[HN Gopher] Arm fills in some gaps and details in server chip ro...
___________________________________________________________________
Arm fills in some gaps and details in server chip roadmaps
Author : rbanffy
Score : 66 points
Date : 2022-09-16 12:42 UTC (10 hours ago)
(HTM) web link (www.nextplatform.com)
(TXT) w3m dump (www.nextplatform.com)
| NelsonMinar wrote:
| I cannot wait until I can buy a good simple ARM server for my
| home. Something like an RPi 4 but with more disk I/O and RAM. Or
| a slightly more baked version of a ROCKPro64 in a nice case.
| wmf wrote:
| Sounds like the Honeycomb.
| 0x202020 wrote:
| I would love to see something akin to an M-based Mac as a dev
| board of sorts, but I'm fairly sure it would never happen. My
| home server has transitioned over time from various 4u rack
| mount options, to a 2u and an M1 Mac Mini, to a Mac Studio with
| a disk enclosure.
|
| I even tried using a Snapdragon phone dev board for awhile for
| something more than an RPi
| rbanffy wrote:
| It's a bit annoying that (with the exception of Apple) ARM
| only exists in the low end and in the high-end cloud server
| space. There is no ARM equivalent to the Core i5 generally
| available.
| smoldesu wrote:
| My (non-professional) opinion is that Intel has invested a
| _lot_ of money into making really good memory and I /O
| controllers. We need someone to do that in the ARM space as
| well, which is a really hard-sell. There are no
| manufacturers out there that want to pay the ARM licensing
| fee, period. Apple can eat the costs because they sell
| direct-to-consumer and like to control as much of their
| stack as possible. Other companies are probably groaning at
| the thought of fixing all of ARM's problems just so they
| can be treated as a second-class platform relative to Apple
| Silicon. Besides Nvidia or Qualcomm, I don't really think
| there are many companies poised to deliver this, much less
| an incentive to provide anything.
|
| RISC-V on the other hand, seems like it could be ripe for
| rapid iteration. We're obviously still in the super-early
| days of the ISA, but the lack of oppressive licensing has
| already been a boon for the dozens of hobbyist boards we've
| seen released in the past few years. As long as Apple
| doesn't try to EEE the ISA like they're trying to do with
| ARM, I think the future is pretty bright for open
| architectures.
| my123 wrote:
| I'd say that there's one: Snapdragon 8cx Gen 3.
|
| (See the ThinkPad X13s)
| jeffbee wrote:
| The 8cx Gen 3 is better than Qualcomm's other stuff but
| it's still generations behind a current Core i5. They're
| about 20% slower than a last-gen Core i5.
| Symmetry wrote:
| The V2 decision to go with 4 128 bit vector pipes instead of 2
| 236 bit ones is a bit odd, maybe due to sharing some of the
| design with the N2? Or maybe they want to keep things compatible
| so code can be migrated from a V2 to a smaller core.
| brigade wrote:
| V1 already had 4 128bit pipes, SVE just used two
| simultaneously. The only real gain in benchmarks on V1 with
| 256bit SVE over 128bit NEON was in load-heavy kernels that were
| bottlenecked by the 5-wide decode, and weren't saved by the MOP
| cache. Improving decode width is easier now that 32-bit is dead
| (or maybe V2's MOP cache was improved enough...), so 256bit
| wide isn't terribly useful unless they commit the transistors
| and power budget for an even wider backend than 512bit/cycle.
| vgatherps wrote:
| It also means that code using smaller vectors from an older
| version just gets faster, as long as the new core can
| schedule/retire instructions fast enough to keep up. If it
| can't then going for 256 bit is the best choice although it
| seems like newer chips are going for wider execution (and that
| each 128 bit unit is in fact individually accessible).
| saagarjha wrote:
| Now that's a relevant username ;)
| rbanffy wrote:
| Can modern chips hold instructions from different processes
| in the reorder buffer tagging them by their context/PID? If
| so, up to four processes could be running 128-bit SIMD ops
| concurrently.
| wmf wrote:
| That's SMT but Neoverse cores don't use it.
___________________________________________________________________
(page generated 2022-09-16 23:01 UTC)