News Linus Torvalds now favors Arm-powered Ampere chip over Apple Silicon Mac for building Linux kernels — says the more powerful system is why he's doi...

[Admin]

Linux creator Linus Torvalds is still using his AMD workstation, but splitting his time between that and Arm64 systems for builds and testing.

Linus Torvalds now favors Arm-powered Ampere chip over Apple Silicon Mac for building Linux kernels — says the more powerful system is why he's doi... : Read more

 

[ezst036]

Has an Ampere Altra ever been benchmarked directly counter ---> Apple ARM? Most likely the M2? (for timing/release date reasons)

Having so many many more cores, I can see why that sort of parallelism would be useful when doing kernel compiling. I shudder to even contemplate how much $$ Apple would charge for a chip with 128 CPU cores in it. Yikes!

 

[DS426]

Linus is surprisingly eccentric for the person that he his (or what he has established and accomplished, rather). Even when Apple Silicon performance was like the holy grain of ARM performance, getting an Apple device was truly a spit-in-the-face for the ENTIRE Linux and even broader FOSS community. Sorry, but making a deal with the devil is... just that, lol. Sure, basically the perfect ARM-based mobile platform from a raw specs/performance standpoint (and I mean including battery life, being lightweight and thin, etc.), but what did they really net him and the grander Linux community?? Really, serious question!

Fortunately, nVidia wasn't allowed to acquire ARM and although they would have made some beastly aarch64 CPU's, the free world was spared some time. And while seeing ARM rise up against the x86 monarchy is great, I'd love to see more RISC-V adoption outside of China. As for Apple... I don't think they'll see that large of a perf gain as they did, with M3, M4, etc. just being incremental gains and requiring that next "quantum leap" in technology to really move the needly. Even if they do, they are still Apple a.k.a. f the rest of the computing ecosystem and those that believe in a technology world that involves more than one freaking vendor/manufacturer. Sounding like a hater but I really can overstate the danger of these monopolies and oligopolies, ESPECIALLY when much to most of the vertical integrations are owned by the same company. It's not even academic; we know what the robber-barons, the Rockefellers and Carnigies and JP Morgans did so long ago and how long it took to break up those empires. FFS, the U.S. federal government is finally tired of Apple's control and therefore is finally ligitating against them. Microsoft needs it as well, to be clear...

 

[jkflipflop98]

LOL Linus. A loser writing crap code for a cult.

 

[bit_user]

Has an Ampere Altra ever been benchmarked directly counter ---> Apple ARM? Most likely the M2? (for timing/release date reasons)

Altra is based on a server version of the Cortex-A76 cores that now grace the Raspberry Pi 5. It's a 4-way issue core, whereas I think M2's P-cores are 8-way (if not more). Also, the M-series are cache-heavy, while Altra has less cache per core than its x86 contemporaries. Considering Apple's other optimizations, it's pretty safe to say the M2 is easily twice as fast, core-for-core.

Where the Altra really wins is on simple core-count. Also, the amount of memory bandwidth it managed to squeeze from 8x DDR4-3200 was pretty impressive.

I guess what surprises me most is that this didn't happen like 2 years ago. You could already buy Altra workstations, back then, and they weren't teetering precariously on the brink of obsolescence, like they are now.

 

[bit_user]

Even when Apple Silicon performance was like the holy grain of ARM performance, getting an Apple device was truly a spit-in-the-face for the ENTIRE Linux and even broader FOSS community.

I feel weird for defending him on this point, because I agree with you in that I wouldn't buy an Apple device purely on principle. But, Linus got it because there's an unofficial community port of Linux for it. Don't think for a second that he actually ran MacOS on it!!

Fortunately, nVidia wasn't allowed to acquire ARM and although they would have made some beastly aarch64 CPU's,

Grace is already pretty beastly, especially if you have two of them as a 144-core "superchip", on one SXM board.

LOL Linus. A loser writing crap code for a cult.

ARM is a cult? He's sure not writing code for MacOS!

 

[Pierce2623]

Linus is surprisingly eccentric for the person that he his (or what he has established and accomplished, rather). Even when Apple Silicon performance was like the holy grain of ARM performance, getting an Apple device was truly a spit-in-the-face for the ENTIRE Linux and even broader FOSS community. Sorry, but making a deal with the devil is... just that, lol. Sure, basically the perfect ARM-based mobile platform from a raw specs/performance standpoint (and I mean including battery life, being lightweight and thin, etc.), but what did they really net him and the grander Linux community?? Really, serious question!

Fortunately, nVidia wasn't allowed to acquire ARM and although they would have made some beastly aarch64 CPU's, the free world was spared some time. And while seeing ARM rise up against the x86 monarchy is great, I'd love to see more RISC-V adoption outside of China. As for Apple... I don't think they'll see that large of a perf gain as they did, with M3, M4, etc. just being incremental gains and requiring that next "quantum leap" in technology to really move the needly. Even if they do, they are still Apple a.k.a. f the rest of the computing ecosystem and those that believe in a technology world that involves more than one freaking vendor/manufacturer. Sounding like a hater but I really can overstate the danger of these monopolies and oligopolies, ESPECIALLY when much to most of the vertical integrations are owned by the same company. It's not even academic; we know what the robber-barons, the Rockefellers and Carnigies and JP Morgans did so long ago and how long it took to break up those empires. FFS, the U.S. federal government is finally tired of Apple's control and therefore is finally ligitating against them. Microsoft needs it as well, to be clear...

I can’t understand why people can’t see that the Zuckerbergs and Musks, and Bezos(s?) and Altmans (maybe even the Huangs and Sus) of the world are absolutely every bit as dangerous and corrupt as the Rockefellers or Carnegies were.

 

[Pierce2623]

I feel weird for defending him on this point, because I agree with you in that I wouldn't buy an Apple device purely on principle. But, Linus got it because there's an unofficial community port of Linux for it. Don't think for a second that he actually ran MacOS on it!!


Grace is already pretty beastly, especially if you have two of them as a 144-core "superchip", on one SXM board.


ARM is a cult? He's sure not writing code for MacOS!

He’s obviously talking about the Linux community when he says “cult”, not a hardware vendor. Also the Grace CPU uses off the shelf ARM cores that are much less potent than Apple core architectures or the new Qualcomm core architecture. Grace is literally just a slightly newer Neoverse core than the Ampere Altra.

 

[ezst036]

LOL Linus. A loser writing crap code for a cult.

Cult of Mac | Apple news, rumors, reviews and how-tos

Cult of Mac | Apple news, rumors, reviews and how-tos

www.cultofmac.com

 

[DS426]

LOL Linus. A loser writing crap code for a cult.

Cult of personality, yes -- not the FOSS community though as most are about the concept, not the person.

 

[bit_user]

He’s obviously talking about the Linux community when he says “cult”, not a hardware vendor.

Okay, whatever. The world runs on Linux... some cult. Just sayin'.

Also the Grace CPU uses off the shelf ARM cores

Yeah, so does Ampere Altra.

that are much less potent than Apple core architectures or the new Qualcomm core architecture.

But you can't get around the disparity in the number of cores! The kernel source tree has tens of thousands of source files, which is a nearly-perfect workload for a massively multi-core machine. I don't know how many cores his Mac has, but let's say it's 16. For kernel builds, you're way better with 8x as many cores, even if they're only half as fast.

Grace is literally just a slightly newer Neoverse core than the Ampere Altra.

There's a pretty massive difference between Neoverse V2 vs. N1 cores. I could easily believe it's close to 2x and higher in vector workloads. If you can find a benchmark comparing AWS Graviton 4 vs. Graviton 2, that's effectively what you're talking about.

 

[bit_user]

Cult of personality, yes -- not the FOSS community though as most are about the concept, not the person.

There's an undeniable cult of personality around Linus, but not hugely and that's not why Linux is popular. It's popular because the tech industry eventually found self-interest in backing it.

 

[Pierce2623]

Okay, whatever. The world runs on Linux... some cult. Just sayin'.


Yeah, so does Ampere Altra.


But you can't get around the disparity in the number of cores! The kernel source tree has tens of thousands of source files, which is a nearly-perfect workload for a massively multi-core machine. I don't know how many cores his Mac has, but let's say it's 16. For kernel builds, you're way better with 8x as many cores, even if they're only half as fast.


There's a pretty massive difference between Neoverse V2 vs. N1 cores. I could easily believe it's close to 2x and higher in vector workloads. If you can find a benchmark comparing AWS Graviton 4 vs. Graviton 2, that's effectively what you're talking about.

You clearly don’t get what I’m saying but that’s cool. Qualcomm is going to put their Oryon cores into server chip or possibly even a CPU/GPU combo chip like everyone is coming out with right now. Their Adreno GPU architecture is basically based on GCN with optimizations towards running on low bandwidth and power. It wouldn’t really take much work to get it back more aligned with compute. Hell CDNA2 or 3 or whatever AMD is on now is basically GCN with more bandwidth and wider registers to execute fp64 at full rate.

 

[bit_user]

You clearly don’t get what I’m saying but that’s cool.

As long as you're not accusing me of a willful misunderstanding, I think we can eventually sort it out.
: )

Qualcomm is going to put their Oryon cores into server chip

Nuvia was founded with that goal, but Qualcomm acquired them only a couple years after cancelling their own internal server CPU effort and has since stated that they're currently focusing on just the client market. I'm not saying it won't happen, but I think Qualcomm wants to gain some traction in the laptop market before trying to take on yet another new market.

Their Adreno GPU architecture is basically based on GCN with optimizations towards running on low bandwidth and power.

Source?

It wouldn’t really take much work to get it back more aligned with compute. Hell CDNA2 or 3 or whatever AMD is on now is basically GCN with more bandwidth and wider registers to execute fp64 at full rate.

Yes, that aligns with my understanding, as well.