S3tc activated and disabled??
Thanks:)

hey despotes.
i made it using the dxtc (this is how S3tc was implemented in the directx).
the problem about it with nvidia gpus is that they will render 16-bit textures when decompressing 32-bit textures using the dxt1. only when using the dxt1. so, if you disable the dxt compressions you will get an amazing difference on visual quality when running 3d applications that use that compression. performance drops a very little bit, not quite noticeable using my mx 32MB.
rivatuner has an available option to disable only the DXT1-DXT5 for both directx7 and 8.

Despotes
01-12-02, 11:49 AM
Thanks Otto, I'll give it a try. I think you mentioned that before in a post a while ago...;)

o770
01-12-02, 01:36 PM
Originally posted by Despotes
Thanks Otto, I'll give it a try. I think you mentioned that before in a post a while ago...;)

can u post a reply about what you thought?

Despotes
01-14-02, 05:15 PM
Originally posted by oTTO


hey despotes.
i made it using the dxtc (this is how S3tc was implemented in the directx).
the problem about it with nvidia gpus is that they will render 16-bit textures when decompressing 32-bit textures using the dxt1. only when using the dxt1. so, if you disable the dxt compressions you will get an amazing difference on visual quality when running 3d applications that use that compression. performance drops a very little bit, not quite noticeable using my mx 32MB.
rivatuner has an available option to disable only the DXT1-DXT5 for both directx7 and 8.

oTTO, I disabled s3tc and didn't notice any changes in visual quality in Q3(no changes in the sky either). I disabled DXT1-DXT5 in addition to S3tc being disabled and did not notice any changes in U.T.
I tried getting screenshots to compare but I was unable to find a program to read the file. Any suggestions on how to see the shots and what particular 3d applications did you notice the increase in visual quality??
Thanks

o770
01-14-02, 05:31 PM
Originally posted by Despotes


the s3tc you mean is the s3tc quality trick isnt it? this is not to be disabled once its just to try fixing the decompression problem with s3tc. leave it enabled.
i dont know about Q3 but those openlg games use to control the opengl settings as they want to, for ex: SOF can use s3tc, DXTC or not use at all any compression and it wont matter how the driver settings r. you just select the method you want in the video settings og the game. if this is the case of Q3 try not using any kind of texture compression for the text formats. instead of using s3 or dx you just select the high or maximum...
the dxt1-dxt5 selection is only for the direct3d applications, im figuring out UT is one of them. if ut has no selection for texture formats you should notice the differences from disabling them in the driver level, as you did.
it can happen that the applications you tried out not use the DXT1, then you wouldnt notice any difference. what that s3tc quality trick does is to force the app that use the DXT1 to use another one. the most surprising results i got was with nfs porsche unleashed.
also, i think you gotta be using 32-bit textures, you wont notice any difference using 16-bit.