Ok, this is odd. When I run the same 2 benchmarks as above through the console, I get some erratic results/behaviour. Here is what happened.
I overclocked my gfx card to see if it would make any difference, and also to se how they were compared to my intial tests at the start of this thread. With the card clocked @ 600/590, running the same 2 benchmarks as above through the HOC program, I got 96.8 fps (up from 95.0 with no oc on the videocard) in the HOC coast demo, and 65.1 fps in d3c171 (up from 62.3 with no videocard oc). It'a about as I expected as the memory on the videocard doesn't clock very much.
No running the benchmarks again through the console (keeping the card @ 600/590) gives me this (a little cut'n'paste from the sourcebench.txt file)
hoc_coast1.dem, 111.2, 8.1, 24.4,1600,1200,2714
hoc_coast1.dem, 112.2, 7.4, 24.2,1600,1200,2714
hoc_coast1.dem, 111.6, 7.4, 24.3,1600,1200,2714
hoc_coast1.dem, 111.9, 7.4, 24.3,1600,1200,2714
As you can see I did 4 runs and the coast benchmark consistently shows more fps through the console. So I'm up from 96.8 to 111-112 fps.... And this time the sound was "on", whereas I didn't hear any sound when I was benching using the HOC utility.
Running the d3c171 demo through the console resulted in HL2 exiting to the desktop every time I ran it.... Results were almost identical to using the HOC utility though.
d3c171.dem, 64.2, 5.9, 30.3,1600,1200,1943
d3c171.dem, 63.8, 5.8, 30.5,1600,1200,1943
So it's 64 fps vs. 65 using the utility. Also here I had sound "on" running it through the console.
It appears that there is some discrepancy with either this utility or benchmarking HL2 altogether....