The Biggest Myth in Speedrunning History

Please check the description if you want links to videos and the Ukikipedia pages featured in this video. i highly recommend checking those videos if you liked seeing those weird anomalies. EDIT: Here's a link to the SM64 TASing and ABC Discord Server for those interested. Many SM64 experts are active there: discord.gg/ECskvyF Also I want to make it clear that cosmic rays have likely caused bitflips in computers like that one election in Belgium. I'm just saying it probably didn't happen to DOTA_Teabag in this specific scenario.

A cosmic ray traveled 92mil miles to flip a bit at Google HQ so this video got recommended to me.

Dies in CCM, gets mad, slaps his N64, then goes to re-enter CCM but ends up in BBH instead. The creepypasta writes itself.

Until I see definitive proof that outside rays didnt cause a bit flip, I'm sticking my N64 in a microwave at the start of TTC and hoping for good IRL RNG.

hey, weighing in with this is a programmer and systems administrator, bit flips happen constantly, be it from heat, radiation or cosmic rays and are usually caught by inbuilt error correction, the n64 used 4mb of RDRAM running at 400mhz on a 9 bit bus, using it's built in error correction bit for the GPU instead. early versions of this RAM standard had no error checking but the standard used a pretty big transistor which would require more energy comparative to the phone in your pocket or your laptop's modern hardware, so a high energy particle is more likely than you may think. changing this bit at the right time seems to me like the most likely cause of the error. weather its ionising radiation from space or because the passively cooled memory was overheating or a solar flare is completely academic and unknowable, but IS more likely to be the case because of the energy requirement to flip a bit in that kind of memory module and it is a compelling story, and one I will continue to tell to comfort people when they can't open their word documents due to data corruption tl;dr we live in a world where ECC memory has made us forget that this isn't even unlikely anymore, it just goes unnoticed because Richard Hamming is an underappreciated genius edit: removed erroneous assumption about hexadecimal conversion because it was incorrect and not really relevant, I couldn't even remember what I meant by it when it was pointed out. I also corrected myself from "a bit flip is more likely" to "A high energy particle is more likely than you might think" because its what I meant and it didn't really make a lot of sense before people have also pointed out that I say in the TL;DR it's because of ECC memory and when they google ECC memory they find the Hamming module and find out that isn't applied to consumer hardware, then assume consumer grade memory has no protection from errors which isn't the case, every stick of memory since DDR4 has had a Cyclic redundancy check (CRC) and parity bits on different parts of the memory module, and DDR5 has a sort of on die ECC even in it's "non ecc" variants, the ecc variants send additional parity bits to the cpu for a second check (unless you're on intel, since I don't think their consumer CPU's even support true ECC), which will catch any errors in transport, because the density of a ddr5 chip is kind of insane now, it bit flips constantly. there's also all of this on the non-volatile storage side which as I understand it wouldn't be a problem for this error, if there was an unhandled error in the non volatile storage it would be consistently reproduced, and I don't think it could happen at all in a variable because the CPU creates and references these in memory

It was originally "speed run", until we found a glitch to drop the space, so now its "speedrun"

As an aside, cosmic ray interactions are in fact very common when working with some equipment! They're a regular nuisance for Raman spectrometry, for example, as they interrupt your spectra by creating 'spikes' at random wavelengths. Speedrunners, please take care while running Mario 64 on your CCD detectors.

The footage you showed comparing DOTA's run to a TAS isn't even identical before the upwarp occurred, and the only difference after the bit flip is a slight change in the camera angle. The fact that, in both cases, DOTA and Panen reach the same height and hit the ground at the same time should be clear evidence that a bit WAS flipped. Even if it wasn't cosmic rays, the height that DOTA got in his speedrun was exactly the amount you would get switching the leading byte of height to from C5 to C4.

"The clip of the bit switch and what happened is actually slightly different so it couldnt have happened" he says as the two clips match up perfectly except for the frame interpolation and fails to give anything else than even resembles the clip.

your opponent getting hit by a solar flare would be one hell of a way to lose a speedrun race

I can’t believe Big Speedrun is trying to gatekeep the common tech of cosmic ray manipulation.

I never expected to see a person so angry at cosmic rays in my life, the internet is truly boundless

The BitFS downwarp being around 900 units doesn't automatically rule out a bit flip, because Mario's position is stored as 32-bit floating-point numbers, meaning the highest bit represents the sign, the next 8 represent the exponent, and the remaining 23 give the mantissa, to make up a number of the form (basically) mantissa * 2^exponent. Flipping one of the mantissa bits will add or subtract a power of 2, but flipping one of the exponent bits will multiply the previous value by some power of 2, meaning if the Y coordinate started as -900 and the last bit of the exponent was flipped from 0 to 1, it would become -1800, resulting in a 900-unit downwarp. I know the lower sections of BitFS are below Y=0 (because of the A press save on Wii VC) but I don't know if the coordinates perfectly match the clip

Hi, Cosmic-Radiation-Bit-Flip researcher here (I am not making that up, that is actually my job). So to keep a long story short: computers are weird and trying to model anything happening on hardware exactly is practically impossible. I disagree with the statement that a cosmic bit flip cannot be the cause, simply as it is very hard to prove that it isn't and it is within the realm of (at least theoretical) possibilities. However, there are two other things we should consider that make the cosmic ray theory very unlikely. The first is the very low amount of radiation we observe on earth. A satellite within the Low Earth Orbit will still only get a small number of radiation events in the mission time, for a device on earth the possibility is negligible. On satellites we still need to consider the effects of radiation if hardware is safety critical, we have a very bad setup (high amounts of radiation expected), or we have long missions and aging effects, but that is satellite mission stuff. Secondly, we shouldn't forget how much hardware just doesn't work. Modern computer chips have a huge percentage of the chip that just doesn't work due to manufacturing problems. If you produce a CPU with 12 cores but only 10 work: sell it as a 10 core CPU, that's how it's done. Thus hardware malfunctions are definitely not unlikely and again really really hard to identify, even if you have the hardware in your hands. TLDR: is the cosmic ray theory possible? Yes, it is possible. Is the theory likely: No, it's not likely. Even though it's not impossible, the odds are very low. So we shouldn't say it's impossible, but also not act as if it was definitely the solution.

that cuphead journalist is never going to live that down.

I'm not very far into this video but just wanted to mention that cosmic rays flipping bits in RAM is not an "out there" concept. Server-grade RAM is literally designed to combat this for data integrity.

"The idea that a cosmic ray flipped a bit in a computer is highly unlikely." And yet, ECC memory is used in servers all around the world for... what, fun?

LJ: "10 years ago..." Me: "Oh so like 2003" LJ "September 21st 2013" Me: Turns to dust

There was an election in Belgium in 2003 where a candidate got 4096 extra votes (2^12) due to a cosmic ray bitflip. It was found and corrected, but strange and unlikely things happen given a large enough sample size and period of time.

"We don't know" not only means "it's not proven it DID happen", it also means "it's not proven it DID NOT happen". It's simply not yet proven on either side.