kano
Legendary
Offline
Activity: 4592
Merit: 1851
Linux since 1997 RedHat 4
|
|
October 19, 2021, 07:36:58 PM |
|
Alas those freightwaves numbers are all more then 6 months ago with hopes they would clear out 4 months ago. They also seem to be written for media attention. This later one says the wait was a 'record' average of 9 days in September ... 9 days doesn't sound as bad as those previous articles were trying to make it out to be ... so it's hard to tell what what's really going on. https://www.freightwaves.com/news/just-how-many-containers-of-cargo-are-stuck-off-californias-coastThis also says that it's a 34% increase in containers so back 6 months ago it was just under 7 days. ... and oddly also says that the container ships are coming in with an average of a lot less containers on them. Would be good if there was another site somewhere with actual daily info - so you could actually see what's going on rather than reports that seem to be written for media dramatisation
|
|
|
|
NotFuzzyWarm (OP)
Legendary
Offline
Activity: 3738
Merit: 2666
Evil beware: We have waffles!
|
For what it's worth, a short while ago we hit a new ATH of $64.34k USD and oddly also says that the container ships are coming in with an average of a lot less containers on them. My guess is that the shipping companies are running smaller loads so the ships can be offloaded faster so they can move on to other ports. Other news items from that site reference that many shippers have had to cancel routes because of the delays...
|
|
|
|
mikeywith
Legendary
Offline
Activity: 2338
Merit: 6504
be constructive or S.T.F.U
|
|
October 19, 2021, 08:47:05 PM |
|
I know this might sound like a conspiracy, but has anyone thought that the mining companies and the rig manufacturers are actively colluding to keep the difficulty as low as possible? Such a scenario might suit all parties if the status quo is maintained, cartels are nothing new!
I do think they are colluding but in regards to keeping the difficulty low, it's obvious that all of these manufacturers make more/easier money selling mining gears than actual mining (the other small algos are all exceptions), of course, they do mine with a small portion of they make, but given that they don't control the majority of the hashrate, not even a considerable amount, I highly doubt they care about difficulty anywhere near to actually selling the mining gears. On the contrary, keeping the difficulty low does allow them to sell their hardware for higher prices.
|
|
|
|
PrimeNumber7
Copper Member
Legendary
Offline
Activity: 1652
Merit: 1899
Amazon Prime Member #7
|
|
October 19, 2021, 10:04:25 PM |
|
I know this might sound like a conspiracy, but has anyone thought that the mining companies and the rig manufacturers are actively colluding to keep the difficulty as low as possible? Such a scenario might suit all parties if the status quo is maintained, cartels are nothing new!
Miners tend to sell their equipment in "sprints". For example, they might sell some number of batches of equipment in the span of a week, then go a month or two (or longer) without selling any gear. The price consumers are willing to pay for mining equipment is based on several factors, namely, the current difficulty, anticipated future difficulty growth, anticipated future price of electricity, and the anticipated future price of the coin the equipment will mine. All of these factors will be as of the day the equipment is purchased/paid for. If there is a situation in which some of the above factors are reflective of conditions that are more favorable to miners at the time the equipment is purchased, consumers might be willing to pay a higher price than they might otherwise be willing to pay. For example, if the difficulty is lower immediately prior to several batches going up for sale, consumers' profitability models might indicate the miner will produce more revenue due to lower difficulty and lower difficulty growth than what might actually is realistic.
|
|
|
|
mikeywith
Legendary
Offline
Activity: 2338
Merit: 6504
be constructive or S.T.F.U
|
|
October 20, 2021, 10:25:28 PM |
|
Latest Block: 705909 (7 minutes ago) Current Pace: 103.5723% (310 / 299.31 expected, 10.69 ahead) Previous Difficulty: 19893045048575.13 Current Difficulty: 20082460130830.84 Next Difficulty: between 20282485247605 and 20805193515870 Next Difficulty Change: between +0.9960% and +3.5988% Previous Retarget: last Monday at 10:25 PM (+0.9522%) Next Retarget (earliest): November 1, 2021 at 10:49 AM (in 11d 10h 31m 35s) Next Retarget (latest): November 1, 2021 at 7:07 PM (in 11d 18h 49m 31s) Projected Epoch Length: between 13d 12h 24m 39s and 13d 20h 42m 35s
https://www.bitrawr.com/difficulty-estimator+3.5% increase in difficulty so far which should continue to increase, while the price is up about 8% from the opening price of last Monday which marked the opening of this difficulty epoch, at current prices and difficulty an S9 makes $5.7 a day, I still remember buying some of them for under a $100 bucks, they were making below a dollar a day back then, it's crazy how things have changed in the past two years, I wonder, if mining bitcoin will ever be as profitable as 2021?
|
|
|
|
philipma1957
Legendary
Offline
Activity: 4228
Merit: 8451
'The right to privacy matters'
|
|
October 20, 2021, 11:02:58 PM |
|
Latest Block: 705909 (7 minutes ago) Current Pace: 103.5723% (310 / 299.31 expected, 10.69 ahead) Previous Difficulty: 19893045048575.13 Current Difficulty: 20082460130830.84 Next Difficulty: between 20282485247605 and 20805193515870 Next Difficulty Change: between +0.9960% and +3.5988% Previous Retarget: last Monday at 10:25 PM (+0.9522%) Next Retarget (earliest): November 1, 2021 at 10:49 AM (in 11d 10h 31m 35s) Next Retarget (latest): November 1, 2021 at 7:07 PM (in 11d 18h 49m 31s) Projected Epoch Length: between 13d 12h 24m 39s and 13d 20h 42m 35s
https://www.bitrawr.com/difficulty-estimator+3.5% increase in difficulty so far which should continue to increase, while the price is up about 8% from the opening price of last Monday which marked the opening of this difficulty epoch, at current prices and difficulty an S9 makes $5.7 a day, I still remember buying some of them for under a $100 bucks, they were making below a dollar a day back then, it's crazy how things have changed in the past two years, I wonder, if mining bitcoin will ever be as profitable as 2021? Once again 2022 will beat 2021 and while we are looking at 2021 as the king 🤴 she will merely be a queen 👸
|
|
|
|
mikeywith
Legendary
Offline
Activity: 2338
Merit: 6504
be constructive or S.T.F.U
|
|
October 21, 2021, 12:20:53 AM |
|
Once again 2022 will beat 2021 and while we are looking at 2021 as the king 🤴 she will merely be a queen 👸
Would love to hear your reasoning, I do agree that prices are likely to continue going higher to sub 100k without the difficulty being able to catch up, but then eventually the price will crash into a bear market like it always does, in other words, the good part of 2022 will likely be bearish, also, the average difficulty will most likely be higher in 2022 than it is in 2021, which suggests that the overall profitability for miners in 2021 is likely to be "as good as it gets".
|
|
|
|
kano
Legendary
Offline
Activity: 4592
Merit: 1851
Linux since 1997 RedHat 4
|
|
October 21, 2021, 02:03:49 AM |
|
The obvious issues are:
1) lotsa large orders 'claimed' in media (though 'media' likes to claim all sorts of crap that's false) that who knows when they will show up. But if some of it does, Diff will of course be going up a lot ...
2) new miners aren't going to suddenly be 50% better H/J I'd be surprised if they can get even 20% better in the coming year, so current new miners will be close to the expected performance, while Diff goes up and up.
|
|
|
|
eectechnology
Newbie
Offline
Activity: 26
Merit: 16
|
|
October 21, 2021, 07:34:26 AM Merited by vapourminer (1) |
|
The obvious issues are:
1) lotsa large orders 'claimed' in media (though 'media' likes to claim all sorts of crap that's false) that who knows when they will show up. But if some of it does, Diff will of course be going up a lot ...
2) new miners aren't going to suddenly be 50% better H/J I'd be surprised if they can get even 20% better in the coming year, so current new miners will be close to the expected performance, while Diff goes up and up.
Your comments about media 'claims' is bang on, unfortunately we live in a society where you can't believe anything the media pumps out. As for new miners, TSMC says that their 5nm process will give you either 30% less power or 15% performance (speed) versus 7nm, I'm guessing the balance for new chips might be somewhere in between. They also claim 1.8x logic density over 7nm but that doesn't really help if power advantage over 7nm is only 15%, especially if the wafer price for 5nm is supposedly 2x that of 7nm. I believe you're right about the difficulty going up and up. Even if Bitmain and their chums are holding back rig supplies, one of them will break ranks or a new competitor might appear. Bitcoins ability to conjure money out of thin air is just too much to resist.
|
|
|
|
NotFuzzyWarm (OP)
Legendary
Offline
Activity: 3738
Merit: 2666
Evil beware: We have waffles!
|
|
October 21, 2021, 02:15:51 PM |
|
As for new miners, TSMC says that their 5nm process will give you either 30% less power or 15% performance (speed) versus 7nm, I'm guessing the balance for new chips might be somewhere in between. Ja, however the S19's are already using 5nm chips so guess what? No newer tech on the horizon for until at least middle of 2022... That said, I've been saying that the node-size race for mining chips is coming to an end since 7nm. The cost/rewards are just too far out of balance to go below 5nm and since TSMC changed their priorities and pricing the 5nm node is now all but shut out for mining chips.
|
|
|
|
eectechnology
Newbie
Offline
Activity: 26
Merit: 16
|
|
October 21, 2021, 07:16:51 PM |
|
As for new miners, TSMC says that their 5nm process will give you either 30% less power or 15% performance (speed) versus 7nm, I'm guessing the balance for new chips might be somewhere in between. Ja, however the S19's are already using 5nm chips so guess what? No newer tech on the horizon for until at least middle of 2022... That said, I've been saying that the node-size race for mining chips is coming to an end since 7nm. The cost/rewards are just too far out of balance to go below 5nm and since TSMC changed their priorities and pricing the 5nm node is now all but shut out for mining chips. Just need to get a bit cleverer with the hashing engine design. There's lot of ways to skin the proverbial silicon cat.
|
|
|
|
mikeywith
Legendary
Offline
Activity: 2338
Merit: 6504
be constructive or S.T.F.U
|
|
October 21, 2021, 07:42:17 PM |
|
As for new miners, TSMC says that their 5nm process will give you either 30% less power or 15% performance (speed) versus 7nm, I'm guessing the balance for new chips might be somewhere in between. Ja, however the S19's are already using 5nm chips so guess what? No newer tech on the horizon for until at least middle of 2022... That said, I've been saying that the node-size race for mining chips is coming to an end since 7nm. The cost/rewards are just too far out of balance to go below 5nm and since TSMC changed their priorities and pricing the 5nm node is now all but shut out for mining chips. Just need to get a bit cleverer with the hashing engine design. There's lot of ways to skin the proverbial silicon cat. There is also a lot of room for improvement in regards to the software level, most of these custom firmware do way better than the stock, getting 10-20% more efficiency out of the same miner isn't something strange, with a good tuning on the S17 pro the difference between it and the S19 pro which uses a newer chip becomes so small, I can't believe that Bitmain does not have the skills that these individual custom firmware devs have, but it looks like Bitmain has not been focusing much on that aspect. My guess is that starting from the next miner version which could very likely be a replica of the 19 series Bitmain will focus more and more on the firmware to try and squeeze every bit of efficiency they can from those chips.
|
|
|
|
NotFuzzyWarm (OP)
Legendary
Offline
Activity: 3738
Merit: 2666
Evil beware: We have waffles!
|
|
October 21, 2021, 08:26:48 PM Merited by vapourminer (2) |
|
Just need to get a bit cleverer with the hashing engine design. There's lot of ways to skin the proverbial silicon cat. Not really. The basic gate structures needed for the logic are extremely simple and were set in stone ages ago. The key point for mining chips is that the signal pathways need to be laid out by hand by someone who knows WTF they are doing and NOT use any sort of autorouting which will *work* but not produce the fastest possible signaling speed. At one time Bitmain had the best person to do that but lost him several years ago... Clunkers like the A1 done by Innosilicon for Bitmine.ch back in 2013 and Canaan's 10nm chip used in the 921/941/951 are perfect examples of how not to do it. The A1 used a lot of pre-designed standard IP blocks for both chip I/O as well as the logic cores and while Canaan has always emphasized using custom blocks tailored for speed they blew it on the signal propagation times within the chip. The end results was that the chips were slower than should have been and used a LOT more power as well.
|
|
|
|
eectechnology
Newbie
Offline
Activity: 26
Merit: 16
|
|
October 21, 2021, 09:28:57 PM |
|
Just need to get a bit cleverer with the hashing engine design. There's lot of ways to skin the proverbial silicon cat. Not really. The basic gate structures needed for the logic are extremely simple and were set in stone ages ago. The key point for mining chips is that the signal pathways need to be laid out by hand by someone who knows WTF they are doing and NOT use any sort of autorouting which will *work* but not produce the fastest possible signaling speed. At one time Bitmain had the best person to do that but lost him several years ago... Clunkers like the A1 done by Innosilicon for Bitmine.ch back in 2013 and Canaan's 10nm chip used in the 921/941/951 are perfect examples of how not to do it. The A1 used a lot of pre-designed standard IP blocks for both chip I/O as well as the logic cores and while Canaan has always emphasized using custom blocks tailored for speed they blew it on the signal propagation times within the chip. The end results was that the chips were slower than should have been and used a LOT more power as well. There's several different ways to construct logic gates at transistor level, especially if you want to trade off speed against power consumption, and design rules have to be observed. This is also true of the power hungry flipflops and adders used to construct the sha256 hashing block. The best designs I've seen were done by mixed signal guys, they have a much better understanding of the analogue nature of high speed logic. I agree with you that most pure digital guys are a bit lazy and use standard cells to make their circuits, but in their defence, most of their designs are vastly more complex than the sha256 block.
|
|
|
|
kano
Legendary
Offline
Activity: 4592
Merit: 1851
Linux since 1997 RedHat 4
|
|
October 21, 2021, 11:01:13 PM Last edit: October 22, 2021, 01:42:29 AM by kano |
|
It's always fun to make claims about how every company who has been making these ASICs for getting close to 10 years and companies like Samsung who are making them, don't know what they are doing and someone already knows better ways to do it Though you'd think an industry like Bitcoin that generates US$54Million a day alone in mining, would prompt at least one better qualified person to do that ... Really makes intel look like a bunch of retards with no 7nm chips on the horizon for more than a year yet Gotta love that the Gold 6258R I bought at the beginning of the year is the most powerful 2nd gen xeon chip they sell, but is only 14nm Edit: and I should add, most experts have no idea about dealing with the amount of power running through miners
|
|
|
|
eectechnology
Newbie
Offline
Activity: 26
Merit: 16
|
|
October 22, 2021, 07:45:03 AM |
|
It's always fun to make claims about how every company who has been making these ASICs for getting close to 10 years and companies like Samsung who are making them, don't know what they are doing and someone already knows better ways to do it Though you'd think an industry like Bitcoin that generates US$54Million a day alone in mining, would prompt at least one better qualified person to do that ... Really makes intel look like a bunch of retards with no 7nm chips on the horizon for more than a year yet Gotta love that the Gold 6258R I bought at the beginning of the year is the most powerful 2nd gen xeon chip they sell, but is only 14nm Edit: and I should add, most experts have no idea about dealing with the amount of power running through miners I always enjoy reading your posts Kano, and agree with much of what you say, but Samsung and the other foundries do not design hashing chips, all they do is make the silicon design rules and process the wafers. If you take some time to look at academic papers you'll find a whole universe of techniques to build relatively simple logic units, and some of them are highly innovative. Most asic designers have been schooled in mainstream techniques and don't usually need to go elsewhere to find solutions. That's no disrespect to them, productivity and research are two very different things. Full custom asic design is a specialised area and there are few designers trained to do it, so I'd take you to task on your initial statement about 'claims'. Mining design is a very niche area and if you believe that every technique possible to increase efficiency has been explored, I'd say you were mistaken.
|
|
|
|
kano
Legendary
Offline
Activity: 4592
Merit: 1851
Linux since 1997 RedHat 4
|
|
October 22, 2021, 08:31:21 AM |
|
always enjoy reading your posts Kano, and agree with much of what you say, but Samsung and the other foundries do not design hashing chips, all they do is make the silicon design rules and process the wafers. If you take some time to look at academic papers you'll find a whole universe of techniques to build relatively simple logic units, and some of them are highly innovative. Most asic designers have been schooled in mainstream techniques and don't usually need to go elsewhere to find solutions. That's no disrespect to them, productivity and research are two very different things.
Full custom asic design is a specialised area and there are few designers trained to do it, so I'd take you to task on your initial statement about 'claims'. Mining design is a very niche area and if you believe that every technique possible to increase efficiency has been explored, I'd say you were mistaken.
Well I do know for a fact that Canaan do their design, their CEO was happy to point out the 'powerful servers' they use for design when I was there in 2017. A billion dollar company like BM would most likely do this also. Both companies started with people who 'very successfully' did the designs 'in their garages' effectively, and since they've continued to evolve the performance and smaller processes, no doubt they do have 'some' understanding of how to make thing perform as fast as possible given the unusual requirements of bitcoin mining hardware. GPUs would be the closest to closing in on these same requirements, but you still don't find a GPU in a box smaller than a typical desktop computer that uses 3kW ... nor one with as many cores ... Samsung do make mining chips, plenty of finds with google, of course in the same way that TSMC do also, for others. But assuming neither company has any intelligent input at all into Canaan and BM is false. Indeed to produce chip designs to be delivered to Samsung or TSMC requires a rather large amount of information from and interaction with them. Also assuming that Samsung and TSMC know nothing about 'useful' techniques when they are the one who make the chips, would seem ridiculous. Aide: You will find all over the forum many attempts in the early years to design ASIC miners that failed dismally. Most of them due to not being able to handle the power requirements of mining, even though they claimed to have 'experts' in the chip design feild. No doubt there may be some techniques to improve performance, however you will probably find that many either are not possible given the power requirements, cannot be done by Samsung or TSMC or have little improvement. If they don't fall under those 3, you'll would find them in use already, or you wont see them for another 5 years. Assuming that there are massive gains to be made and no one uses them seems a rather unlikely situation. Also realise that the 5nm process took .. around 19 years to go from 'demonstation' to 'production', so I'll stand by my original statement: 2) new miners aren't going to suddenly be 50% better H/J I'd be surprised if they can get even 20% better in the coming year, so current new miners will be close to the expected performance, while Diff goes up and up.
|
|
|
|
philipma1957
Legendary
Offline
Activity: 4228
Merit: 8451
'The right to privacy matters'
|
|
October 22, 2021, 12:12:55 PM Last edit: October 22, 2021, 12:25:14 PM by philipma1957 |
|
Kano is correct miners are not gong to alter greatly from 36 th and 900 watts = 25watts a th. This is a firmware modded after market s17pro that is the best of all my 17 s17pros. Even if there is a silicon lottery winning s19pro with great legal after market firmware at best it would do 80th at 1800 watts = 22.5 watts a th. So realistically 20 watts a th may be the best that you get. It is kind of like the automobile and miles per gallon 50 MPH = very close to the best any standard car can do. Now maybe just maybe you could heat regenerate and use the miners heat recapture to preheat liquid for a steam turbine then boosting to under 20 watts a th. If lucky maybe you get 16 watts a th and it would need to be a very large operation. Realistically I do not expect 20 watt per th miners anytime soon back to the diff ... https://www.bitrawr.com/difficulty-estimatorLatest Block: 706145 (a minute ago) Current Pace: 103.8601% (546 / 525.71 expected, 20.29 ahead)Previous Difficulty: 19893045048575.13 Current Difficulty: 20082460130830.84 Next Difficulty: between 20439568967247 and 20860709856586 Next Difficulty Change: between +1.7782% and +3.8753% Previous Retarget: last Monday at 4:25 PM (+0.9522%) Next Retarget (earliest): November 1, 2021 at 3:55 AM (in 9d 19h 53m 39s) Next Retarget (latest): November 1, 2021 at 10:34 AM (in 10d 2h 32m 1s) Projected Epoch Length: between 13d 11h 30m 43s and 13d 18h 9m 6s ... it is good, but I suspect more upwards flow. If you go back to back of 2017 the diff numbers lagged far behind the price raise. from Oct 1 to Jan1 oct 5 2017 = 1.112 th Jan 9 2018 = 1.931 th oct 6 2017 = $ 4425 jan 5 2018 = $17098 so 1931/1112= 1.73x diff jump and 17098/4425 = 3.86x price jump. numbers from google search and bitcoincharts we were 51,487 on Oct 5 so do we go to $198,739 on Jan 5 diff was 19.89 on Oct 5 I do not see us going to a 34.41 diff by Jan 5 but those are the matching numbers
|
|
|
|
NotFuzzyWarm (OP)
Legendary
Offline
Activity: 3738
Merit: 2666
Evil beware: We have waffles!
|
|
October 22, 2021, 07:19:56 PM Last edit: October 23, 2021, 02:55:54 AM by NotFuzzyWarm |
|
... I always enjoy reading your posts Kano, and agree with much of what you say, but Samsung and the other foundries do not design hashing chips, all they do is make the silicon design rules and process the wafers. If you take some time to look at academic papers you'll find a whole universe of techniques to build relatively simple logic units, and some of them are highly innovative. Most asic designers have been schooled in mainstream techniques and don't usually need to go elsewhere to find solutions. That's no disrespect to them, productivity and research are two very different things.
Full custom asic design is a specialized area and there are few designers trained to do it, so I'd take you to task on your initial statement about 'claims'. Mining design is a very niche area and if you believe that every technique possible to increase efficiency has been explored, I'd say you were mistaken.
You pretty much hit the nail on the head about mining chips needing to be viewed as mixed-signal devices when designing them and doing routing. Look up posts by 2112 as he is the local Forum guru on it. He and I have had several lengthy discussions about this several years ago. Correct that Foundries do not design chips. However, Kano is correct in saying that Foundries DO work together with their customers when new chips are being designed and even more so when the chips are a bleeding-edge node size. Fact is, Apple, Cisco, Broadcom et al paid for most if not all of the development of equipment and techniques used to produce the various nodes - not the Foundries. BUT it was Bitmain and other miner chip makers who gave the Foundries practical experience in actually producing gazillions of chips for the ever shrinking nodes. As noted many times before, mining chips are extremely simple circuits not coupled with other complex data support things, not subject to loop closure time errors, and which operate 1:1 with clock cycles. If a core or reasonable multiple cores screw up - no big thing, the chips can still be used (and the production process tested). That makes them perfect test articles. Using the actual chips that Apple, Cisco et al have are far too complex and expensive to test like that - if they have a failure odds are the much higher value chip is trash. As an aside, As Kano said, mining chips are most like GPU's - each chip has a massive number of cores. For example, per jstefanop the chips (from BitFury) in his Apollo have ~4,000 cores in them. My guess would be 4096... Having gone rather OT here, I now return us back to our regular scheduled programming: Heh - sold a bit of BTC when it hit $65,550 -- considering the recent drop, ja I missed out on selling at ATH but nonetheless great timing eh?
|
|
|
|
alh
Legendary
Offline
Activity: 1846
Merit: 1052
|
|
October 25, 2021, 08:01:43 PM |
|
Well past the half way point for this epoch and it's looking like a jump: Latest Block: 706679 (4 minutes ago) Current Pace: 107.4275% (1080 / 1005.33 expected, 74.67 ahead) Previous Difficulty: 19893045048575.13 Current Difficulty: 20082460130830.84 Next Difficulty: between 21235366408839 and 21575738823623 Next Difficulty Change: between +5.7409% and +7.4357% Previous Retarget: October 18, 2021 at 3:25 PM (+0.9522%) Next Retarget (earliest): Sunday at 4:11 PM (in 6d 1h 12m 50s) Next Retarget (latest): Sunday at 9:11 PM (in 6d 6h 13m 18s) Projected Epoch Length: between 13d 0h 46m 8s and 13d 5h 46m 35s
BTC price: $62,758 Minor note to @Fuzzy: Your good sell might have been just a bit more perfect if it had been $65536 (or $65535). Either way it's better than now!
|
|
|
|
|