bitcoindaddy
|
|
April 27, 2012, 03:36:11 PM |
|
Is anyone sucessfully running their BFL Singles against p2pool? I understand the frequent LONGPOLL's cause a problem with BFL's 5 second calculation. (By successful, I don't just mean that it works, I mean it works efficiently and you are making the normal amount of coin as expected). If so, can you share your settings for CGMiner?
|
|
|
|
SgtSpike
Legendary
Offline
Activity: 1400
Merit: 1005
|
|
April 27, 2012, 03:39:07 PM |
|
Is anyone sucessfully running their BFL Singles against p2pool? I understand the frequent LONGPOLL's cause a problem with BFL's 5 second calculation. (By successful, I don't just mean that it works, I mean it works efficiently and you are making the normal amount of coin as expected). If so, can you share your settings for CGMiner?
It's impossible to run it with p2pool right now...
|
|
|
|
Epoch
Legendary
Offline
Activity: 922
Merit: 1003
|
|
April 27, 2012, 03:46:23 PM |
|
Is anyone sucessfully running their BFL Singles against p2pool? I understand the frequent LONGPOLL's cause a problem with BFL's 5 second calculation. (By successful, I don't just mean that it works, I mean it works efficiently and you are making the normal amount of coin as expected). If so, can you share your settings for CGMiner?
As SgtSpike mentioned, the issue cannot be fixed with cgminer settings (or any mining software settings) ... BFL must create/release new firmware (which they have not yet done) for the Singles to be able to work 'successfully' with p2pool. Until that happens, you must choose a different pool. Or choose different mining hardware.
|
|
|
|
kano
Legendary
Offline
Activity: 4592
Merit: 1851
Linux since 1997 RedHat 4
|
|
April 28, 2012, 04:24:57 AM |
|
Is anyone sucessfully running their BFL Singles against p2pool? I understand the frequent LONGPOLL's cause a problem with BFL's 5 second calculation. (By successful, I don't just mean that it works, I mean it works efficiently and you are making the normal amount of coin as expected). If so, can you share your settings for CGMiner?
As SgtSpike mentioned, the issue cannot be fixed with cgminer settings (or any mining software settings) ... BFL must create/release new firmware (which they have not yet done) for the Singles to be able to work 'successfully' with p2pool. Until that happens, you must choose a different pool. Or choose different mining hardware. Yeah I've explained it a few times already (elsewhere?) but I'll do it again The BFL does not report shares until it finishes processing the full nonce range. So about 5.2 seconds (assuming 830MH/s) after it starts on work it replies with the answer(s). If during that 5.2 seconds you get an LP then anything it did during that 5.2 seconds is wasted. What that means is that on average for every LP your BFL will have wasted 2.6 seconds processing. The miner code should/will then of course abort the work so that it doesn't continue to finish the remainder of the 5.2 seconds, but the work done up to the abort is wasted since you cannot ask it if it found a share (and it didn't tell you even though it would have been a valid share if it had already found it) Since LP's occur on expected average of every 10 seconds ... and on expected average you waste 2.6 seconds processing ... that's 26% of your processing is thrown away (and you can't avoid it)
|
|
|
|
amazingrando
|
|
April 28, 2012, 04:46:06 AM |
|
Is anyone sucessfully running their BFL Singles against p2pool? I understand the frequent LONGPOLL's cause a problem with BFL's 5 second calculation. (By successful, I don't just mean that it works, I mean it works efficiently and you are making the normal amount of coin as expected). If so, can you share your settings for CGMiner?
As SgtSpike mentioned, the issue cannot be fixed with cgminer settings (or any mining software settings) ... BFL must create/release new firmware (which they have not yet done) for the Singles to be able to work 'successfully' with p2pool. Until that happens, you must choose a different pool. Or choose different mining hardware. Yeah I've explained it a few times already (elsewhere?) but I'll do it again The BFL does not report shares until it finishes processing the full nonce range. So about 5.2 seconds (assuming 830MH/s) after it starts on work it replies with the answer(s). If during that 5.2 seconds you get an LP then anything it did during that 5.2 seconds is wasted. What that means is that on average for every LP your BFL will have wasted 2.6 seconds processing. The miner code should/will then of course abort the work so that it doesn't continue to finish the remainder of the 5.2 seconds, but the work done up to the abort is wasted since you cannot ask it if it found a share (and it didn't tell you even though it would have been a valid share if it had already found it) Since LP's occur on expected average of every 10 seconds ... and on expected average you waste 2.6 seconds processing ... that's 26% of your processing is thrown away (and you can't avoid it) Does this mean that you would expect with a BFL FPGA to generate 26% fewer shares at a given hash rate than a GPU?
|
Bitbond - 105% PPS mining bond - mining payouts without buying hardware
|
|
|
DeathAndTaxes
Donator
Legendary
Offline
Activity: 1218
Merit: 1079
Gerald Davis
|
|
April 28, 2012, 05:00:27 AM |
|
Is anyone sucessfully running their BFL Singles against p2pool? I understand the frequent LONGPOLL's cause a problem with BFL's 5 second calculation. (By successful, I don't just mean that it works, I mean it works efficiently and you are making the normal amount of coin as expected). If so, can you share your settings for CGMiner?
As SgtSpike mentioned, the issue cannot be fixed with cgminer settings (or any mining software settings) ... BFL must create/release new firmware (which they have not yet done) for the Singles to be able to work 'successfully' with p2pool. Until that happens, you must choose a different pool. Or choose different mining hardware. Yeah I've explained it a few times already (elsewhere?) but I'll do it again The BFL does not report shares until it finishes processing the full nonce range. So about 5.2 seconds (assuming 830MH/s) after it starts on work it replies with the answer(s). If during that 5.2 seconds you get an LP then anything it did during that 5.2 seconds is wasted. What that means is that on average for every LP your BFL will have wasted 2.6 seconds processing. The miner code should/will then of course abort the work so that it doesn't continue to finish the remainder of the 5.2 seconds, but the work done up to the abort is wasted since you cannot ask it if it found a share (and it didn't tell you even though it would have been a valid share if it had already found it) Since LP's occur on expected average of every 10 seconds ... and on expected average you waste 2.6 seconds processing ... that's 26% of your processing is thrown away (and you can't avoid it) Does this mean that you would expect with a BFL FPGA to generate 26% fewer shares at a given hash rate than a GPU? No it will produce the expected # of shares however you will have a 26% stale rate. Actually a little higher because you have the normal stales due to network propogation too. This is borne out in observations by users reporting a DOA & orphan combined rate of something like 30% to 40% of hashing power.
|
|
|
|
kano
Legendary
Offline
Activity: 4592
Merit: 1851
Linux since 1997 RedHat 4
|
|
April 28, 2012, 05:04:00 AM |
|
Is anyone sucessfully running their BFL Singles against p2pool? I understand the frequent LONGPOLL's cause a problem with BFL's 5 second calculation. (By successful, I don't just mean that it works, I mean it works efficiently and you are making the normal amount of coin as expected). If so, can you share your settings for CGMiner?
As SgtSpike mentioned, the issue cannot be fixed with cgminer settings (or any mining software settings) ... BFL must create/release new firmware (which they have not yet done) for the Singles to be able to work 'successfully' with p2pool. Until that happens, you must choose a different pool. Or choose different mining hardware. Yeah I've explained it a few times already (elsewhere?) but I'll do it again The BFL does not report shares until it finishes processing the full nonce range. So about 5.2 seconds (assuming 830MH/s) after it starts on work it replies with the answer(s). If during that 5.2 seconds you get an LP then anything it did during that 5.2 seconds is wasted. What that means is that on average for every LP your BFL will have wasted 2.6 seconds processing. The miner code should/will then of course abort the work so that it doesn't continue to finish the remainder of the 5.2 seconds, but the work done up to the abort is wasted since you cannot ask it if it found a share (and it didn't tell you even though it would have been a valid share if it had already found it) Since LP's occur on expected average of every 10 seconds ... and on expected average you waste 2.6 seconds processing ... that's 26% of your processing is thrown away (and you can't avoid it) Does this mean that you would expect with a BFL FPGA to generate 26% fewer shares at a given hash rate than a GPU? No it will produce the expected # of shares however you will have a 26% stale rate. Actually a little higher because you have the normal stales due to network propogation too. This is borne out in observations by users reporting a DOA & orphan combined rate of something like 30% to 40% of hashing power. Actually no. It will report 26% less shares since it aborts 26% of shares before you see them.
|
|
|
|
amazingrando
|
|
April 28, 2012, 04:15:43 PM |
|
Does this mean that you would expect with a BFL FPGA to generate 26% fewer shares at a given hash rate than a GPU?
No it will produce the expected # of shares however you will have a 26% stale rate. Actually a little higher because you have the normal stales due to network propogation too. This is borne out in observations by users reporting a DOA & orphan combined rate of something like 30% to 40% of hashing power. Actually no. It will report 26% less shares since it aborts 26% of shares before you see them. So 830mh/s of bfl will produce same number of shares as 830mh/s of GPU? Then what's the big deal? That there is a theoretical loss of hashing power because some shares are aborted?
|
Bitbond - 105% PPS mining bond - mining payouts without buying hardware
|
|
|
eleuthria
Legendary
Offline
Activity: 1750
Merit: 1007
|
|
April 28, 2012, 04:37:52 PM |
|
Does this mean that you would expect with a BFL FPGA to generate 26% fewer shares at a given hash rate than a GPU?
No it will produce the expected # of shares however you will have a 26% stale rate. Actually a little higher because you have the normal stales due to network propogation too. This is borne out in observations by users reporting a DOA & orphan combined rate of something like 30% to 40% of hashing power. Actually no. It will report 26% less shares since it aborts 26% of shares before you see them. So 830mh/s of bfl will produce same number of shares as 830mh/s of GPU? Then what's the big deal? That there is a theoretical loss of hashing power because some shares are aborted? Just to be clear since it wasn't said in the last few posts: This flaw in the BFL units is exclusive to p2pool due to p2pools rapid long polling. While the flaw DOES affect all other mining, its only ~0.43% when mining a pool/solo since regular long polls are 10 minutes apart, not 10 seconds apart. An 830 MH/s BFL will produce the same number of shares as an 830 MH/s GPU. However, due to the way the BFL unit reports shares, you're losing ~26% of what is produced on p2pool. So on p2pool you'd only be getting an effective rate of ~614.2 MH/s with a BFL unit.
|
RIP BTC Guild, April 2011 - June 2015
|
|
|
amazingrando
|
|
April 28, 2012, 04:42:11 PM |
|
Ok, now it's clear to me. Thank you!
|
Bitbond - 105% PPS mining bond - mining payouts without buying hardware
|
|
|
Vorksholk
Legendary
Offline
Activity: 1713
Merit: 1029
|
|
April 28, 2012, 06:36:32 PM |
|
Read through about 10 pages of this thread so far, and was just wondering, how's the current shipping lengths these days? Thinking of getting a single.
|
|
|
|
bitcoindaddy
|
|
April 28, 2012, 06:42:43 PM |
|
I thinks it's down to about 2 months but it's improving.
|
|
|
|
Epoch
Legendary
Offline
Activity: 922
Merit: 1003
|
|
April 28, 2012, 07:01:46 PM |
|
I thinks it's down to about 2 months but it's improving.
No, more like 11 weeks right now; they are still working on orders placed mid-February: https://bitcointalk.org/index.php?topic=77796.0
|
|
|
|
abeaulieu
|
|
April 30, 2012, 01:32:53 PM |
|
Not sure if anyone else noticed, but BFL posted another production update a couple days ago. It shows the pre-production Mini-Rig (with a Raspberry Pi host!) and more info on the water block that they are making for the singles. http://www.butterflylabs.com/production-update/
|
|
|
|
|
heavyb
|
|
April 30, 2012, 08:17:25 PM |
|
my 2 singles came in today that I ordered on 2/14/12. I have been using poclbm gui miner with my GPU's and have switch to CGminer. I have a single plugged in and cgminer running, and it detects my 2 gpu's but not my single. Could someone please provide detailed instructions on how to get these running? Thank you!
|
|
|
|
Epoch
Legendary
Offline
Activity: 922
Merit: 1003
|
|
April 30, 2012, 08:34:48 PM |
|
my 2 singles came in today that I ordered on 2/14/12. I have been using poclbm gui miner with my GPU's and have switch to CGminer. I have a single plugged in and cgminer running, and it detects my 2 gpu's but not my single. Could someone please provide detailed instructions on how to get these running? Thank you!
HB, take a peek here: https://bitcointalk.org/index.php?topic=28402.msg874683#msg874683Read through the above post plus the various replies; that should get you straightened out with your Singles and cgminer. You may or may not need to run a separate instance of cgminer for your gpus and Singles. Keep in mind that under Windows, cgminer does NOT autodetect Singles ... you need to specify them explicitly on the command line as shown in the above link. And to do that, you'll need to know the COM port that Windows has assigned to them.
|
|
|
|
heavyb
|
|
April 30, 2012, 08:46:14 PM |
|
thank you very much epoch, unfortunatly, my single does not show up in the COM section, but as an other device and lists it has no driver info. Any ideas?
|
|
|
|
Inaba
Legendary
Offline
Activity: 1260
Merit: 1000
|
|
April 30, 2012, 08:58:52 PM |
|
Do you have an FTTI driver installed?
|
If you're searching these lines for a point, you've probably missed it. There was never anything there in the first place.
|
|
|
heavyb
|
|
April 30, 2012, 09:06:05 PM |
|
Do you have an FTTI driver installed?
I do not. this is probably the problem. any advice for this?
|
|
|
|
|