Category:
Video Cards
|
|
Date:
03/10/2005
|
Author:
Giacomo Usiello
|
Manufacturer:
Crucial
|
Comparing the cards' specifications
In the following table you will find values obtained from official specifications that the producers gave for the graphic processors used for building the graphic controllers: though in some cases the card producer can set different working for the hardware, especially for the core and local memories rate (tables 1 and 2).
Videocard |
Core
Clock |
Pixel
Pipeline |
Memory
Bus |
Radeon
X800 Pro |
475MHz |
12 |
256
bit |
Geforce
6800 GT |
350MHz |
16 |
256
bit |
Radeon
9800 XT |
412MHz |
8 |
256
bit |
Table 1
Memory Clock |
Pixel
Fill Rate |
Texel
Fill Rate |
Memory
BW Peak |
900MHz |
5.7GPixel/s |
5.7GTexel/s |
28.8Gb/s |
1000MHz |
5.6GPixel/s |
5.6GTexel/s |
32Gb/s |
730MHz |
3.3GPixel/s |
3.3GTexel/s |
23.4Gb/s |
Table 2
In our case, for instance, the Crucial's Radeon works, by bios default, with a core clock rate of 472.50MHz and memories clock rate of 905.04Mhz (2x452.25MHz). We got these measures with the ATI Tool utility: here follows its output in the same operative conditions of the testing (figure 10).
Reading thoroughly the table above we may anicipate the hardware testing results; it is clear that the Radeon 9800 XT can't assure a response comparable with the other two controllers because of the generation gap: the core clock, the pipelines number and memories operative rate give low nominal values for the
Pixel Fill Rate, Texel Fill Rate e Max Peak Bandwidth parameters.
More interesting it is comparing the Radeon X800 Pro and the Geforce 6800 GT specifications: they have the same target: high performances, but they use two deeply different architectural approach to get it. On the one hand the ATI's vpu based card works at a higher core clock rate (475MHz vs 350MHz), on the other hand that has lower both memory clock rate (900MHz vs 1000MHz) and pixel pipeline number (12 vs 16). This way, on the whole the two cards, that are competing because of sharing the same segment of the market, roughly give alike nominal values.(ho sintetizzato un po', vedi se va bene il senso) In particular, the bigger architecture complexity (see the pipeline number) and the local memories speed of the nVIDIA card compensate the lower processor's operative rate, enough to get over the Peak Bandwidth nominal value of the Radeon card. But, on the contrary, the Radeon vpu has a simplified temperature control system because of its working at lower rates (you know that the temperature control system usually is required to have an increasing efficiency with the clock rate).
In the next page we start the
performance analisys with 3DMark tools suite.
|