Topic: What short lived or lesser known ASICs are worth noting? (Read 1870 times)

 Intel has announced a 10nm process they are working on - but they're having some major issues getting it to work at production-scale (and they have had major issues that are NOT all cured yet getting it to work at ALL even on lab scale).
 If it does show up next year:

 1) It will be Intel-specific - Intel doesn't share it's fabs with anyone else as they need all the capasity they have.
 2) It probably will have the usual "low yield" issues of ANY new process for the first year or two. Possibly longer given all the issues they are having getting it to work at ALL.
 3) It would probably not show up 'till very LATE next year, if at all.

 One thing that is certain about it - it WILL be the last "pure silicon" process ever, according to both Intel and IBM "10nm is the end of the road for pure silicon" - and THAT change is going to make for a lot of infrastructure changes needed for future processes to ever achieve production.

 I believe TSMC and GF have also both announced they are doing work on "next generation" fab processes - but very few details yet.

 IBM has announced it's working on a 7nm process using a MIXED germanium/silicon wafer, but not much in the way of details yet, and it's estimated "soonest to production" timeframe is  2019 IIRC.


 One other thing to keep in mind - there was a 22nm process around (Intel made their CPUs on it among other usage, and at least one other company probably TSMC also had at least one process at that feature size) - but it was very little more efficient than 28nm was despite the significant feature size shrink that SHOULD have made it ballpark 40% more efficient.

 It's also interesting to note that 14/16nm is ALSO nowhere near as great an improvement in effficiency vs. 28nm as previous "half size" generations achieved - commonly when you drop the feature size in half the efficiency almost quadrupled, the current 14/16nm process only seems to offer somewhere between double and triple the efficiency.
 Hello, quantum physics effects getting to be a MAJOR issue at current feature size, instead of just a minor but noticeable factor to be designed around in previous generations of semiconductor tech.

 No sooner do you say that and Litecoin starts climbing big-time.

 8-)


 Though to be fair, you're not earning much from that Zeus because it's fairly low hashrate even for when it came out (around the same time or a bit AFTER the Innosilicon A2 original 88Mhs units showed up).


 to others:


 specialised chips for altcoins started YEARS ago with the Gridseed GC3355 - they're not a new concept, and Scrypt-based ASIC is into at least it's 3'd generation now.


 There will be no S11 this year as the only reason to create an S11 would be a new chip based on a new process node for much higher efficiency.
 There is NO NEW PROCESS FOR SIGNIFICANT EFFICIENCY IMPROVEMENT due for well over a year at soonest and more likely 2-3 years from now.
 
 I can see a faint chance of BitMain moving to a more-efficient 14/16nm process, but the efficiency gain would be minor at best and it likely would not be worth the $millions for them to redesign the BM1387 for such a SMALL improvement.

 The "few months" gaps between the S5/SP20 generation to the S7/Avalon 6 generation to the current S9/Avalon 741/BF6 generation were possible because those earlier generations were NOT being build on state-of-the-art semiconductor technology.
 THIS IS NO LONGER THE CASE, there is no more "catch up to state of the art" AVAILABLE - the S9 generation will not see any significant improvement POSSIBLE for years as it will be years before a more-eficient state-of-the-art is AVAILABLE.
 Future mining chip generations will last for YEARS, not MONTHS, as "most efficient or very very close" as a result - you CAN'T compare the past as we hit a sea-change with the current 14/16nm mining chip generation. This also means that needing a year or two to achieve ROI on a particular miner is now going to be entirely practical, and mining gear makers will probably price their gear to make that THE NORM.

 ASIC aren't dying - they're just going through a major shift in paradigm to where the 4-6 month expected payoff isn't going to happen any more.

 The current big runup in altcoin pricing is a LOT more over the issues of BitCoin lack of scaleability, the Chinese regulatory crackdown on the Chinese exchanges, and the FUD/HYPE/resulting uncertainty about the future direction of Bitcoin driving investors into alternate Crypto investements, NOT about "ASIC is dying" - and without that runup the last couple weeks, expected ROI on altcoin mining rigs would be in the "ballpark a year" range TOO just like the ASIC gear is.



 Keep in mind that by the scale of Nvidia, Intel, and AMD the cryptocoin mining gear market is SMALL potatoes.
 The total market cap of Etherium ALONE probably dwarfs the total cost of ALL GEAR USED TO MINE ALL NON-ASIC CRYPTOCOINS COMBINED by a fairly wide margin.
 Some of the second-tier manufacturers like ASUS and EVGA have paid attention to crypto-related markets (the various "BTC-line" motherboards and EVGA's "EVGA bucks" for folding program) but even for THOSE folks it's still not a major factor (I suspect EVGA uses the EVGA bucks program as a charitable writeoff since it's supporting scientific research).

 Most case makers do NOT win from GPU mining, as most GPU mining rigs seem to either not use a case at all, or use a very-specialised mining-specific type of rack/"case" setup that most case makers don't make anything similar to.

 Similarly, most Power Supply makers don't profit as folks with GPU rigs mostly use PS from about 5 or perhaps 6 makers (EVGA, Seasonic, Corsair, hmmmm perhaps it's mostly 3 as I can't think of any others that are widely mentioned for use in mining rigs).

 CPU makers only profit in SMALL amounts, as the CPU used in most mining rigs is a very low-end low-cost LOW PROFIT model even when it is bought new.




 

"reasonable" would imply completeness. There are now specialized chips in existance that are targetting pure altcoin algos. philipma1957 made a point about changing landscapes. I would guess specialized chips are always outperforming generalized ones, just a feeling from a nontechnician. But the algos are a wide variety nowadays.


One renowned pool operator openly tried to make me believe those are the same. Even nontechnicians are capable of googling some facts. That left me flabbergasted about honesty in the scene.

Nope no s11.

What you are seeing is asics being killed off by gpus.

This is a fundamental problem for asics.

How many gpu builders are out there.

Asus Msi sapphire EVGA zotac gigabyte.

How many mobo builders?

Lots Asus asrock bio star gigabyte Evga
How many cpu builders
Intel AMD
Gpu chip makers
Nvidia
Amd
Psu
Corsair
Seasonc
Antec
FSP
EVGA
RAM
SAMSUNG
CRUCIAL

CASE MAKERS.

ALL OF THE ABOVE WIN IF GPUS WHALE.

THEY have a shit ton of money.
More then one trillion between them.

They will simply prop up Zec or eth or xmr or a new coin

Asics are in real trouble  and a s-11 is not the way to go.


There is 16 billion in btc mined as I type,  and the easy way to get btc is mine an alt and trade for btc.


The 16 million coins mined at 1000 each is 16 billion.
I round.

The 5 million coins left at 1000 each are hard and expensive to mine with an ASIC.

The coins mined already are easy to trade alt coins for.

I can pay 300 in power to get a btc with the s9 and with 7 cent power

I can pay 150 in power or maybe less with a gpu rig to trade for a Btc

The btc asics may never recover.

The math favors gpus.  Because all those companies I mention can add value to the alts by buying them on exchanges.

Btc cost too much for  bitmain or Avalon or bitfury to prop price.

I sold a btc today I am fleeing to cash as I see a lot of shift to alts.

every day I see on ASICs ASIC miners, from time to time I wondered if I new up-dap- migrants may need on record.

This is true for the ASICS and S9 not going so well. You can make 0.012 btc a day through Nicehash miner with a PC that has 6 GPU-s of AMD RX 480 which is a very nice amount without deducting electricity fees but still a very good amount. Why should I tire myself configuring the Antminer S9 when I can point my GPU-s to the Nicehash miner and enjoy the profits for the moment. S11 is needed and I predict it will be around during August of this year.

2-3 years for the next asic? isn't that too much, i don't remember such a void between s7 and s9, and the same between s5 and s7

if they keep waiting so long s9 will be rendered useless, it's already earning below 0.008, and it's inefficiency would increase over time

intel for example is near the 10nm in the next year they should present(icelake) something that run with that efficiency

 There is no new process available for the next year at least, more likely 2-3 years, to MAKE such a miner on.

 They might be able to eek out some small improvements on the S9 using updated 14/16nm process, but the improvements available will be too small to matter (ref the announced specs on the upcoming AMD RX 580 vs the RX 480 - 5% or so improvement in clock rate with same to a hair less power usage) and likely would NOT be worth it to Bitmain to go through the major cost of redesigning the 1387 for the upgraded process that have become available through TSMC or GlobalFounderies in the last few months, especially when they're STILL the most efficient chips on the market.


 To NotFUzzyWarm:

 To be technical, all so-called "ASIC miners" should be properly called "ASIC-based miners" but that's splitting VERY fine hairs for no reason.

 The term is NOT interchangeable with FPGA miner at all though - very different concepts and implimentations and efficiency potentials.

 NO current ASIC_based miner uses a FPGA - been a few years since any FPGA-based gear has been made for SHA256 mining, and I am not aware of ANY that was ever build for any other algorythm (other than one-shot experiments).

Thank you. I bought mine in mid-October last year for the same price. Just wanted to make I didn't overpay for mine. 

I agree, when I said they are used interchangeably nowadays, I meant in the mining community. For that matter, there's no such thing as an "ASIC Miner." There is mining hardware that use ASIC chips, but that does not make an antminer an "ASIC." However, ask many people on this forum what an "ASIC" is and they'll say it's a specialized type of mining equipment.

What I meant about the old days, was that the foundation had not yet been laid. Today you don't have to start from scratch, and additionally, computer simulations help a great deal. You can now design ASIC circuitry in a software suite in a fraction of the time. My father once made a simple chip for something having to do with image capture (he was an engineer for NBC, but this was just a side project of his back in the day.) I recently found some schematics relating to it. All hand drawn, hundreds of pages of notes, pages upon pages of scrapped designs. Being able to use simulators and design equipment on a modern computer has greatly simplified the process. Not that the actual programming is not still difficult or has changed.

Now, as for FPGAs in mining equipment, while ASICs are utilized in most SHA256 mining equipment, due to (as you mentioned) efficiency and the requirement for extremely fast chips at extremely low power usage in order to create a profitable product. Not to mention it's generalized, widespread use as a cryptological method means there exists a lot more research and documentation on the subject, making design (slightly) easier. But more recent algorithms used for altcoins that have gone the way of "ASICs" (in quotes because) often employ FPGA chips. FPGAs come in a wide variety, most have some hard wired logic already programmed in, which make them prime candidates for the algos that were initially designed to take advantage of instruction sets present in CPUs. I don't own one, but the elusive and desired Baikal Mini for example. There's not much information available regarding it's design, but it looks a lot to me like a FPGA on a design board.

More like "hundreds of millions" and probably soon into the billions range for current high-end CPU and GPU chips.


 Anyone that equates a FPGA with an ASIC is clueless - they are completely different in form and function, though sometimes a FPGA can be used to "validate" the intended circuit of a very small ASIC design before you set it in stone as part of the process of putting it on specific-design silicon.



 FPGA technically is a subclass of PGA, but the "field" varient of the concept took over the entire market over time as the non-FPGA version was a lot more cumbersome to use and eventually offered NO performance advantage.

 This is VERY similar to how EPROMS eventually killed off PROMS when the cost got close enough to not matter (with a VERY few narrow exceptions involving security-related stuff).
 

FPGA and ASIC are used in the same breath by folks who have no idea what they are talking about...   A FPGA and ASIC are 2 very different things.

FPGA is more than a blank chip you program. It is a chip architecture that can be repeatedly erased and re-programmed to perform different functions. For what it's worth, at one time there were PGA's - one-time programmable gate arrays but I have not seen them in decades since that tech was long ago replaced by FPGAs and custom factory mask-programmed gate arrays.

An ASIC is hard-wired at its birth and can never be changed. An ASIC is generally much faster and lower power: The signal routes are optimized for minimal delay and there is zero unneeded circuitry in them. An FPGA has banks of addressable memory, different com circuits, etc. ASIC has them too but again, they are hard-wired and only what is needed is put in them, no spare or unused function blocks.

Is still that way and current chips can have millions of gates.

Bought a few months ago for around $350.

Hello, if you don't mind me asking: what did you pay for the SP31 and when was it?

I've been through all the top ones (the ones that aren't deemed likely fraud) already lol. But it's pretty outdated. I don't think there's been an update in over a year. I Own 2 S7-F1 Final Batches (I can run them ar 5.3TH each with less than .002% HW Error and a comfortable 54C temp for days on end with no fluctuation) Great machines. I own 3 ASICMiner Tubes (one only runs on 3 boards, I need to replace a capacitor on the last board so I just disabled it in the mean time) I get about 2.2TH/s out of that, and I have a wreck of a Spondoolies SP31 (bad eBay purchase that I fixed as best I could, when I got it it didn't work at all.) I get 3TH/s from that. I'm just below 16TH/s. I manage to stay above cost of electricity by selling coins or contracts on eBay (where I can usually get twice market value for the convenience factor.) I also have a ZEUS Thunder X6 which gives me 22MH/s Scrypt, but honestly, that is a crap paying algo.

I should have studied computer science at Stanford. But, I was thinking, ASIC just stands for Application-Specific Integrated Circuit. Nowadays it's used pretty interchangeably with FPGA, or field-programmable gate array. FPGAs are field-programmable, meaning  you get a blank chip and burn the programming into the circuit. In the old days that meant coordinating thousands or even hundreds of thousands of gates in order to make a chip work the way you wanted it to. It meant sketching it out on paper, programming in the most basal binary code. Nowadays, they sell chips pre-soldered to USB linked programming boards, complete digital simulation of ICs, and even python modules set to program common routines. And the ones today don't require a "hard burn," you can reset and restart (on some.) Aliexpress has starter kits for like $20. Not that you'd build the next gen chip with that module, but you could start to learn. Bitmain buys their blank chips from Taiwan. You can buy those same chips. And a lot of this is open source, I'm talking even the schematics of the boards and the chip routines. Anyone, with enough motivation, intelligence, free time or a combination of, could create an ASIC Miner, or improve on previous design.

Back to the main discussion though, I do agree, it's time for the next gen (or has the difficulty finally surpassed technological developments?) Probably not, but there is certainly less competition out there in the ASIC market today...

You can find a reasonable list of ASICs on the Wiki.  Generally people tend to go for ASICs by Bitmain (Antminers), because they are generally regarded as the best, but you could look around at different specs and prices for you.

they should be work on antminer 10/11 for now, but aside from the old generation like antminer s5 and s7 that can be good if you have very chep electricity like less than 5 cent, because they have a very low value if bought on second hand

the primarely asic is always the s9, which i think it's too damn inefficent already and need to be substituted, just compare it to other kind of asic that mine somethign else, they have far better efficiency in comparison to what they can make

i hope the s11 will be something like 1000watt maximum but double the hashrate, then you avalano which are on average worse, and are not worth mentioning for me

It seems every day I'm learning about anew ASIC miner. Lately there isn't much competition and there is less variation in the SHA256 market. But what ASIC miners, over time do you think were/are worth noting? Are there any new up-and-comers that might be worth noting?