My results seem to contradict yours.
Test Machine:
g73jh with i720, HD5870 and 8 gigs ddr3 1333
Cat version 11.8 early preview
Win 7 hp x64
benchmark 3DMark 11 Basic
test 1:
Twinturbo off, timer = TSC (default), no overclocking
RESULTS Click Me (3DMark Score P2233 - Physics Score 3781)
test 2:
Twinturbo on, timer = TSC (default), no overclocking
RESULTS Click Me (3DMark Score P2242 - Physics Score 3936)
test 3:
Twinturbo off, timer = HPET (fix), no overclocking
RESULTS Click Me (3DMark Score P2222 - Physics Score 3592)
test 4:
Twinturbo on, timer = HPET (fix), no overclocking
RESULTS Click Me (3DMark Score P2229 - Physics Score 3724)
test 5:
Twinturbo on, timer = HPET (fix), GPU overclocked to 800/1100MHz & CPU overclocked by editing pll registries with setfsb (1755MHz on all 4 cores with full load)
RESULTS Click Me (3DMark Score P2490 - Physics Score 3828)
test 6:
Twinturbo off, timer = TSC (default), GPU overclocked to 800/1100MHz & CPU overclocked by editing pll registries with setfsb (1755MHz on all 4 cores with full load)
RESULTS Click Me (3DMark Score P2495 - Physics Score 3944)
As seen here by my results the best performance is obtained by leaving TT off, the timers on the normal TSC, GPU @ 800/1100MHz and CPU overclocked via increased blck. Although it only gave a 5 point improvement in total score 2490 < 2495 it gave a 116 point increase in physics (cpu) score 3944 > 3828.
I am aware that some people mainly setfsb users are experiencing timing issues and if that is the case changing to HPET timing might help. As for me and I will assume other people who have the same rig and are not experiencing sync issues IMO it is better to leave the timer alone and forget about Asus TT.