[Info-vax] OpenVMS Potential Platform of Interest: Intel NUCs

[Arne Vajhøj]

On 7/3/2022 8:13 AM, John Dallman wrote:
> In article <t9rtod$kfv$1 at gioia.aioe.org>, chris-nospam at tridac.net (chris)
> wrote:
>> HP are just about to release a Proliant class server on multicore
>> arm, which will probably be competitive on price compared to X86.
>> Looks interesting, as arm base servers have never achieved critical
>> mass, but that could change with commodity machines like the RL380.
>>
>> https://www.nextplatform.com/2022/06/29/hpe-is-the-first-big-oem-to-
>> adopt-ampere-computing-arm-chips/
>>
>> RiscV laptops soon as well, so starting to get interesting again...
> 
> ARM64 servers have significant traction in cloud services. I ported the
> mathematical modeller that I work on to ARM Linux late last year, because
> some customers want to run it on Amazon Cloud. When you have datacentres
> on the Amazon scale, cooling them becomes a major problem. The better
> performance-per-watt of ARM has seen Amazon design and deploy three
> generations of their "Graviton" cores for their own use.
> 
> They get people to use ARM-based cores by simply charging less money for
> them per hour. They also charge less for running their own Linux than
> they do for running RHEL or SLES. A lot of what gets run on those ARM
> cores is Python, Java, Javascript, or other languages that don't need
> specialised software ported to ARM, but people are starting to do that
> porting.
> 
> /If/ VMS becomes popular in cloud-based deployments, expect customer
> demand for an ARM64 version quite rapidly. This should be easier than the
> x86 port, because LLVM is fully available on ARM, and not much in the way
> of new methods will need to be created.

It is absolutely possible that ARM and RISC-V becomes huge
in server market.

ARM is already making some inroads. But there has also been talk
about ARM for servers for many many years. It is not moving so fast.

And VSI cannot afford to invest in a port to a platform that does
not become mainstream, so they need to see actual big market share
before jumping on the train.

So let us say that ARM reaches a critical mass in 2024 or 2026. Add
the 8 years and we have 2032 or 2034.

Would it be nice to see a much shorter porting time? Sure. But VSI
is not a company in the DEC/CPQ/HP/HPE size.

Arne