FRAPS stole my frames!

Well that didn't go as planned. I discovered a performance-draining issue with the program FRAPS, and some new tests later, we see what the effects were!

Platform specs:
Athlon XP-M 2400+ @ 2.4GHz
512MB PC3200 @ 266MHz
Asus A7S333 (SiS 745)
Aureal Vortex 2
Western Digital WD800 80GB HDD
Windows 98SE Unofficial SP3
Drivers:
Rage Fury MAXX: A6.40CD15
Voodoo5 5500: v1.04.01
GeForce 256 32MB: FW 45.23
GeForce DDR 32MB: FW 45.23
Radeon DDR 32MB: Cat 4.1

○Follow me on Twitter! @pixelpipes
○Donate via Paypal: @pixelpipes

[PixelPipes]

Hello! Nathan here with PixelPipes, and just a heads up, this video is a follow-up to my previous Card Battles video featuring the Voodoo5 5500 versus the Rage Fury MAXX. Please watch that first or you won't understand what's happening here:
For those that just want the numbers and don't want to hear any explanation, go to the timestamps:
11:15 for redone tests
21:13 for new tests with additional bonus cards

Thank you all for being patient with me! We'll be back on track with the next video!

[anasevi]

Fraps has been known for its overhead since the 2000s, i find it more remarkable it is used at all these days.

[gloriousnait]

Hey Nathan,
I'm sure i speak for other people but often times these "mistakes" can end up being even more interesting than the original video content. I think this was awesome and I appreciate you going back and trying to understand exactly what happened. This is the kind of content that I think there just isn't enough of.

[mr]

What this really tells me is that we need a new tool for old Windows that's more reliable than FRAPS. Everyone knows that FRAPS has issues with overhead and such, but if you're testing old software/hardware on Windows 98 or something your options are somewhat limited.

[MaTtRoSiTy]

I'm just glad to see you making vids again tbh

[infinityzr]

For someone like me who was pretty late to the retro party (and would be lost without YouTube help), it is seeing an entire journey, mistakes and all, where I learn the most. Like you, I tended to just let fraps run on my 98se machine (Voodoo3 w 600mhz coppermine). I actually believed the fraps FAQ where it claimed the performance penalty was minimal. Gonna try my own benchmarks now. Big thanks for the update bro!!

[ErraticPillow]

fraps has been known for a long time to have a performance decrease just having it run not doing anything.

[daehxxiD]

Perfect reaction to your discovery. Of course people in the comments go all Dothraki like "It is known...!". Wise-assing about some supposed overhead that FRAPS has had (I for one only remember it being very heavy when capturing video).
But what you found is actually much more specific. If it was an inherent overhead of Fraps, then it would affect all your results the same way, but it did affect the results differently! So that in itself is already an interesting discovery. I certainly didn't think Fraps would have a preference in regards to GPU architecture or drivers (and quite honestly, it would be very interesting to figure out why it impacts the Voodoo so much).


The new Bench-results make it all the more intriguing... Somewhere there is some weird bottleneck.

[Mineav]

Fantastic video. I think you're doing a great job. Everyone goofs up; it's part of being human. It's obvious that you're very sincere about rectifying any such mistakes with more accurate info, and that matters most in the end.

[BadManiac]

D3D being faster in UT on the V5 5500 is almost guaranteed to be because of your very fast CPU. If you run the same benchmark on a period correct CPU, The glide score would very likely be much faster. The biggest strength of Glide was the low CPU overhead.As for asshats on the internet, best to ignore them. Your videos are some of the best available on the topic and a huge influence on my collecting and testing methodology.
Keep on keeping on, youtube needs more quality retro content!

[shadesofbeige]

I can't believe I missed this whole ordeal. Lots of arm chair quarterbacks out there. Don't sweat it Nathan.

[fedeterra]

As far as im aware, you are the best retro channel on Youtube, i was thinking yesterday that i dont understand how you dont have over 100k subs, its all good Nathan, we are all human.

[Saturn]

Dang dude, what an awesome video! Not only do I get to see these charts again, but now it makes way more sense based on your empirical data :).

[the_freebeard]

I've caught only a couple of your videos in the past that showed up in my recommended feed. For whatever reason, those weren't compelling enough for me to subscribe. However, the transparency and and candor you display here was what finally got me to click that subscribe button. We need more people like you here on YouTube. There's a big problem on this platform of people covering up their mistakes and denying their audience the opportunity to learn something from it. Often it's far more helpful to know what obstacles a person encountered than just trying to replicate their successes. Thank you.

[yuri]

You’ve done a good job with this mea culpa. As someone who didn’t have any follow up interaction with the previous video, I really appreciate your corrections and discussion about them and what went wrong. Owning the mistake and making good says much more about you and the channel than the original mistake ever could.

[ianweber]

16:20

Am I reading that right, the _PCI_ version of the Voodoo 5 is 10fps faster than the AGP variant? How in the hell does that work?

[RetroGPUsandBuilds]

Very interesting, going to take this into account when benching

[CrunchyMcDugals]

Hey, Nathan. I've really been enjoying your content. I think you handled this situation really well. Keep up the good work, I'm gonna keep watching!

[Arivia]

Hey Nathan, you’re right: Episode 1 Racer is an undeniable classic. Eeee kuuntaa!!

[kevinjennings]

As many people have stated here, don't sweat it. There are so many variables that can come into play between systems, even with the same hardware, even down to what updates have been installed on the operating system. We have to keep in mind that no one is saving the world by having more accurate benchmarking results. (Not everyone can get as picky as Steve from G.N...nor do we have to be.)