[HN Gopher] ARM's Cortex A710: Winning by Default
___________________________________________________________________
ARM's Cortex A710: Winning by Default
Author : ingve
Score : 26 points
Date : 2023-08-11 20:22 UTC (2 hours ago)
(HTM) web link (chipsandcheese.com)
(TXT) w3m dump (chipsandcheese.com)
| packetlost wrote:
| I wish I could buy a A710 in a SBC. I'd love to cluster up as
| many ARM cores as I can.
| CameronNemo wrote:
| Don't SBCs have a lot of peripherals that aren't particularly
| useful for a generic compute cluster but balloon up the cost?
|
| Anyway you can find A78s these days if you look hard:
|
| https://www.ipi.wiki/pages/i-pi-smarc-1200
|
| And Rockchip is pushing A76s in their rk3588. Bit of a joke,
| but it seems like that is the best we got.
| packetlost wrote:
| I mean, sure. Less than all the extras that a phone would
| have. I already have a few RK3588 boards, and they're pretty
| good.
| modeless wrote:
| I wish someone, anyone, would attempt to compete with Apple on
| single threaded performance in phone SoCs. Nobody is even trying.
| They all have business strategies that prioritize other things.
| sliken wrote:
| Or memory bandwidth (800GB/sec).
|
| Or iGPU perf, which of course needs the above mentioned
| bandwidth.
|
| Or ML inference performance (5 tokens/sec with llama 65B).
| MikusR wrote:
| They are trying, but ARM is suing and demanding that the cores
| that are faster than ARM cortex ones be destroyed.
| brucethemoose2 wrote:
| Huge, GPU heavy SoCs are not economical without the huge
| volume, software ecosystem _and_ margins of Apple.
|
| Smaller, higher clocked SoCs make sense for Android.
|
| And seperate CPUs+dGPUs are what users want on PCs. Intel and
| AMD _tried_ to sell GPU heavy designs (eDRAM Broadwell, Vega M,
| Van Gogh (the Steam Deck SoC)) and PC OEMs unequivocally
| rejected them. And they are trying to make better cores, but
| again they have to balance die area and target servers and
| cheap consumer PCs with the same cores.
| lnsru wrote:
| The question is if customers care about that. Crazy engineering
| is cool for sure, but it must be profitable too. In my
| environment people stopped buying newest phones years ago
| anyway.
| makapuf wrote:
| People around me didnt buy the newest phones for benchmarks
| then either. They just wanted to have the last one, maybe a
| better screen or better camera, more storage or radio but cpu
| was way down the list.
| beebmam wrote:
| Who cares? I genuinely haven't needed my phone to be faster for
| at least 5 years. The only reason my phone becomes slower is
| because of operating system bloat that they keep introducing on
| phones.
|
| Just give me an open OS, an old phone, and a web browser and
| I'm good. If I need serious computation I'll use a 400-core
| ephemeral cloud computer with 12 TB of RAM, or a modern GPU,
| depending on workload.
| wyldfire wrote:
| > Nobody is even trying.
|
| Qualcomm has a $1.5B wager that they can, with the same
| engineers from Apple who helped them get where they are now.
|
| > . Qualcomm, and then Samsung decided licensing ARM's cores
| would be easier than trying to outdo them.
|
| ...
|
| > ARM has a firm grip on the Android market. Samsung, Qualcomm,
| and MediaTek may develop their own SoCs, but all use CPU core
| designs from ARM (the company).
|
| But they've now reversed that decision, so we'll see whether
| they can change it.
| mschuster91 wrote:
| The problem that Qualcomm and all the other ARM manufacturers
| have: they're _impossible_ for tinkerers to get ahold of,
| outside of Raspberry Pi and a truckload of shitty, barely
| supported clones. And even then, the RPi _still_ doesn 't
| have the basics of PCIe working - in 2023 [1]. What. The.
| Fuck. Yes, Apple theoretically offers ARM devices, but they
| are not cheap, not extendable _at all_ beyond USB-C and a ton
| of stuff doesn 't work on Linux.
|
| In contrast, say I want to develop something on Intel? No
| problem, I head to Amazon, buy a CPU, a motherboard and try
| if that old ATX power supply is still working. I plug in
| whatever card I need and it Just Works.
|
| Steve Ballmer was right on track with "developers,
| developers, developers" - because if the ecosystem is crap or
| impossible to use for creative people on a low budget, guess
| what, they won't and go for the alternative. Linux started
| out on x86 for a reason, and Android blasted Windows Mobile
| (and everyone else but Apple) to pieces despite it being a
| solidly established player. A large part was due to neglected
| developers: outdated APIs, expensive and half-broken dev
| tools (developing for WinCE was a _real_ pain in the
| proverbial arse), and a complete inability to even try to
| match Apple 's innovation. Apple had a monopoly on capacitive
| touchscreens for years!
|
| If the ARM ecosystem players actually want to throw punches
| towards the unholy duopoly of Intel/AMD, they need to
| standardize on a common core of UEFI and get modular, working
| components out on the market.
|
| [1] https://www.jeffgeerling.com/blog/2022/external-graphics-
| car...
| geerlingguy wrote:
| To be fair, the PCI Express bus on the current Pi was kind
| of an afterthought, only meant to work with a very limited
| set of devices, so I'm pretty sure nobody at Broadcom, and
| few in the design stages at Raspberry Pi, had ever tested
| more complex devices with it.
|
| It works fine in _most_ cases for simple devices (USB
| controllers, SATA, NVMe, WiFi, and the like), but really
| falls apart for more advanced devices (hardware RAID, GPU,
| TPU, etc.).
|
| And all Arm processors have to deal with cache coherence
| issues (which aren't a problem on X86), meaning some
| drivers (notably, AMD still) need to program for the
| different architecture (some patches exist but they're not
| perfect yet, and not in mainline Linux).
| mschuster91 wrote:
| > To be fair, the PCI Express bus on the current Pi was
| kind of an afterthought, only meant to work with a very
| limited set of devices, so I'm pretty sure nobody at
| Broadcom, and few in the design stages at Raspberry Pi,
| had ever tested more complex devices with it.
|
| The first Raspberry Pi was sold over a decade ago and I
| 'member people actually using them as embedded boards for
| whatever stuff they were working with eight years ago
| (especially once the GPU performance became powerful
| enough to run digital signage). Sorry but that a company
| like Broadcom can't be arsed to develop a standards-
| compliant PCIe interface is a joke, and with this kind of
| attitude the ARM world complains that no one buys their
| chips?!
|
| (Edit: Oh, just noticed whom I replied to - the person
| who wrote the article I referred to. HN is a small world
| indeed, and I guess we share at least some of our
| frustrations)
|
| > meaning some drivers (notably, AMD still) need to
| program for the different architecture (some patches
| exist but they're not perfect yet, and not in mainline
| Linux).
|
| Drivers... oh don't get me started on _that_ front.
| Everyone in the x86 space seems to have learned over the
| last two decades that it is a good idea to submit drivers
| to the Linux kernel _early_. Intel and AMD both do that
| for CPUs and also for a lot of their other stuff. In
| contrast, the entire embedded world _still_ locks away
| drivers behind years-old kernel forks, ridiculous NDAs,
| absurdly expensive dev boards, completely whack u-boot
| forks and even more whack BSPs.
___________________________________________________________________
(page generated 2023-08-11 23:00 UTC)