this post was submitted on 07 Sep 2023
157 points (100.0% liked)
technology
23293 readers
53 users here now
On the road to fully automated luxury gay space communism.
Spreading Linux propaganda since 2020
- Ways to run Microsoft/Adobe and more on Linux
- The Ultimate FOSS Guide For Android
- Great libre software on Windows
- Hey you, the lib still using Chrome. Read this post!
Rules:
- 1. Obviously abide by the sitewide code of conduct. Bigotry will be met with an immediate ban
- 2. This community is about technology. Offtopic is permitted as long as it is kept in the comment sections
- 3. Although this is not /c/libre, FOSS related posting is tolerated, and even welcome in the case of effort posts
- 4. We believe technology should be liberating. As such, avoid promoting proprietary and/or bourgeois technology
- 5. Explanatory posts to correct the potential mistakes a comrade made in a post of their own are allowed, as long as they remain respectful
- 6. No crypto (Bitcoin, NFT, etc.) speculation, unless it is purely informative and not too cringe
- 7. Absolutely no tech bro shit. If you have a good opinion of Silicon Valley billionaires please manifest yourself so we can ban you.
founded 4 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
Kirin 9000S is equivalent to the Snapdragon 865, GPU is equivalent to 888. Only mobile chip to support multi-threading, probably because it's a modified server chip.
Really impressive. It's faster than the Google Tensor G2 when power consumption is equalized. Tensor sucks but still the best that Google has to offer despite having no sanctions to deal with.
This is way better than the earlier domestic produced Chinese silicon I've seen (full sized desktop GPUs). The domestic GPUs seemed really bad so this a huge leap in the last 6 months. The GPU in this phone is probably better than the full size GPU I saw lol.
On a side note I want to know how well China's AI chips are doing. Nvidia seems to have an impossible lead right now over AMD, AI model training seems like a huge priority for national security and tech so it would be very interesting to see how they're doing compared to consumer chips.
Lmao it's cool af they put a server CPU into a phone
Modern ARM processors are nuts. My three year old phone has a GPU, in raw tflop numbers at least (I know) more powerful than an Xbox One. 1.4 tflops
The overclocked Adreno 740 in the Samsung Galaxy S23 has the equivalent tflops to an Xbox Series S(3.8 I think).
This isn't even mentioning Apple's ARM laptop chips. M1 macs and that.
And I haven't even mentioned CPUs.
Obviously comparing tflops across different architectures is more of a fun exercise than a meaningful comparison, but it's still very interesting to see.
haha it's cool to know these exponential leaps are being wasted what is, in terms of general use, an MP3 player glued to a flip-phone
ARM is the future for sure, the west will be seething. Nintendo is using an UNDERCLOCKED Nvidia tablet from 2014 as their main console and getting amazing results from it. As more developers move towards ARM platforms because of the apple jump, we're gonna start seeing X86 fall.
I think you mean hyperthreading. Multi threading has been a thing on ARM mobile chips forever. It being based on server architecture has advantages like that, but also disadvantages in terms of thermals and efficiency at higher clock speeds. (It only clocks up to 2.6GHz, and energy efficiency and thermals really suffer at that clock speed). It has the largest vapour chamber I've seen on a phone, the size of the entire screen. Single core performance being right in between the 865 and 888 is still highly impressive. Less efficient than the 865, which was made on the older TSMC process, but more efficient than the 888 and 8 gen 1, which was on Samsungs process. The 8+ gen 1 and 8 gen 2 blow it out if the water on efficiency and raw power with the new TSMC process, but that's to be expected.
ARM GPUs are a whole different ballgame compared to desktop GPUs. Raw numbers and benchmarks aren't always the best to compare performance because of different levels of driver support for certain features. There's the Mali approach, which treats the GPU and CPU as an APU with its approach, and has poor driver implementation of features such as dual source blending. Then you get Adreno, which goes for a more traditional approach in having the GPU and CPU more seperate, and has much better implementation on the driver side of things. This is why it's always recommended to go for an Adreno GPU in GPU bound tasks on Android, like high performance game emulation (think PS2 games at 1080p or higher resolution) or maxing out genshin impact. I think Apple ARM GPUs are still based off of what PowerVR did back in the day, and PowerVR made very good ARM GPUs. Such as the ones in the early iPads, and the Samsung Galaxy S4 international model.
It's positive that Huawei have gone for their own GPU implementation, Mali keeps dropping the ball, and Adreno needs competition. As for how it works in intensive real world tasks, only time will tell.
The less said about tensor, the better. It's not even a Google design, it's years old Samsung designs that even Samsung ditched. Purely a stop gap measure until Google can actually make their own chips.
Huawei's Ascend AI processors are actually quite good. They were arguably world leading before the sanctions. There are lots of others but so far they are a big mediocre group without real standouts.
So what's an AI chip? Is it sort of like how graphics chips are specially made for graphing calculations? But for running, like, several tables of nodes?