This article is more than 1 year old
NVIDIA boxes clever with Intel and AMD
Optimise this. And that
Generally speaking, GPU driver sets are geared toward a united goal, to improve the performance of each feature on a graphics card under a certain operating system, such as Windows Me or 2000.
Optimisations on a per-processor basis are somewhat unusual, and when they do occur, they are frequently lop-sided, favouring one system over the other. This was particularly true of graphics cards released when the Athlon was starting to make its presence known just over a year ago.
The press releases are fairly similar, with identical trumpet-blowing NVIDIA information, and individualised processor-specific spiel about features and support.
Dan Vivoli, VP of marketing at NVIDIA, for instance pointed out that "the Pentium 4 processor introduced many exciting capabilities for increasing system throughput and the GeForce3 is ideally suited to take advantage of them".
Meanwhile, Ned Finkle at AMD explained that "the NVIDIA GeForce3 provides an excellent, high-performance video option for AMD’s processor customers...NVIDIA’s use of DDR memory in the GeForce3 perfectly complements the AMD Athlon processor’s 266MHz front-side bus and DDR memory speeds".
According to the Intel release, the GeForce 3 GPU takes advantage of the new Intel SSE2 instruction set, "including 144 new instructions for 128-bit Single Instruction Multiple Data (SIMD) integer arithmetic and 128-bit SIMD double-precision floating point". The system also takes advantage of the "Advanced Transfer Cache for higher data throughput, a 400MHz system bus, and a Rapid Execution Engine for higher execution throughput".
"SSE2" refers to Streaming SIMD Extensions 2. SIMD, in turn, is lengthened to Single Instruction Multiple Data, and is a way of applying a single instruction to multiple datasets simultaneously. With so much repetitive data manipulation involved in gaming, it's not hard to see how the GeForce 3 could be trained to take advantage of this.
The Advanced Transfer Cache is merely lower latency, higher bandwidth cache on the processor, and the 400MHz system bus will join it in helping memory-bandwidth heavy tasks like high resolution/32-bit display mode games.
It's actually more interesting to see the Intel optimisation than it is the AMD - because based on this, one would suspect that 3D games which use DirectX 8 and the GeForce 3 on a Pentium 4 platform would vastly outscore games running on the same platform but with an equivalent AMD chip.
Memory bandwidth limitation may even become something of a moot point; after all, the Pentium 4 isn't exactly low on this front. ®
Copyright © 2001, EuroGamer.net, all rights reserved.
NVIDIA GeForce3 Graphics Architecture Optimized for AMD Athlon Processors
NVIDIA GeForce3 GPU Optimized for Intel Platforms; NVIDIA GeForce3 Architecture and Drivers Take Full Advantage of the Pentium 4 Processor