Arm Talks Up Their Open-Source Contributions, Adding Support For Panfrost

[schmidtbag]

And AMD is chopped liver or a dead company that doesnt exist and never contributed anything....

To be fair, Intel has done far more for the Linux kernel than AMD. Even where AMD contributes, it's often a lot later than it should be. Take Ryzen temperature sensor drivers - those took a surprisingly long time to work correctly.
Where AMD deserves recognition is with Mesa.


As for RISC-V... it's probably going to be another decade before that starts to be of a real threat to ARM. Wake me up when we get RISC-V boards with better capabilities and/or a better price than Nano Pi.

[ayumu]

I was just going to say...

RISC-V is coming for you guys with Jim Keller* at the helm of at least one large project.

* Chief architect of the Athlon 64 and Ryzen processors.

You're definitely talking about Ascalon. Yet there are two problems there.

1. Jim Keller is very accomplished, but he was not chief architect of Ryzen. That's Mike Clark.
2. The Chief Architect of Ascalon is Wei-han Lien, who was lead architect of M1 at Apple.

It helps your case when you do the research and get your points right: Jim Keller is CEO at Tenstorrent.

The conclusion is still the same: Tenstorrent is going to surprise the world (which isn't paying attention) by launching a very competitive RISC-V processor: Ascalon.

[ayumu]

As for RISC-V... it's probably going to be another decade before that starts to be of a real threat to ARM.

That was five years ago already, as proven by ARM themselves.

https://www.phoronix.com/news/ARM-RISC-V-Facts

[schmidtbag]

That was five years ago already, as proven by ARM themselves.

https://www.phoronix.com/news/ARM-RISC-V-Facts

Indeed, though to their credit, RISC-V has gained some traction since then. Not a lot, but it isn't the vaporware that it was back then.

[ayumu]

Indeed, though to their credit, RISC-V has gained some traction since then. Not a lot, but it isn't the vaporware that it was back then.

With RISC-V shipping 10b+ cores in 2022, it could be said RISC-V already has momentum.

There's very high performance architectures coming from not just Tenstorrent (which Ascalon is presented as competitive with projected Zen5), but SiFive, Rivos, Ventana, MIPS, as well as others. They can't all fail at the same time. This is happening, and it is unavoidable.

Some of these (Tenstorrent Ascalon and Ventana Veyron) have announced nearby release date targets (2024 and before 2023 end, respectively).

It'll be fun to see what happens once these very high performance implementations begin to hit the market.

[NeoMorpheus]

To be fair, Intel has done far more for the Linux kernel than AMD. Even where AMD contributes, it's often a lot later than it should be. Take Ryzen temperature sensor drivers - those took a surprisingly long time to work correctly.
Where AMD deserves recognition is with Mesa.


As for RISC-V... it's probably going to be another decade before that starts to be of a real threat to ARM. Wake me up when we get RISC-V boards with better capabilities and/or a better price than Nano Pi.

Still the point stands and the omission of AMD is wrong.

[wswartzendruber]

You're definitely talking about Ascalon. Yet there are two problems there.

1. Jim Keller is very accomplished, but he was not chief architect of Ryzen. That's Mike Clark.
2. The Chief Architect of Ascalon is Wei-han Lien, who was lead architect of M1 at Apple.

It helps your case when you do the research and get your points right: Jim Keller is CEO at Tenstorrent.

The conclusion is still the same: Tenstorrent is going to surprise the world (which isn't paying attention) by launching a very competitive RISC-V processor: Ascalon.

Now that I think about it, you're right. Jim Keller was the Chief Architect of the entire Zen, Zen+, and Zen 2 line-up for all form-factors, not just Ryzen.

[discordian]

Scared about Risc-V?

Not quite for a while. Trying to steer against Qualcomm Oryon more likely.

[uid313]

With RISC-V shipping 10b+ cores in 2022, it could be said RISC-V already has momentum.

There's very high performance architectures coming from not just Tenstorrent (which Ascalon is presented as competitive with projected Zen5), but SiFive, Rivos, Ventana, MIPS, as well as others. They can't all fail at the same time. This is happening, and it is unavoidable.

Some of these (Tenstorrent Ascalon and Ventana Veyron) have announced nearby release date targets (2024 and before 2023 end, respectively).

It'll be fun to see what happens once these very high performance implementations begin to hit the market.

Tenstorrent isn't going to put a high-performance RISC-V CPU in your hands. Not for your desktop, your laptop, your tablet, or your phone. They're focused on RISC-V based AI and ML accelerators to be sold to big tech companies like Facebook and Amazon so they can apply AI to better track you and analyze harvested user data.

SiFive so have only made development boards that costs more than ten times more than a Raspberry Pi and still manage to perform less than half as good at best.

Rivos same as Tenstorrent, they wont make any consumer CPU, they just making AI and ML stuff for big tech and Fortune 500 companies.

Ventana, no idea what company that, but I doubt they making anything for consumers. Will probably get bought up by some other company.

MIPS haven't been relevant for decades. I think they only have old customers who haven't bothered to change, no new customers. I think its a company ran by old people, they failed to adopt to change, and got outran and their lunch eaten by ARM. ARM was more open, while MIPS was closed and they kept running the company like it was still in the 80s. They couldn't embrace change or openness, or didn't see the world was changing or was scared of the change.

None of the companies buying RISC-V CPUs and microcontrollers have any interest at all to bring any consumer desktop, laptop, tablet or phone to the market. None of them have any interest in bringing in ideas such as freedom or openness, or to bring an open architecture without binary blobs to consumers. Their only interest in RISC-V is as an cheaper, low-cost alternative to ARM where they don't have pay any royalties so they can increase revenue by cutting costs in the supply chain.

[okias]

Android is the market with the most ArmSIPs, but Lima and Panfrost have absolutely nothing to do with Android, ...

I must respectfully disagree, as the objective of Mesa3D also includes providing support for Android builds. For instance, a few months ago, we began testing of Android containers with git Mesa in our CI pipeline. Although currently, Mesa is only utilized in specific use cases on Android, the ultimate goal is to extend its usage to mass-produced Android devices in the future.