Jump to content

Raytsou

Member
  • Posts

    217
  • Joined

  • Last visited

Awards

This user doesn't have any awards

Profile Information

  • Biography
    Just your average software engineer / tech enthusiast

System

  • CPU
    R9 5900X
  • Motherboard
    ASUS X470 Prime PRO
  • RAM
    2 x 32GB GSkill DDR4-3600 cl18
  • GPU
    Gigabyte 6800XT
  • Case
    NZXT H510
  • Storage
    ADATA SX8200 PRO 1TB, Samsung 850 EVO 500GB, Crucial BX200 240GB SSD, ADATA SP920 256GB SSD, 1TB WD Black
  • PSU
    EVGA 750G3
  • Display(s)
    acer ed347ckr 3440x1440 100Hz
  • Cooling
    black ice nemesis 240GTS + 120GTS 30mm with noctua nf-a12x15
    ekwb quantum velocity waterblocks for AM4 and 6800XT
    ek quantum kinetic tbe 120ddc
  • Keyboard
    DURGOD taurus k320
  • Mouse
    Glorious Model O Wireless
  • Sound
    Sennheiser HD 8XX, Sennheiser HD6XX, HiFi Man HE4XX, Beyerdynamic DT 1990 Pro, Meze 99 Noir
  • Operating System
    Windows 10 Pro, Ubuntu 18.04

Recent Profile Visitors

2,419 profile views
  1. so anything that's printed in your application's output, such as logs or error messages, are going to go to stdout, if you're in the terminal, then it will display within the console. You can actually pipe this std out to go somewhere, such as using >> to overwrite a file or > to append to a file. So if this is a short script one might do `./script.sh >> log.txt`.
  2. Newegg is cheaper but if anything goes wrong (and they have a track record of fucking up half their orders nowadays), customer service will give you the run around and do everything in their power to blame you, up to and including call you a liar. Stay away from them, they used to be good until they got bought out.
  3. Apple's marketing strategy is to create a walled garden ecosystem. As long as you exclusively use apple for your tech products, everything "just works". To discourage people from trying other things, apple generally makes the user experience absolute dog shit (sometimes intentional, but usually by just not putting any engineering resources there). The way they gain users is by partnering with schools to teach kids how to use apple products from a young age, and part of this partnership is that the schools get free ipads/MacBooks in exchange for not allowing the kids to be exposed to non-apple products. The downside of this is that it's extremely hard for people to get into this ecosystem if you're already used to non-apple devices. You have three options: fully commit to apple devices, never touch apple devices ever again, or just live with being frustrated all the time.
  4. Xm4 are great. Also consider Bose qc45, they have differences but basically fill the same role. If you don't care about ANC, consider ATH-m50xBT, although they're quite lacking in base imo.
  5. Oh I didn't even realize that's a feature. I think you should figure out first whether it's a software issue or hardware issue. But assuming it's software, try disabling windows from automatically installing drivers, then go into control panel and uninstall all audio devices, then try installing your mobo realtek drivers again. Remember to re-enable windows driver updates.
  6. Have you downloaded the audio drivers from your motherboard's support page? Also not sure why you used ddu? It manages display drivers, the only audio drivers it touches would be the HDMI output from your GPU.
  7. You should read up about python deep copy vs shallow copy. Basically python variables work like references so setting array A equal to array B would copy the pointers to each sub array of B into A. Using .copy would copy the references one level deeper. If you actually want to clone the values of arbitrary n-dimensional lists you need to recursively .copy or import deepcopy.
  8. Yeah considering what's "interesting" is very subjective, youd have to train your own algorithm. Yes it is just programming, but it's machine learning engineering. What's you have to do is scrap together a database of a few thousand videos and manually label the "interesting" parts of each video. Then you feed train the ml algo on this data ( basically you feed the data into a bunch of linear algebra and have it shit out a function that is able to take a video and output the label). Once you have the model you're now ready to run it on newer videos and it should hopefully output what you want. You don't need a GPU at all if you're not doing this real time, you can set a video to be processed and come back a few days later to the labels generated.
  9. So you could make one script that contains the commands you need to run on each individual machine, then make a top level script that iterates over different IPs and runs that script over ssh. https://stackoverflow.com/questions/305035/how-to-use-ssh-to-run-a-local-shell-script-on-a-remote-machine
  10. I'm gonna only discuss m2 vs 4090 bc using a CPU for anything ML is using the wrong tool for the task. As for GPU performance, I don't think they are comparable and that's probably why there are no benchmarks. The performance of the m2 ultra shouldn't even come close to a single 4090. Quick Google search shows 4090 has 83 tflops of compute at fp32, while the m2 ultra has 27. While it usually isn't that good of an indicator of real world performance, Nvidia having much more mature software support would only widen this gap. Actually I found a tomshardware article that's shows the m2 to trade blows with the 4070ti. The 4090 would be at least as least 30% faster. Having a lot of vram shouldn't make the GPU significantly better at ML. At a certain point you have already loaded everything into memory and there will be no more speed benefits from having more empty space available to use. You could argue that apple has their neural engine going for it, which is a dedicated ml accelerator that reportedly performs at 31.6 TOPS. But the 4090 has the same tech, with tensor cores coming in at 1321 TOPS, only further exaggerating the gap. I genuinely think that if you're serious enough to want to spend 6k+ on a machine just for ml, it might make more sense to spin up a ec2 instance with multiple a100s, which will be orders of magnitudes faster, for just a few dozen dollars per hour.
  11. I suppose a more satisfying answer would be to also explain a bit about anti aliasing, or the problem of representing curves within a grid. I mentioned SSAA and that it's very computationally inefficient. Most of the later methods try to achieve the same effect by trading off accuracy for speed. Most popular is multisampling anti aliasing, or MSAA. Your GPU will take look at only the parts of your screen where you'll notice a difference (where there are curves) and selectively scale those parts to look better. No discernable difference at all from SSAA but is much faster. We also have FXAA, which is just as popular. In this method, we have the CPU blur the image outputted by the GPU. Very inaccurate but extremely fast. We next have TXAA, which instead of rendering at a higher resolution, will look at the few frames leading up to the current frame. Looked okay bit was computationally expensive. Often would be used in conjunction with MSAA. We then have deep learning super sampling, or DLSS. DLSS 1.0 is pure AI image upscaling. 2.0 conceptually combined 1.0 with TXAA for best accuracy and speed.
  12. Image scaling is traditional scaling, also known as super sampling anti-aliasing (SSAA), which renders the frame at a higher virtual resolution, then fits it to your screen. Nothing will look better than this, but this is very computationally inefficient. DLSS is AI upscaling with temporal filtering. In other words, it has an algorithm that looks at the current frame, along with the few frames leading up to it, and guesses what a higher resolution version of it might look like. This is usually not noticeably less accurate but significantly faster.
  13. I disagree, without this merger there isn't a single company large enough to compete with tencent
  14. I was thinking it sounds like he was describing fediverse, but I hear it's closer to a decentralized reddit. I'm not too familiar with the technical implementation but I suspect a decentralized twitter might actually not be possible.
  15. I believe raspbian is Linux so this should work: https://superuser.com/questions/168316/how-can-i-block-all-internet-access-on-ubuntu
×