"Doom3 is up to 112% slower than the 6600GT"
- Nvidia PR doing the math. The X700 apparently renders at a negative framerate in Doom3.
More pages: 1 2 3 4
New dynamic branching demo
Tuesday, April 25, 2006 | Permalink

There's a new demo showing off the benefits of dynamic branching. The demo should run on Radeon 9500 and up and GeForce FX and up, but performance improvements by enabling dynamic branching is only expected to happen on Radeon X1300 and up and possibly also GeForce 6 series and up (haven't tested) since earlier chips do not support this feature.

Name

Comment

Enter the code below



AnarchX
Tuesday, May 16, 2006

The higher fps are also reported by the framecounter of your demo.

Nitro
Wednesday, May 17, 2006

The most timers in windoze use performance timer. Frequency of this timer is set on startup of system and it won't change when you overclock or downclock CPU until you restart the system. Because the function QueryPerformanceCounter usually uses CPU instruction rdtsc, on lower-clocked CPU it'll take less CPU ticks to draw a frame (if the demo is GPU limited as most are). Divided by the unchanged timer frequency you'll get higher FPS. Hope this helped

Jackis
Thursday, May 18, 2006

AFAIK, rdtsc counts ticks, and ticks can't "flow" with CPU frequency changing. If your function needs 1500 ticks for it's fininsh, then it takes 1500 ticks on all CPUs like yours one, not being depended on it's actual speed.

Correct me, if I'm wrong.

Nitro
Thursday, May 18, 2006

Why couldn't it flow with frequency? CPU is just like clock, isn't it? If I lower frequency, it will "tick" slower. And I know the function will take 1500 ticks, but only if it depends on CPU and not GPU. When it depends on GPU, it'll take less CPU ticks on lower clocked CPU, but the same time. There's only one problem that the value returned by QueryPerformanceFrequency will not change with the change of CPU frequency.

Jackis
Friday, May 19, 2006

Ah, well, you're right, Nitro, I've meant CPU clocking only )))

More pages: 1 2 3 4