[XPM] [ANN] Primecoin High Performance | HP14 released!

[roy7]

What's the default keypool size? Is a key only "used up" when a block is found?

[notme]

What's the default keypool size? Is a key only "used up" when a block is found?

A key is the magic number that allows you to spend the coins at an address.  When you find a block it uses up an address/key.

[masterOfDisaster]

What's the default keypool size? Is a key only "used up" when a block is found?

The default keypool size is 100. A key is used each time a block is solved or a transaction is incoming.

[gigawatt]

What's the default keypool size? Is a key only "used up" when a block is found?

A key is the magic number that allows you to spend the coins at an address.  When you find a block it uses up an address/key.

But I've mined multiple blocks with the same address... so wouldn't that prove the "used up" aspect incorrect?

[blackboy]

You could also use your wallet until addresses diverge or increase the keypool.

Could you elaborate on this?

primecoind --daemon --keypool=10000

I get that part.  What did you mean about the addresses diverging?  As long as each node has the exact same wallet, that shouldn't be a problem, correct?

Initially, wallets have a limit of 100 addresses. Once you generate more than 100 blocks, trouble shows up.

[blackboy]

Well... did some tuning and now it seems that everything is running at 3M PPS.
Will post the config details after I confirm the block rate is better. So in an hour or so...

[superfluouso]

No tricks. 500 8-core opteron servers running mikhaels hp8.
I simply took the best scripts from this thread and used some recommendations from members.

500!

Self funded? Or do you work in a data centre ?

Self funded.

How does one even set up 500 servers?  Just that task alone to order that many sounds massive.  And at $70/month for the cheapest dedicated 8core opteron...$35,000?I must be missing something somewhere.  Can you elaborate?

[blackboy]

So this is the configuration that works best for me

Sieve size 1M
Sieve percentage 3
Round sieve size 45

It gives 6.2k PPS per machine on average

And about 850 5-chains/h

[blackboy]

No tricks. 500 8-core opteron servers running mikhaels hp8.
I simply took the best scripts from this thread and used some recommendations from members.

500!

Self funded? Or do you work in a data centre ?

Self funded.

How does one even set up 500 servers?  Just that task alone to order that many sounds massive.  And at $70/month for the cheapest dedicated 8core opteron...$35,000?I must be missing something somewhere.  Can you elaborate?

You build once and clone 499 times. Pretty simple actually.

Then you put a script on each server that makes it mine automatically.

I didn't get the 35K right away, I just took some of the money I was making while mining. I starting mining with 500 Amazon EC2 instances when diff was about 8, and then took the winnings to double down on dedicated servers with opteron processors (that perform about 20x better).

So Amazon practically financed my server farm.

[superfluouso]

No tricks. 500 8-core opteron servers running mikhaels hp8.
I simply took the best scripts from this thread and used some recommendations from members.

500!

Self funded? Or do you work in a data centre ?

Self funded.

How does one even set up 500 servers?  Just that task alone to order that many sounds massive.  And at $70/month for the cheapest dedicated 8core opteron...$35,000?I must be missing something somewhere.  Can you elaborate?

You build once and clone 499 times. Pretty simple actually.

Then you put a script on each server that makes it mine automatically.

I didn't get the 35K right away, I just took some of the money I was making while mining. I starting mining with 500 Amazon EC2 instances when diff was about 8, and then took the winnings to double down on dedicated servers with opteron processors (that perform about 20x better).

So Amazon practically financed my server farm.

I see..thanks!  So you're the reason XPM fell to .60 from 1.50

[Lauda]

You build once and clone 499 times. Pretty simple actually.

Then you put a script on each server that makes it mine automatically.

I didn't get the 35K right away, I just took some of the money I was making while mining. I starting mining with 500 Amazon EC2 instances when diff was about 8, and then took the winnings to double down on dedicated servers with opteron processors (that perform about 20x better).

So Amazon practically financed my server farm.

I see..thanks!  So you're the reason XPM fell to .60 from 1.50

He definently is 

[roy7]

You build once and clone 499 times. Pretty simple actually.

Then you put a script on each server that makes it mine automatically.

I didn't get the 35K right away, I just took some of the money I was making while mining. I starting mining with 500 Amazon EC2 instances when diff was about 8, and then took the winnings to double down on dedicated servers with opteron processors (that perform about 20x better).

So Amazon practically financed my server farm.

I use EC2 for my TRC pool, and other Amazon goodies like SQS/SMS etc. Wish I'd known about Primecoin a bit sooner. I tried running some CC2 instances last night mining primecoin (about 8500 pps each) but even with 20 going, only found one block in a couple hours. Don't think I can break even with that, even with the spot price so low. I'd need a couple blocks per hour.

Still, Amazon is pretty amazing and I did finally learn about the @reboot date trick with crontab to make rolling out the auto-starting servers easier.

Edit: And actually, after this experiment, I rolled out some virtual Windows Server 2012 machines to run the ypool miner (about 20 val/h using latest client they provide) but that's way below break even, sadly.

[blackboy]

You build once and clone 499 times. Pretty simple actually.

Then you put a script on each server that makes it mine automatically.

I didn't get the 35K right away, I just took some of the money I was making while mining. I starting mining with 500 Amazon EC2 instances when diff was about 8, and then took the winnings to double down on dedicated servers with opteron processors (that perform about 20x better).

So Amazon practically financed my server farm.

I use EC2 for my TRC pool, and other Amazon goodies like SQS/SMS etc. Wish I'd known about Primecoin a bit sooner. I tried running some CC2 instances last night mining primecoin (about 8500 pps each) but even with 20 going, only found one block in a couple hours. Don't think I can break even with that, even with the spot price so low. I'd need a couple blocks per hour.

Still, Amazon is pretty amazing and I did finally learn about the @reboot date trick with crontab to make rolling out the auto-starting servers easier.

Edit: And actually, after this experiment, I rolled out some virtual Windows Server 2012 machines to run the ypool miner (about 20 val/h using latest client they provide) but that's way below break even, sadly.

Send a message to gigawatt. He's got some cool settings to optimize EC2 instances.

[gigawatt]

Send a message to gigawatt. He's got some cool settings to optimize EC2 instances.

You say that, but it's more of a setup script.  It's in my signature if anybody is interested.
I'll probably post an update in a bit.

[teknohog]

Sounds like somebody needs to make a swapfile.

This reminds me of building bitcoind on ARM machines with 128 MB of RAM. These days primecoind compiles and runs fine on a 256 MB Buffalo Linkstation. Of course, ccache and distcc help a lot.

OT: It's nice that cryptocurrencies make people learn unix, but I wonder if there should be separate threads/FAQs about these. After all, the same issues seem to come up every time a new coin is released...

[matt4054]

Yes obviously you can no longer break even with VPS, it was only the case when the difficulty was not catching up. But it's actually a good thing for individual miners that are pooling and who now get the full pie, although only tiny slices.

[altsay]

It's much better to trade than mining now. Both VPS and rig owners barely break even if they're lucky enough.

[mikaelh]

Version 0.1.2-hp9 is finally out! (Download links are on the first page as usual.)

Here's the list of notable changes:
 * Sunny's mining optimizations from 0.1.2.
 * Minor mining optimizations from mtrlt and gatra.
 * Small sieve optimizations for 32-bit processors.
 * Sunny's new chains/day estimate.
 * Sunny's new 'listprimerecords' RPC command.
 * New default value for roundsievepercentage is 70 on mainnet (better chains/day, equal 5-chains/h).

This release should hopefully provide big performance improvements on mainnet. I don't have the resources to confirm the block rate on mainnet but I have done extensive testing on testnet. The new chains/day estimate is not very accurate currently. Therefore it should not be used to estimate mining profitability etc. but it should be useful for finding better mining settings. Currently the estimate is about 20% bigger than the observed chain rate on testnet. The difference may be even greater on mainnet.

I wrote a couple of posts about chains/day earlier:
https://bitcointalk.org/index.php?topic=255782.msg2847844#msg2847844
https://bitcointalk.org/index.php?topic=255782.msg2848092#msg2848092

[Kalshnikov]

"primespersec" 12-15 times down?

G1610 - 1500 on hp8/ 116 on hp9?

Any comments?



{
"blocks" : 94903,
"chainspermin" : 0,
"chainsperday" : 0.05914120,
"currentblocksize" : 1000,
"currentblocktx" : 0,
"difficulty" : 9.21033770,
"errors" : "",
"generate" : true,
"genproclimit" : -1,
"roundsievepercentage" : 70,
"primespersec" : 116,
"pooledtx" : 0,
"sievepercentage" : 10,
"sievesize" : 1000000,
"testnet" : false
}

[mikaelh]

"primespersec" 12-15 times down?

G1610 - 1500 on hp8/ 116 on hp9?

Any comments?

Yes, primes/sec is likely to go down with HP9. Sunny's optimizations and my new default settings actually reduce it. But as I have written before, primes/sec does not correlate with your mining performance. If you still want more primes/sec, you can add 'roundsievepercentage=30' to your config file.