IWETHEY v. 0.3.0 | TODO
1,095 registered users | 1 active user | 0 LpH | Statistics
Login | Create New User
IWETHEY Banner

Welcome to IWETHEY!

New That's why it probably looked like a great idea at the PowerPoint stage :D
New Plus, people wanted 64-bitness
(on more than just servers, eventually,) and it took AMD to show the world that sensible 64-bit X86 extensions worked just fine, and that one didn't need to move to ginormously expensive VLIW processors to get it.

(At least that's my recollection.)

If Intel is still around then, I won't be at all surprised if they try to pull the same stunt on moving to 128-bit.

("But nobody will ever need 128-bit..." - never is a very long time!!)

Cheers,
Scott.
New There isn't a technical case for 128-bit right now or in the foreseeable future
Whereas there was for 64-bit. A primary reason for increasing word size is to address more memory without the kind of segmented-memory shenanigans we had in the 80s/90s.

It's worth writing this stuff out, to truly appreciate the difference.

2^32 bits of memory = 4 GB (4,294,967,296 bits)
2^64 bits of memory = 17,179,869,184 GB (18,446,744,073,709,600,000 bits)

Context: to hit the ceiling at current RAM prices, you'd need about 134 million 128 GB memory modules, the largest available now. They're about £1500 a pop. So let's call it about £200 billion, yeah?

There is a lot of headroom in 64-bit.
New We'll use it all up trying to implement IPv6
--

Drew
Expand Edited by drook June 25, 2020, 10:10:27 AM EDT
New +1
Plus, there's going to be a huge amount of data moving around as transport becomes more automated.

https://www.tuxera.com/blog/autonomous-cars-300-tb-of-data-per-year/

[...]

More sensors equals more data

Today, even at lower levels of autonomy, connected cars generate around 25 Gigabytes of data per hour. And as more self-driving features appear inside connected cars, the architecture required to make it all possible will become increasingly complex. This directly correlates to the number of sensors needed for an autonomous system to operate.

It’s important to point out that not all sensors are the same. A broad spectrum of various sensors exists, each having a special purpose and quantity in a car. Depending on the sensor setup, the total amount of data generated can vary substantially. As presented by Stephan Heinrich from Lucid Motors, here are some estimates on sensor-generated data:Combined, the total bandwidth can reach up to 40 GBit/s (~19 TB/h). Even the lowest possible figure of 3GBit/s (~1.4TB/h) is a very substantial amount of data to maintain.

To give you an example of how much data that actually is, a basic laptop with 240 GB of storage on board could hold around 30 DVD movies. But the laptop would run out of storage capacity in less than a minute in this environment. A phone with 32 GB of storage would be full in under 7 seconds, assuming the flash storage would even be able to store data at the required speed.

On a yearly estimate, the amount of data is even more staggering. According to AAA, an average American spends 17,600 minutes driving annually. When combined with the amount of sensor data estimated above, one car could produce between 380 TB to 5,100 TB of data in just one year.

[...]


Still orders of magnitude less than what 64-bits can provide, but that's just one example.

Yeah, people are clever and have been able to do 128-bit math for a long time. But eventually those tricks aren't enough.

Cheers,
Scott.
New I don't have an intuitive feel for that
In my mind, "data" is smaller than "multimedia". So you can fit hundreds of books on a DVD, but only one movie.

But sensor data for a car, that's essentially 3D video. So even when I tell myself that's bigger than video (which still doesn't feel right) I have no sense for how much bigger.
--

Drew
New Friend sent a link to perzackly: This.
Honda e Urban EV 1st Look with Hidden Easter Eggs

Haven't more than scanned the blurb, but noted further evidence that Koan-666 applies:
The best laid plans of mice and men ..run afoul of the dread [n!] terminator ..of all guesses.

That is: this sucker nas [n+1] cameras, arrayed 360º/3-D, in places you'd never 'suppose', plus (apparently.. AI presumably used properly) not-in Life/Death 'decisions'--but that's a w.a.g. Friend asserts that they are also gestating a less over-the-Top version to sell ~ low-$20K range) ...



Shan't lust over either model, may settle for a proffered Subaru (while: the more I learn of their strange Engine-periodically Eats OIl screw-up, their vastly overpriced periodic 'services'--oft replacing fluids already replaced well before sell-date), etc. Thence the more I deem a pox on their House; they've become Corporate in worst sense: using over-charged $$service-income to replace engines of bad-design, then pretend, ~'it's a gift to You'.
tl;dr: Subaru's fan-bois&gals appear not unlike the Menace's groupies--most of such being pretty unaware of how engines, cars actually 'work;'.
Now that's real-'Murican style.

OK maybe a v.-old Tesla, replacing-self the aged batteries at my leisure--as those prices necessarily decline.
Recreation too! Can oscilloscope the guts-in-action, for dessert. 'Lectric motors, so very-much simpler /more reliable than petroleum's thermodynamic-awfulness, at any price..

Vroo .... oom.
     Apple's starting ~ 2(+) year Mac CPU transition. - (Another Scott) - (14)
         That's progress! - (a6l6e6x) - (12)
             Microsoft can already go there. - (static) - (11)
                 This will hopefully push everything towards ARM - (pwhysall) - (10)
                     Intel did try to solve the complexity problem of x86. - (static) - (9)
                         I think they started at the wrong end - (pwhysall) - (8)
                             The margins are higher at the server end. :-) -NT - (static) - (7)
                                 That's why it probably looked like a great idea at the PowerPoint stage :D -NT - (pwhysall) - (6)
                                     Plus, people wanted 64-bitness - (Another Scott) - (5)
                                         There isn't a technical case for 128-bit right now or in the foreseeable future - (pwhysall) - (4)
                                             We'll use it all up trying to implement IPv6 -NT - (drook) - (3)
                                                 +1 - (Another Scott) - (2)
                                                     I don't have an intuitive feel for that - (drook)
                                                     Friend sent a link to perzackly: This. - (Ashton)
         I’m assuming that, absent - (rcareaga)

Able to chew and walk gum at the same time.
128 ms