The whole thing is a giant mess.

Futurehome has fought efforts to crack its firmware, with CEO Øyvind Fries telling Norwegian consumer tech website Tek.no, per a Google translation, “It is regrettable that we now have to spend time and resources strengthening the security of a popular service rather than further developing functionality for the benefit of our customers.”

Fighting your own customers right after a bankruptcy is going to work out great!

The Nvidia supplier expects to double high-bandwidth memory (HBM) chip sales for the full year compared to 2024, the company said. SK Hynix overtook Samsung Electronics as the world’s top memory chip maker in the first quarter due to its leadership in HBM chips, a crucial component of AI chipsets designed by the likes of Nvidia that assist the processing of vast amounts of data to train AI models.

That’s an insane growth forecast.

f that came to pass, it would potentially spell the end of Intel’s ability to follow Moore’s Law - the observation turned must-hit target made by Intel co-founder and pre-teen explosives enthusiast Gordon Moore

Pre-teen explosives enthusiast?

To be honest, I still don’t understand how share buybacks are legal, but what do I know?

I guess we’ll see what happens. I remain unconvinced about ARM replacing x86.

Risc-V is indeed very interesting. Although the performance numbers I’ve seen require a lot more work.

I definitely have less exposure to enterprise CPUs. From the little that I have read, their MT performance (or even TCO) isn’t really as great as some of the early previews would lead one to believe…

Mind you, I am not saying ARM isn’t an excellent platform. Just look at Apple’s M and A series, but their approach also comes with its own set of tradeoffs.

I just don’t see ARM being a universal silver bullet (a straight line upgrade from x86) and with SoftBank trying to extract more cash out of ARM, things could get interesting.

Not that great?

The ones listed on our local second hand platform have no 3D capabilities (beyond monitor output), Ivy Bridge/Sandy Bridge CPUs that are slowly entering the “not good enough for even basic tasks” zone and miniscule SSDs (120 GB) by modern standards.

They are cheaper of course, but these two options don’t seem comparable.

Is that really true though?

I believe there are only two high performance cores with ARM (Apple M series and ARM’s own X series), real-world benchmarks for Oryon are shit.

In terms of smartphone SoCs, you’re stuck with Qualcomm or Apple A series.

There is honestly not that much choice even though there may be many licensees.

And with ARM building out their own SoC, you’re going to have even more challenges with openness moving forward.

I am personally not believer in the magical capabilities of ARM. It has its pros it has its cons.

I use LLMs as a complement to search and Luma is far worse than even Le Chat from Mistral for moderately complex prompts.

Luma is also notable slower (to an unacceptable level).

I would they rather they focused on existing services. I use their email services and it’s pretty good. Based on reviews, it seems that their cloud storage offering isn’t on that level.

Almost certainly, especially with AMD being relatively competitive with 9060/9070 GPUs.

clear debt and mortgage

What don’t you do? You’re like a genie.

“Make dick long and thick” not on your shitty list?

I wonder if people working in Intel’s Network and Edge group see some upside in becoming independent from Intel.

That’s pretty good win for the Samsung Foundry. Right after winning the Switch 2 SoC as well.

Hopefully this helps them compete against TSMC.

Agreed on both counts! :)

“HBF is set to revolutionize edge AI by equipping devices with memory capacity and bandwidth capabilities that will support sophisticated models running locally in real time. This advancement will unlock a new era of intelligent edge applications, fundamentally changing how and where AI inference is performed.”

This would actually be very helpful.

Although they do seem primarily focused on enterprise data centres.

If you have 10,000+ users that are a representative sample of all internet users (or pretty close), your estimates should be pretty good for popular sites.

Your sample might still be small for lemmy.ml.

I don’t think they can make a direct competitor. An LLM is expensive to run and getting a good LLM model requires lots of spend.

I like the idea of a private cloud LLM, but based on my experience with Mistral (an EU based LLM service), it is noticeably less useful than ChatGPT and especially Gemini for work use cases.

Bases on some basic test prompts (related to finding sources and documents related to specific parts of government budgets in different countries all around the world), Lumo did not perform well. Only some general suggestions for sources were provided.

It would be great to see Lumo improve, but I have my doubts.

I don’t believe they will be able to compete with Google/OpenAI in a direct battle by having a 1:1 LLM product copy but with privacy. The costs are likely too high for an organisation like Proton and their LLM is likely to have significantly subpar output.

Don’t get me wrong, I am all for a private, cloud LLM, but I would rather they came up with novel usability features, a better front-end for evaluating sources (and faster identification of errors and hallucinations) and so on.

I am not seeing any of that.

Agreed, that’s pretty fucked up.

However, on some level it’s to be expected that 3rd parties may report you if they feel you are engaging illegal activities (especially on their premises).

While I don’t support technological backdoors, there are legitimate for society to engage in surveillance. It’s the responsibility of voters to make sure that this is done in a responsible and transparent manner.