Jump to content

DarkSwordsman

Member
  • Posts

    304
  • Joined

  • Last visited

Everything posted by DarkSwordsman

  1. I forgot to update this thread. After a little bit of prodding around and shutting down random programs one day, I realized it was Elgato Camera Hub that is causing the problem. I re-tested it and ran my system for 7 days. I was able to consistently see the entire system rise in CPU usage (as if it's throttling), and then as soon as I close Camera Hub, it goes from 10-20% idle down to 4-6% idle. I was able to reproduce it a few times now in various tests. I will file a bug report with Elgato at some point. With the fix, here is what the Task Manager looks like after 6 days. Usually at this point, the CPU experiences 25-40% idle where all progarms on the entire system are all using more CPU. It was insane finding this out and seeing my CPU finally come back to normal ever since I got this webcam over a year ago.
  2. Hello, An issue that has plagued me for the past few years has been random high CPU usage. No, it's not just one single program or an svchost that has high CPU usage. What often happens is that the entire system ends up using more CPU, as if the CPU is throttling, but it isn't (which I can often verify with HWiNFO). All the programs start to use 2-10x more CPU as the system stays on for 2, 3, 4+ days. A restart fixes the problem, but then it just comes back after a day or two. I know I should probably restart my system anyways, but it's incredibly frustrating to do so when I need to pick up where I left off the next day most times. I have done countless reinstalls of Windows 10 to try and solve the problem, but it always reappears. I do notice that, as far as I can tell, it doesn't happen with a fresh install of Windows, but I do have a plethora of programs I do need to install on a new system. These include: Logitech G HUB MSi Afterburner (without Riva Statistics Tuner) Voicemeeter VRCX Spotify Discord Steam Process Lasso (I often don't download it until I need it, but it does happen without it) Elgato Wave Link Elgato Camera Hub others that I may not remember, or have just downloaded over time, such as Blender Photoshop Clip Studio Paint Substance Painter Oculus VSCode Unity Hub + Unity 2019/2022 Escape from Tarkov Below are the required specs of my entire system: CPU: Ryzen R9 3950X GPU: EVGA RTX 3080 RAM: 32 GB (4x8 GB) 3000 Mhz CL15 Corsair Vengeance LPX Motherboard: ASRock X570M Pro4 - BIOS: Last Latest @ 3.90 (I see they finally released a new one: 5.50 as of 5 days ago, so I will try that at some point, but it does not mention a major issue like this) CPU Cooler: NZXT X52 240mm PSU: EVGA 750 B2 Boot SSD: ADATA SX8200 Pro (first gen, before they did that change I guess where they are worse now) Other SSD: Samsung 840 Evo 500 GB Other SSD: Crucial P2 2TB HDD: Seagate Barracuda 2TB (this thing is 10 years old) Ext. HDD: Seagate BP Slim 2TB NAS: TrueNAS hosted SMB, 4 shares connected to my pc via local network Let me know if I can provide any other information. Possible theories I have: I wanted to present a couple theories I have that may be related to or the cause of this problem, but have not been able to properly test. However, I will spoiler them, as to not bias any questions or suggestions, so look at your own risk:
  3. I thought this was a pretty good video that nicely summarizes everything. I figured it should have it's own thread, but I understand if it needs to be merged.
  4. I don't think these conflict at all. She could have been in the final stages of an interview, or had a few decent chances at other jobs before she quit. She was probably waiting to resign, but decided to just quit since she had some possibilities open already.
  5. Colin (former LMG employee) confirms partially what Madison said about her experience there. https://twitter.com/DubMFG/status/1692018343476441144?t=Cc5gefhFzdFEAsNJuYG1FQ&s=19
  6. So I don't want to make any assumptions from this. However, it is at least clear that Madison is aware of legal recourse. So here's to hoping something comes out of it?
  7. Criminally? For sure. I don't think anyone should be prosecuted off allegations. But considering the context of what LTT is doing, how they are responding incredibly poorly, and the weight of the allegations, I am at least morally obligated to support Madison here. Why would Madison make up such crazy claims over nothing? Especially when she wrote the same stuff in her glassdoor review back then.
  8. I'm not sure where 18 came from. She was 20 at rig reboot, so presumably 21/22 when she worked at LMG.
  9. You do understand that each of those social media posts were to be more than a "funny haha" that takes 5 seconds to come up with, right? And same with the floatplane exclusives?
  10. Hmmm, yes. Lets play individual paying out of pocket for a lawyer where the only evidence is not recorded vs. a multi-million dollar company whose image to the public is paramount. Surely that will go over well, especially if the lawsuit's existence were to leak.
  11. I did put it in its own thread, but it was merged into this one.
  12. Madison explains in a Twitter thread her working experience at LMG/LTT. A notable piece:
  13. I will say, despite how I feel about LMG's faults of accuracy and content quality over the years as a general velocity/target, this is completely true. People need to stop it with the football games. Nothing is completely black and white. The moment you feel angry about someone else's business, you have lost and should shut the fuck up.
  14. All I can say is that there have been a lot of inconsistencies that I've called out in comments and I believe on this forum before. The biggest one I remember was the SATA SSD vs M.2 SATA vs M.2 NVMe test and how the drives used were random and old, plus there was a chance there for explanation of why this happened that was never explored, or at least clarification that was just skipped. I get that LTT is trying to push a lot of content and be accurate, but you can't hit two birds with one stone. You can certainly find safe shortcuts and improve processes for accuracy and speed (oftentimes, accuracy IS speed), but over the years, so many video have had small imperfections that seemed glaringly obvious to fix. It seems like the focus has been on bandwidth and not quality. Your company has grown how big and how many channels do you have now, but the accuracy has barely improved, if not gotten worse in terrible ways? It seems many aspects are being skipped and small errors are falling through the cracks. I do know you mentioned possible "beta" program through Floatplane for community members to review videos for accuracy, so I do hope that works out. But I can count on more than two hands the number of times I saw some glaring error or oversight that should have been noticed. I think this is a time for LTT to focus on quality over quantity, even if it takes a hit to revenue. There needs to be a fix for this, as soon as possible.
  15. You could probably get away with just some more (and better) RAM, and a new GPU. The prices of everything right now are pretty amazing. You can even get 32 GB of DDR4 3600 Mhz CL18 for $65, or 64 GB for about $115. RAM prices are amazing right now. You can also get something like a 3070 Ti for about $320, or 3080 for about $420. I would probably say to also go for a full upgrade if you can afford it. The 13700K is a decent price, and supports DDR4 RAM. I personally avoid AMD just because of my bad experiences with the 1700X and 3950X, but even a 5700X or 7700X is also worth it, maybe except for DDR5 prices.
  16. If I were building this, I'd go 1200W if it isn't too much more expensive. I know the 13900K has been seen to spike around 350W at times, and the 4090 surely will peak sometimes around 550-600W even. So it's probably a good idea to just be safe rather than sorry. I'm also the guy running a 3080 off a 750W, and was running a 3950X, 2080 Super, and 1070 off a 650W. But if I had the option I'd prefer some headroom.
  17. If you don't want to spend a lot, you can buy a few cards like the GTX 1660 (especially if they're a single-slot variant) which have a Turing encoder (highest quality of the NVENC chips) and use this custom nvidia firmware to unlock NVENC (common for jellyfin/plex servers). You'd have to go 4000 series if you need AV1 (but I doubt you do). I believe they should be able to do about 200-300 FPS of 1080p, depending on your RAM and NVMe speed and if there's any filtering or anything you need to do on CPU. At least with a 2080 Super, which has the same encoder, I can often transcode a 1080p video at about 200-300 FPS in ffmpeg, which scales a little more than linearly with multiple transcodes (2x at 120-200, 4x at 80-120, etc.). I remember doing 4 at the same time with Jellyfin and it could theoretically manage 250 FPS of throughput on a 1070. Alternatively, AMD just released the Alveo MA53D, a 64 decode, 32 encode card. Bonus points for AV1 support, but I am not sure if AME support it yet, and it's about $1,500. There's also a ton of other dedicated transcoding cards you can get, but IDK about software support, and they're often enterprise products that are overpriced. Quadro/Tesla cards often have the same capability as consumer cards as far as NVENC goes. I just know that some of them have more chips, but that was back on Pascal and earlier, which has worse encoding quality.
  18. ChatGPT is not the only AI that exists. And yes, the entire point of a neural network is to make weighted decisions based on inputs and outputs. That's why they're so good at what they do. It isn't about ChatGPT taking over the world. It's about the maturity and application of neural networks and machine learning to do complex tasks.
  19. I don't think the scary part is the scammy part. The scary part isn't that they are intelligent or have a conscious (because they aren't and don't). The scary part is how quickly and easily they can process information with accuracy, and as a result, what humans will do with that. The scary part isn't the text chat or image generation, the scary part is the speed and efficiency that it can identify things or make decisions, and what that could mean with the context of nefarious use cases. People have already brought up the idea of AI being used to do pen testing. What if you were to train an AI on attack vectors, feed it info about a system, give it API and other access/functionality, and then have it quickly, across multiple instances, attack a system? We haven't really seen what an AI can do if it has unrestricted access to a system, beyond Bing's Chat and the supposed tests that OpenAI did where it tricked a human with logic and reasoning into getting it to pass a captcha. That's already scary in itself.
  20. I see. Either way, it should work fine. Your non-primary card being in the chipset slot shouldn't affect performance that much considering this type of workload. Most of the work is done on-GPU so any latency penalties from a chipset slot shouldn't cause problems. And if anything it should help with cooling the top card.
×