|
Advice Beginners BIOS Guide CPUs Links Mainboards Memory Network Storage Video/Sound Cards Contact Forum SiteMap Sponsors WebNews Home |
. | . |
Prices: Mainboards ABIT ASUS Chaintech Shuttle Soyo Tyan CPU Intel P4 2.4C-800 P4 2.6C-800 P4 2.8C-800 P4 3.0-800 P4 3.2-800 AMD AthlonXP XP 1700+ XP 2000+ XP 2400+ XP 2500+ XP 2700+ XP 3000+ XP 3200+ Athlon64 Athlon64 3200+ Athlon64 FX-51 Opteron Opteron 240 Opteron 242 Opteron 244 Opteron 246 Memory Corsair Crucial Kingston Mushkin OCZ |
LOSTCIRCUITS |
|
| ASUS V9280 A Case Study of AGP 3.0 | |
| (Review by MS, Dec. 9, 2002) |
AGP 8X, Is it Worth It?
The current grain of wisdom states that the performance increase that can be achieved from AGP 8X is hardly measurable or almost non-existent. We have looked into a few benchmarks earlier and came to similar conclusions. A recent article on [H]ard|OCP showed some differences depending on resolution and level of antialiasing, suggesting that there may be more to look out for in the future.
It is necessary, though, to keep in mind that changes from AGP 2.0 to AGP 3.0 go beyond the simple doubling of raw bandwidth. For once, abandoning the pipelined requests in favor of an exclusive sideband addressing scheme and the associated data streaming could easily account for differences in benchmark results. Some other factors that could influence the outcome of the benchmarks include the possibility of transmitting a complete request bundle using all four different types (T4-T1) within a single clock cycle through the SBA channel. The same request bundle would have needed 2 cycles in AGP 4X and 4 cycles in AGP 2X. Bottomline is that there are a number of important differences beyond the simple move from QDR to ODR that need to factored in to come up with a correct answer or at least an approximation.
Another issue that we have not touched upon yet is the definition of the Graphics Address Remapping Table (GART). Briefly, the operating system allocates several pages* of memory as the AGP aperture. These pages are usually discontiguous, however, AGP requires contiguous memory space for the graphics aperture. Using the remapping table, a physically discontiguous memory space can be remapped to the AGP controller to look like a contiguous memory block. Historically, VIA has had some problems with their GART drivers which was amongst the reasons why AGP 4X originally did not show any benefit over AGP 2X. The key issue, without pointing fingers here is that disclaiming of performance benefits based on a single chipset platform is not necessarily valid.
The last point is that there is only a certain range of setting where AGP 8X will bring out the most improvement. That is, with level of detail and resolution set too low, the system performance will cap the AGP performance. With a lot of filtering like AA and AF enabled, GPU and LFB limitations will mask any improvement in AGP performance. Keep this in mind for the following.
* AGP 1.0 and 2.0 only offered support for 4K page size that will be standard up to 512 Mbit memory chips. AGP 3.0 supports multiple page sizes for future generations of DRAM technology.
Next Page: => ASUS V9280S =>