Initial Impressions On 16 May
Ok. Have added a $4.00 fan (originally a 586 cooling fan) to
the Viper 770's heatsink and did more benchmarking using Q3Test again. Here's the
Case open / ambient Motherboard temp of about 27 deg C. Overclocked it to 140/160 as
base-line for all benchmarking before performing tests (comes default at 125/150).
- Celeron 300A-450
- 128MB PC100 LGS RAM
- CL V2 12MB SLI
- Viper 770 + 32MB SDRAM (Eltrontech 6ns)
- VL1025 ATM card
- SBLive! Deluxe
- Viper770 latest drivers (12May) - version 4.10.01.0401.
I'm not really sure which TNT reference drivers these are based on (1.73 / 1.76 /
- Liveware 2.0 drivers
- Q3test version 1.05
Did the proposed "cg_drawfps 1; timedemo 1; demo q3test1" commands (thanks BK).
Of course, do take this with a pinch of salt as it may not be truely reflective as
rampantly pointed out in the net...Also, it was noted that the first timedemo always seems
to be higher than subsequent timedemos performed without first exiting Q3test. Hence, the
benchmarks below show the first timedemo run after restarting Q3test for each new clock
Based on Tom's highest settings (ie. 32 bit
mode / 32 bit textures / markings on walls enabled / Sky = high / Trilinear filtering /
highest textures / 1024 x 768 / Lightmap / etc), received the following for the first
timedemo for the various clock settings:-
1) At 140/160 = 29.7 fps
2) At 150/175 = 32.1 fps
3) At 150/183 = 33.7 fps
4) At 160/190 = 35.2 fps
5) At 160/200 = 36.7 fps
6) At 165/200 = 37.1 fps (with fan attached to V770)
7) At 170/190 = 35.1 fps (with fan attached to V770)
8) At 160/200 = 30.1 fps (resolution 1152 x 864 / all other settings being the same)
However, at 170 core, it was unstable and only
managed to run 1 round of Q3Test timedemo - it hung the screen when I retired the
timedemo. Hence, I guess the TNT2 chip has probably reached its max, stable core value
achievable at about slightly < 170 core.
Again, I couldn't go beyond the 200 memory
timing as Powerstrip restricts this. But it does seem that overclocking the memory
does offer a decent jump in fps...
It is interesting to note how overclocking the
memory seems to offer pretty significant jumps in frame-rate as compared to overclocking
the core. Also, increasing the resolution to 1152 x 864 doesn't seem to drop the
frame-rate that badly - so perhaps, I'm seeing a CPU limitation now....
Anyway, confirmed that the Viper770 doesn't
support overlay with the Iomega Buz. However, BK just sent me a mail saying that it may be
possible to do so at lower res of 640x480 (16bit).
Will probably run my TNT2 at 165/200 for now.
Not a bad deal - for the price of a regular TNT2 chip, managed to clock it slightly
above an Ulta-TNT2 chip spec...Guess I'm pretty satisfied with it....
Okay, guess it's back to Q3Test for now...!!