[Info-vax] report of the last "rendez-vous autour de VMS" (2-FEB-2024)

[Arne Vajhøj]

On 4/19/2024 5:44 PM, John Dallman wrote:
> In article <uvu9t1$343t5$1 at dont-email.me>, arne at vajhoej.dk (Arne Vajhøj)
> wrote:
>> Requiring ARM for VMS would mean introducing a new CPU type. And
>> with todays multi-multi-core CPU's that would typical mean
>> either having a lot of wasted resource by only using it for VMS
>> or having to move other workloads from x86-64 to ARM to accomodate
>> VMS.
> 
> Probably not, actually. The common ARM servers have Ampere Altera
> many-cores processors with 64 or 80 cores. Those cores aren't very fast,
> because their design prioritised low power usage. That's because their
> target market was huge cloud datacentres, where their selling point is
> their power efficiency reducing the cooling requirements. The square-cube
> law means that in a big enough datacentre, cooling becomes the main
> problem.
> 
> Those Ampere-based servers aren't terribly expensive. If VMS can handle
> 80 cores, it might be quite responsive running on one.

Maybe. But VMS applications are traditionally not very CPU hungry.
I suspect 80 cores @ GHz speed will be overkill for the VMS application
developed on like dual CPU Alpha 200 MHz or VAX 30 MHz. It is not
like the new Java/Python/PHP/whatever applications that use
CPU power like a sponge suck in water.

And even if the cores are so slow that the combined power of
all cores are needed, then the VMS application would
typical be a few single threaded processes that could not
practically use that many cores.

But I believe that DCL would be very responsive.

:-)

Arne