AGP-4x vs AGP-8x

chrisdent

Newcomer
Does anyone know if there is any real world (or even a theoretical benchmark) difference in speed between the 4x and 8x AGP bus with the Radeon 9700?
 
Probably only if you play at 1600*1200 with 4aa and textures have to come over the agp bus :)
Even if it would be usefull it still wouldnt make any difference if you dont have rdram or dual channel ddr.
 
I'm sure right now there isn't but in a year or two when they are ready to put out agp 10 or whatever there will be a 5% increase. That will most likely be because of driver increases in both video card and agp drivers and of course more advance games and what not. It wont be a big deal but hell everything little speed increases helps right ?
 
I think pci express graphics port will be 2 times faster than agp8.
4GB/s?
I think you need at least dual channel 400mhz ram for that to have any improvement over agp8, but the graphics cards wont be fast anough for it anyways.
There will be anough memory for textures on the card so its only for the polygons.
How many can you do on agp8?
I think its fast anough for 1-2 million at 100fps?
 
ye . .1st thing i thoough of before i got to Dave's response was the PCI express.

I wonder though when it'll be implemented.

US
 
On somewhat related topic: When r9700 was launched, there were some articles investigating the issue with AGP upstream speeds, i.e. framebuffer upload to PC for storage. The performance was abysmal.
This might become increasingly important, when people begin to understand what video and image processing capabilities r300/nv30 gen cards have.
I wonder how the PR people touting AGP8X benefits would react when asked about this issue.
 
GOOD AGP BENCHMARK

I find a good benchmark is Intel's AGP Stress Test that you can download off there site.

URL http://www.intel.com/technology/agp/downloads/agp_stress.ZIP

It uses OpenGL and DirectX 7.0 APIs to test throughput across the AGP bus.

I have a Geforce4 and an Intel i850chipset and have seen throughputs approaching 1017 Megabytes/s which is close to the AGP 4x theoretical maximum throughput of my graphics card/chipset.

It would be interesting to see no_way run this benchmark to see if the benchmark gets close to the the theorectical throughput of AGP 8x.

QUOTE FROM INTEL AGP STRESS TEST PDF
Games are not a good benchmark for AGP as most games spend a lot of processing power on artificial intelligence , collison detection, physics calculations etc which are irrelevant to the AGP bus. These overheads inhibit the maximum use of the AGP bus. Also games provide no ability to log errors or provide statistical information that are needed for regression testing.

:p
 
Mmmm, thanks for the link.

I've finally got an ASUS/SIS AGP8x board and an ASUS NV18 (as well as 9700) so perhaps I can get round to testing that.
 
AMDMB has a comparison between KT400 and nforce2 which delves into AGP 4x vs AGP8x

About a 4% difference but there is something wrong with the KT400 chipset at the moment which exxagerates that to something like 15%.
 
Re: GOOD AGP BENCHMARK

Diesel said:
I find a good benchmark is Intel's AGP Stress Test that you can download off there site.

It would be a better benchmark if it'd support non-Intel chipsets...
 
I believe PCI Express will support simultaneous 4GB/s uploads and downloads which should help apps that want to copy framebuffer data back to the PC.
 
All the major IHVs are focussing on games as an indicator of performance of their chips (note to all : every IHV is focussed on DOOM3 performance... ask Kristof :) ). That is also because that's what most reviews have focussed/are focussing on. Until that changes (absolutely not likely), all non-gaming-related AGP tests (like Intel's) are irrelevant.

There is no need to tell/prove to everyone that the AGP8X spec provides greater bandwidth than the AGP4X spec (because it is true). What reviewers need to provide are real-world situations where this is evidently true. Please let me know which app (or *shock, horror* game) is such evident proof. Until then , all AGP8X/AGP4X "comparisons" should be read with the thought of "Move along, nothing to see here".
 
I haven't tested it yet cause i don't have an AGP 8X board, but UT2K3 may benefit from agp 8x over 4x and show noticeable gain if using the UltraHigh Textures and quality settings on cards with 128MB or lower local memory... maybe, un-tested though yet :D

Doom 3 I have heard will benefit from AGP 8X

so a couple games out there at least that may help to have agp 8x with
 
Back
Top