Reply 140 of 145, by `Moe`
- Rank
- Oldbie
I guess then it's a rounding error - seems like ATI and nvidia do slightly different math. I think I can find a way around that. Are you using full scene antialiasing? That might worsen the problem as well.
I guess then it's a rounding error - seems like ATI and nvidia do slightly different math. I think I can find a way around that. Are you using full scene antialiasing? That might worsen the problem as well.
No, I don't use FSAA. However, setting it to Quincunx, 4x 9-tap or 8xS and higher helps to get rid of this problem. But brings with it even more - 2xQ and 4x9T are blurry, besides in 4x 9-tap and >=8xS mouse cursor behaves badly. 😀 It seems to become "inertional", i.e. stops not when I want it to, but slightly later 😮 .
I'm experiencing an odd performance issue when software HQ2x is enabled in gulikoza's build. When I run a game with Adlib music, the sound lags, whatever the cycle/frameskip config. In fact, with frameskip at 10, the effective cycles are around 5-10% below the requested cycles, whatever the requested cyle amount (even around 100). The CPU usage is far from 100%.
This happens on both my Duron 700 and on a 2.8 GHz P4 laptop.
Hmm.. I know that issue, though not as bad as you describe it. Up to now, I blamed the CPU cache (hq2x and adlib are both quite cache-intensive, thus interfering with each other). If you're on linux, you could use oprofile to count cache misses (or find other hot spots)
Note that Durons and P4's indeed have less cache than other CPUs
I'm in Windows, so, no oprofile for me.
ATM I've found a sweet spot on the P4, with 3000 cycles, the dynamic core and 1 frame skipped (with more, the sound lags, go figure).
A few hours ago, I had smooth sound too (with 4 frames skipped), the I let the computer go to sleep mode. When it woke up, I couldn't get the sound right. It was the same DOSBox session, and I didn't quit the game (The Colonel's Bequest).
You're probably right about the cache.. Do you know which core is the less cache hungry ?
You have to experiment yourself. I am an AMD person, but there was a time when normal core was faster than dynamic on an old Athlon. (that issue was fixed, I think)
If it gets worse with resume, try updating video drivers and using a different output method. Perhaps updating all the other drivers would be a smart thing to do as well. Such suspend-resume related problems could even be BIOS bugs.