Skip Navigation

mapumbaa

@ sith @lemmy.zip

Posts
15
Comments
73
Joined
2 yr. ago

  • HP Z2 Mini G1a Review: Running GPT-OSS 120B Without a Discrete GPU

    Jump

  • I promise. It's not possible. But things change quickly of course.

    (Unless you're lucky/pro and get your hands on some super cheap used high end hardware..)

  • HP Z2 Mini G1a Review: Running GPT-OSS 120B Without a Discrete GPU

    Jump

  • It will probably kick the ass of that desktop. $2000 won't get you far with a conventional build.

  • GPT-OSS 20B and 120B Models on AMD Ryzen AI Processors

    Jump
  • LocalLLaMA @sh.itjust.works
    mapumbaa @lemmy.zip

    GPT-OSS 20B and 120B Models on AMD Ryzen AI Processors

    www.amd.com /en/blogs/2025/how-to-run-openai-gpt-oss-20b-120b-models-on-amd-ryzen-ai-radeon.html

  • Do you use models developed in China?

    Jump

  • I believe the full size DeepSeek-R1 require about 1200 GB of VRAM. But there are many configurations that require much less. Quantization, MoE and other hacks. I don't have much experience with MoE, however I find that quantization tend to decrease performance significantly. At least with models from Mistral.

  • Removed

    Searching through massive files

    Jump

  • Try ripgrep instead of grep. Unless you go for index etc.

  • [Debian Stable] Which Static Blog Generator: blag, Jekyll, Hugo, Lektor, Pelican, staticsite?

    Jump

  • [Debian Stable] Which Static Blog Generator: blag, Jekyll, Hugo, Lektor, Pelican, staticsite?

    Jump
  • Europe @feddit.org
    mapumbaa @lemmy.zip

    Greenland ‘Freedom City’? Rich donors push Trump for a tech hub up north

    www.reuters.com /world/europe/greenland-freedom-city-rich-donors-push-trump-tech-hub-up-north-2025-04-10/

  • Trump cuts funding to FOSS projects.

    Jump

  • What languages and patterns will be considered woke?

    For sure memory safety and dependent types. C++ will probably be shoved down our throats.

  • Deleted

    Permanently Deleted

    Jump

  • Fuck Reddit.

  • Rant! 100GB Log file in Nextcloud.

    Jump

  • I stopped using Nextcloud a couple of years ago after it corrupted my encrypted storage. I'm giving it a try again because of political emergency. But we sure need a long term replacement. Written in Rust or some other sane language.

  • Elon Musk’s X blocks links to Signal, the encrypted messaging service

    Jump

  • Because its a more user friendly experience. Your grandma can use it with no hustle. Matrix is more complicated UX and also much slower and buggier (at least if you use Element and matrix.org). Ive actually started to prefer Delta Chat over Matrix.

  • Removed

    Does the Fediverse give you hope?

    Jump

  • Yes. But I'm worried about Big Tech hijacking ActuvityPub and that they enforce the end of net neutrality. EU regulations and investments is our only hope right now.

  • Nvidia falls 14% in premarket trading as China's DeepSeek triggers global tech sell-off

    Jump

  • Thanks for high effort reply.

    The Chinese companies probably use SIMC over TSMC from now on. They were able to do low volume 7 nm last year. Also, Nvidia and "China" are not on the same spot on the tech s-curve. It will be much cheaper for China (and Intel/AMD) to catch up, than it will be for Nvidia to maintain the lead. Technological leaps and reverse engineering vs dimishing returns.

    Also, expect that the Chinese government throws insane amounts of capital at this sector right now. So unless Stargate becomes a thing (though I believe the Chinese invest much much more), there will not be fair competition (as if that has ever been a thing anywhere anytime). China also have many more tools, like optional command economy. The US has nothing but printing money and manipulating oligarchs on a broken market.

    I'm not sure about 80/10 exactly of course, but it is in that order of magnitude, if you're willing to not run newest fancy stuff. I believe the MI300X goes for approx 1/2 of the H100 nowadays and is MUCH better on paper. We don't know the real performance because of NDA (I believe). It used to be 1/4. If you look at VRAM per $, the ratio is about 1/10 for the 1/4 case. Of course, the price gap will shrink at the same rate as ROCm matures and customers feel its safe to use AMD hardware for training.

    So, my bet is max 2 years for "China". At least when it comes to high-end performance per dollar. Max 1 year for AMD and Intel (if Intel survive).

  • Nvidia falls 14% in premarket trading as China's DeepSeek triggers global tech sell-off

    Jump

  • Yeah. I don't believe market value is a great indicator in this case. In general, I would say that capital markets are rational at a macro level, but not micro. This is all speculation/gambling.

    My guess is that AMD and Intel are at most 1 year behind Nvidia when it comes to tech stack. "China", maybe 2 years, probably less.

    However, if you can make chips with 80% performance at 10% price, its a win. People can continue to tell themselves that big tech always will buy the latest and greatest whatever the cost. It does not make it true. I mean, it hasn't been true for a really long time. Google, Meta and Amazon already make their own chips. That's probably true for DeepSeek as well.

  • Nvidia falls 14% in premarket trading as China's DeepSeek triggers global tech sell-off

    Jump

  • I really don't believe the technological lead is massive.

  • Nvidia falls 14% in premarket trading as China's DeepSeek triggers global tech sell-off

    Jump

  • The thing is that they can no longer sell insanely overpriced hardware. The VC fucktards and coin bros begin to understand they're being scamed. And that murdering competition from China is around the corner. All this is very good news for consumers.

  • Nvidia falls 14% in premarket trading as China's DeepSeek triggers global tech sell-off

    Jump

  • Nice. Fuck you Nvidia.

  • Nvidia falls 14% in premarket trading as China's DeepSeek triggers global tech sell-off

    Jump

  • It will probably not reduce demand. But it will for sure make it impossible to sell insanely overpriced hardware. Now I'm looking forward to buying a PC with a Chinese open source RISCV CPU and GPU. Bye bye Intel, AMD, ARM and Nvidia.

  • If you could improve PeerTube, what would you improve?

    Jump

  • Better federation and search. Make it easier for content creators and instance runners to monetize. Make it easier to script your own feed/recommendations.

    All of the above is true for all software on the fediverse. But monetization especially for PeerTube. Pay to watch, donations and ads should all be options in the official implementation. Because of the high cost of running a large high bandwidth instance (if that wasn't obvious).

    Seems like the only ones who really benefit from PeerTube right now are right-wing extremists. The only large Swedish instances are far-right. And they are big because of content supply and demand.

  • Fediverse @lemmy.world
    mapumbaa @lemmy.zip

    Will 2025 be the year that venture capital discovers the Fediverse?

  • Fediverse @lemmy.world
    mapumbaa @lemmy.zip

    I need a Facebook replacement

  • Guix @lemmy.ml
    mapumbaa @lemmy.zip

    Is it possible to get pretty and precise error messages from Guile?

  • politics @lemmy.world
    mapumbaa @lemmy.zip

    New research shows the massive hole Dems are in

    www.politico.com /news/2024/12/22/democrats-2024-election-problem-focus-group-00195806
  • RISC-V @lemmy.ml
    mapumbaa @lemmy.zip

    Gaisler to build 7nm RISC-V chip for space

    hardwarebee.com /electronic-breaking-news/gaisler-to-build-7nm-risc-v-chip-for-space/
  • Guix @lemmy.ml
    mapumbaa @lemmy.zip

    How to configure Emacs for Guix development?

  • Fediverse @lemmy.world
    mapumbaa @lemmy.zip

    Don't reply "just Google it"

  • Selfhosted @lemmy.world
    mapumbaa @lemmy.zip

    How do you solve dynamic DNS?

  • Technology @lemmy.world
    mapumbaa @lemmy.zip

    Elon Musk wanted an OpenAI for-profit

    openai.com /index/elon-musk-wanted-an-openai-for-profit/
  • Technology @lemmy.world
    mapumbaa @lemmy.zip

    China targets Nvidia with antitrust probe

    www.reuters.com /technology/china-investigates-nvidia-over-suspected-violation-antimonopoly-law-2024-12-09/
  • Guix @lemmy.ml
    mapumbaa @lemmy.zip

    Missing dependencies in/with Steam (nonguix)