Uploaded October 2017 | Updated November 2026, 3 hours ago
Is this even a thing? My rules are to time from end of name entry to the end of the race that unlocks ultimate parts. Nothing else, really, until such a time that dubious tricks show up. Which will probably never happen given the obscurity of this game. Don't you just love when you can apparently hold a world record with a casual run with limited planning and plenty of mistakes? There is one other speedrun on YouTube that isn't competitive and doesn't even unlock ultimate parts youtube.com/watch?v=yGrcsYQsjbA and there's this which has unclear rules and a potentially different Japanese version in addition to a longer time: speedrun.com/chq1
The hardest part about this game is definitely the bad controls along with that affront to physics. Both slow to react and really touchy. Makes it look like I intended to run into some walls if you don't know what's going on.
Unlike what some people have said, you don't really have to grind in this game if you get the right parts and use them well. I won every single one of the required races without having to do others to get more money. It only takes an hour that's about as varied as the game gets, which is why I ran it now that I learned that. This time should be in the ballpark of optimality as well.
This is the PAL version running on a somewhat modified PCSXR 1.9.95. No CD audio because something about the Mac version of PCSX or my custom image format/plugin (images weren't natively supported at the time I ripped it) doesn't support it, not to mention the physical disc was badly scratched. I was lucky the data track could be read in its entirety by a computer; a physical PlayStation often froze while loading. The music wouldn't be a good experience with all the interruptions anyway.
Most modifications are to graphics. This is basically DFXVideo with a new GTE Accuracy path and a homemade clipless homogeneous polygon rasterizer. The result is perspective correction, support for vertices outside the usual 2048x2048 guard band including behind the camera (edit: vertices were clamped to z≥0 here before being fed to the rasterizer), the equivalent of approximately 11 subpixel bits, and the occasional polygon looking ridiculous when different vertices end up in the same bucket or only some vertices are sourced from the GTE. And although it isn't literally perfectly precise like my custom 2D rasterizer which even applies the fill rule to edges between texels as if with infinite precision, it is perfectly robust. Unlike standard DFXVideo which isn't even accurate, I should add.
I also modified the video output to remove an unnecessary frame of latency of the Mac version (as if uploading on the order of 100000 pixels to the GPU warrants a multi-frame pipeline), and to add biquintic resampling in a linear color space. Biquintic probably isn't the best idea, but it's cool to two-up bicubic. It improves acuity (although not as much as windowed sinc with the same support) while representing cubic polynomials perfectly and improving smoothness to C². If the outer lobe were more pronounced than it is with coefficients focusing on these properties, ringing would be quite abhorrent.
Maybe I should've investigated the aspect ratio. Everything has a fixed aspect ratio in DFXVideo (usually 4:3), which only uses parameters to determine which pixels to use. This game should be 8:5 or so to have the least distortion. I can change that if I just think of it before starting the recording, the same way I didn't think to put my traditional rasterizer back.
Doing a full disclosure here to make clear that I haven't done anything nefarious. I didn't even bother to check if #define BIAS 1 works with this game. Would be nice not to have those sometimes severe drops in frame rate. Mind that the real hardware also had many of those drops; this emulation is fairly purist-friendly for something that isn't cycle-accurate. Either way, I'm probably slightly disadvantaged compared to real hardware, and this is too beatable for bickering.
I had to delete a rant on less relevant details of my rasterizer and ideas for changes. Now you'll never know. Thank YouTube for saying my description was too long. Thank people who fill their description with garbage tags for most likely being what encouraged the implementation of this limit.
Is this even a thing? My rules are to time from end of name entry to the end of the race that unlocks ultimate parts. Nothing else, really, until such a time that dubious tricks show up. Which will probably never happen given the obscurity of this game. Don't you just love when you can apparently hold a world record with a casual run with limited planning and plenty of mistakes? There is one other speedrun on YouTube that isn't competitive and doesn't even unlock ultimate parts youtube.com/watch?v=yGrcsYQsjbA and there's this which has unclear rules and a potentially different Japanese version in addition to a longer time: speedrun.com/chq1
The hardest part about this game is definitely the bad controls along with that affront to physics. Both slow to react and really touchy. Makes it look like I intended to run into some walls if you don't know what's going on.
Unlike what some people have said, you don't really have to grind in this game if you get the right parts and use them well. I won every single one of the required races without having to do others to get more money. It only takes an hour that's about as varied as the game gets, which is why I ran it now that I learned that. This time should be in the ballpark of optimality as well.
This is the PAL version running on a somewhat modified PCSXR 1.9.95. No CD audio because something about the Mac version of PCSX or my custom image format/plugin (images weren't natively supported at the time I ripped it) doesn't support it, not to mention the physical disc was badly scratched. I was lucky the data track could be read in its entirety by a computer; a physical PlayStation often froze while loading. The music wouldn't be a good experience with all the interruptions anyway.
Most modifications are to graphics. This is basically DFXVideo with a new GTE Accuracy path and a homemade clipless homogeneous polygon rasterizer. The result is perspective correction, support for vertices outside the usual 2048x2048 guard band including behind the camera (edit: vertices were clamped to z≥0 here before being fed to the rasterizer), the equivalent of approximately 11 subpixel bits, and the occasional polygon looking ridiculous when different vertices end up in the same bucket or only some vertices are sourced from the GTE. And although it isn't literally perfectly precise like my custom 2D rasterizer which even applies the fill rule to edges between texels as if with infinite precision, it is perfectly robust. Unlike standard DFXVideo which isn't even accurate, I should add.
I also modified the video output to remove an unnecessary frame of latency of the Mac version (as if uploading on the order of 100000 pixels to the GPU warrants a multi-frame pipeline), and to add biquintic resampling in a linear color space. Biquintic probably isn't the best idea, but it's cool to two-up bicubic. It improves acuity (although not as much as windowed sinc with the same support) while representing cubic polynomials perfectly and improving smoothness to C². If the outer lobe were more pronounced than it is with coefficients focusing on these properties, ringing would be quite abhorrent.
Maybe I should've investigated the aspect ratio. Everything has a fixed aspect ratio in DFXVideo (usually 4:3), which only uses parameters to determine which pixels to use. This game should be 8:5 or so to have the least distortion. I can change that if I just think of it before starting the recording, the same way I didn't think to put my traditional rasterizer back.
Doing a full disclosure here to make clear that I haven't done anything nefarious. I didn't even bother to check if #define BIAS 1 works with this game. Would be nice not to have those sometimes severe drops in frame rate. Mind that the real hardware also had many of those drops; this emulation is fairly purist-friendly for something that isn't cycle-accurate. Either way, I'm probably slightly disadvantaged compared to real hardware, and this is too beatable for bickering.
I had to delete a rant on less relevant details of my rasterizer and ideas for changes. Now you'll never know. Thank YouTube for saying my description was too long. Thank people who fill their description with garbage tags for most likely being what encouraged the implementation of this limit.







![Minecraft SkeleTNT mod: brutally devastating [1.8.8 download]
I wanted to experiment with ballistics. So I gave skeletons TNT and some excellent aim. What more can I say?
Class file for 1.8.7 and 1.8.8 (client and server): https://drive.google.com/file/d/0B96nn91pYjmia2JxQTNBakczWG8/view?usp=sharing Minecraft SkeleTNT mod: brutally devastating [1.8.8 download]](https://i.ytimg.com/vi/VunWO_-JPP0/mqdefault.jpg)


