Dann Corbit wrote:There was a local bubble in memory prices.
It's going away now. Micro trends are just noise on the curve.
You're delusional if you think it's not going to happen.
Sorry Dann, but it's you who is delusional. You have to first understand technology and its limits to see what will happen with prices.
NAND flash is currently going from TLC to QLC which at best is 25% increase in capacity but in reality it is less because of increased overprovisioning.
We are now just starting with 128 layers, compared to 64L the price per GB didn't drop yet at all and in 2 years time might drop 20%.
In 5 years we'll be at 256L and at best with further 20% decrease.
So if you look overall price per GB might drop by 50% in an optimistic scenario by 2023, but most probably it'll drop much less.
So even if we reach 1000TB discs for consumer market (which is another SF, we won't even have 50TB discs by 2023) the price would be over 100k$.
Since it is already possible to make it, the density is feasible. And I guess that it costs little more to make it than to make a 10 TB SSD.
You can already buy 1TB for $200 and 2TB for $350. The masks will continue to shrink until they are a few atoms thick.
Now, it is always possible for a technology to dry up. When that happens, a new one will be invented.
Of course I can easily be wrong. Time will tell.
Taking ideas is not a vice, it is a virtue. We have another word for this. It is called learning.
But sharing ideas is an even greater virtue. We have another word for this. It is called teaching.
koedem wrote:Cool, I just looked at the stats of the most recent one ftp://112.73.74.24/pub/syzygy/7men_testing/KQQRNvKQ.txt and found it interesting that there actually are cursed wins for white. So black has 50 checks but yet at some point runs out.
In these cases white cannot avoid a black capture into a cursed KQRNvKQ win (KQQRvKQ and KQQNvKQ not having any cursed positions).
KQRNvKQ seems to have some very long (cursed) wins without captures (before the 50-move counter runs out), but also positions where black can force a capture into a cursed KQRvKQ position.
I'm loving these. Here another interesting position, white wins in a lot of moves with black to move giving checks. https://lichess.org/analysis/k7/2QR4/8/ ... 2r4Q/1K6_b_-_-
SF for some time even showed 0.00 (after initial +22) and only later found mate in 43 or whatever.
Any idea how to access the tb files? SFs TB option is limited to 6 men.
Thanks a lot, it works.
At first I was wondering why it didn't work with the two 7 men files and the possible 6 men tables. However after getting the 3-4-5 in that folder too it worked. (so tablebase lookups not only require the next smaller tables but go down recursively?!)
Not sure if I set something up incorrectly but SF only shows TB win score (+132 or whatever) in analysis, not the DTZ... (I guess I could just use that d command for the DTZ however that's a lot less comfortable to go through the lines)
koedem wrote:(so tablebase lookups not only require the next smaller tables but go down recursively?!)
Correct. A TB probe tries out all captures to see if there is a winning one or a drawing one (or a capture giving a cursed win or loss) and combines that information with the value stored in the TB file for the position to reconstruct the correct value. Trying out a capture means probing the position after the capture, and there you have the recursion.
Not sure if I set something up incorrectly but SF only shows TB win score (+132 or whatever) in analysis, not the DTZ... (I guess I could just use that d command for the DTZ however that's a lot less comfortable to go through the lines)
syzygy wrote:
SF's search indeed does not output the DTZ value.
Was that changed recently? I remember SF used to give DTZ values.
It would output all moves and give the DTZ score for each with something like +3, or -2.
Or maybe that was my GUI, not 100% sure.
syzygy wrote:
SF's search indeed does not output the DTZ value.
Was that changed recently? I remember SF used to give DTZ values.
I don't
It would output all moves and give the DTZ score for each with something like +3, or -2.
Or maybe that was my GUI, not 100% sure.
Probably your GUI. Which GUI?
It seems syzygy-tables.info needs an old version of aiohttp which I can't get installed at the moment. (But I'm not very familiar with the python packaging system.)
With a recent aiohttp I get the complaint that aiohttp.web.run_app() does not recognise the "loop" argument. If I remove that argument the server starts but returns a 500 Internal Server Error.
Since it is already possible to make it, the density is feasible. And I guess that it costs little more to make it than to make a 10 TB SSD.
The fact that somebody put together a 100TB SSD is not really that significant. SSDs are just a bunch of NAND flash dies packaged together and connected through a controller. The dies are the important parts, and only a few companies make them. The largest NAND dies in real products right now are 96GB afaik, but in terms of physical size, they're tiny. Size of a fingernail. So to make a 100TB SSD, just buy 1000 of those, glue them together and make a controller to talk to them all. But 1000 dies is just expensive as fuck. They're getting 20-30% cheaper every year on average, which is nice, but 20x cheaper SSDs is not even on the horizon,
