Differences between the vertex shaders of the NV3x and ATi's R300 are primarily down to the NV3x's advantage of higher actual number of executable instructions through its 256 max static instructions via loops and branching (65,536 versus R300's 1024, also the NV3x flow control of loops and branching is data-dependent versus the R300's constant-based; the loops and branching abilities greatly reduces bandwidth consumption and may mean both the R300 and NV30's maximum vertex processing performance will be limited by current CPUs), an additional address vector (2 versus R300's single one), additional temporary vectors (16 versus R300's 12) and the availability of high-precision trigonometric math functions (SIN, COS, both not available in the R300).