[XPM] [ANN] Primecoin High Performance | HP14 released! - page 89.

nmersulypnem

full member

Activity: 238

Merit: 100

Quote from: Tamis on July 19, 2013, 12:55:32 PM

Quote from: mikaelh on July 19, 2013, 12:16:40 PM

Quote from: Tamis on July 19, 2013, 12:08:09 PM

Didn't the consensus show that 1M sievesize was some kind of sweet spot ?

Well, I'm not sure if I'm part of that consensus because I haven't seen any statistically significant data to support that.

What would you think is the best for finding blocks mikaelh ?
I know this is more of a guess than anything on your side but being the coder you're probably the best person to ask.

How do you write the size ? sievesize=2000000 ?

Based on the code:

nSieveSize = (int)GetArg("-sievesize", nDefaultSieveSize);

... I believe that will work, yes.

ReCat

sr. member

Activity: 406

Merit: 250

You're welcome. Tongue

Feel free to donate. XD

Wild7

newbie

Activity: 8

Merit: 0

Quote from: ReCat on July 19, 2013, 11:43:13 AM

https://i.imgur.com/ih8Gano.png

Mined four blocks in the past 24-ish hours.

Holy. Shit.

(Win 2008, HP4, specs down in siggy)

Thanks for your guide.. got 1 block now.. Grin

xyzzy099

legendary

Activity: 1063

Merit: 1048

Quote from: nmersulypnem on July 19, 2013, 12:40:50 PM

Quote from: xyzzy099 on July 19, 2013, 12:35:55 PM

Quote from: nmersulypnem on July 19, 2013, 12:33:15 PM

Quote from: Tamis on July 19, 2013, 12:08:09 PM

Didn't the consensus show that 1M sievesize was some kind of sweet spot ?

Can I pass in the sieve size, or do I need to recompile?

You can pass it on the command line with the '-sievesize' parameter, or you can put sievesize= in your primecoin.conf.

Is that the only tuning parameter I can pass in at the moment?

I don't know. Maybe mikaelh will chime in and answer your question.

Tamis

sr. member

Activity: 476

Merit: 250

Quote from: mikaelh on July 19, 2013, 12:16:40 PM

Quote from: Tamis on July 19, 2013, 12:08:09 PM

Didn't the consensus show that 1M sievesize was some kind of sweet spot ?

Well, I'm not sure if I'm part of that consensus because I haven't seen any statistically significant data to support that.

What would you think is the best for finding blocks mikaelh ?
I know this is more of a guess than anything on your side but being the coder you're probably the best person to ask.

How do you write the size ? sievesize=2000000 ?

ReCat

sr. member

Activity: 406

Merit: 250

Quote from: Tamis on July 19, 2013, 12:08:09 PM

Didn't the consensus show that 1M sievesize was some kind of sweet spot ?

No. I think it was just the failsafe.

mumus

sr. member

Activity: 291

Merit: 250

Quote from: mikaelh on July 19, 2013, 12:16:40 PM

Quote from: Tamis on July 19, 2013, 12:08:09 PM

Didn't the consensus show that 1M sievesize was some kind of sweet spot ?

Well, I'm not sure if I'm part of that consensus because I haven't seen any statistically significant data to support that.

2M on testnet performs better on my i7-3770 quad (8HT).I think Hypertheading really helps here.

oroqen

sr. member

Activity: 280

Merit: 250

Quote from: airtreb on July 19, 2013, 12:38:49 PM

I have 60 computers 30 i5-3550 and 30 core2duo 8200...for the last 2 days i just mined 3 blocks OMG each pc have different wallet, now i try all of them using just 1 wallet, are those 60 cpu add their primespersec when mining? because when i try to getprimespersec they give different value.

1.Theres no way to corralate the work of each cpu without a pool, some will do the same work nothing you can do about it right now unless you want to start coding.
2.primespersec is a poor measure and changes quite often, best too take an average for afew minutes too compair performace.

blackboy

member

Activity: 109

Merit: 16

Anyone using VPSs to mine XPM? Which one would you recommend?

nmersulypnem

full member

Activity: 238

Merit: 100

Quote from: xyzzy099 on July 19, 2013, 12:35:55 PM

Quote from: nmersulypnem on July 19, 2013, 12:33:15 PM

Quote from: Tamis on July 19, 2013, 12:08:09 PM

Didn't the consensus show that 1M sievesize was some kind of sweet spot ?

Can I pass in the sieve size, or do I need to recompile?

You can pass it on the command line with the '-sievesize' parameter, or you can put sievesize= in your primecoin.conf.

Is that the only tuning parameter I can pass in at the moment?

airtreb

full member

Activity: 217

Merit: 100

I have 60 computers 30 i5-3550 and 30 core2duo 8200...for the last 2 days i just mined 3 blocks OMG each pc have different wallet, now i try all of them using just 1 wallet, are those 60 cpu add their primespersec when mining? because when i try to getprimespersec they give different value.

xyzzy099

legendary

Activity: 1063

Merit: 1048

Quote from: nmersulypnem on July 19, 2013, 12:33:15 PM

Quote from: Tamis on July 19, 2013, 12:08:09 PM

Didn't the consensus show that 1M sievesize was some kind of sweet spot ?

Can I pass in the sieve size, or do I need to recompile?

You can pass it on the command line with the '-sievesize' parameter, or you can put sievesize= in your primecoin.conf.

nmersulypnem

full member

Activity: 238

Merit: 100

Quote from: Tamis on July 19, 2013, 12:08:09 PM

Didn't the consensus show that 1M sievesize was some kind of sweet spot ?

Can I pass in the sieve size, or do I need to recompile?

#Darren

sr. member

Activity: 784

Merit: 250

DIA | Data infrastructure for DeFi

Quote from: romerun on July 19, 2013, 11:44:25 AM

i got no shit for the last 24

xyzzy099

legendary

Activity: 1063

Merit: 1048

Quote from: xyzzy099 on July 19, 2013, 11:14:12 AM

Quote from: mikaelh on July 19, 2013, 11:10:51 AM

Quote from: ReCat on July 19, 2013, 11:02:21 AM

Quote from: mikaelh on July 19, 2013, 10:51:54 AM

I posted a new compilation guide for Linux:
https://bitcointalksearch.org/topic/xpm-primecoin-high-performance-linux-compilation-guide-259022

It shows you how to compile your own libgmp and everything else.

Mikaelh. I have a question for you.

My CPU's have 12MB of cache each. Is that advantageous for primecoin mining? I have noticed I have been getting significantly higher primespersec than most of my friends when we use larger seivesizes, even when they have newer processors!

Well, it might run a tiny bit faster if you have a large L3 cache. I wrote most of the code to run fast using the L1 and L2 caches, so the L3 cache is actually not that important. But of course it never hurts to have a big L3 cache. Wink

GPU architecture also has a big impact.

Running hp4 on a CPU with a 10M L3 cache, I found that if I set the sievesize any larger than 4M, the program would crash within seconds of startup. What should be the maximum sievesize?

I am using hp5 now, but I have not checked to see if it still crashes.

Is this a question so stupid that it just doesn't deserve consideration?

I've asked about this crashing with large sieve sizes twice now, but have had no reply.

I've verified that it happens with hp5 as well. I wish I could provide more details about the crash, but Windows just pops up a dialog saying the program stopped working, and doesn't include much useful debug info.

mikaelh

sr. member

Activity: 301

Merit: 250