To: Elmer who wrote (3594 ) 1/8/1998 10:10:00 AM From: Adrian Wu Read Replies (1) | Respond to of 6843
Elmer: Both systems had identical configurations other than the motherboard and the processor. The new socket 7 boards with the VIA VP3 chipset all have 2x AGP. In fact, the VP3 chipset is more advanced than the 440LX since it supports DDR SDRAM and 100MHz frontside bus, features not available for slot 1 until the 440BX ships. Both machines had the RIVA128-based AGP boards and both had 64MB of SDRAM. "It seems to me that an awful lot of bandwidth will be eated up between the cpu and memory moving all this data back and forth." I am not quite sure what you meant. The data moves between the frame buffer memory on the graphics card and the pixel engine. The RIVA128 is able to move blocks of data to and from the main memory, hence they have no need for a Z-buffer and is still able to support high resolution with just 4MB of video memory. This makes no difference whether the board is attached to the PCI or AGP bus. In fact, the only video processor on the market that takes advantage of the AGP is the Rage Pro. No other graphics chip perform better on the AGP than on PCI. If you have 8 MB of video memory, AGP makes no difference anyway unless you are dealing with extremely large texture maps. The K6-3D's extra instruction set helps with the initial geometry setup. This is floating point intensive work. Hence, a graphics processor such as the Rendition V2x00, with it's own RISC floating point engine, is almost completly independent of the main processor. THe Diamond Stealth II (V2100) performs equally well on a Pentium 166 or on a PII-300. Other graphics chips such as the RIVA128, Rage Pro and even the Voodoo can take advantage of the K6-3D instruction set. The critisism of the K6 is the fact that it's floating point unit is not pipelined. AMD answered this with the new 3D instruction set. The oricessor core: Deep-pipelined, 3-way superscalar RISC engine is actually more efficient than the PII for interger function. The PII, of course, has a faster L2 cache. Adrian