China recently moved that direction. That would be nice collaboration to see between EU and China.

China to publish policy to boost RISC-V chip use nationwide, sources say https://www.reuters.com/technology/china-publish-policy-boos...

If we want this to go anywhere, not just super computing, the first step is to get devices, useful devices, in the hands of enthusiast. That means funding projects similar to the Raspberry Pi, but for RISC-V, and perhaps mini-itx boards.

We need these cheap-ish computers in the hands of people who will port software to the platform. Without a good selection of ready to go software, the hardware is pretty irrelevant.

It rather should be for general computing too starting with government office computers.

Yeah they'll be slow but nothing can be slower than an x86 loaded with a Windows 11 or something on it.

There's nothing to celebrate here. This is another sad moment for Europeans everywhere.

> The first phase of this six-year endeavor is backed by €240 million (£200 million, $260 million) in funding.

For this to be a serious effort it would take another two zeros at the end of that number. This is 100x too small.

In 6 years, we'll have spent a pittance, to realize that we got basically nothing for it, and we're even further behind the US whose companies are spending tens of billions to develop new accelerators.

Let's take one US company at random, Groq, they've raised 10x this amount of money. That's one startup. Never mind Cerebras, SambaNova, Tenstorrent, etc. How is this effort going to compete? And they're giving the money to "38 leading partners" instead of one focused entity. It won't compete. It's just a waste.

The EU is still thinking too small. In an era where the US is no longer a reliable partner (maybe even a rival), and where Taiwan could disappear overnight, this is extremely stupid and dangerous.

I don't understand why the EU can't get serious about tech. Why does every investment need to be peanuts? Why can't we pay people well so they don't all leave to the US/Canada? Why can't we seriously invest in startups?

Here are some relevant slides from hpcasia25: https://github.com/RISCVtestbed/riscvtestbed.github.io/blob/...

I also found this report on their FPGA Emulation Platform: https://www.riser-project.eu/wp-content/uploads/2024/11/RISE...

So from these resources it seems like they develop a vector processor with Semidynamics out-of-order Atrevido core as a scalar core and their Vitruvius VPU.

There is a paper about a previous iteration of the VPU: https://dl.acm.org/doi/10.1145/3575861

In the more recent report they have a vector length of 16,384 bits, with 16 lanes (8 in FPGA, 16 in the diagram, final version could be more), so total of 16*64=1024 bits of ALUs.

Slide 15 seems to indicate that they want to create a chip with 32 of those cores, a shared L3 cache, and access to HBM.

Naturally the next step is to also rely in OSes and programming languages not controlled by export regulations.

There’s a certain irony here as ARM is 100% British European.

When I read

> Europe bets on supercomputing sovereignty

I'm laughing and dying inside. Europe has forfeited all possibilities on creating their own chips, partially because of production regime, partially because we were never good in this subject. At the moment the war is already lost (yes, I consider any negotiation for resources a war, whether in law creation or in movement of forces). Therefore, we're condemned to rely on China's supplies, and chips supplied by China will have this

https://www.techspot.com/news/107073-researchers-uncover-hid...

which defies idea of sovereignty at all.

> Axelera's chips follow a similar formula as other AI ASICs, such as Google's tensor processing units. The Dutch outfit's current silicon feature four accelerator cores, each with a matrix multiply-accumulate (MAC) unit, a RISC-V control core to make the accelerator programmable, and some digital signal processors which handle neural network activation functions.

This seems very focused on current architecture, which could be replaced with something more novel without the fundamental limits of matrix multiplication.

RISC-V is interop as the ISA level. No wonder EU and even china are moving towards this US standard.

Given the stunningly low performance of risc-v chips that make a raspberry Pi look fast, I'm wondering how soon this is supposed to pay off.

Everybody who "bets on RISC-V for supercomputing sovereignty" will probably end up buying Chinese anyway.

RISC architecture is going to change everything

Surely arm is a better choice?