Originally posted by: Mister Dark
I dropped four frames (from 118 to 114) in the source video stress test and couldn't figure out why ... Finally remembered that I had disabled fastwrites the last time I was tweaking my MB BIOS ... I re-enabled fastwrites and went back up to 118 ... I'm leaving it on.
Originally posted by: Squally Leonharty
In general, ATI cards usually are perfectly fine with FastWrite enabled, and, in fact, it should be enabled. It's the nVidia cards that don't like FastWrite in most cases. That's what I've read a long time ago on a different board. *shrugs* Might not be true with the newer cards, but meh. Just felt like saying this. 😛
Originally posted by: otispunkmeyer
damn those 4 un-noticable frames!
Originally posted by: Peter
Can we stop guesstimating what it is? CPU-to-GPU transfers, without putting the data in system RAM, have always been possible. The difference is that w/o FW, they're normal 66 MHz PCI cycles, whereas with FW enabled such a CPU-driven pushing of data can utilize AGP transfers at 2/4/8x that speed.
So what it does is accelerate the transfer of data (geometry, typically) that have been computed by the system CPU. Apart from when you're viewing HDTV, the amount of data isn't enough to have normal PCI cycles form a bottleneck, so normally, there is no mentionable difference.
lolSpunkmeyer, you smug bastard! I had a digital sensor implanted in my left eye that can detect variations in plus or minus a SINGLE frame all the way up to 300 FPS! I CAN tell the difference. So before you go making JUDGEMENTS about someone you don't really know, you should remember that when you assume ...
Looking at the Nvidia's Technical Brief titled "AGP 4x With Fast Writes - A Unique Nvidia GeForce 256 Feature," I quote:Can we stop guesstimating what it is? CPU-to-GPU transfers, without putting the data in system RAM, have always been possible. The difference is that w/o FW, they're normal 66 MHz PCI cycles, whereas with FW enabled such a CPU-driven pushing of data can utilize AGP transfers at 2/4/8x that speed.
So what it does is accelerate the transfer of data (geometry, typically) that have been computed by the system CPU. Apart from when you're viewing HDTV, the amount of data isn't enough to have normal PCI cycles form a bottleneck, so normally, there is no mentionable difference.
Originally posted by: Peter
Sideband addressing is either supported by your hardware or it isn't. Either the signalling of the target address works in standard PCI fashion - on the same signal lines that on the next clock start transporting data - or on their own "sideband" signals alongside the 1st data transmission.
Originally posted by: Gamingphreek
Yes they are a good thing in theory. THere are no real tangible performance gains. That is why you should just leave it on unless you have problems.
-Kevin