Arm Looks to Laptops: Cortex-A78C Processor for PCs Announced | Hacker News Reader

Arm Looks to Laptops: Cortex-A78C Processor for PCs Announced

tomshardware.com

41 pointsrbanffy28 comments

As neat as this is....I have a few major pain points for ARM that will keep me away from them as a general purpose platform:

- Can I even install my own OS, or am I locked down?

- Can I use mainline linux, or some proprietary vendor kernel that will not be updated at all?

- Can I do a manual install, or is the actual install process going to be some maze of partially installing, figuring out where proprietary vendoor boot images go, and hoping that when I press power that it actually turns on? Or do I just have to resort to some .img file and have no control over encryption or how my partitions look?

This is "just" the CPU. Not the SoC/motherboard.

Whoever designs the SoC/motherboard incorporating this CPU shall determine the answers to the above 3 questions...

Common reasons for proprietary kernel / vendor blobs are BT/WiFi, GPU, PMU. So, whether the end-users will have access to the source or ability to choose/customise, will depend on whether open or closed variants are chosen for each of these controllers in the final product.

BTW, have you checked-out the Pinebook? https://pine64.com/product/14%e2%80%b3-pinebook-pro-linux-la...

> This is "just" the CPU. Not the SoC/motherboard.

With Nvidia soon owning ARM, what do you think will happen?

Will ARM be a good investment for people looking for a general purpose CPU?

The problem is more that ARM is way less standardized about things like the boot process or how to find and initialize every hardware.

So I am reminded of something that I heard when in a masterclass from Andy Martin (he is a world class trombonist). He said when he is performing, no one cares how much time it took him to prepare a piece, what his mood is that day, how much he played that day, etc. Those are all things that can positively and negatively affect his performance. They only care about the end product, the performance.

I relay that story because it is much the same with an ARM chip. I know it is "just" the CPU, and the SoC is what integrates it. I don't really care. The end effect is still all of the problems I listed above, which makes me steer clear of using ARM in a laptop. If ARM really wants to break into this space, these are problems that need to be fixed, and it is in ARMs interest to fix them.

I have not tried the Pinebook (though it uses the same RK3399 has the rockpro64, and the annoying install experience I spoke of was about the rockpro64), and frankly, after my experience with the Rockpro64, I am hesitant to really buy more products from them again, which is sad because I have two pinephones and I really believe in what the Pinephone stands for. The RockPro64 had bad memory, and there are a few reports of other people with the same issue. I bought it in early September, it took me three weeks to get an RMA (and that was only after I filed a dispute with Paypal), and it is in week two of me waiting for me to have them ship it back.

The pinebook pro has a pretty dim, mediocre screen, only 4GB of RAM, and opengl ES.

ARM itself only makes the cores and licenses IP, and doesn't even sell physical chips, so you can't really ask it directly, but rather the other companies who will produce SoCs with those cores.

My bet given the direction things have been heading in the computing industry, and ARM platforms in general, is that the answer to all of those questions will be some variant of "no". Even x86 and the PC is heading in that direction. It is unfortunate.

Very sad that unlocked bootloader's have gone away. Those couple years where it was required to have unlocked bootloaders & alternative Android OS spins were getting popular were so neat.

There is a Server Base Boot Requirements specification that should make installing os'es much more normal, requiring acpi support & uefi, the x86 standards. Alas it has the word "server" in the spec. A pity. We can only hope consumer devices end up with similar user-respecting treatment.

I'm shocked to find that the maiome kernel situation is getting better. Few devices that, like x86, are general purpose computers that will let us install our own os or kernel, but kernel support has been getting much better.

ARM is working on a complementary spec called Embedded Base Boot Requirements. While I haven't analyzed it in detail, it sounds like something that might be used in consumer devices.

https://github.com/ARM-software/ebbr

One correction: Unlocked boot loaders are still present. All the platforms that I have seen that support Windows on Arm have boot loaders that can be unlocked.

This is probably the best place to follow machines that have unlocked boot loaders, as currently Linux enablement requires it: https://github.com/aarch64-laptops/build

late reply sorry.

this article is about laptops, & weirdly, thanks to Microsoft, you are right; many arm laptops are do have secure boot systems that respect the user!

I was mainly thinking of every other breed of arm consumer device, like phones & tablets. For a couple years we were getting unlockable devices- I think maybe by law?- but now all these Android & other devices are usually locked down & wont let the user install an os on to the device. which is especially brutal & butter when devices often stop receiving major is updates after a two or three years.

The Arm laptop community for these higher-end cores in laptops is currently active here: https://github.com/aarch64-laptops/build

+ You can install your own OS. + Mainline Linux boots, with some caveats. + There is an Ubuntu 18.04 patched installer that works as expected, but support for the platforms isn't in the main distro installers yet.

- Are these going into reasonable price/capability machines or just "premium ultra portables".

Look, I know you don't want to create a market perception that ARM = "cheap & slow" but I'm not paying $1500 for something as slow and limited as a Surface Pro X. I'm not even paying $900 for it, though some might.

See the recent announcement of Arm SystemReady SR and ES, which is a standard mandating UEFI across the board.

Yes, Secure Boot can be disabled on Windows on Arm systems in the same way as Windows on x86_64 systems.

For Windows on Arm laptops, you cannot quite use upstream kernels at this point w/ full functionality, see https://github.com/aarch64-laptops/build. (being rebased continuously).

For other Arm machines, like servers, you can just use regular kernels though.

And over custom partitions and other things, it's just UEFI now, the firmware itself is in a secondary SPI flash or an UFS namespace.

* you might always need to run a script to extract firmware for the GPU/modem and such, which are OEM-signed on Qualcomm Windows/arm64 devices. That firmware is then put to /lib/firmware/.

Very confusing. How does this compare to the ARM X1 core? That, I thought, was going to be their "big" ARM core (but not server class).

The X1 core is indeed an upcoming larger core. This core is just a variant of the same A78 for upcoming phones with the interface to the rest of the system modified a bit to work better in laptops and servers.

Compared to A78, the X1 has twice the SIMD units, 20% wider instruction fetch, 33% wider macro-op fetch/rename, 2x the macro-op cache, 40% larger instruction window, 2x the L2, 2x the L3, 2x the L0 BTB entries, and 66% larger L2 BTB capacity.

All of that should provide something like 25-30% faster performance than A78.

Despite that, the X1 is still projected to be significantly slower than A13 and A14.

ARM should be putting their best foot forward instead of this A78 attempt.

Been using the Pinebook Pro which using an arm chip. The linux experience has been fine for the most part. VScode just finally released an arm port, and I think the only thing that I don't have that'd I want are desktop clients for Discord/Spotify

Wow, this writing sucks. What even is the A78C?

The writer implies it’s a processor competing with Qualcomm’s 8cx, but the branding and Arm’s slides point to it being a variant of the A78 core. If that’s the case there should be a chart showing exactly how it diverges.

Additionally, the author seems not to know much about the 8cx. Half of its Kryo 495 cores are variants of Arm’s A76 and the other half are A55s. They definitely aren’t all custom big cores as the author claims.

How do these compare to AMD/Intel processors? I'm sure they are more efficient, but can they compete in speed and power?

> can they compete in speed and power?

Almost certainly not. For you that might be a deal breaker, but for a lot of users, this isn't.

Especially since Microsoft is now "serious enough" about Windows on ARM, this is more about competing on price for low end machines. Think about all the devices sold that have an i3 or less. I'd bet that this processor is very competitive in terms of speed and power, so how much does it cost the OEM?

There are already benchmarks showing (sometimes, for certain workloads) mobile phones outperforming x86 laptops.

You can read claims like "Geekbench has the iPhone 7 beating the $6,500 12-core Mac Pro in single-thread" [1] and "new iPad Pros are faster than 92% of all laptops, tablets, and convertible PCs sold in the last year" [2]

Of course, I doubt that generalises to all workloads.

[1] https://daringfireball.net/linked/2016/09/14/geekbench-andro... [2] https://www.howtogeek.com/393139/mobile-cpus-are-now-as-fast...

Highly doubtful since heat dissipation will become a huge bottle bottleneck on mobile devices. There's no free lunch.

If you're only needing that compute power in burst mode, it's fine - and a laptop/PC with actual fans and heat sinks can push quite a lot more out of the same chipset.

No, it's true. What Apple does is partially pay more square mm of silicon for the same performance to reduce power, and partially things like restrict the use of small page sizes to allow VIPT L1 cache. And partially the ARMv8 64 bit architecture is a ground up redesign intended to used in modern high end architectures, though that's a smaller factor I think.

I have a custom todo app written in clojure, and I ran it on one of those "linux on android" apps on my new phone the other day... it ran WAAAY faster on my smartphone than on my laptop, to my shock (admittedly my laptop has a Celeron processor, but still...)

Probably somewhere around where the Snapdragon and Graviton 2 compete in Anandtech's recent article:

https://www.anandtech.com/show/16214/amd-zen-3-ryzen-deep-di...

That is to say, not great on the single threaded performance but not unusable.

A78 is a 4-wide OoO design so it's quite a ways smaller vs current big cores in x86 or Apple's lineup.

Reply on news.ycombinator.com