mdude77
Legendary
Offline
Activity: 1540
Merit: 1001
|
|
January 21, 2015, 12:50:54 AM |
|
So besides noise & heat, does it really pay to underclock vs. stock on the SP20? I tried to run the numbers allowing a total of 4200 watts at the units (not wall). I used 700w for underclock at 1200gh/s and 1200w for 1600gh/s. My calc shows about $697 higher price for 6 underclocked units vs. 4 stock units. The underclocked higher hashrate would take over 600 days to return the higher price of underclocking.
Am I missing something?
Depends on your electricity price. At current BTC/USD value, I'm losing money running at stock. M
|
I mine at Kano's Pool because it pays the best and is completely transparent! Come join me!
|
|
|
bgibso01
Legendary
Offline
Activity: 1218
Merit: 1001
|
|
January 21, 2015, 12:55:09 AM |
|
So besides noise & heat, does it really pay to underclock vs. stock on the SP20? I tried to run the numbers allowing a total of 4200 watts at the units (not wall). I used 700w for underclock at 1200gh/s and 1200w for 1600gh/s. My calc shows about $697 higher price for 6 underclocked units vs. 4 stock units. The underclocked higher hashrate would take over 600 days to return the higher price of underclocking.
Am I missing something?
Depends on your electricity price. At current BTC/USD value, I'm losing money running at stock. M That's why I used a fixed wattage so that the electric would be the same in the comparison. It's just additional hashrate at lower w/gh but higher $/gh.
|
|
|
|
Epoch
Legendary
Offline
Activity: 922
Merit: 1003
|
|
January 21, 2015, 01:33:18 AM Last edit: January 21, 2015, 01:47:58 AM by Epoch |
|
So besides noise & heat, does it really pay to underclock vs. stock on the SP20? I tried to run the numbers allowing a total of 4200 watts at the units (not wall). I used 700w for underclock at 1200gh/s and 1200w for 1600gh/s. My calc shows about $697 higher price for 6 underclocked units vs. 4 stock units. The underclocked higher hashrate would take over 600 days to return the higher price of underclocking.
Am I missing something?
Depends on your electricity price. At current BTC/USD value, I'm losing money running at stock. That's why I used a fixed wattage so that the electric would be the same in the comparison. It's just additional hashrate at lower w/gh but higher $/gh. You are not missing anything. mdude77 is correct in that it depends on your local electricity price. But it also depends on whether you are planning to buy units or already have them. If you already have the units, then your goal might be to maximize net income. See https://bitcointalk.org/index.php?topic=521520.msg10146041#msg10146041 for an example. The higher your electricity rate, the more it makes sense to underclock existing units. Your net hashrate will decrease due to the underclock, but your electricity usage will decrease even more with the result that your net profit (income minus expense) actually increases. In your example you have used a fixed power budget (4200W) and have taken into account the purchase of either 6 underclocked versus 4 stock units. That is a different calculation and, in this case, you are correct in that buying 6 units only to underclock them may not make sense in terms of return on investment. i.e. buying 4 units and running them at stock speeds may recover your initial investment cost sooner than buying 6. But even here it depends on your electricity costs; if your costs are high enough it is possible that stock units will run at a loss (i.e. you will never recover the purchase price), while underclocked units will run at a profit ... meaning that buying 6 underclocked units make more sense than buying 4 stock units. Or, as the artificial intelligence WOPR concludes in the the 1983 movie WarGames, sometimes "the only winning move is not to play."
|
|
|
|
PeaMine
|
|
January 21, 2015, 02:23:59 AM |
|
What temperature should I aim for to prevent harming the SP20? I am reaching 78C on Back T,B with 23C intake. Also does anyone have a calculator or chart showing effeciency ? Clocking it down to around 1050GH gets me around 0.51 watt per GH.
Take a look at your chip temps. They are more important than the temp of the air leaving the machine. 70-100c only 1 was 100c
|
Datacenter Technician and Electrician. If you have any questions feel free to ask me as I am generally bored looking at logs and happy to help during free time.
|
|
|
bgibso01
Legendary
Offline
Activity: 1218
Merit: 1001
|
|
January 21, 2015, 03:15:51 AM |
|
Or, as the artificial intelligence WOPR concludes in the the 1983 movie WarGames, sometimes "the only winning move is not to play."
Okay, that I understand So I was just using an arbitrary $.10/kwh rate. You are saying there may be a sweet spot where rates are high enough that it would make more sense to run underclocked rather than stock. I'd have to see hard numbers to get it when trying to compare apples to apples. For the additional capital outlay, I don't see the underclocking making up the difference in a reasonable time. Since in my example I'm going to pay the same amount for electricity whether I'm using underclocked units or stock units I'm having a hard time seeing the underclock advantage for best dollars spent.
|
|
|
|
bgibso01
Legendary
Offline
Activity: 1218
Merit: 1001
|
|
January 21, 2015, 03:17:23 AM |
|
What temperature should I aim for to prevent harming the SP20? I am reaching 78C on Back T,B with 23C intake. Also does anyone have a calculator or chart showing effeciency ? Clocking it down to around 1050GH gets me around 0.51 watt per GH.
Take a look at your chip temps. They are more important than the temp of the air leaving the machine. 70-100c only 1 was 100c Those are good solid numbers. >115 it starts alerting and I think at 125 it adjusts the chip <?>.
|
|
|
|
DiCE1904
Legendary
Offline
Activity: 1118
Merit: 1002
|
|
January 21, 2015, 03:44:27 AM |
|
Or, as the artificial intelligence WOPR concludes in the the 1983 movie WarGames, sometimes "the only winning move is not to play."
Okay, that I understand So I was just using an arbitrary $.10/kwh rate. You are saying there may be a sweet spot where rates are high enough that it would make more sense to run underclocked rather than stock. I'd have to see hard numbers to get it when trying to compare apples to apples. For the additional capital outlay, I don't see the underclocking making up the difference in a reasonable time. Since in my example I'm going to pay the same amount for electricity whether I'm using underclocked units or stock units I'm having a hard time seeing the underclock advantage for best dollars spent. This is what helped me make up my mind on where to set my miners settings at.
|
|
|
|
bgibso01
Legendary
Offline
Activity: 1218
Merit: 1001
|
|
January 21, 2015, 04:05:10 AM |
|
Or, as the artificial intelligence WOPR concludes in the the 1983 movie WarGames, sometimes "the only winning move is not to play."
Okay, that I understand So I was just using an arbitrary $.10/kwh rate. You are saying there may be a sweet spot where rates are high enough that it would make more sense to run underclocked rather than stock. I'd have to see hard numbers to get it when trying to compare apples to apples. For the additional capital outlay, I don't see the underclocking making up the difference in a reasonable time. Since in my example I'm going to pay the same amount for electricity whether I'm using underclocked units or stock units I'm having a hard time seeing the underclock advantage for best dollars spent. This is what helped me make up my mind on where to set my miners settings at. Got it now. Thanks. At current price of btc, the extra 400gh/s return is higher than the additional wattage used for me.
|
|
|
|
Epoch
Legendary
Offline
Activity: 922
Merit: 1003
|
|
January 21, 2015, 04:05:20 AM |
|
Or, as the artificial intelligence WOPR concludes in the the 1983 movie WarGames, sometimes "the only winning move is not to play."
Okay, that I understand So I was just using an arbitrary $.10/kwh rate. You are saying there may be a sweet spot where rates are high enough that it would make more sense to run underclocked rather than stock. I'd have to see hard numbers to get it when trying to compare apples to apples. For the additional capital outlay, I don't see the underclocking making up the difference in a reasonable time. Since in my example I'm going to pay the same amount for electricity whether I'm using underclocked units or stock units I'm having a hard time seeing the underclock advantage for best dollars spent. It is good that you are going through the mental exercise. Consider this scenario using an SP30 as an example (SP20 values will of course be different, but the concept is the same) at various underclock settings: BTC exchange: $220USD/BTC electricity charge: $0.10/kWh difficulty: 44 billion 4400GH/s; 3000W; $332.14/month gross; $216.00/month electricity cost; $116.14/month profit 4000GH/s; 2320W; $301.94/month gross; $167.04/month electricity cost; $134.90/month profit3350GH/s; 1760W; $252.88/month gross; $126.72/month electricity cost; $126.16/month profit 2870GH/s; 1392W; $216.64/month gross; $100.22/month electricity cost; $116.42/month profit 2590GH/s; 1232W; $195.51/month gross; $88.70/month electricity cost; $106.80/month profit So if you happen to have an SP30 and you are being charged $0.10/kWh, it should be clear that you are best off underclocking it to around 4000GH/s. Your monthly electricity charge will be $167.04 on an income of $301.94, netting you a monthly profit of $134.90 per month. That's about the best you can do under those conditions with an SP30. Let's say you buy 4 SP30's and run them at stock. That is, 4400GH/s using 3000W each (12000W). You can expect each one to net you $116.14 monthly; so for all 4 you will earn $464.56 per month. 4 SP30's will cost you roughly 4x $2000 or $8000. So (assuming BTC value stays the same and difficulty remains the same) it will take you $8000/$464.56 or 17.2 months to recover your investment. Not particularly good. Now let's say you buy 5 SP30's and underclock them to 4000GH using 2320W each (11600W ... very close to the 12000W above). You can expect each one to net you $134.90 monthly; so for all 5 you will earn $674.50 per month. Now, 5 SP30's will cost you 5x $2000 or $10000. So it will take you $10000/$674.50 or 14.8 months to recover your investment. This is better than what the 4 stock SP30's can do. So for this example, with the assumptions and costs as above, 5 underclocked SP30's will recover their $10,000 cost sooner than 4 stock SP30's can recover their $8000 cost. In other words: you are better off underclocking in this scenario. I haven't run the numbers for an SP20; I'll leave that to you. But it seems you've already run the analysis. Most folks will have an electricity rate higher than $0.10/kWh so their net monthly profits will be lower, their recovery time will be longer, and the "sweet spot" will be an even more aggressive underclock. I see that DiCE1904 has provided a handy table of SP20 values.
|
|
|
|
dimitri111
Newbie
Offline
Activity: 24
Merit: 0
|
|
January 21, 2015, 04:15:40 AM |
|
Hey guys,
I got my 3 SP20 Jackson's about a week ago and i love them, since they were in my living room - I only ran them at night, now I moved them to an empty room. So when i came home today and power on my rigs, i could not connect to myminer.io. Is Spondoolies-tech server down?
Thank you,
|
|
|
|
MrTeal
Legendary
Offline
Activity: 1274
Merit: 1004
|
|
January 21, 2015, 04:35:17 AM |
|
I actually made up a formula that spits out a target voltage that maximizes profit given your parameters. I'm looking at automating the setting of voltage on a daily basis to maximize profitability. For those interested, I whipped up a quick formula for calculating your maximum profit with an SP20 using 2GOOD's efficiency numbers. LN((20116.26*(BR/D)*P)/(0.00417093689255513*C))/0.001368351 BR = Block reward D = Difficulty P = Bitcoin Price C = cost of electricity (in $/kWh)
There's no bound on it, so obviously if it gives you >1700 or <800, you'd have to limit it.
There's a little discussion after about it, but it's based on 2GOOD's efficiency/voltage numbers. I used it to create a formula for profit as a function of voltage, then took the derivative of that and set it to zero to find the maximum. As always, YMMV.
|
|
|
|
DiCE1904
Legendary
Offline
Activity: 1118
Merit: 1002
|
|
January 21, 2015, 04:55:03 AM |
|
Or, as the artificial intelligence WOPR concludes in the the 1983 movie WarGames, sometimes "the only winning move is not to play."
Okay, that I understand So I was just using an arbitrary $.10/kwh rate. You are saying there may be a sweet spot where rates are high enough that it would make more sense to run underclocked rather than stock. I'd have to see hard numbers to get it when trying to compare apples to apples. For the additional capital outlay, I don't see the underclocking making up the difference in a reasonable time. Since in my example I'm going to pay the same amount for electricity whether I'm using underclocked units or stock units I'm having a hard time seeing the underclock advantage for best dollars spent. It is good that you are going through the mental exercise. Consider this scenario using an SP30 as an example (SP20 values will of course be different, but the concept is the same) at various underclock settings: BTC exchange: $220USD/BTC electricity charge: $0.10/kWh difficulty: 44 billion 4400GH/s; 3000W; $332.14/month gross; $216.00/month electricity cost; $116.14/month profit 4000GH/s; 2320W; $301.94/month gross; $167.04/month electricity cost; $134.90/month profit3350GH/s; 1760W; $252.88/month gross; $126.72/month electricity cost; $126.16/month profit 2870GH/s; 1392W; $216.64/month gross; $100.22/month electricity cost; $116.42/month profit 2590GH/s; 1232W; $195.51/month gross; $88.70/month electricity cost; $106.80/month profit So if you happen to have an SP30 and you are being charged $0.10/kWh, it should be clear that you are best off underclocking it to around 4000GH/s. Your monthly electricity charge will be $167.04 on an income of $301.94, netting you a monthly profit of $134.90 per month. That's about the best you can do under those conditions with an SP30. Let's say you buy 4 SP30's and run them at stock. That is, 4400GH/s using 3000W each (12000W). You can expect each one to net you $116.14 monthly; so for all 4 you will earn $464.56 per month. 4 SP30's will cost you roughly 4x $2000 or $8000. So (assuming BTC value stays the same and difficulty remains the same) it will take you $8000/$464.56 or 17.2 months to recover your investment. Not particularly good. Now let's say you buy 5 SP30's and underclock them to 4000GH using 2320W each (11600W ... very close to the 12000W above). You can expect each one to net you $134.90 monthly; so for all 5 you will earn $674.50 per month. Now, 5 SP30's will cost you 5x $2000 or $10000. So it will take you $10000/$674.50 or 14.8 months to recover your investment. This is better than what the 4 stock SP30's can do. So for this example, with the assumptions and costs as above, 5 underclocked SP30's will recover their $10,000 cost sooner than 4 stock SP30's can recover their $8000 cost. In other words: you are better off underclocking in this scenario. I haven't run the numbers for an SP20; I'll leave that to you. But it seems you've already run the analysis. Most folks will have an electricity rate higher than $0.10/kWh so their net monthly profits will be lower, their recovery time will be longer, and the "sweet spot" will be an even more aggressive underclock. I see that DiCE1904 has provided a handy table of SP20 values. Thank you very much for your post Epoch, it helps running through the numbers like that. The more ya know
|
|
|
|
philipma1957
Legendary
Offline
Activity: 4158
Merit: 8055
'The right to privacy matters'
|
|
January 21, 2015, 04:55:11 AM |
|
Hey guys,
I got my 3 SP20 Jackson's about a week ago and i love them, since they were in my living room - I only ran them at night, now I moved them to an empty room. So when i came home today and power on my rigs, i could not connect to myminer.io. Is Spondoolies-tech server down?
Thank you,
I can not sign on so you may be correct.
|
|
|
|
dimitri111
Newbie
Offline
Activity: 24
Merit: 0
|
|
January 21, 2015, 04:58:46 AM |
|
Hey guys,
I got my 3 SP20 Jackson's about a week ago and i love them, since they were in my living room - I only ran them at night, now I moved them to an empty room. So when i came home today and power on my rigs, i could not connect to myminer.io. Is Spondoolies-tech server down?
Thank you,
I can not sign on so you may be correct. Thank you very much for your response.
|
|
|
|
qumqats
Newbie
Offline
Activity: 12
Merit: 0
|
|
January 21, 2015, 08:35:00 AM |
|
Thank you for this research and posting it. This reinforces my impression that the sweet-spot or elbow of the speed vs power is around or slightly less than 1300Gh/s
|
|
|
|
johnyj
Legendary
Offline
Activity: 1988
Merit: 1012
Beyond Imagination
|
|
January 21, 2015, 08:42:21 AM |
|
Or, as the artificial intelligence WOPR concludes in the the 1983 movie WarGames, sometimes "the only winning move is not to play."
Okay, that I understand So I was just using an arbitrary $.10/kwh rate. You are saying there may be a sweet spot where rates are high enough that it would make more sense to run underclocked rather than stock. I'd have to see hard numbers to get it when trying to compare apples to apples. For the additional capital outlay, I don't see the underclocking making up the difference in a reasonable time. Since in my example I'm going to pay the same amount for electricity whether I'm using underclocked units or stock units I'm having a hard time seeing the underclock advantage for best dollars spent. It is good that you are going through the mental exercise. Consider this scenario using an SP30 as an example (SP20 values will of course be different, but the concept is the same) at various underclock settings: BTC exchange: $220USD/BTC electricity charge: $0.10/kWh difficulty: 44 billion 4400GH/s; 3000W; $332.14/month gross; $216.00/month electricity cost; $116.14/month profit 4000GH/s; 2320W; $301.94/month gross; $167.04/month electricity cost; $134.90/month profit3350GH/s; 1760W; $252.88/month gross; $126.72/month electricity cost; $126.16/month profit 2870GH/s; 1392W; $216.64/month gross; $100.22/month electricity cost; $116.42/month profit 2590GH/s; 1232W; $195.51/month gross; $88.70/month electricity cost; $106.80/month profit So if you happen to have an SP30 and you are being charged $0.10/kWh, it should be clear that you are best off underclocking it to around 4000GH/s. Your monthly electricity charge will be $167.04 on an income of $301.94, netting you a monthly profit of $134.90 per month. That's about the best you can do under those conditions with an SP30. Let's say you buy 4 SP30's and run them at stock. That is, 4400GH/s using 3000W each (12000W). You can expect each one to net you $116.14 monthly; so for all 4 you will earn $464.56 per month. 4 SP30's will cost you roughly 4x $2000 or $8000. So (assuming BTC value stays the same and difficulty remains the same) it will take you $8000/$464.56 or 17.2 months to recover your investment. Not particularly good. Now let's say you buy 5 SP30's and underclock them to 4000GH using 2320W each (11600W ... very close to the 12000W above). You can expect each one to net you $134.90 monthly; so for all 5 you will earn $674.50 per month. Now, 5 SP30's will cost you 5x $2000 or $10000. So it will take you $10000/$674.50 or 14.8 months to recover your investment. This is better than what the 4 stock SP30's can do. So for this example, with the assumptions and costs as above, 5 underclocked SP30's will recover their $10,000 cost sooner than 4 stock SP30's can recover their $8000 cost. In other words: you are better off underclocking in this scenario. I haven't run the numbers for an SP20; I'll leave that to you. But it seems you've already run the analysis. Most folks will have an electricity rate higher than $0.10/kWh so their net monthly profits will be lower, their recovery time will be longer, and the "sweet spot" will be an even more aggressive underclock. I see that DiCE1904 has provided a handy table of SP20 values. Nice calculation, if there is also a variable of electricity cost, then it will make a contour map to show the sweet spot of the voltage settings under different electricity charge And, another variable is the amount of mined bitcoins that is being sold immediately. Pro-bitcoin miners never sell more than 10% of mined coins per year, but pro-fiat-money miners might sell majority of their coins right away
|
|
|
|
Newar
Legendary
Offline
Activity: 1358
Merit: 1001
https://gliph.me/hUF
|
|
January 21, 2015, 09:54:13 AM |
|
Hey guys,
I got my 3 SP20 Jackson's about a week ago and i love them, since they were in my living room - I only ran them at night, now I moved them to an empty room. So when i came home today and power on my rigs, i could not connect to myminer.io. Is Spondoolies-tech server down?
Thank you,
I can not sign on so you may be correct. Simply browsing to the IP of the miner works for me.
|
|
|
|
mdude77
Legendary
Offline
Activity: 1540
Merit: 1001
|
|
January 21, 2015, 12:15:15 PM |
|
And, another variable is the amount of mined bitcoins that is being sold immediately. Pro-bitcoin miners never sell more than 10% of mined coins per year, but pro-fiat-money miners might sell majority of their coins right away
That's an odd statement. M
|
I mine at Kano's Pool because it pays the best and is completely transparent! Come join me!
|
|
|
lulu2003
|
|
January 21, 2015, 02:08:17 PM |
|
push gathered firmware suggestions in a list and edit that list and extend
current firmware: 2.6.7 TEST
- adding start fan speed for auto-mode - tail option for showing last x rows of log files. - using help screen for sort of change log in the firmware - may be watchdog checking network state. had hung system after network outage, not reachable due to DHCP. can be probably worked around by fixed IP. - scheduling certains mining speeds(voltages) <- will try to implement - individual voltage limits as right now for the back chips <- will consider / note: can be done quite well with max. PSU watts at least per loop - Timezone "bug"? first page graphs are always UTC? <- will consider - enable API in own subnet/net by web interface <- common! change /etc/cgminer.conf.template with ssh it seems to works not straight forward at least not for all - include fan speed in API datasets <- not likely, use remote ssh to change /etc/mg_custom_rate fan speed OUTPUT is obviosly meant - other pool options than failover only <- common! change /etc/cgminer.conf.template with ssh - Graphs on the first webguy page, there should be a button to disable some/all of them. <- you can reset them with a button. Disable them in /etc/cron.d with ssh
- Poll page could be managed on the front page with up&down buttons on each pool - GUI switches to turn off SSH would also be a good thing, plus for advanced users some sort of iptables in the OS for security. <- Use big random password from UI to disable it. Will consider adding iptables.
- true fan speed control by target temp
- change host name in web interface
- sum of PSU(unknown) input watts in addition to DC2DC watts, either on ASIC Stats page or main page, as for estimate of power at wall or PSU efficiency <- will implement in 2.6.7. For sp20 I assume 90% efficiency
- show ASIC Stats
|
|
|
|
aclass
|
|
January 21, 2015, 02:11:35 PM |
|
Tried but it often goes above it and the back chips get too hot push - individual voltage limits as right now for the back chips <- will consider / note: can be done quite well with max. PSU watts at least per loop
|
|
|
|
|