Most active commenters
  • segmondy(3)
  • Workaccount2(3)

←back to thread

Alibaba Cloud says it cut Nvidia AI GPU use by 82% with new pooling system

(www.tomshardware.com)
521 points hd4 | 12 comments | 20 Oct 25 12:31 UTC | HN request time: 0.939s | source | bottom
Show context
hunglee2 ◴[20 Oct 25 12:59 UTC] No.45643396[source]▶
The US attempt to slow down China's technological development succeeds on the basis of preventing China from directly following the same path, but may backfire in the sense it forces innovation by China in a different direction. The overall outcome for us all may be increase efficiency as a result of this forced innovation, especially if Chinese companies continue to open source their advances, so we may in the end have reason to thank the US for their civilisational gate keeping
replies(17): >>45643584 #>>45643614 #>>45643618 #>>45643770 #>>45643876 #>>45644337 #>>45644641 #>>45644671 #>>45644907 #>>45645384 #>>45645721 #>>45646056 #>>45646138 #>>45648814 #>>45651479 #>>45651810 #>>45663019 #
1. segmondy ◴[20 Oct 25 13:23 UTC] No.45643618[source]▶
may backfire? it's a bit too late for that.

go to 2024, western labs were crushing it.

it's now 2025, and from china, we have deepseek, qwen, kimi, glm, ernie and many more capable models keeping up with western labs. there are actually now more chinese labs releasing sota models than western labs.

replies(4): >>45643764 #>>45646364 #>>45650725 #>>45650819 #
2. hunglee2 ◴[20 Oct 25 13:37 UTC] No.45643764[source]▶
too early to call a winner, though it is disappointing to see US withdrawal from open source. Still the main outcome of open source is distribution / diffusion of the idea, so it will inevitably mean US open source will come back, hopefully via some grass roots maniac, there will be a Linus-like character emerge at some point
replies(2): >>45643897 #>>45644425 #
3. segmondy ◴[20 Oct 25 13:48 UTC] No.45643897[source]▶
i'm not calling a winner, i'm just saying that the chinese have caught up despite the embargo. google, openai & anthrophic have phenomenal models. i stopped using openai & anthropic after they called for open weight/source regulation. i use google because they offer gemma and i got a year gemini-pro subscription for free, use openai gpt-oss-120b since i can run it at home, and the only model i currently pay for is a chinese model.
4. mixologist ◴[20 Oct 25 14:39 UTC] No.45644425[source]▶
user growth has slowed. the technology that should help users is only being pushed from the top, while users refuse to use it. openai pivoted to porn.

does it really feel like they have a chance to recover all the expenses in the future?

crypto grifters pivoted to ai and, same as last time, normal people don’t want to have anything to do with them.

considering the amount of money burned on this garbage, i think we can at least declare a looser.

5. Workaccount2 ◴[20 Oct 25 17:10 UTC] No.45646364[source]▶
But they aren't keeping up

They are lauded for the ability to cost ratio, or their ability to parameter ratio, but virtually everyone using LLMs for productive work are using ChatGPT/Gemini/Claude.

They are kind of like Huffy bicycles. Good value, work well, but if you go to any serious event, no one will be riding one.

replies(2): >>45646880 #>>45647584 #
6. segmondy ◴[20 Oct 25 17:49 UTC] No.45646880[source]▶
they are keeping up. i have been using just chinese models for the last 2 years. chatgpt/gemini/claude have marketing. there's nothing that you can do with those models that can't be done with deepseek, glm or kimi. if there is, do let us know.
replies(1): >>45649214 #
7. MSFT_Edging ◴[20 Oct 25 18:45 UTC] No.45647584[source]▶
The downside of their efficiency and cost-ratio is that they undermine the circular economy of massive data centers, GPU sales, and VC money that is constructing an extremely wasteful bubble.
replies(1): >>45649237 #
8. Workaccount2 ◴[20 Oct 25 20:53 UTC] No.45649214{3}[source]▶
They can't attract a large contingent of users. Because despite being able to do everything the big name models can do, they cannot do it as well.

This aligns with the benchmarks as well; they benchmark great for what they are, but still bottom of the barrel when competing for "state of the art."

And yes, it's great you daily Chinese models, but the vast majority of people try them, say "impressive", then go back to the most performant models.

replies(1): >>45653431 #
9. Workaccount2 ◴[20 Oct 25 20:54 UTC] No.45649237{3}[source]▶
The bubble is there in China too, it's just on the governments books instead of private investors books.
10. rasz ◴[20 Oct 25 23:31 UTC] No.45650725[source]▶
Have you tried using those models? qwen for example cant even do something as basic as clustering analysis on a list of integers, hell it goes off the rails when just reading said integers from a file - starts babbling about determining number of digits, indexes, tries concatenating numbers together into one big string, no idea wtf is going on with that model.
11. NSPG911 ◴[20 Oct 25 23:44 UTC] No.45650819[source]▶
way too early to say that

while qwen, deepseek and kimi are opensourced and good, they are preferred because of their insane token ratio, they use a lot less for more, but a by product is that they are less accurate it is amazing progress by the chinese companies, but they definitely can improve a lot more

12. vachina ◴[21 Oct 25 07:43 UTC] No.45653431{4}[source]▶
I'm not sure if you understood what OP meant by "marketing".