JustAnEngineer wrote:With the inevitable cryptocoin mining crash beginning to take hold (coin values are down to a third of their mid-December peak), a few graphics cards are actually in stock at your favorite e-tailers.
Ethereum is down to where it was mid-November now, back when you -could- buy graphics cards for only slightly inflated MSRP and I *almost* did.
Lo and behold: now cards are not only on the shelves again (albeit with sticker shock, but prices are infamously sticky in the reverse direction), but people just this week have bought direct from Nvidia for only slightly inflated MSRP.
How about that!
Chrispy_ wrote:The price drop is because there's tangible evidence that vastly superior Etherium ASICs are releasing soon.
There are lot of things going on, and I'm not even sure if I'd hazard that being the biggest one.
The Government is now not just suing the ICOs that are predominantly performed on the Eth network, but they've actually started arresting outright. That's pretty serious, usually they start with civil sanction and build that into criminal if they are so inclined. Heck, the Canadian Government has already sentenced one.
In other words, the deluge of SEC lawsuits about how these are obviously securities and no, you can't just say "crypto" (or anything, it literally doesn't matter what incantation bamboozle yourself & others with:
Howey gonna
Howey) and get a free pass for your shenanigans, aren't even dramatic enough. We're in the big leagues now.
JBI wrote:Anything that can be done with a GPU can be done with a sufficiently sophisticated ASIC. Ethereum made things difficult by requiring lots of RAM, forcing the ASIC designers to incorporate high-speed DRAM interfaces onto the ASICs and lots of RAM onto the mining cards. This doesn't make it impossible, it just drives cost and complexity up.
Exactly, the trade-off was economic and it just so happened that GPUs already were in the sweet-spot for semi-related technical reasons.
But this ASIC is going take take 800 watts to do 180MH/s at $800 USD. All of those could change for the worst, and almost certainly will, and they are not even pretending to that they are going to ship until 3 months from now.
In other words, even as optimistically stated, it's only roughly 5 times faster than a good GPU for likely 2.5x or so more energy. The cost is obviously competitive (at inflated prices, like 6-7 times less, but at MSRP, maybe 2-4 depending on card choice?), but since actual availability is going to be non-existent, the cost really isn't all that relevant anyway
Basically, it's on the order of a doubling of the typical watt-efficiency. That's, uh, really, really paltry.
When the original Antminer S1 was released (also by bitmain, just like the upcoming ASIC eth miner) ~5 years ago, it was 180GH/s for ~350 watts.
My i7-8700k, the most recently released CPU by Intel, is probably around 100MH/s for BTC, and the power usage at full bore is maybe 1/2 of that S1 at best?
In other words, like the first BTC ASIC was like a 3 orders of magnitude of increase in hashrate and efficiency.
--
Do you see the difference Aranarth? They "did it", but uh, Ethereum can hardfork at any time (unlike bitcoin core they show no hesitancy in doing so) and change the PoW to break this ASIC one way or another.
Meanwhile, Bitmain won't even falsely promise to ship you one for another 3 months, it'll obviously be out of stock (bitmain and S9s chronically are, for instance), and it's really not that fundamentally more powerful than a GPU.
In fact, if you could buy AMD's 580s at actual MSRP, it would only take 7 to reach the hashrate, for 1,400 USD, at ~1100 watts.
In other words, now that GPU prices are rapidly returning to being somewhat normal, you are getting like a 80% improvement in cost for a 40% improvement in efficiency. For a card that was release a year ago that was blatant refresh of a card from the year before that.
And Ethereum can break it at any time, they're not actually selling it, and the specifications probably won't be as good as promised as virtually every miner announcement->release has demonstrated.
Again, yes, they "did it".