From: cr88192@gmail.com

On 12/26/2025 1:56 AM, Lawrence D’Oliveiro wrote:
> On Fri, 26 Dec 2025 01:52:15 -0500, Paul wrote:
>
>> On Fri, 12/26/2025 12:42 AM, Lawrence D’Oliveiro wrote:
>>>
>>> On Thu, 25 Dec 2025 23:25:39 -0500, Paul wrote:
>>>
>>>> My CPU happens to have a random number generator running at
>>>> 500MB/sec.
>>>>
>>>> ...
>>>>
>>>> The Linux people happen not to like those, but, they exist anyway,
>>>> chugging away.
>>>
>>> Because it’s difficult to see how you can trust them.
>>>
>>> Not without thoroughly mashing them through something like this
>>> , anyway.
>>
>> The claim was, that x86 processors didn't have anything.
>
> Not what I was responding to. I said nothing about such a claim, either
> way.

In my case, I was not aware of such a feature having been added, but I
haven't really been keeping up with every new feature being added to x86
in recent years.




Some years ago, I got sorta burned on AVX, as it wasn't until comparably
recently that I got a CPU that could actually run it (I tend to build
PCs with parts that are a few generations behind, to keep cost more
reasonable). And, it wasn't until very recently (a few months ago) that
I got something where using AVX didn't make things actively slower (my
main PC can run AVX, technically, but doing so performs poorly).

Much after AVX, I sorta lost motivation to keep up on newer additions to
the ISA; as often it would be a painfully long time before I would
actually be able to use it (and I still don't have any PC's with CPUs
made in this decade).


Well, and then it has started seeming like in a longer term sense, x86
may be doomed. Granted, its end has been predicted for a long time,
though its main threat may be in part the end of Moore's law, which in a
mostly steady-state it is likely that performance per area and
performance per watt will become the dominant factors, and conventional
x86 processors haven't been great on either front here (combined with a
longer term timeframe making JIT compilation and eventually abandonment
the more likely endgame).

Well, and after Moore's law hits its limit, it may backslide slightly as
things settle on whichever process node is most economical in a perf/$
sense.

Though, this is unlikely to happen quickly (more likely over a period of
decades).

Well, and then there are more near-term things, like the mess of things
the whole "AI" thing is creating, and MS repeatedly shooting itself in
the foot (maybe not enough to dethrone themselves, but they are pushing it).

...

--- SoupGate-Win32 v1.05
 * Origin: you cannot sedate... all the things you hate (1:229/2)