zimmie

crowley said:

Seems odd that the better card is only available with the lesser amount of memory.

All of the new options are 32 GB per GPU. The ones which show 64 GB have two GPUs on one card.

The single-card prices are right in line with retail for the Radeon Pro line. It's more than a little weird that one the pairs are all $400 more expensive than buying two cards individually, though. Maybe that's the cost of the "Infinity Fabric Link connector"?

crowley said:

michelb76 said:

imagladry said:

mike54 said:

After all the unlimited praise youtubers, tech sites, Apple fanboius, etc gave the M1, I just hope Apple is not taking advantage of this praise, milking as much revenue as they can from it, thereby delaying advancement. Apple does have a bad habit of releasing something great and then sitting on it past its use-by date.

ah, they produce a new A Series chip every year. My guess is that will be the plan for the M Series, also. Is that "sitting on it past its use-by date?"

Probably, as the M1 is simply an A14X. 

It's "simply" something that doesn't exist?  The definition of "simply" must have changed since I was a kid.

If you look at the A12 versus A12X and A12Z, the changes are almost identical to the changes from the A14 to the M1. CPU goes from two fast cores to four fast cores, GPU goes from four cores to eight cores.

If you look at the packaging, they're almost identical there, too. The big reason for the M1's limited memory performance and capacity is that it only has space for two RAM packages.

Thus, the M1 can reasonably be called an A14X.

crowley said:

Intel processors are competitive with Apple's in many ways, superior in some, but with the notable exception of power consumption, and they have numerous avenues to improving that, if they can just whip their engineering into touch.

Have you been following the Atom cores? They started as a minimal implementation of the amd64 instruction set without speed boosts like speculative execution and branch prediction. Once they had the basics working, they refused to accept changes unless each percentage point of additional power budget produced a greater number of percentage points of performance improvement. The recent cores are really capable. Power consumption is still a bit higher than ARM's, but dramatically better than other amd64 cores. They're also small enough to fit 50+ cores on a die, which is what the Xeon Phi line was.

loopless said:

HPC runs on Intel. Software vendors are slow to move in this field as it isn’t a simple recompile to run on Apple silicon. Apple would need to show a massive performance advantage…

Not so much. Fugaku, the top HPC cluster in the world, is ARM. Summit and Sierra, the number 2 and 3 clusters are POWER9. Sunway TaihuLight is a custom, decidedly-non-x86 architecture. 5 through 10 are all amd64 plus some specialized hardware (mostly Nvidia A100 cards). The fifth place system isn't even particularly close to fourth place, let alone to first. In Rmax TFLOPS:

1. Fugaku - 442,010
2. Summit - 148,600
3. Sierra - 94,640
4. TaihuLight - 93,014
5. Perlmutter - 64,590 - This is the top system with amd64 processors involved

On a percentage basis, Perlmutter is almost as far behind TaihuLight as Sierra is behind Summit, and Fugaku's lead is just as stark as it was when it was introduced in 2020.

Commercially-pure titanium is stronger than steel and significantly lighter. The perception of finish issues is mostly because titanium oxide is white, less shiny, and much softer than the unoxidized titanium underneath it. Since the oxide layer isn't protective, titanium machine parts are often coated to prevent oxidation (titanium implants are a different story; they're stored in an oxygen-free environment until implantation to improve integration with bone). For parts which aren't subject to a lot of metal-on-metal wear, a few resins have been popular for a while. I have a nice pen machined out of titanium, flame-anodized (giving it bands of vibrant colors), then resin-coated for protection. It looks really good, and doesn't show fingerprints or wear.

In contrast, aluminum oxide is far, far harder than the unoxidized aluminum under it (emeralds, rubies, and sapphires are aluminum oxide with various impurities), and it's slightly porous which lets it accept dyes nicely. In aluminum, anodization is used to accelerate the growth of the oxide layer and optionally dye it.

beowulfschmidt said:

Asimov himself explicitly tied together the Foundation universe with the Robots universe near the end, so it wouldn't surprise me at all if one of the "extended lifespan" characters is R. Daneel Olivaw.  The others might turn out to be robots as well.

In the trailer, they're clearly using some kind of cloning to have multiple copies of the emperor (Brother Dawn, Brother Day, and Brother Dusk are all clones at different ages). Rather like Duncan Idaho in the Dune series.

The new Battlestar Galactica already did the "A bunch of these characters ... are robots!" thing. It would seem derivative. I could see it with one or two beyond Olivaw (which would give them opportunities to explore different interpretations of the Laws), but not more.