Jump to content
Search In
  • More options...
Find results that contain...
Find results in...

Mister Snow

Member
  • Content Count

    421
  • Joined

  • Last visited

Reputation Activity

  1. Informative
    Mister Snow reacted to tim0901 in RX 6800 Ray Tracing Performance Leaked   
    As someone that's personally done research into this field (and spent many hours writing their own ray tracer): no, not really.
     
    Ray tracing is a game of statistics. Each pixel's data has an intrinsic uncertainty, due to raytracing having a heavy reliance on random number generators. This uncertainty manifests in the final image as noise: the larger the uncertainty across the image, the noisier it gets. We've known this since the 90s and a large amount of ray tracing research since has been about how we can reduce this uncertainty.
     
    And reducing uncertainty is, essentially, the same as making it faster in a time-limited scenario. If you can get the same uncertainty using less rays, you'll likely have a shorter overall render time.
     
    One of the ways to reduce this uncertainty, as you mentioned, is to use large amounts of rays per pixel. But this isn't ideal - more rays means longer render times - and this technique rapidly falls against the brick wall of diminishing returns. The difference in uncertainty between a 100 rays per pixel (rpp) image and a 200 rpp image is far larger than that between a 200 and 300 rpp image, and eventually you'll get no improvement at all. Many commercial renderers try to reduce the impact of increased ray counts by only applying them where they're needed. Rather than requiring all pixels to use the same ray count, they instead target an uncertainty level for the pixels to reach. Easy pixels like plain, matte surfaces might reach that after 10 rays, but your caustics and reflections might need 100. No need to waste ray calculations on pixels that don't need it. I haven't heard of this technique being used in games (although I admit finding information about this is rather difficult as game companies are far less open about their craft than computer graphics companies) but I imagine that's partly due to the incredibly low ray count they're using anyway (1-2 rpp I believe is common in gaming).
     
    So what else can we do to improve uncertainty? Well some techniques don't apply to real time graphics, such as rendering multiple images and multiplying them together. Turns out the image obtained by merging two 100rpp images can often be less noisy than a single 200rpp image. But there are also many things we can do within the renderer itself.
     
    One such way is through use of a Monte Carlo renderer. Named after the casinos and invented to design nuclear bombs, Monte Carlo Methods are mathematical tools that use random numbers to find the answer to complex mathematical equations, just like the ones that are generated while doing ray tracing. By using a probability distribution function (PDF) one can change how rays reflect within the scene, so as to reduce the probabilty of generating ray paths that will not contribute to the final image (eg, ones that will leave the scene and never hit anything) or to increase the probabilty of the opposite (eg, more rays directed towards light sources), which is known as Importance Sampling. This is a tricky process but massively improves the quality of an image with only a small performance impact. As far as I'm aware, no games use this technique.
     
    Another such technique is Path tracing - Quake II RTX uses this (but renders the whole scene using it rather than just shadows and lighting, which massively changes performance) - which requires changes to the raytracing algorithm that slows it down slightly. In return for these changes, a lot of effects that you would otherwise have to code in manually are suddenly done automatically eg. soft shadows, caustics, ambient occlusion and indirect lighting. But this is just the tip of the iceberg when it comes to algorithmic tweaks you can make. Light tracing, for example, reverses the raytracing algorithm and instead shoots rays from light sources, rather than the camera. Bidirectional path tracing combines both - shooting rays from both lights and cameras, where they eventually meet mid scene. Each completed ray traversal may be more difficult with these techniques, but the uncertainty of that traversal is lower as well. As far as I'm aware, no games use any of these.
     
    We can also make efficiencies regarding how the rays traverse the scene. DirectX raytracing uses a Bounding Volume Heirarchy (BVH) - which essentially encapsulates the scene in a complex series of boundary cubes, kinda like a Russian doll. If a ray doesn't hit a big box, then it definitely wont hit all the boxes inside it, meaning you don't need to bother calculating intersection tests for them. This is a common technique and pretty easy to code, but is a far cry from being the only thing we can do to improve performance. For example Metropolis Light Transport caches 'nodes' along a ray path, which can be reused later to massively reduce the amount of computation along common ray paths.
     
    All of the above is about CPU raytracing, and all of it can be applied to GPU ray tracing as well. All RT cores do is accelerate the process of navigating a BVH, which is a tiny fraction of the code that's involved in writing a ray tracer. The rest is all still handled by the CPU or in an HLSL shader, and as such can benefit from the techniques I've described above. And this is all before you even think about applying a denoiser to your output.
     
    So no, saying there's nothing that can be done to improve the raytracing algorithm is very much incorrect. There are many, many ways of doing this, a lot of which Nvidia invented themselves - they've been publishing a lot of papers about raytracing for years, from long before RTX.
     
    My guess as to why games aren't using most of these techniques? They take time to implement well and the companies can't justify spending that time on something that barely anyone is using. The return on investment just isn't there vs a simple solution that they can use to slap 'Ray Tracing Support' on their list of advertising points.
     
    (Sorry for the wall of text)
    As much as I want AMD's solution to be amazing, we have to remember that DLSS leverages the tensor cores available to RTX GPUs. The RX 6000 series has no such acceleration structures and so I wouldn't be surprised if the effectiveness of their solution was hampered by this. It will have to run on the GPU itself, thereby reducing how much headroom is gained using the technique. Until AMD has hardware that can accelerate their version of DLSS, I doubt it will truly be able compete with Nvidia's solution.
  2. Agree
    Mister Snow reacted to Ashley xD in US Considering Banning TikTok   
    so wait hold up, they wanna ban chinese apps because of data concerns, but they also want a law that bans the use of encryption? or at least gives them a backdoor into anything? hold up. 
  3. Agree
    Mister Snow reacted to captain_to_fire in COVID-19 - READ THE RULES BEFORE REPLYING   
    I’m gonna copy and paste what I’ve written on an earlier thread that got locked: 
     
    Just a PSA to everyone, now is not the time to brag how immune you are. You are risking the lives of many people so please stay at home, and if you really need to buy necessities like food or medication, don’t bring your entire family outside just to buy those. Don’t be that asshole who will shout at a cop for politely asking you to return home because you brought three other family members in your car. You’re not part of the solution, you’re part of the problem!
     
    You can do stuff at home like play with your dog, watch that Netflix series (sorry 480p Netflix in Europe, we in Asia still get 1080p and 4K), read those books, learn how to cook, workout using your body weight, whatever to take your stress away. 
     
    Also right now is not the time to be disgusting because so wash your hands for 20 seconds before rinsing and for the love of god, take a bath everyday regardless of your weather condition. No evidence shows that antimicrobial soaps are more effective than regular bath soaps so it doesn’t matter what brand.
     
  4. Funny
    Mister Snow reacted to WereCatf in ( Edit: Disproven ? ) According to a study, Excessive Phone Usage Results in the Formation of Bone Spurs in the Back of the Head   
    Nope, I rushed to the basement to check all of my skulls, but I can't find any tubes on any of them!
  5. Agree
  6. Agree
    Mister Snow reacted to Delicieuxz in Cyberpunk's creator slams critics who claim Cyberpunk 2077 is racist / inaccurate to the source, refutes their claims   
    Last week, gaming news site Rock Paper Shotgun ran an article written by a gaming journalist who is transgender and who emphasizes pronouns on their Twitter profile, who saw the depiction of minorities in Cyberpunk 2077, including "transhumans", black people, and Latinos as uninspired and as reinforcing stereotypes and racism. The author also threw in some swipes at Cyberpunk's gunplay in the same breath, just to make sure readers didn't miss the fact that the article has an objective of taking down the game by any means possible.
     
    I personally suspect the author's criticism is disingenuous and basically made to swipe at CDPR and Cyberpunk in hopes of causing damage to the company and the game's sales because Cyberpunk 2077 wasn't the personal soap box platform the RPS author wanted it to be for transgenders and other groups of people, with game characters' attitude toward modifying themselves in Cyberpunk being the trigger for it all.
     
    Well, the Rock Paper Shotgun article caused a sizable stir from commentators about whether Cyberpunk 2077 was plagued by racist ignorance and cultural insensitivity. One comment made in a reddit thread featuring a video interview with Cyberpunk's creator, Mike Pondsmith, summarized the comments on the RPS article like this:
     
    That comment sparked a response from Mike Pondsmith, himself, who categorically and succinctly refuted each of those assertions being made by commentators and also in the RPS article.
     
    The response made by Mike (who, in case you didn't know, is himself a black person) reads:
     
    So, Mike Pondsmith says the critics are wrong and are interpreting things in a way that fuels their agendas rather than how they actually are supposed to be understood.
     
     
    Another article I read said it's unacceptable to call a gang of (allegedly) mostly-black people The Animals when there's a history of negative association. There was also some criticism over an in-game poster showing a chick with a dick:
     
    Deeply transphobic ad in ‘Cyberpunk 2077’ represents everything wrong with cyberpunk
    Cyberpunk 2077 art criticized as transphobic is meant to reflect corporate exploitation, says the artist
     
     
    I kind of see the RPS article as the author thinking that the world and everyone in it is to be judged based on whether they're using every opportunity as a soap box to further that author's own agenda - including when the author's agenda is unrelated to the projects people are working on. Like, as though the RPS author thinks their personal goal overrides everyone else's and the purposes of other people's work.
     
    I also noticed that the Rock Paper Shotgun article's author takes aim at Cyberpunk 2077's presentation of "transhumans" and black people as allegedly being "unimaginative stereotypes", but then also takes aim at the game's presentation for the opposite, alleging the "clumsy and inauthentic presentation of a Latino character". So, some characters are supposed to be less stereotypical, while others are supposed to be more stereotypical? Or, are the author's attacks on the game disingenuous?
     
     
    Here's the interview with Cyperbunk creator Mike Pondsmith that the reddit thread was made for:
  7. Like
    Mister Snow reacted to rcmaehl in A-Waze-ing Directions - New algorithm from Toshiba 10x faster at finding the best solution   
    Source:
    Phys.org
    Science Advances (White Paper)
     
    Summary:
    A break through in combinatorial optimization by Toshiba has greatly increased the speed and best selection in combinations of patterns. This could improve computation in which hundreds or thousands of options have to be considered at once, such as navigating streets, finance, and drug development.

    Media:

     
    Quotes/Excerpts:
     
    My Thoughts:
    I'm going to focus on something this will affect a large amount of people daily, Google Maps and Waze. This algorithm allows more highly scalable and multi-core processing of potential routes. While the majority of users will use the major roads such as highways and interstates, this will allow a lot better and faster alternative directions to avoid things such as congestion. It should not be understated however, how this could potentially influence finance portfolios and drug creation. It's nice to see major advances in traditional computing to continually rival quantum computing as we have it today.
  8. Agree
    Mister Snow reacted to PlayStation 2 in *New* Edge (Chromium based) Preview is here! - Get it now!   
    I'm not even gonna lie, I'm gonna miss old Edge.
    I always loved how it actually used your hardware to a good extent, especially compared to fucking Chrome. Hell, I could play 720p60 video just fine through YouTube on Edge on my mom's old laptop compared to not even being able to do 480p properly on Chrome.
  9. Like
    Mister Snow reacted to rcmaehl in It's Rewind Time - Scientists revert three qubits 'fractions of a second back in time'   
    Source:
    Phys.org (Best read)
    Newseek (Simplified)
    Whitepaper (Original)
     
    Summary:
    Scientists successfully returned three qubits to their original state . Results show naturally occurring reversal and reversal of atoms or electrons is exponentially unlikely

    Media:

     
    Quotes/Excerpts:
     
    My Thoughts:
    Allons-y, it will definitely be interesting to see additional research into this. Is it only certain quantum information such as spin, or is it all of it? Is it actually making them simple from their complex state and is this something that will naturally occur in some sort of Big Crunch? Will this be another Faster Than Light Neutrino incident? Regardless, it's definitely interesting to see considering many scientific laws are time-agnostic.
  10. Funny
    Mister Snow reacted to NvidiaIntelAMDLoveTriangle in Ryzen Price Drops   
    What a great time to build a system, you can get a good processor for less than you would pay for RAM.

     
  11. Agree
    Mister Snow reacted to Tech_Dreamer in Forum usage survey (official)   
    i thought that pin colonel mentioned on the top navigation bar was a static logo until now.
  12. Agree
    Mister Snow got a reaction from matrix07012 in UK home secretary Amber Rudd thinks real people don't care about encryption   
    Yes, I'm sure Pinocchio cared about encryption before he became a real boy. Now he doesn't bother with frivolous things like privacy.
     
    On a serious note, UK has become a complete surveillance country and just when I think there's nothing else they can do, they think of something new.
  13. Agree
    Mister Snow got a reaction from Trippik in UK home secretary Amber Rudd thinks real people don't care about encryption   
    Yes, I'm sure Pinocchio cared about encryption before he became a real boy. Now he doesn't bother with frivolous things like privacy.
     
    On a serious note, UK has become a complete surveillance country and just when I think there's nothing else they can do, they think of something new.
  14. Funny
    Mister Snow got a reaction from RumSwift in UK home secretary Amber Rudd thinks real people don't care about encryption   
    Yes, I'm sure Pinocchio cared about encryption before he became a real boy. Now he doesn't bother with frivolous things like privacy.
     
    On a serious note, UK has become a complete surveillance country and just when I think there's nothing else they can do, they think of something new.
  15. Agree
    Mister Snow got a reaction from Jinchu in UK home secretary Amber Rudd thinks real people don't care about encryption   
    Yes, I'm sure Pinocchio cared about encryption before he became a real boy. Now he doesn't bother with frivolous things like privacy.
     
    On a serious note, UK has become a complete surveillance country and just when I think there's nothing else they can do, they think of something new.
  16. Agree
    Mister Snow reacted to tom_w141 in Hard OCP Blind test of Vega vs 1080Ti   
    Blind testing is bad news. When a company blind tests instead of revealing impressive numbers it means the product is inferior. They are basically saying its not as good but hey it feels the same! This is because at higher frame rates for example, we can't notice the difference between 120 and 150 fps.
     
    Also the fact that they keep stressing vega + freesync is cheaper than NVidia +gsync just to me says they know they aren't competitive on price for the gpus and are relying on the large premium carried by gsync panels. Well that's crap logic because for a start someone upgrading their gpu might already have a decent freesync monitor and therefore it would be better for them to get a 1080Ti and just not use the freesync functionality,
  17. Informative
    Mister Snow reacted to captain_to_fire in See the first movie uploaded to DNA of living cells   
    Primary Source: Nature
    Secondary Source: CNET
     
    ABSTRACT:
    I'm probably the only one geeking out to this but storing a movie into DNA is awesome. The scientists worked on bacterial DNA instead of mammalian DNA for a couple of reasons like bacterial DNA is circular with a single origin of replication and it's much simpler to work with. 
    Prior to CRISPR, genetic engineering is a hit or miss by using restriction enzymes to snip a portion of DNA and insert a desired gene or remove an unwanted gene into bacteria. But CRISPR changed all that with extreme accuracy and precision. CRISPR (Clustered regularly interspaced short palindromic repeats) is how prokaryotic cells defend themselves from virus attacks. A bacterial enzyme snips a portion of viral DNA and inserts it to their own genome so that once the virus attacks again, the bacteria becomes immunes. This has the potential to end life threatening diseases like cancer and HIV, reverse aging, or even make real life Captain Americas which the idea creeps me a little.
     
    But at this moment, the Harvard scientists were able to save a small video into bacterial DNA. 
    So I don't think anyone for the next ten years people would be making computers with DNA as a means of storage and run Windows 20. DNA is harder to maintain than NAND cell of a SSD. But according to the researchers...
    This could either be a life saver for patients with debilitating degenerative neurological diseases like Alzheimer's to kids in school who wanted to excel and remember better. It's kinda like taking steroids or doping but instead for competitive sports, it can be used to beat every other kid in school when it comes to academics. This proof of concept isn't new exactly as Linus even pointed it out on a Techquickie episode.
    There's so many factors to consider just yet when will human genome be our own mass storage. Immune system cells (NK and Tc cells specifically) might recognize these edited cells as foreign and kill it in the process. What would happen if those CRISPR edited cells containing instructions how to assemble a gaming PC becomes mutated or cancerous? Another concern is that gene edits are permanent. One mistake like a single random deletion, insertion, substitution can result to something fatal or nasty. It raises the question, it it worth it to tamper with the human genome using CRISPR just to store data or we just need to make solid state drives more efficient like NVMe and Intel Optane? 
     
    The possibilities are endless at the moment since the technology as at its infancy stage. What would be more exciting other than a fully automated restaurant is how would a computer read DNA. At the moment, in order to know how a DNA segment works is by transcribing the DNA segment by an enzyme RNA pol into a single stranded RNA segment which goes into the cell's ribosome for translation into protein.

    Scientists are able to replicate transcription process by PCR or Polymerase Chain Reaction involving an enzyme Taq polymerase

     
    I would love to know how the scientists encoded the video from that DNA segment. I would love to read the actual journal article and provide my own comments on it especially on the methodology but unfortunately, it's behind a paywall of $200 which I am not paying. This will just keep getting better once other scientists be able to reproduce this experiment. Who knows, maybe Apple would be selling movies and have them stored in our DNA to watch after payment.
     
     
  18. Funny
    Mister Snow got a reaction from VicBar in 7900X reviewed!   
    I love how "mega-tasking" is a thing now.
  19. Agree
    Mister Snow got a reaction from shadowbyte in 7900X reviewed!   
    I love how "mega-tasking" is a thing now.
  20. Agree
    Mister Snow got a reaction from Belgarathian in 7900X reviewed!   
    I love how "mega-tasking" is a thing now.
  21. Agree
    Mister Snow got a reaction from Menzenski in 7900X reviewed!   
    I love how "mega-tasking" is a thing now.
  22. Funny
    Mister Snow got a reaction from Techstorm970 in 7900X reviewed!   
    I love how "mega-tasking" is a thing now.
  23. Funny
    Mister Snow reacted to Energycore in 7900X reviewed!   
    Next year we'll have 64 Core CPUs for "Gigatasking"
  24. Funny
    Mister Snow reacted to Majestic in 7900X reviewed!   
    EXTREME MEGATASKING
  25. Agree
    Mister Snow got a reaction from Nicnac in 7900X reviewed!   
    I love how "mega-tasking" is a thing now.
×