I have a ryzen apu, so I was curious. I tried yesterday to fiddle with it, and managed to up the “vram” to 16gb. But installing xformers and flash-attention for LLM support on igpus is not officially supported and was not possible to install anything past pytorch. It’s step further for sure, but still needs lots of work.
Awesome recap. Telegram dev channels and xda used to be a place to spread the message so that users would pressure the company to fulfill their gpl duty.
On a side note Fuck Asus for screwing bootloader unlocking for the zenfone.
Linux, thanks to steam, is better at gaming than windows, esecially for older games. Proton ftw
Is titback still functional? Had many issues on A13
Another reason to root. Then Neobackup.
I don’t understand why linux doesn’t get all the love patches like Windows.
ROCm? Is that even supported now? Last time I checked it was still a dumpster fire. What are the RAM and VRAM reqs for the Mixtral8x7b?
I stumbled upon a website through DDG, and after a long intro, the main section supposedly where the thing I was searching for had “Sorry I can’t fulfill your request right now”. Basically a fully generated page to match my search with some parasitic seo tactics. The web be chaging. Front page of DDG.
Confuse the consumer and make it harder to make informed decisions. “Just trust our recommendations, choom”.
But basing a recommendation on a ballpark anectdotal evidence is eidiculous.
Doing the maths, that’s ~200$/chip. Even Nvidia 4060 is more expensive.
Holy shit, that’s the rookiest mistake.
Actually there is a russian guy who made that happen. Youtube it.
But it’s not deterministic.
Interesting read, basically the demonstrated that Gpt_4 can understand causality, using random graphs. Interesting take-away though is this excerpt:
“And indeed, as the math predicts, GPT-4’s performance far outshines that of its smaller predecessor, GPT-3.5 — to an extent that spooked Arora. “It’s probably not just me,” he said. “Many people found it a little bit eerie how much GPT-4 was better than GPT-3.5, and that happened within a year.”
But idle still would run much more than 15w. There a very good compilation google sheets for the most efficient X86 cpus, but once you start factoring hdds and ssds, it’s only natural to go higher (20w-30w) at least. That’s at least double than rpis
The mian issue with Mini/used PCs is the power efficiency. It’s just a waste of wattage and performanve/Watt is very bad, especially at idle.
Link for the lurkers https://github.com/evilsocket/pwnagotchi
Pwnagotchi is an A2C-based “AI” leveraging bettercap that learns from its surrounding WiFi environment to maximize the crackable WPA key material it captures (either passively, or by performing authentication and association attacks). This material is collected as PCAP files containing any form of handshake supported by hashcat, including PMKIDs, full and half WPA handshakes.
The landscape is changing so fast thanks to LLMs, everything is becoming gated behind logins. Thanks ChatGPT.
It had lots of bugs and crashes back last year, but considerably improved. Flawless backups/restore, A13. It’s almost a replacement for titanium now.