Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!


New on LowEndTalk? Please Register and read our Community Rules.

All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.

RackNerd’s BLACK FRIDAY TAKEOVER! NEW DEALS, NEW TECH, AdminBolt HOSTING + HUNDREDS OF GIVEAWAYS!

1…2728303233…104

Comments

  • Count me in! Thanks for the Black Friday giveaway — good luck to everyone!

  • @noob404 said:

    @PuDLeZ said:

    @noob404 said:

    @PuDLeZ said:

    @gritty said:
    will RackNerd still be hosting flash sales this bf?

    Chef always bakes some tasty treats on BF!

    Hey dude. You still here?

    yes/no, glancing at the thread every now and then :)

    Did you get the Toronto VPS giveaway? I missed it.

    Nope, I wasn't at home at that time. All good though, giveaways are nice but that's not the real reason I here. Just here to chat, work on my f5 game, and then hopefully snack on some of chef's tasty treats bf when he makes an appearance!

  • @Jamie96 said:

    @DaDeveloper said:

    @noob404 said:

    @DaDeveloper said:

    @op23 said:

    @noob404 said:
    At this point, is it better to get gaming GPUs or NPUs like Ryzen AI Max for local LLM inference?

    At this point you need to get lucky.

    Any hardware with capability to run something decent including strix halo is crazy expensive.
    I wish I bought the 128gb one when it was $1800 USD late 2025. Dumb mistake. Now it’s 3-4x.

    Gaming GPUs or Macs are even more expensive for same memory capacity.

    Yea I panic bought in the 4k range for the sparks now they are more expensive, 7k if you can find them cheap, 9-11k otherwise XD

    Dang! Anyone has any idea if/when AI hardware prices are going back to sane?

    prob not for a good few years if ever, companies tend to be greedy like that

    Hate to agree here but this is it. As long as people cave and buy or companies too, they prove this is a viable market with these prices, so they are here to stay. They don't care if we start taking something like auto loans to buy our PCs, as long as someone is buying.

    Absolutely. Sometimes I wonder if it was all pre-planned. And the latest Huang episode doesn't sit right with me. There are rumours that the new NVIDIA management might pull out of supporting local open weights

    Thanked by 1op23
  • @DaDeveloper said:

    @op23 said:

    @DaDeveloper said:

    @op23 said:

    @noob404 said:
    At this point, is it better to get gaming GPUs or NPUs like Ryzen AI Max for local LLM inference?

    At this point you need to get lucky.

    Any hardware with capability to run something decent including strix halo is crazy expensive.
    I wish I bought the 128gb one when it was $1800 USD late 2025. Dumb mistake. Now it’s 3-4x.

    Gaming GPUs or Macs are even more expensive for same memory capacity.

    Yea I panic bought in the 4k range for the sparks now they are more expensive, 7k if you can find them cheap, 9-11k otherwise XD

    You did the right thing. Many of us didn’t.

    Wanted 4, wish I had the cash at the time XD

    Be grateful that if you want now you can sell yours at 4x prices. :lol:

  • @op23 said:

    @noob404 said:

    @op23 said:

    @noob404 said:
    At this point, is it better to get gaming GPUs or NPUs like Ryzen AI Max for local LLM inference?

    At this point you need to get lucky.

    Any hardware with capability to run something decent including strix halo is crazy expensive.
    I wish I bought the 128gb one when it was $1800 USD late 2025. Dumb mistake. Now it’s 3-4x.

    Gaming GPUs or Macs are even more expensive for same memory capacity.

    Also op23 which 27b model did you try?

    Qwen 3.6-27B

    yea the closer to full weights the better the models get, this one a few weeks ago was right up there with the best closed weights, now its fallen a bit, but damn it still is so much better than when it came out and still so much better than 4o

    Thanked by 1op23
  • op23op23 Member

    @noob404 said:

    @Jamie96 said:

    @DaDeveloper said:

    @noob404 said:

    @DaDeveloper said:

    @op23 said:

    @noob404 said:
    At this point, is it better to get gaming GPUs or NPUs like Ryzen AI Max for local LLM inference?

    At this point you need to get lucky.

    Any hardware with capability to run something decent including strix halo is crazy expensive.
    I wish I bought the 128gb one when it was $1800 USD late 2025. Dumb mistake. Now it’s 3-4x.

    Gaming GPUs or Macs are even more expensive for same memory capacity.

    Yea I panic bought in the 4k range for the sparks now they are more expensive, 7k if you can find them cheap, 9-11k otherwise XD

    Dang! Anyone has any idea if/when AI hardware prices are going back to sane?

    prob not for a good few years if ever, companies tend to be greedy like that

    Hate to agree here but this is it. As long as people cave and buy or companies too, they prove this is a viable market with these prices, so they are here to stay. They don't care if we start taking something like auto loans to buy our PCs, as long as someone is buying.

    Absolutely. Sometimes I wonder if it was all pre-planned. And the latest Huang episode doesn't sit right with me. There are rumours that the new NVIDIA management might pull out of supporting local open weights

    Time to make huggingface backups for all the good models on the storage idlers lol

  • @op23 said:

    @noob404 said:

    @op23 said:

    @noob404 said:
    At this point, is it better to get gaming GPUs or NPUs like Ryzen AI Max for local LLM inference?

    At this point you need to get lucky.

    Any hardware with capability to run something decent including strix halo is crazy expensive.
    I wish I bought the 128gb one when it was $1800 USD late 2025. Dumb mistake. Now it’s 3-4x.

    Gaming GPUs or Macs are even more expensive for same memory capacity.

    Also op23 which 27b model did you try?

    Qwen 3.6-27B

    If you can, try 3.8-35B. I also came under pressure after articles praising the 27B model. But then tried out 35B which I found to be better.

    Thanked by 1op23
  • LOL just might do that!

  • @spinel said:
    Count me in! Thanks for the Black Friday giveaway — good luck to everyone!

    Welcome. Gl to you as well

  • op23op23 Member

    @DaDeveloper said:

    @op23 said:

    @noob404 said:

    @op23 said:

    @noob404 said:
    At this point, is it better to get gaming GPUs or NPUs like Ryzen AI Max for local LLM inference?

    At this point you need to get lucky.

    Any hardware with capability to run something decent including strix halo is crazy expensive.
    I wish I bought the 128gb one when it was $1800 USD late 2025. Dumb mistake. Now it’s 3-4x.

    Gaming GPUs or Macs are even more expensive for same memory capacity.

    Also op23 which 27b model did you try?

    Qwen 3.6-27B

    yea the closer to full weights the better the models get, this one a few weeks ago was right up there with the best closed weights, now its fallen a bit, but damn it still is so much better than when it came out and still so much better than 4o

    My problem is I need to run a smaller quant which kneecaps it a lot.

    Well, may be I’ll eventually find a strix halo for ok enough price and just pull the trigger.

  • @noob404 said:

    @Jamie96 said:

    @DaDeveloper said:

    @noob404 said:

    @DaDeveloper said:

    @op23 said:

    @noob404 said:
    At this point, is it better to get gaming GPUs or NPUs like Ryzen AI Max for local LLM inference?

    At this point you need to get lucky.

    Any hardware with capability to run something decent including strix halo is crazy expensive.
    I wish I bought the 128gb one when it was $1800 USD late 2025. Dumb mistake. Now it’s 3-4x.

    Gaming GPUs or Macs are even more expensive for same memory capacity.

    Yea I panic bought in the 4k range for the sparks now they are more expensive, 7k if you can find them cheap, 9-11k otherwise XD

    Dang! Anyone has any idea if/when AI hardware prices are going back to sane?

    prob not for a good few years if ever, companies tend to be greedy like that

    Hate to agree here but this is it. As long as people cave and buy or companies too, they prove this is a viable market with these prices, so they are here to stay. They don't care if we start taking something like auto loans to buy our PCs, as long as someone is buying.

    Absolutely. Sometimes I wonder if it was all pre-planned. And the latest Huang episode doesn't sit right with me. There are rumours that the new NVIDIA management might pull out of supporting local open weights

    any direct sources on that?

  • @PuDLeZ said:

    @noob404 said:

    @PuDLeZ said:

    @noob404 said:

    @PuDLeZ said:

    @gritty said:
    will RackNerd still be hosting flash sales this bf?

    Chef always bakes some tasty treats on BF!

    Hey dude. You still here?

    yes/no, glancing at the thread every now and then :)

    Did you get the Toronto VPS giveaway? I missed it.

    Nope, I wasn't at home at that time. All good though, giveaways are nice but that's not the real reason I here. Just here to chat, work on my f5 game, and then hopefully snack on some of chef's tasty treats bf when he makes an appearance!

    Sure. I am looking forward to scoring atleast 1x VPS this time during the flash sales.

  • @op23 said:

    @noob404 said:

    @Jamie96 said:

    @DaDeveloper said:

    @noob404 said:

    @DaDeveloper said:

    @op23 said:

    @noob404 said:
    At this point, is it better to get gaming GPUs or NPUs like Ryzen AI Max for local LLM inference?

    At this point you need to get lucky.

    Any hardware with capability to run something decent including strix halo is crazy expensive.
    I wish I bought the 128gb one when it was $1800 USD late 2025. Dumb mistake. Now it’s 3-4x.

    Gaming GPUs or Macs are even more expensive for same memory capacity.

    Yea I panic bought in the 4k range for the sparks now they are more expensive, 7k if you can find them cheap, 9-11k otherwise XD

    Dang! Anyone has any idea if/when AI hardware prices are going back to sane?

    prob not for a good few years if ever, companies tend to be greedy like that

    Hate to agree here but this is it. As long as people cave and buy or companies too, they prove this is a viable market with these prices, so they are here to stay. They don't care if we start taking something like auto loans to buy our PCs, as long as someone is buying.

    Absolutely. Sometimes I wonder if it was all pre-planned. And the latest Huang episode doesn't sit right with me. There are rumours that the new NVIDIA management might pull out of supporting local open weights

    Time to make huggingface backups for all the good models on the storage idlers lol

    Absolutely. I hope someone archives it. I guess the guys at datahoarders subredditors might have already begun.

    Thanked by 1op23
  • @noob404 said:

    @PuDLeZ said:

    @noob404 said:

    @PuDLeZ said:

    @noob404 said:

    @PuDLeZ said:

    @gritty said:
    will RackNerd still be hosting flash sales this bf?

    Chef always bakes some tasty treats on BF!

    Hey dude. You still here?

    yes/no, glancing at the thread every now and then :)

    Did you get the Toronto VPS giveaway? I missed it.

    Nope, I wasn't at home at that time. All good though, giveaways are nice but that's not the real reason I here. Just here to chat, work on my f5 game, and then hopefully snack on some of chef's tasty treats bf when he makes an appearance!

    Sure. I am looking forward to scoring atleast 1x VPS this time during the flash sales.

    I got more than enough last time, still riding the credits from last years bf, XD I think i was rivaling fat32 on the sniping of the giveaways

  • @DaDeveloper said:
    LOL just might do that!

    You got storage? If yes, a good time to download atleast a few wuantisations of top models each.

  • @DaDeveloper said:

    @noob404 said:

    @Jamie96 said:

    @DaDeveloper said:

    @noob404 said:

    @DaDeveloper said:

    @op23 said:

    @noob404 said:
    At this point, is it better to get gaming GPUs or NPUs like Ryzen AI Max for local LLM inference?

    At this point you need to get lucky.

    Any hardware with capability to run something decent including strix halo is crazy expensive.
    I wish I bought the 128gb one when it was $1800 USD late 2025. Dumb mistake. Now it’s 3-4x.

    Gaming GPUs or Macs are even more expensive for same memory capacity.

    Yea I panic bought in the 4k range for the sparks now they are more expensive, 7k if you can find them cheap, 9-11k otherwise XD

    Dang! Anyone has any idea if/when AI hardware prices are going back to sane?

    prob not for a good few years if ever, companies tend to be greedy like that

    Hate to agree here but this is it. As long as people cave and buy or companies too, they prove this is a viable market with these prices, so they are here to stay. They don't care if we start taking something like auto loans to buy our PCs, as long as someone is buying.

    Absolutely. Sometimes I wonder if it was all pre-planned. And the latest Huang episode doesn't sit right with me. There are rumours that the new NVIDIA management might pull out of supporting local open weights

    any direct sources on that?

    Unfortunately no. All hearsay. But, once you connect the dots, it seems believable. Or, I am just overthinking.

  • @DaDeveloper said:

    @noob404 said:

    @PuDLeZ said:

    @noob404 said:

    @PuDLeZ said:

    @noob404 said:

    @PuDLeZ said:

    @gritty said:
    will RackNerd still be hosting flash sales this bf?

    Chef always bakes some tasty treats on BF!

    Hey dude. You still here?

    yes/no, glancing at the thread every now and then :)

    Did you get the Toronto VPS giveaway? I missed it.

    Nope, I wasn't at home at that time. All good though, giveaways are nice but that's not the real reason I here. Just here to chat, work on my f5 game, and then hopefully snack on some of chef's tasty treats bf when he makes an appearance!

    Sure. I am looking forward to scoring atleast 1x VPS this time during the flash sales.

    I got more than enough last time, still riding the credits from last years bf, XD I think i was rivaling fat32 on the sniping of the giveaways

    Yup, you went against the pro and came back victorious. This year, I am guessing the competition will be much tighter. Hope bots don't get them all.

  • Heh I said I wouldn't spend a lot of time here and now I am starting to look like a liar... Man I thought this thread was popping off more last year...

  • 26923659

    thanks for your double ;)

  • Alright guys. Gotta go. Keep the party alive. Good talking to the squad.

  • sh97sh97 Member, Host Rep

    Woot woot! Let's go!!

  • @sh97 said:
    Woot woot! Let's go!!

    Ayy! How are ya?

  • @noob404 said:
    Alright guys. Gotta go. Keep the party alive. Good talking to the squad.

    Cya you big noob that can't be found when needed :trollface:

    Thanked by 1noob404
  • LET’S GO! RACKNERD BLACK FRIDAY 2026 IS LIVE: https://www.racknerd.com/BlackFriday/ ⚡

  • BTW, @DaDeveloper @op23 have any suggestions for good scriptwriting models preferably anything under 35b?

  • OI! its @sh97 XD, you gotta take over as my talking buddy what you been up to since last year

  • @noob404 said:
    BTW, @DaDeveloper @op23 have any suggestions for good scriptwriting models preferably anything under 35b?

    I honestly have not touched the smaller models besides Qwen's

  • @DaDeveloper said:

    @noob404 said:
    Alright guys. Gotta go. Keep the party alive. Good talking to the squad.

    Cya you big noob that can't be found when needed :trollface:

    Lol! I am 404 unless the chef is here.

  • @PuDLeZ what have you been up to as well?

  • @DaDeveloper said:

    @noob404 said:
    BTW, @DaDeveloper @op23 have any suggestions for good scriptwriting models preferably anything under 35b?

    I honestly have not touched the smaller models besides Qwen's

    I have been using Qwen 35b, but I do feel it hallucinates. There's only much that prompting can do.

Sign In or Register to comment.