drakoin, how about a middle ground between (B2) and (C) above?
Perhaps 0.95V on the core with your optimized Stilt bios allows core ~950Mhz and VRAM 1500Mhz, leading to ~650 Kh/s and ~3.3 Kh/W
Topic: SAVE money - undervolting 7970 XFX, self made VBE7 bios = TABLE OF CONTENTS = - page 6. (Read 56567 times)
January 30, 2014, 04:20:04 AM
January 30, 2014, 12:35:15 AM
If you cannot donate, and have no business opportunities to share,
please let you rig run a bit for me cgminer --scrypt with this cgminer.conf file:
or
or
on any other autochangingmultipool like wafflepool, clevermining, etc. with my address BTC 15UipUK3kGcArJzRyHbVKrwG5TsqYi1b9N
Thx a 10^6
please let you rig run a bit for me cgminer --scrypt with this cgminer.conf file:
Code:
{"pools" : [{
"name" : "support mining for drakoin at middlecoin.com",
"url" : "stratum+tcp://eu.middlecoin.com:3333",
"user" : "15UipUK3kGcArJzRyHbVKrwG5TsqYi1b9N",
"pass" : "x", "pool-priority" : "0"
},{
"name" : "support mining for drakoin at doge.crypto49er.com",
"url" : "stratum+tcp://doge.crypto49er.com:9555",
"user" : "DFPA5UXRH8jzJGEGWtJQVq27jyf1Nd4KEt",
"pass" : "x", "pool-priority" : "1"
}]}or
Code:
cgminer.exe --scrypt -u 15UipUK3kGcArJzRyHbVKrwG5TsqYi1b9N -p x -o stratum+tcp://eu.middlecoin.com:3333 -u DFPA5UXRH8jzJGEGWtJQVq27jyf1Nd4KEt -p x -o stratum+tcp://doge.crypto49er.com:9555
or
on any other autochangingmultipool like wafflepool, clevermining, etc. with my address BTC 15UipUK3kGcArJzRyHbVKrwG5TsqYi1b9N
Thx a 10^6
January 30, 2014, 12:19:22 AM
undervolting not only saved me on my powerbill, but I also went down 10 degrees or so celcius with each 7950.
Undervolted from 1.25 to 1.09
Super, happy to hear that! Lots of energy saving as a result. And much less heat, yes.Undervolted from 1.25 to 1.09
Very good thread!
Thank you very much. That's always nice to read :-)
Support me if you can.
January 30, 2014, 12:16:18 AM
I've tried a lot of combinations and I’ve realized that if I don’t set -g 2, they works on low performance. But If I set -g 2, one of the three card doesn't work.
Lower the thread-concurrency for that card, and then try again with -g 2At the end, I've set -g 2 for two cards, and -g 1 for one card.
How do you do that?With starting several instances of cgminer?
January 27, 2014, 11:33:10 PM
undervolting not only saved me on my powerbill, but I also went down 10 degrees or so celcius with each 7950.
Undervolted from 1.25 to 1.09
Undervolted from 1.25 to 1.09
January 27, 2014, 08:54:32 PM
Very good thread!
Not entirely sure, that is TheStilt's domain. Memory? Memory controller? But you have Elpida, too. So it should work.
Elpida is a memory manufacturer from Japan or Taiwan, not sure January 27, 2014, 08:33:01 PM
I've tried a lot of combinations and I’ve realized that if I don’t set -g 2, they works on low performance. But If I set -g 2, one of the three card doesn't work.
At the end, I've set -g 2 for two cards, and -g 1 for one card.
Does anybody know which is the reason regarding this behavior?
At the end, I've set -g 2 for two cards, and -g 1 for one card.
Does anybody know which is the reason regarding this behavior?
January 27, 2014, 12:47:57 PM
I don't understand. What does a reboot change about the voltage?
Have you undervolted the BIOS?
Yes, I've undervolted my BiOS following your steps.Have you undervolted the BIOS?
Choose the "(B) CC=1025 MC=1635" settings instead if you are using my mod3-BIOS (from posting #45), they work better there.
I've try also the above values and i get 500khas/s
Try -g 2
or "gpu-threads" : "2" in the .conf
I using -g 1 because i have 3 cards and i used -g 2 just runing 2 cards
Thanks a lot
January 27, 2014, 12:23:18 PM
I answer not as a PM because other people might benefit, too.
Have you undervolted the BIOS?
It's mentioned in posting #67 "2) then modified it myself to enable undervolting (like in posting #45), and flashed that into my card with atiflash."
Because if not, the voltage is locked, and cgminer cannot control it.
Those "(B2) CC=1054 MC=1500" are optimized for really high khash with the TheStilt-mod5-BIOS (posting #67).
Choose the "(B) CC=1025 MC=1635" settings instead if you are using my mod3-BIOS (from posting #45), they work better there.
And if not happy, always search a bit to the left and right of the given values, your card might be slightly different. You can vary clockspeed & voltage from within the running cgminer - you know that, right?
Try -g 2
or "gpu-threads" : "2" in the .conf
Good cgminer.conf with these settings are in the download in posting #45
or a similiar one is mentioned in posting #40 already.
That does the trick, doesn't it?
:-)
I can vary the voltage but not using cgminer, if i try to change the voltage through cgminer i must reboot the rig.
I don't understand. What does a reboot change about the voltage?Have you undervolted the BIOS?
It's mentioned in posting #67 "2) then modified it myself to enable undervolting (like in posting #45), and flashed that into my card with atiflash."
Because if not, the voltage is locked, and cgminer cannot control it.
Now i get 500 khas/s per card with:
cgminer --gpu-engine 1054 --gpu-memclock 1500 --gpu-vddc 1.065 --temp-target 75 --temp-overheat 85 --temp-cutoff 93 --gpu-powertune 20 --thread-concurrency 8192 -I 13 -g 1 -w 256 --lookup-gap 2 --scrypt -o stratum+tcp://xxx -u x.x -p x
cgminer --gpu-engine 1054 --gpu-memclock 1500 --gpu-vddc 1.065 --temp-target 75 --temp-overheat 85 --temp-cutoff 93 --gpu-powertune 20 --thread-concurrency 8192 -I 13 -g 1 -w 256 --lookup-gap 2 --scrypt -o stratum+tcp://xxx -u x.x -p x
Those "(B2) CC=1054 MC=1500" are optimized for really high khash with the TheStilt-mod5-BIOS (posting #67).
Choose the "(B) CC=1025 MC=1635" settings instead if you are using my mod3-BIOS (from posting #45), they work better there.
And if not happy, always search a bit to the left and right of the given values, your card might be slightly different. You can vary clockspeed & voltage from within the running cgminer - you know that, right?
Now i get 500 khas/s per card
Try -g 2
or "gpu-threads" : "2" in the .conf
Good cgminer.conf with these settings are in the download in posting #45
or a similiar one is mentioned in posting #40 already.
That does the trick, doesn't it?
:-)
January 23, 2014, 06:51:13 PM
Edit: Just had a longer look. Very nice. So for 80 euros I would have the data in my computer? Or only in some proprietary online platform?
Still, it's overkill. I just want a clamp and a board with a USB connection - and a programmable API.
Every hint welcome!
Been looking for such a device too. Efergy has been promissing an API fornages now, but at least it is something. But indeed, a clamp, usb driver, and go... Hmm, might be possible with an Arduino... http://openenergymonitor.org/emon Let the reading begin
January 23, 2014, 06:38:56 PM
Just an idea for your gaming: Perhaps it makes sense to optimize an Afterburner gaming profile, starting with the cgminer optimal settings, and try that for gaming, too? Please do a comparison. For gaming the FPS is probably the order parameter? Try 1050, then try 1100 MHz, and see which one results in more FPS. Perhaps we can also optimize for gaming experience now?
Have fun!
And ... bump to the next page
Now that I bought the 2 Dual x 280x Sapphires more than 2 weeks ago, I wish I didn't buy the Gigabyte, they hash faster and don't have any crashing problems nor do they use as much power as the Gigabyte (even when on default voltages and overclocked they use only slightly more power than the Gigabyte). The Gigabyte also suffers from artefacts while mining, I get grey squares all over the screen when mining, quiet annoying (it's a known problem with many of them). Got a 7970 Asus DCT comming tomorrow I'll let you know what the performance is of that one.
January 23, 2014, 06:29:40 PM
Nice numbers!
Very nice to see that there is a huge difference in powerdraw between super underclocked and max settings. Didn't expect that much of a difference!
For you the running cool is absolutely the best choice now that I see your numbers. You electricity is slightly more expensive, you probably don't have green power (save the planet yay! ;-)) and the difference in power draw from your card is great! I draw a lot less power at my highest kH/s rate (undervolted as far as I can) than you do. And that's including the system power for 2 cards when I am about do add another card (psu can handle 1 more) meaning I can divide the system power draw by 3 instead of 2.
I guess it depends on the card, as my card runs at only 65°C on max settings and isn't all that loud. The other 2 are in my shed outside so they run very cool aswell :-). But it is very interesting stuff, for other people with lower earnings it might be very profitable indeed if they have cards like yours, to lower their voltages and clock speeds. Also when the summer comes and my 2 cards in the shed heat up I will probably remind myself to find that sweet spot, because they wont run that cool anymore :-( if I don't.
Very interesting indeed! You got me wondering right now how much memory speed actually helps in hashrate, as I think I will try and see what hashrates I can get with my mem @ 1800 and underclocking/volting the core. If the memspeed is very important I could potentially underclock/volt my card and lower my powerdrawings massively while still holding a high memory speed and therefore a high hashing rate. Must check somewhere next week.
Very nice to see that there is a huge difference in powerdraw between super underclocked and max settings. Didn't expect that much of a difference!
For you the running cool is absolutely the best choice now that I see your numbers. You electricity is slightly more expensive, you probably don't have green power (save the planet yay! ;-)) and the difference in power draw from your card is great! I draw a lot less power at my highest kH/s rate (undervolted as far as I can) than you do. And that's including the system power for 2 cards when I am about do add another card (psu can handle 1 more) meaning I can divide the system power draw by 3 instead of 2.
I guess it depends on the card, as my card runs at only 65°C on max settings and isn't all that loud. The other 2 are in my shed outside so they run very cool aswell :-). But it is very interesting stuff, for other people with lower earnings it might be very profitable indeed if they have cards like yours, to lower their voltages and clock speeds. Also when the summer comes and my 2 cards in the shed heat up I will probably remind myself to find that sweet spot, because they wont run that cool anymore :-( if I don't.
Very interesting indeed! You got me wondering right now how much memory speed actually helps in hashrate, as I think I will try and see what hashrates I can get with my mem @ 1800 and underclocking/volting the core. If the memspeed is very important I could potentially underclock/volt my card and lower my powerdrawings massively while still holding a high memory speed and therefore a high hashing rate. Must check somewhere next week.
January 23, 2014, 02:46:30 PM
So now I am off for a while.
Did you like this nerdy detective story?
Then have a quick look at posting #80 to see how happy you can make me as ...
... my2nd 3rd supporter 
Please
tip me if you can:
BTC 1EyyjBMMHjMfx6M3Ngu4sn5M4QJ6HAtWFG
[AUR] AevVj2wQSsM2TeAn8oACZY51fzLU7JYSF3
[BC] BPz72JMFkw4ymWSDe2Cdcp5mBZKSXBuk9i
[C2] SAYYYy9e92xGVuWdgCi995R1kywxmkvBWn
[DGB] DPxNvaRZLoeQGZpAmRRae2AdCLWNiQnzqm
[DOGE] DLE2HhboCsx5JbeLhRAGW7TQJze41dCgiK
[EAC] eaoPAMndxjTkvvFUUq5SuSxMVg6KGoEQYW
[LTC] LQt3Wky3b1v7xEaatVYw25n22hujKMa2yq
[MAX] mKyVKaKoRetNnEVZLoouogyS7iz7qj74uu
[MINT] Mfxyu7TRCDxco5sGn8EV1hNqF2A7Ut9oo4
[NOBL] 9XJfp3mUrrmmdus6HWPtmFftoEZ7vBeYfA
[PRT] PnKS4C8euVMzN4YZXEkYwyUXxpdGugsmQm
[QRK] QhfXtCetfQE6pnup59MW45au65vcMZYK88
[VTC] Vx2iYor2ir1Pq9QTALs7TRdE5kmW9aABD6
...then PM me so that I can THANK YOU ;-)
If you cannot donate, please let you rig run a while for me, or PM me a business opportunity, which looks profitable to invest.
Thx
Did you like this nerdy detective story?
Then have a quick look at posting #80 to see how happy you can make me as ...
... my
Please
tip me if you can:
BTC 1EyyjBMMHjMfx6M3Ngu4sn5M4QJ6HAtWFG
[AUR] AevVj2wQSsM2TeAn8oACZY51fzLU7JYSF3
[BC] BPz72JMFkw4ymWSDe2Cdcp5mBZKSXBuk9i
[C2] SAYYYy9e92xGVuWdgCi995R1kywxmkvBWn
[DGB] DPxNvaRZLoeQGZpAmRRae2AdCLWNiQnzqm
[DOGE] DLE2HhboCsx5JbeLhRAGW7TQJze41dCgiK
[EAC] eaoPAMndxjTkvvFUUq5SuSxMVg6KGoEQYW
[LTC] LQt3Wky3b1v7xEaatVYw25n22hujKMa2yq
[MAX] mKyVKaKoRetNnEVZLoouogyS7iz7qj74uu
[MINT] Mfxyu7TRCDxco5sGn8EV1hNqF2A7Ut9oo4
[NOBL] 9XJfp3mUrrmmdus6HWPtmFftoEZ7vBeYfA
[PRT] PnKS4C8euVMzN4YZXEkYwyUXxpdGugsmQm
[QRK] QhfXtCetfQE6pnup59MW45au65vcMZYK88
[VTC] Vx2iYor2ir1Pq9QTALs7TRdE5kmW9aABD6
...then PM me so that I can THANK YOU ;-)
If you cannot donate, please let you rig run a while for me, or PM me a business opportunity, which looks profitable to invest.
Thx
January 23, 2014, 01:56:20 PM
Summary, the "bottom line":
How to choose the best settings for your GPU
#99 A bit of math, and some new jargon to disentangle the observables
#101 cgminer parameters, and resulting measurements: Temp, fanspeed, hashrate, Watt usage --> and thus the monthly contribution to the electricity bill.
#102 Profit calculations of these settings for different market situations.
Main insights
:
Moreover (with calculations not shown here, just trust me
):
As I am willing to flash the BIOS
:
And finally, my personal decision
:
Done.
Thanks for all the help I got on the way. Amazing community of selfish AND helpful people *g* !
And thanks a lot for all your questions and doubts. They have helped a lot to make this better!!! Knowing now exactly that it's not much profit that I miss out on during those happy high earning days, with "cool" (C) instead of "hot" (B2) ... but I would always spend 66% more on electricity with "running hot" ... I have decided to go for (C) "running cool" almost always, and have the surplus of longer living hardware, lower sound pollution, less heat.
And a greener conscience :-)
How to choose the best settings for your GPU
#99 A bit of math, and some new jargon to disentangle the observables
#101 cgminer parameters, and resulting measurements: Temp, fanspeed, hashrate, Watt usage --> and thus the monthly contribution to the electricity bill.
#102 Profit calculations of these settings for different market situations.
Main insights
:- Depending on the market, different hardware usage is profitable:
- High earning days ask for a "hot machine", low earning days for a "cool machine".
- The threshold for my card is at 6€/MH/day, on days above that I gain more if I "run hot".
- The extra gain for "running hot" is not immense, so I might even consider "running cool" on slightly more profitable days.
- Because of less sound, less heat, less pollution.
- I would expect that "running cool" might result in less tear on the hardware, and prolong its life?
Moreover (with calculations not shown here, just trust me
):- an unoptimized card (0) must be switched off on bad days - it can actually cost you more electricity than it produces coins,
- for my card that happens if the market output drops below 3.46€/MH/day in coins
- So at least overclock to settings like (A), on locked cards that's all you can do without flashing the BIOS.
- But if the market drops below 3.11€/MH/day, it's also game over for settings (A). Then rather switch off.
- Or ...
As I am willing to flash the BIOS
:- if I go for "running hot", I choose (B2) the optimized psw scrypt kernel and TheStilt BIOS + my undervolting from #67
- because of the 18khash more it's (B2) not (B)
- and for "running cool", I choose the almost classical and now also Ferengi-proven sweet spot settings (C) praised in #37.
And finally, my personal decision
:- On very profitable days (like a crazy coinstart), it's just plain stupid to try to save energy by all means. So then (B2) for sure!
- But otherwise, I stay with my efficient ecological quiet cool sweet spot settings (C) most of the time.
Done.
Thanks for all the help I got on the way. Amazing community of selfish AND helpful people *g* !
And thanks a lot for all your questions and doubts. They have helped a lot to make this better!!! Knowing now exactly that it's not much profit that I miss out on during those happy high earning days, with "cool" (C) instead of "hot" (B2) ... but I would always spend 66% more on electricity with "running hot" ... I have decided to go for (C) "running cool" almost always, and have the surplus of longer living hardware, lower sound pollution, less heat.
And a greener conscience :-)
January 23, 2014, 01:16:15 PM
Saving energy might be second, but making profit is first. Let's study the
Profit for different tech settings, and for different market situations
My profit F depends on the earnings C ~ m*H (with m ~ market situation, H ~ my machine)
minus the electricity costs E, so let's look at that in four different market situations now:
with the same row meanings (0), (A), (B), (B2), (C) as in posting #101 = different optimization constraints of the same card.
How to read that table?
Bad days
Let's say we are in a sad 4€/day/MH phase with low earnings, because e.g. the BTC exchange rate might have crashed, or your favorite coin has just raised its difficulty. There are many reasons for a low mining conversion rate m (for new jargon see posting #99) ... but they all result in a low eurovalue of coins per megahash. You spend the same amount of electricity (see last column in posting #101 above), but you just get out less coins C, and thus less profit F.
Each field reads like "profit F (F/C% of C coins)" so (C) first column "34€ (49% of 70€ coins)" means
I make F=34€ profit because I mine C=70€ coins, with the electricity E = 36€ for that setting (C).
In that market situation, my (C) underclocked undervolted sweet spot clearly saves the day,
with better electrical efficiency AND MORE PROFIT than any other setting! And it's less volume, and heat.
Good days
Now, in contrast, let's look at a very happy 7€/day/MH phase, the same harvest is paid better. On those days, I don't want to loose out on a single khash, regardless how much electricity I have to spend for that. Now settings (B2) are better, have a look at the last column at the right, second last line: "92€ (61% of 151€ coins)" means: I make F=92€ profit from 151€ coins which renders the spent E=60€ electricity more profitable - than my setting (C) in which I would spend 24€ less on electricity, but make 29€ less coins.
Here's the same table again, with a wider range of market situation values, and dropping (A) and (B1) which I wouldn't choose anymore, anyways - because by now I know how to flash a BIOS and undervolt :-)
Middle days
The threshold between the regimes is for my card at around 6€/day/MH market mood - Above it pays off more to go for settings (B2), and below 6€/day/MH my sweet-spot-parameter set (C) is the most profitable ... just because ... especially in difficult times, better efficiency does pay off.
P.S.: Rounding errors contribute to +/-1 Euro, the actual calculations are done in much higher precision.
Profit for different tech settings, and for different market situations
My profit F depends on the earnings C ~ m*H (with m ~ market situation, H ~ my machine)
minus the electricity costs E, so let's look at that in four different market situations now:
| profit when 4€/day/MH | profit when 5€/day/MH | profit when 6€/day/MH | profit when 7€/day/MH | |
| (0) | 8€ (14% of 61€ coins) | 23€ (31% of 76€ coins) | 39€ (42% of 91€ coins) | 54€ (51% of 107€ coins) |
| (A) | 17€ (23% of 75€ coins) | 36€ (38% of 93€ coins) | 54€ (48% of 112€ coins) | 73€ (56% of 131€ coins) |
| (B) | 26€ (30% of 84€ coins) | 47€ (44% of 105€ coins) | 68€ (54% of 126€ coins) | 89€ (60% of 147€ coins) |
| (B2) | 27€ (31% of 86€ coins) | 48€ (45% of 108€ coins) | 70€ (54% of 130€ coins) | 92€ (61% of 151€ coins) |
| (C) | 34€ (49% of 70€ coins) | 52€ (59% of 87€ coins) | 69€ (66% of 105€ coins) | 86€ (71% of 122€ coins) |
How to read that table?
Bad days
Let's say we are in a sad 4€/day/MH phase with low earnings, because e.g. the BTC exchange rate might have crashed, or your favorite coin has just raised its difficulty. There are many reasons for a low mining conversion rate m (for new jargon see posting #99) ... but they all result in a low eurovalue of coins per megahash. You spend the same amount of electricity (see last column in posting #101 above), but you just get out less coins C, and thus less profit F.
Each field reads like "profit F (F/C% of C coins)" so (C) first column "34€ (49% of 70€ coins)" means
I make F=34€ profit because I mine C=70€ coins, with the electricity E = 36€ for that setting (C).
In that market situation, my (C) underclocked undervolted sweet spot clearly saves the day,
with better electrical efficiency AND MORE PROFIT than any other setting! And it's less volume, and heat.
Good days
Now, in contrast, let's look at a very happy 7€/day/MH phase, the same harvest is paid better. On those days, I don't want to loose out on a single khash, regardless how much electricity I have to spend for that. Now settings (B2) are better, have a look at the last column at the right, second last line: "92€ (61% of 151€ coins)" means: I make F=92€ profit from 151€ coins which renders the spent E=60€ electricity more profitable - than my setting (C) in which I would spend 24€ less on electricity, but make 29€ less coins.
Here's the same table again, with a wider range of market situation values, and dropping (A) and (B1) which I wouldn't choose anymore, anyways - because by now I know how to flash a BIOS and undervolt :-)
| . | profit when 3€/day/MH | profit when 6€/day/MH | profit when 9€/day/MH | profit when 12€/day/MH |
| (0) | -7 € | 39€ (42% of 91€ coins) | 84€ (62% of 137€ coins) | 130€ (71% of 183€ coins) |
| (B2) | 5€ (8% of 65€ coins) | 70€ (54% of 130€ coins) | 135€ (69% of 194€ coins) | 200€ (77% of 259€ coins) |
| (C) | 17€ (32% of 52€ coins) | 69€ (66% of 105€ coins) | 121€ (77% of 157€ coins) | 174€ (83% of 210€ coins) |
Middle days
The threshold between the regimes is for my card at around 6€/day/MH market mood - Above it pays off more to go for settings (B2), and below 6€/day/MH my sweet-spot-parameter set (C) is the most profitable ... just because ... especially in difficult times, better efficiency does pay off.
P.S.: Rounding errors contribute to +/-1 Euro, the actual calculations are done in much higher precision.
January 23, 2014, 01:09:32 PM
Updated numerical results
Reasons:
First all the optimal parameters, and the resulting measurements:
The last column is calculated to 30*24*(WATT+40)/1000*0.256 euros, because there
are 2 cards in that rig, each one sharing 40W of that "non-mining part of total power use".
The five settings are all optimized ones, but each one in different constraints:
(0) factory settings = out of the box; unoptimized 2.07 khash/Watt, no overclocking nor undervolting. Optimized for selling the card
(A) overclock = my best parameters before I started this thread a hundred postings ago, whew.
(B) overclock undervolt = saves energy & allows higher clockspeed at lower temp. Maximum attainable H=702 khash. Yes, that's an XFX :-(
(B2) overclock undervolt + TheStilt + psw (posting #67) = all pushed to the limits; hot, loud & high consumption. Much coin so wow.
(C) underclock undervolt sweetspot = non negligible loss of hashrate, superb efficiency 3.78 khash/Watt; low sound, temperature, very low Watt.
As you can guess by my comments, I am biased towards (C) - also because I keep the machine in the room where I am working & sleeping, to use it instead of oil heating. Looking at the new results now, I might switch it to (C) for sleeping and bad income days, but drive in it loud'n'hot (B2) on good income days.
Why? You can see best if you look at the costs calculations:
Reasons:
- I moved my cards from an old to a new rig, from 166W to 81W baseline offset (precisely called "non-mining part of total power use" by JCviggen).
- The new rig situation interestingly demanded slightly higher voltages necessary for a stable system.
- To also include the recent TheStilt BIOS experiments.
- To see the relation between the technical/ecological efficiency khash/Watt, and the economical profit = coins - electricity.
- To understand in which regions of the optimization landscape which approach gives the best results.
First all the optimal parameters, and the resulting measurements:
| Volt | CC | MC | Celsius | % fan | khash/s | Watt GPU | khash/Watt | euros electricity bill | |
| (0) | 1.175 | 925 | 1375 | 71 | 55 | 508 | 246 | 2.07 | 53€ |
| (A) | 1.175 | 907 | 1450 | 74 | 58 | 623 | 274 | 2.27 | 58€ |
| (B) | 1.053 | 1025 | 1635 | 74 | 64 | 702 | 278 | 2.53 | 59€ |
| (B2) | 1.065 | 1054 | 1500 | 74 | 62 | 720 | 283 | 2.54 | 60€ |
| (C) | 0.899 | 854 | 799 | 67 | 39 | 582 | 154 | 3.78 | 36€ |
The last column is calculated to 30*24*(WATT+40)/1000*0.256 euros, because there
are 2 cards in that rig, each one sharing 40W of that "non-mining part of total power use".
The five settings are all optimized ones, but each one in different constraints:
(0) factory settings = out of the box; unoptimized 2.07 khash/Watt, no overclocking nor undervolting. Optimized for selling the card
(A) overclock = my best parameters before I started this thread a hundred postings ago, whew.
(B) overclock undervolt = saves energy & allows higher clockspeed at lower temp. Maximum attainable H=702 khash. Yes, that's an XFX :-(
(B2) overclock undervolt + TheStilt + psw (posting #67) = all pushed to the limits; hot, loud & high consumption. Much coin so wow.
(C) underclock undervolt sweetspot = non negligible loss of hashrate, superb efficiency 3.78 khash/Watt; low sound, temperature, very low Watt.
As you can guess by my comments, I am biased towards (C) - also because I keep the machine in the room where I am working & sleeping, to use it instead of oil heating. Looking at the new results now, I might switch it to (C) for sleeping and bad income days, but drive in it loud'n'hot (B2) on good income days.
Why? You can see best if you look at the costs calculations:
January 23, 2014, 12:54:59 PM
I just solved my problem too that I had with my powerdraw not going down even when I undervolted. Turns out I had a driver problem on my windows machine.
Well I was on the beta drivers of AMD, and they apparently gave me problems or perhaps I had driver corruption I don't know rly. When I undervolted my card, it helped to draw less power, but only up to a certain point, after that lowering the voltages more didn't seem to do much. Thing is I installed the latest non-beta drivers now and now when I undervolt more it actually goes down some, not all that much, but still some.
I am currently running on 1.04V for 1040Mhz core and 1500Mhz ram. The system that I am undervolting draws 83W with the gpu disabled and 300W with the card hashing away.
Same here when disabled, 81W. All the other values below.
I have to say though, I can't play games with that vcore. It crashes, hashing is fine, but games need at least 1.15V but that's probably due to the fact that in games it runs at its full default speed of 1100Mhz instead of 1050MHz
I am not a gamer, but I would probably try to solve it by having an Afterburner profile for gaming with higher VDDC.that I use for hashing (If I increase the MHz above 1050 for hashing my hashrate actually drops, same for the memory).
Same here, my max khash rate I get at CC=1025 with tweaked original BIOS, and at CC=1054 with tweaked TheStilt BIOS.Just an idea for your gaming: Perhaps it makes sense to optimize an Afterburner gaming profile, starting with the cgminer optimal settings, and try that for gaming, too? Please do a comparison. For gaming the FPS is probably the order parameter? Try 1050, then try 1100 MHz, and see which one results in more FPS. Perhaps we can also optimize for gaming experience now?
Have fun!
And ... bump to the next page
January 23, 2014, 11:02:21 AM
So, finally I am back on this :-)
Very nice. I earned less. Much less.
And here is also the explanation for our different opinions.
You have been optimizing for a different region of the overall problem than me.
So we are both right, but in the two different realities "low / high earnings"
I know.
Profit is the only thing that counts.
Sounds a bit Ferengi, though :-)
But it's good that you insisted, thank you very much - now I see clearer as
I recalculated the whole thing, all the numerical empirics is published here in #101.
I think we should consider both.
Or put differently: For you not, but for me, khash/Watt does matter.
The used electrical power P is very physical. I am talking efficiency.
Let's try to see clearer:
There is a "physical" or "hashing conversion ratio" h ~ H/P which makes hashrate H out of Watts P
and a "mining conversion ratio" m ~ C/H which makes coins C out of hashrate H, depending on pool choice & luck.
In the end, we get coins C from electricity E like this
And profit F (for Ferengi ) is coins C minus electricity E
Now we have a new proportionality, between profit F and electricity E, and the factor is (m*h-1)
Our profitability limit is at m*h == 1. Below 1 means paying more for electricity than earning in coins.
Personally, I would really like to see much more F coming out than energy E put in, so
- but that's my personal taste.
Look at an extreme case:
Even with m*h = 1.1, you still make profit - but me, I wouldn't want to anymore, because I cannot responsibly do that, put in 100 euros electricity to make 10 Euros profit.
Your m*h is 3.49, isn't it?
We cannot really choose E freely - just in jumps of buying more cards :-)
Within one card, E is simply proportional to P
... which is a nontrivial function of (CC, MC, VDDC),
and P and H are interdependent.
And what can we say about the absolute hashrate H? Actually,
both H and P are nontrivial functions of our cgminer settings:
{H, P} = function (CC, MC, VDDC)
with h ~ H/P in khash/Watt my efficiency ratio
and H the absolute hashrate in khash/s
Ideally we would actually like to optimize efficiency h AND the absolute hashrate H, right?
At your highly successful m ~ 7.84 euros/day/MH, I can well understand that you rather want to push your H as high as possible, even if your P goes up even more, and thus your efficiency h = H/P decreases. With e.g. m ~ 5 Euros/day/MH, the situation already looks different. Real numbers are in in #101.
C ~ H
Then we can calculate differences in absolute earnings,
and differences in relative earnings (in relation to energy costs).
Let's optimize both; depending on the current m, different strategies might be profitable.
On that way, I had lost track that I am also lowering my absolute earnings C if I lower my hashrate H, even if at the same time I succeed in increasing the efficiency h of the machine. And if I am in the high earnings realm, I actually loose profit. I appreciate your reminder!
I had always hoped that we together find another sweet spot (CC, MC, volt) with a higher absolute hashrate H, and still very good efficiency h = H/P but that hasn't happened (yet). The search is still on, people - go and test your cards in unusual clock settings!
If you want to please check my math. As I was swapping variables a bit to make it clearer, there might be a typo still.
I have added 40W, that's half of the baseline consumption, because I have two cards in there.
For 4 or even 6 cards, it's slightly better, of course.
Enough theory, we want to see numbers now!
I have 5.1 euros per day for electricity (slightly less, but I rounded it up), I have 2269kH/s which nets me on average per 24h 17.8 euro/day with current ratings
Thanks a lot for the numbers - that is 7.84 euros/day/MH. Very nice. I earned less. Much less.
And here is also the explanation for our different opinions.
You have been optimizing for a different region of the overall problem than me.
So we are both right, but in the two different realities "low / high earnings"
You need to compare earnings vs spendings and nothing more.
I know.
Profit is the only thing that counts.
Sounds a bit Ferengi, though :-)
But it's good that you insisted, thank you very much - now I see clearer as
I recalculated the whole thing, all the numerical empirics is published here in #101.
I see all of you in this forum talking about kH/W, but that doesn't matter.
I think we should consider both.
Or put differently: For you not, but for me, khash/Watt does matter.
The used electrical power P is very physical. I am talking efficiency.
Let's try to see clearer:
There is a "physical" or "hashing conversion ratio" h ~ H/P which makes hashrate H out of Watts P
and a "mining conversion ratio" m ~ C/H which makes coins C out of hashrate H, depending on pool choice & luck.
In the end, we get coins C from electricity E like this
Code:
C = m *h * E
And profit F (for Ferengi
) is coins C minus electricity ECode:
F = C - E
= m*h*E - E
= (m*h -1) * E
Now we have a new proportionality, between profit F and electricity E, and the factor is (m*h-1)
Our profitability limit is at m*h == 1. Below 1 means paying more for electricity than earning in coins.
Personally, I would really like to see much more F coming out than energy E put in, so
Code:
F >> E or m*h >> 2
Look at an extreme case:
Even with m*h = 1.1, you still make profit - but me, I wouldn't want to anymore, because I cannot responsibly do that, put in 100 euros electricity to make 10 Euros profit.
Your m*h is 3.49, isn't it?
I have 5.1 euros per day for electricity ... which nets me on average per 24h
17.8 euro/day with current ratings (average over last 17 days ...)
(17.8-5.1)/5.1 = 2.49 That's cool. Very cool. Good m, and good h, I guess.17.8 euro/day with current ratings (average over last 17 days ...)
We cannot really choose E freely - just in jumps of buying more cards :-)
Within one card, E is simply proportional to P
... which is a nontrivial function of (CC, MC, VDDC),
and P and H are interdependent.
And what can we say about the absolute hashrate H? Actually,
both H and P are nontrivial functions of our cgminer settings:
{H, P} = function (CC, MC, VDDC)
with h ~ H/P in khash/Watt my efficiency ratio
and H the absolute hashrate in khash/s
Ideally we would actually like to optimize efficiency h AND the absolute hashrate H, right?
At your highly successful m ~ 7.84 euros/day/MH, I can well understand that you rather want to push your H as high as possible, even if your P goes up even more, and thus your efficiency h = H/P decreases. With e.g. m ~ 5 Euros/day/MH, the situation already looks different. Real numbers are in in #101.
Meaning when you lower your total hashrate 10% your total earnings will lower 10%.
True.C ~ H
Then we can calculate differences in absolute earnings,
and differences in relative earnings (in relation to energy costs).
Let's optimize both; depending on the current m, different strategies might be profitable.
For every euro you earn less per day, you have to spend more than a euro less per day or it's less profitable.
And that's exactly where my thinking lead me to khash/Watt being the perfect measure to optimize.On that way, I had lost track that I am also lowering my absolute earnings C if I lower my hashrate H, even if at the same time I succeed in increasing the efficiency h of the machine. And if I am in the high earnings realm, I actually loose profit. I appreciate your reminder!
I had always hoped that we together find another sweet spot (CC, MC, volt) with a higher absolute hashrate H, and still very good efficiency h = H/P but that hasn't happened (yet). The search is still on, people - go and test your cards in unusual clock settings!
If you want to please check my math. As I was swapping variables a bit to make it clearer, there might be a typo still.
And when you do your calculations, make sure you use TOTAL system power, because only your card's powerdraw is useless.
When you pay your bills, you have to pay for total powerdraw, not just your cards.
Sure.When you pay your bills, you have to pay for total powerdraw, not just your cards.
I have added 40W, that's half of the baseline consumption, because I have two cards in there.
For 4 or even 6 cards, it's slightly better, of course.
So lowering powerdraw by 25% for every 10% hashpower lost means lowering total power draw by 25%.
I understood you ... but have a look at the numerics in posting #101 and #102 please.Enough theory, we want to see numbers now!
January 23, 2014, 11:01:05 AM
@Dracoin, look at this link: http://efergy.com/eu/ I have some of their products (Efergy Engage and E2), and they work quite well. That is, if you don't have any solar panels, because the device can't differentiate between power going from the grid and power going to the grid. Quite an eye opener to have one installed..... (And there is always pvoutput, again, if you have solar panels and a smart meter installed: http://www.pvoutput.org/intraday.jsp?id=22911&sid=20798 )
Cool, thanks for that hint. Edit: Just had a longer look. Very nice. So for 80 euros I would have the data in my computer? Or only in some proprietary online platform?
Still, it's overkill. I just want a clamp and a board with a USB connection - and a programmable API.
Every hint welcome!
January 23, 2014, 10:34:01 AM
Slightly off topic:
A cool idea, to use the powermeter of the whole flat. :-)
@Dracoin, look at this link: http://efergy.com/eu/
I have some of their products (Efergy Engage and E2), and they work quite well. That is, if you don't have any solar panels, because the device can't differentiate between power going from the grid and power going to the grid. Quite an eye opener to have one installed..... (And there is always pvoutput, again, if you have solar panels and a smart meter installed: http://www.pvoutput.org/intraday.jsp?id=22911&sid=20798 )
A cool idea, to use the powermeter of the whole flat. :-)
@Dracoin, look at this link: http://efergy.com/eu/
I have some of their products (Efergy Engage and E2), and they work quite well. That is, if you don't have any solar panels, because the device can't differentiate between power going from the grid and power going to the grid. Quite an eye opener to have one installed..... (And there is always pvoutput, again, if you have solar panels and a smart meter installed: http://www.pvoutput.org/intraday.jsp?id=22911&sid=20798 )
Jump to: