Alibaba Crafts A 16-Core RISC-V Chip @ 2.5GHz

[starshipeleven]

I don't even post on Arch forums. I only care about their wiki. Try guessing my homeland again.

The Arch forums.

[schmidtbag]

I really don’t understand this point of view. How is a chip developed for internal use a negative for the overall user community? If anything, with something like RISC-V, you want to see as much development as possible take place. In the end without enough hardware to drive compiler development the architecture isnt going anywhere.

It's negative if we Linux users want to see anything useful for ourselves. Companies making their own downstream variants aren't going to help others.

Isn’t that the whole point of RISC-V anyways? At least the way I understand it one big target area for the initial designs is the embedded world. That means chips with RISC-V designed in to support special function hardware.

Yup pretty much. But that brings me back to my anarchy analogy: as much as it's nice to be able to do what is necessary without any restrictions, you are equally as vulnerable to those who have selfish motives.

Right so if a company builds a variant like this we are seeing RISC-V quickly moving into more advanced solutions. This is a very positive thing if the resultant hardware proves to be effective.

That's only true if they intend their variants to be adaptable and modular.

Mic you are looking for a workstation solution nothing RISC-V based and suitable has shown up yet. Nothing bad there really. Think about it it took ARM decades to get to the point where they are almost workstation capable. Frankly it looks like Apple was a big driver in getting them there. Even after all of those years we are still waiting on viable ARM based laptops and desktops So I wouldn’t expect a viable workstation solution from the RISC-V camp anytime soon. That even if a team is working on such a chip right now.

I'm not looking for anything. I don't really care what happens with RISC-V. I agree we're not going to see a workstation solution for a while.
As for ARM, that's slightly different. ARM CPUs were utter crap for desktop use up until ARMv7hf came out. And even then, they shot themselves in the food with their poor kernel support (particularly with their Mali GPUs, which were also at a disadvantage for only supporting GLES). It took them decades because all they prioritized were things like routers and ATMs. There are several ARM laptops, it's just that most of them are running Chrome OS. There isn't enough of a market to make Windows-based ARM laptops due to poor binary compatibility. As for desktops, I just don't see the point.

The problem here here is that you can’t let the reality of long lags in development stop you from taking an interferes in RISC-V. RISC-V will never get anywhere if people don’t take an interest in the beginning.

I won't take an interest in RISC-V because it doesn't fulfill any of my needs. Every other existing architecture does something I want/need. Maybe not in the most preferential way, but not in an especially detrimental way either. RISC-V is not immune to the problems I have from other architectures.

[dungeon]

There isn't enough of a market to make Windows-based ARM laptops due to poor binary compatibility. As for desktops, I just don't see the point.

There is this Lenovo Yoga C630 running Windows 10:

https://www.lenovo.com/us/en/laptops.../p/88YGC601090

To be honest i heard of it and some else hardware during DebConf's arm ports bof:

http://gensho.ftp.acc.umu.se/pub/deb...ports-bof.webm

They say it runs Gnome shell on Ubuntu it seems and because of big vendor starts doing it... they were suggesting Debian should start supporting ARM laptops and desktops.

Another one have $99 PineBook or whatever and some server mentioned for $2.5K, etc...

Or this recent PineBook Pro for preorder

https://www.pine64.org/2019/07/05/ju...-pinebook-pro/

I mean there are some products for interesed if you search better

[schmidtbag]

I mean there are some products for interesed if you search better

I said enough of a market, implying there is still some market. Those handful of devices (plus a few others, that you actually forgot) are not a big market.

[dungeon]

I said enough of a market, implying there is still some market. Those handful of devices (plus a few others, that you actually forgot) are not a big market.

Of course probably not a big, i don't have any numbers how much they sell these things around.

Who knows, maybe it is big enough per particular market segment When i heard that Lenovo, Asus, Samsung... also have these laptops with ARM, that does not sound to me like very tiny market anymore.

[schmidtbag]

Of course probably not a big, i don't have any numbers how much they sell these things around.

Who knows, maybe it is big enough per particular market segment When i heard that Lenovo, Asus, Samsung... also have these laptops with ARM, that does not sound to me like very tiny market anymore.

That's what MS wants you to think. It's their incentive, so they probably contracted a bunch of different brands (including HP, to add to your list) to make it seem like the market is much bigger than it really is. But these laptops don't seem to be selling all that well, because they're slow at running x86 programs and most people who use them don't understand why or what to do about that. Add to the fact they're not much cheaper than x86 alternatives and people just simply look elsewhere. The long battery life and SIM card slot are the only distinct selling points for Windows users.

Personally, I'm hugely interested in them, but I haven't bought one since their Linux compatibility is very lacking.

[dungeon]

That's what MS wants you to think. It's their incentive

Of course, since all of these recommends Windows

(including HP, to add to your list)

Of course and Huawei too

https://www.anandtech.com/show/14202...snapdragon-850

Whatever, don't you think all these Snapdragon laptops from various vendors should get desktop support on Linux?

[schmidtbag]

Whatever, don't you think all these Snapdragon laptops from various vendors should get desktop support on Linux?

Absolutely - I'd have bought one of these laptops a year ago if they did.

[dungeon]

Absolutely - I'd have bought one of these laptops a year ago if they did.

Well that is WIP still... but according to this, basics should work since february... currently sans WiFi sans accelerated graphics

https://github.com/aarch64-laptops/build

GPU require fw blobies of course

https://github.com/aarch64-laptops/b...a4d042f1df2c7b

[audir8]

... I won't take an interest in RISC-V because it doesn't fulfill any of my needs. Every other existing architecture does something I want/need. Maybe not in the most preferential way, but not in an especially detrimental way either. RISC-V is not immune to the problems I have from other architectures.

The SiFive unleashed dev board with the expansion board (PCIe x4) will work with an AMD GPU and open source drivers. This is about the most open system you can build yourself currently. From the demo I saw, the software was of beta quality, but getting better fast with performance being similar to ARM. Competing with licensed ARM in the embedded space, then doing it in the mobile and desktop spaces are all vastly different objectives with different requirements. I haven't seen any indication that anyone is focusing on the mobile or desktop RISC-V for the consumer space. I want to support RISC-V more, but it's hard to do without hardware running Linux priced for the consumer market: something like the RPi or the RPi clones running Android. This is a big ask, but I think it's possible.

Software and hardware patents are not a tool that fosters innovation, they're a way to maintain monopolies (Qualcomm, Intel) or to use defensively/offensively (Oracle, Facebook's React) when needed. Fair patent licensing doesn't exist in software or hardware. It's also fair to say open source wouldn't be where it is without the ability to get around patents in creative ways like almost all GNU utilities and AV1. Copy-left for patents is: you can given everyone a license, but at that point, why file for a patent anyway?

wizard69 coder 's comments on GPL/patents

There can be a balance between using GPL and non-GPL licenses, just add GPL exclusions and modular software design as needed I think. This is in-effect how the GPL works currently: people/companies write wrappers, whole new APIs to unmodified GPL code, and wall off source/binaries as needed. Maybe maintainers should be thinking about or doing some of this in ways they think are reasonable in the first place.

Adding proprietary SIMD instructions to a RISC-V probably has already been done. If you can't do it with a GPL RISC-V, it means either there aren't enough GPL exclusions or RISC-V needs to be made a bit more modular. The GPL does require this additional work, but then also lets you require reciprocity for the parts of the project you want to keep open. It's all more than fair I think.