R4,099! Better than the Take2 price for sure, but still a little too much for my liking... :/ R3,600 would have been a little better at least we'd only be paying R800 more than retail U$..
Either way Aquamark3 GFX and CPU score are not exclusive to each other. The CPU can either drive the GPU score up or down and so can the GPU drive the CPU score up or down. It's primarily a DX8.1 benchmark with DX9 extensions. There is very little PS2.0 processing going on and it's mostly VS/PS 1.0/1.3. It behaves much like 3DMark01 in this way and that's why it's not ideal for testing gfx cards anymore.
eg of shader instructions in this benchmark
================================
ps_1_1
tex t0
tex t1
tex t2
mul r0 , t1 , v0.wwww
mad r0 , t2 , v1.wwww , r0
mad r0 , r0 , t0.wwww , t0
(PS427)
vs_1_1
dcl_position0 v0
dcl_normal0 v1
dcl_color0 v2
dcl_texcoord0 v3
dcl_texcoord1 v4
m4x4 oPos , v0 , c3
dp3 r0.w , v1 , c55
max r0.x , r0.wwww , c0
min r0.y , r0.wwww , c0
mov r1 , c51
mad r1.xyz , c53 , r0.xxxx , r1
mad oD0.xyz , c54 , -r0.yyyy , r1
dp4 oT0.x , v0 , c72
dp4 oT0.y , v0 , c73
mov oD0.w , c0.xxxx
mov oT1.xy , v3
add r0 , v0 , -c2
dp3 r0.w , r0 , v1
mul r1.xyz , r0.wwww , v1
mad r2.xyz , r1 , c8.xxxx , r0
m3x3 oT2.xyz , r2 , c10
(VS430)
These were dumped randomly from shader instruction out as you can see they are really simple fragment programs and the vertex program is far more elaborate but still of DX8.0 complexity :/