RISC-V: Free Core, Some Assembly Required | Hacker News Reader

RISC-V: Free Core, Some Assembly Required

eetimes.com

58 pointsnkurz21 comments

It will be a long time yet, but in my lifetime I want to see a completely open source, open hardware laptop that is competitive in speed with closed source alternatives, and it will hopefully be a good deal safer.

Security Researcher Joanna Rutkowska just released a concept for a stateless laptop, http://blog.invisiblethings.org/2015/12/23/state_harmful.htm....

A fully Trusted computing base all and a secure architecture for application in the user level.

Lets make the future happen.

I didn't know a board was on the way with 28nm chips. That's exciting. The only thing that threw me is that Grey could fit 400 RISC-V cores on one FPGA. Makes me wonder what type of core and FPGA he was using...

I believe Jan Gray was only implementing RV32I, but he was manually laying out the core and the router himself!

You can find his slides here (http://riscv.org/workshop-jan2016.html).

Thanks for the link!

re cores

They don't appear to have floating point. That a pre-requisite for quite a few accelerator cases. Also could help explain how many he fits in there at what performance.

re FPGA

BOOM! I was thinking it was an UltraScale or Achronix given the numbers. So, that was a hit.

re NOC

Probably the most exciting thing there. I'm guessing he's going to license it. Might be able to use it in open projects or even ASIC's depending. I lack the expertise to speak on the latter except to say that the existence of EDA companies specializing in NOC generation means such a thing might be useful.

Also, good to know we can hit this kind of performance on an UltraScale. Makes me more hopeful about what can be achieved with an open 45nm or 28nm FPGA once we get them. Archipelago is progress in that area at 45-65nm already.

That's the 32 bit integer subset, floating point is available.

Plenty of stuff can get away with integer-only. Massive numbers of Xilinx DSP48 on a FPGA is pretty common.

"Ports of a handful of Linux variants including FreeBSD are well underway..."

When did FreeBSD become a Linux variant?

> When did FreeBSD become a Linux variant?

I would imagine my answer is the same as yours: never.

Just add the phrase to your indicators of poor and non-fact checked journalism. It helps tell if the author, article, and publication have any real value.

I do hope whoever puts out the first RISC-V powered motherboard makes it in a standard size that fits in existing cases. I hope it very much the same as an Intel motherboard with a RISC-V instead of an Intel CPU. Please don't make the funny development boards that ARM and MIPS vendors are making. Give us something we can put in a case, buy some memory and hard drives / SSDs, and get hacking on.

Development boards are not meant to be "funny", they're meant to expose all the features of a given CPU (or, more commonly SoC).

This is for embedded users, who typically add the processor chip to their own products, i.e. they are not intended to be end-user personal computers for general purpose use.

I know they are for embedded systems and sorry if "funny" offended you, but I keep hearing these companies tell me how good their ARM, MIPS, POWER are going to be in my server then they ship embedded specific boards. I can go buy a Intel or AMD board for a PC and write / port software. I'm still waiting for the same from other manufactures. This is one reason I'm really not sold on the ARM server boards since every company makes it a pain in the rear to buy one.

The Nvidia Jetson TX1 comes in mini-itx. Would go great in a nice case like the m350.

Is the memory expandable on that board?

Nope, the memory is soldered onto the board.

I mean, ATX (or other standards like MiniITX) boards are still really nice for initial embedded development. Even past that, it's really nice for a test farm for CI of embedded code. The weird boards are just them cheaping out. When they pass the savings onto you ala the Raspberry PI it's great; when they don't ala most FPGA dev boards, it really sucks.

That's why ARM has the Intergrator/Versatile series of dev boards. Back in the day, MIPS had the Maltas.

Even so, why couldn't they be made in some standard form factor? It seems like that would be easier, even for those who want to use the board for their intended use.

Can this be used in production? To make NSA-free servers?

Eventually, in a few years, once hardware and software is actually available… and assuming neither the NSA nor anyone else introduces hardware backdoors in the factory. Have fun looking for sabotaged transistors in a 28 or 14nm chip.

At the moment we have so many other problems that security in the production is barly on the radar. If we are at a point were production is the major secuirty problem, we are in a good place.

You have to use a secure lifecycle for the ASIC development and distribution. That requires a lot of work and money. So, no, these chips won't be NSA free. NSA will own them if they want to with a variety of attack points.

[he] sported a beard that is a sign he is well along in working on his next-generation Epiphany processor.

Thats where I'm going wrong..

More seriously, a RISC-V Epiphany would be quite cool, and potentially quite a bit more mainstreamish than the current one.

Reply on news.ycombinator.com