19.03 22:18:40   3h 53m    966   1584231     0.02957334
19.03 21:50:28   3h 24m    350   582220     0.02915565
19.03 21:39:36   3h 14m    239   384413     0.03015377
19.03 21:29:55   3h 04m    59   93375     0.03064525
19.03 21:28:24   3h 02m    1244   2004247     0.03010308
19.03 20:49:40   2h 24m    731   1259703     0.02814433
19.03 20:25:33   1h 59m    330   500802     0.03195874
19.03 20:16:17   1h 50m    431   698913     0.02990859
19.03 20:02:52   1h 37m    2641   4351399     0.02943617

there seems to be a problem o.o look at the times

If you have, what hashes are you getting?

That is correct.  Higher clocked card has a higher chance of getting the extra few shares in most of the time.  But the parallel cards can beat it too.  It all depends on the luck.

Is there any other way to recieve block broadcasts besides the web socket?  I'm trying to impliment it low-level and I keep getting "GET requests are not supported on this server" etc.

Statistics don't seem to be updating o.o


and from deepbit

Ok, will try.  I underclock my cards while I sleep for courtesy for my roommate.  Is there any way to find out which card it is?

I clock my 5970 down to 675/300 @ .95v
All temps are kept between 60 and 70

Its crazy underclocked though :/

Is there any way to check logs to see why my system hung?  Sometimes I need to cycle it, because it stops mining.  Its like it just freezes :/

I also get this:

Just upgraded to 0.5c today, working great!

However, I just stuck my sapphire 6870 in and I can't adjust the clocks unless I go into atitweak.  Also, My voltage is rock solid.  I have no idea how to lower it o.o

inb4 the license file can and will be cracked.

If you can afford a $30k unit, you can afford a cluster of GPU's to crack it.

Gah.  They changed it.  Oh well.  Know of any other cards that you can water cool without losing warranty?

Generally you will have a higher quality ASIC in the black edition.

Other than that, there is a lifetime warranty on the black edition, while the core edition has 2 years.

And they're water cooling friendly.

Same.  I've been eyeing that card for awhile now.  Lifetime warranty + water cooling friendly.

he may be talking about this card:

I found a block today too.

Would be nice to see the block # users have found!

No reason to leave! Our debate was just getting heated!

This is probability and odds my friend.

Say you have 10 rounds:

A and B both get new work and round ends in 30 second.


Multiplication Rule 1:      When two events, A and B, are independent, the probability of both occurring is: P(A and B) = P(A) · P(B)

A: 42.9 s/share: Extrema: 10, if every card finds a share before 30 seconds.  Will average to ~0.7 shares / card.
1 share: 0.7^1 x 100%  70%
2 shares: 0.7^2 x 100% = 49% (70% chance on two cards)
3 shares: 0.7^3 x 100% = 34.3%
4 shares: 0.7^4 x 100% = 24.0%
5 shares: 0.7^5 x 100% = 16.8%
6 shares: 0.7^6 x 100% = 11.8%
7 shares: 0.7^7 x 100% = 8.23%
8 shares: 0.7^8 x 100% = 5.76%
9 shares: 0.7^9 x 100% = 4.03%
10 shares: 0.7^10 x 100% = 2.82%

*note: this does not repeat.  For one card to find 2 shares in that time is exponential growth.

B: 4.29 s/share: Extrema: 10 if it finds a share < 3 seconds every time.  Will average to ~7 sometimes 4 sometimes 10

In order for that gap to be filled A would have to have the same # of shares submitted as B.

It is almost certain that AT LEAST 6 shares will be found in 30 seconds.

Shares 7 8 9 10

s 4.29 3.75 3.33 3.00

P(i) 0.999 0.874 0.777 0.699

P(i)^n 99.3% 34.1% 10.3% 2.8%

Conclusion:
Given n trials of this:
If every trial ends 30 seconds after each card submits their hashes -
A: 70% of the time, the cluster will submit 1+ share.
    49% of the time, the cluster will submit 2+ shares.
   and so on.
B: 99.3% of the time, the cluster will submit 7+ shares
    34.1% of the time, the cluster will submit 8+ shares
   and so on.

As you see, the odds are against the single card submitting under 6 hashes in that allocated time.  Whereas the 10 card setup odds suggest that will happen 11.8% of the time.

Everything I have said in this thread has been theory.  I've stated that it's an estimate.  If you read the last part of my first post, I touch upon the randomness of the subject.  But as more and more shares are found (limit as shares->infinity) it will converge at 1/2^32.  Plain and simple. 

I understand what you are saying, but that would only be true as the number of shares/round -> infinity. 

What I'm stating is purely theoretical.

The real life model of this analysis will obviously depend on randomness.

In some cases it will win, in some cases it will lose.  But due to the evidence I've stated about submitting shares,  The single more powerful card will almost always submit between 0 and (1cardhash/10cardhash)) hashes per round if shares->infinity.  In most cases this is negligible, but in cases where it takes 30 seconds to solve a block, those extra shares provide the extra income.

What are the odds of a card taking an average of 42.9 seconds to submit a share, submitting one in < 4.29 seconds?

Now lets say you have 30 seconds till the round ends.  Not all rounds end as you submit your shares.

Your logic is correct, but the reason for the round down is because you can't submit 1/9th of a share.  At that period of time *if you average 1 share / 2^32 hashes* your share would not be submitted in time.  The idea behind this, the chances are, with more firepower, you will find shares more quickly.

Lets say a block is found 30 seconds in the future:

if there is 30 seconds left and all of your cards just submitted a share, they're beginning to work on new shares.

A: The cards average 42.9 seconds per share, so what are the odds 5 of them find a share and submit it?

B: Cards average 4.29 seconds per share.  Let's round down to 6 shares submitted.  Even if one share takes 8 seconds to find, you're still submitting 1 more than you would have in A.

This takes advantage of the rounding errors and can be seen in extreme cases such as 2x 5870's vs 1x BFL single.

127 views and 0 replies? o.o