Fraps and games

[jameswhite1979]

I have been checking out you tube videos of peoples machines running Crysis with fraps and a lot say fraps lowers there FPS. Is this true and if out of interest why?

J

 

[jameswhite1979]

I think its the FRAPS recorder that does this but if no I am interested in peoples input.

 

[zorrt]

Personally it feels like any kinda of FPS measuring software lowers the framerate including built in ones. Thats why I never keep them on.

 

[Lonyo]

I think he means when capturing videos with FRAPS.
And yes, it does reduce your frame rate. FRAPS requires CPU power to record the video, which takes system resources away from the game, so of course you are going to lose performance.
If you mean just the FPS counter, then it shouldn't have any impact, but since you mention youtube I assume you are talking about people capturing the action with FRAPS.

 

[skace]

FRAPS is hell on your hard drive when recording.

In the past, just having FRAPS running but not recording has caused slow downs in certain games. It's easy to test for, simply run the same game without it started and see if you are still having whatever slowness issue you were seeing.

 

[chizow]

I've never experienced any tangible slowdowns while having only the FPS counter in FRAPs running when checked against in-game FPS counters. There are slight differences due to sampling intervals but FRAPs is pretty transparent on system performance when simply checking frame rate.

Recording is a totally different beast though. I posted on this before and didn't get many responses, I guess not many people use FRAPs for recording game footage. Its incredibly disk intensive especially at higher resolutions and frame rates and in Vista, its also very CPU intensive. I don't recall recording being nearly as CPU intensive under XP however, but I was also running RAID 0 at a much lower resolution (1280x1024).

This is one of the few gaming applications where quad core provides a boost over dual/single core. In most of my games in Vista, trying to record with FRAPs would peg both cores on a C2D leading to choppy gameplay and lower frame rates. With quad cores, you can spread that load over 2 or even 3 cores and still have some leg room.

The lower frame rates I believe are due to FRAPs sync'ing your frame rate to your record rate. For example, with Vsync off, Sync off in FRAPs, full-size at 1920 and recording set to 30 FPS in a game where I get 30-45 FPS, I still get 20-25 FPS when recording in FRAPs. Some of this I'm sure is due to the high overhead from recording, but with a quad core its no longer choppy with low frame rates while recording even though FPS still drops to 20-25.

When examining the output video file, its pretty obvious why frames drop as the bitrate is ~700Mbps for 1920x1200 @25FPS. Some quick math shows this is ~90MB/s which is pretty close to the maximum write speed for consumer-level single drive performance. As a quick frame of reference, the HD VC-1 codec is only 135Mbps max or ~17MB/s. I'm pretty sure FRAPs is internally sync'ing rendered frames to written frames which is why there's such high CPU utilization resulting in the observed drop in FPS (even when 'sync every frame' is disabled).

So if you're getting low frame rates at high resolutions with FRAPs with a fast video card and CPU, I don't think there's too much you can do aside from getting faster storage or dropping resolution/recording quality. I'm planning to add a 750GB or 1TB disk for recording but it probably won't help much as the write times are similar to my Raptors. I could try RAID 0 with the Raptors after I get the large storage drive up and running but I've typically found running the OS off a RAID array is more of a PITA than its worth.

 

[KeithTalent]

I've never noticed a slowdown just using the FPS counter, but I certainly do when I try to record something using it.

KT