Right, I think I should clarify that pooled mining is different. In pooled mining you're always targeting the network difficulty but you're allowed to submit shorter chains to the pool. The ratios between lengths 7, 8, 9 and 10 should always be approximately 10 (at least for CPU miners). That's because each number has nearly the same probability of being prime after it's being sieved. Since the ratio is 10, the probability is about 0.1.

However, you can find a lot more 7-chains if you change the sieve parameters. This is what I did in my previous post. That's why the ratios are more than 10.

I haven't checked the numbers, but the eventual total supply of 55 million sounds reasonable. Primecoin is definitely intended to sustain a healthy mining market, since only about 11% of that has been mined in the first year.

I did a quick benchmark with regards to the difficulty ratio between chain lengths. I used the CPU miner in my Primecoin High Performance wallet (latest HP14 release). It allows targeting different chain lengths using the -sievetargetlength parameter.


So 11-chains should be about 28.7 times harder to find than 10-chains. So that number is definitely closer to 30 rather than 10. Sunny King originally assumed that the ratio would be about 32. Also note that the ratio keeps growing as the chain length increases.

I also compared these numbers to some earlier benchmarks. It seems that the ratios are smaller with more optimized code. So my guess is that for GPU miners the ratios may be slightly smaller. (Anyone want to confirm that?)

My network charts page at http://xpm.muuttuja.org/charts/ tries to estimate the total network computation power in terms of chains per minute. It shows that chains/min peaked at 20 chains/min on May 13th, which is definitely not enough to push difficulty above 11. Last December we actually needed about 40 chains/min to go above 10.

Currently the network is mining at about 8.5 chains/min, which is equivalent to about 1500 GPUs mining at 8.2 chains/day.

Good to hear that the issue is solved. I think your wallet wasn't done synchronizing with the network. Your getinfo output showed that you only had 237209 blocks while currently the network is at 601597 blocks.

HP14 released! This is a big release with a huge speed boost for CPU miners. Download links are in the first post.

Changes in -hp14:
 * Huge CPU mining speedup. Up to 73% higher chains/day on Intel's Haswell.
 * Developer fee is now 1% (-donationpercentage defaults to 1.0).
 * Sieve now uses the BTS instruction (same as RapidPrime miner).
 * Added compile-time support for using BMI2 instructions on Haswell.
 * Added compile-time support for SSE2 and AVX2 bitwise operations.
 * New extended sieve algorithm similar to rdebourbon's beta4 code.
 * New default values for the tuning parameters:
   . sievefilterprimes defaults to 14000 (up from 7849)
   . sievesize defaults to 1376256 (up from 917504)
   . sieveextensions defaults to 10 (up from 9)
 * Lots of other minor improvements.

I fixed the OS X build and it's now available for download.

HP13 released! Download links are on the first page. Links point to my own server because SourceForge is being especially slow for me.

Changes:
 * Binaries compiled with OpenSSL 1.0.1g which addresses the Heartbleed vulnerability
 * Faster blockchain download
 * Bonus: Also compiled with GMP 6.0.0a

Mac OS X version will be delayed because it failed to build this time.

I'm working on a new release with updated OpenSSL. AFAIK, there's two ways to exploit the OpenSSL bug in the wallet:
1.) Using -rpcssl with a wallet daemon. This turns the RPC port into an SSL server which can be exploited. If the port is open to the internet, your wallet may be compromised.
2.) Opening payment protocol links which makes the wallet connect to an SSL server. If the server is malicious, your wallet may be compromised. So don't open any links starting with primecoin: .

Also, it's possible to check that you're using your own custom libgmp by typing "ldd primecoind". The output should say something like this:

That means it's using the libgmp installed in /usr/local.

Yes, that will install the latest version of GMP in /usr/local. However, you then need to use some special flags to link against it. Here's a snippet from my Linux compilation guide which modifies makefile.unix.


After that run "make -f makefile.my clean" and then "make -f makefile.my".

To donate 1% of the blocks you will find, add this line to the .conf file:

Any other command-line option also works in the .conf file the same way. The source code tarball for HP12 is still the latest code. There's nothing new on Github yet.

If you really want to optimize, you can try getting the latest version of libgmp which was released a few days ago.

I checked the link and it's the latest version. The binaries in that archive say they are version v0.1.2xpm-hp12-beta. Make sure you don't have an old binary in /usr/local/bin or something.

Just to be clear: that feature is disabled by default. Still, if someone wants to donate, it would be very much appreciated.

HP12 released! The download links are on the first page as usual. Since people keep running into the bugs in older releases, I have finally decided to make a new release. This is a recommended upgrade for all users. Note that this release does not contain my very latest experimental optimizations. Nevertheless, many smaller optimizations and improvements are included.

Release highlights:
 * Fix for shared wallet files. If the same wallet.dat file is copied to multiple machines, they should all try to solve unique blocks now.
 * Fix for getwork protocol. Some versions of jhPrimeminer support solo mining using a patched getwork protocol. That patch is included in this release.
 * OS X binaries of the Qt client.
 * Fix for the assertion failure crash.

Changes:
 * IMPORTANT: Fix for multiple machines doing the same work if wallet is shared.
 * New chain statistics in debug.log.
 * Removed 'primespersec' from the output of getmininginfo.
 * Removed 'chainspermin' metric.
 * Replaced -sievepercentage parameter with -sievefilterprimes.
 * New parameter -l1cachesize for specifying the size of sieve segments.
 * Added new blocks/day metric which tries to estimate how many blocks will be found at current difficulty.
 * Better automatic adjustment of primorial factor.
 * New parameter -primorial for manual adjustment of primorial.
 * Implemented early support for v0.2 proposal (needs to be enabled with -miningprotocol=2).
 * David's fix for getwork.
 * Updated to Bitcoin v0.8.6 codebase.
 * Improved the checking of BiTwin chain candidates. As a consequence, short BiTwin chains (1 or 2 primes) are now counted correctly in the mining statistics. Primes/h is now much higher due to this.
 * The Qt application is now titled Primecoin High Performance.
 * Fixes for Mac OS X and the Clang compiler.
 * Revised default sieve parameters.
 * Added an automatic donation feature when a new block is found. The -donationpercentage allows a percentage of the block reward to be donated to an address. E.g. passing the command-line parameter -donationpercentage=1.5 will donate 1.5% of every block to the developer. Alternatively, this value can be set in primecoin.conf. Also, the default address can be changed using -donationaddress.

I have only tested my code on x86 and x86-64, so some issues will likely arise on other platforms. You would probably need to do some porting to get the code running on SPARC.

That's normal. The miner chooses an unused address from the wallet.

No plans for that currently. I've been hoping that the stand-alone miner devs would eventually implement either getwork or getblocktemplate (GBT). GBT is working and getwork should be fixed but I haven't tried it. There's a few forks of jhPrimeminer out there that do that but it looks like the code hasn't spread to the more popular forks. Also, I don't know the specs of the pool protocols so I don't know exactly what the advantages are.

Oh yeah, I forgot to add prime.h to the commit. It's fixed now. It was a problem with commit b6a92cd8b9.

Well, more support is always welcome. However, I see two problems with advertising VPS mining:
1) The AUP of many providers forbids mining. For example, Linode pretty clearly states that you're not allowed to use excessive CPU resources. So if you try mining there, they may terminate your account at any time citing violation of AUP.
2) VPS mining isn't always profitable. Many VPS providers are actually quite expensive, which means that you'll only be able to profit when XPM prices are high.

As far as I know, Amazon and DigitalOcean are the only big providers that allow mining. And for what it's worth, here's my referral link for DigitalOcean:
https://www.digitalocean.com/?refcode=a285e54b94da

The cheapest droplet costs $5/month and mines at about 0.01 chains/day. At current difficulty, that's about 0.006 blocks/day. So in a month you could earn (in theory):
0.006 blocks/day * 30 days/month * 9.2 XPM/block * 3.20 USD/XPM = 5.30 USD/month

So VPS mining at DigitalOcean should still be profitable (in theory). Note that if you are solo mining, then the expected time to block is 166 days. With 100 droplets, the time will be 1.66 days. So solo mining only makes sense if you have lots of droplets. If you don't like waiting for blocks, check out beeeeer.org for pooled mining.

I haven't done the math for Amazon's EC2 recently but I'd expect it to not be profitable. Even the spot prices are usually too high.

Also my BTC, LTC, PPC and XPM donation addresses are in my signature below.

They are coming out when I'm done with testing and I'm fairly confident there are no more bugs lurking around. Or in other words, when it's ready™.

The latest code already shows blocks/day when running getmininginfo.

On another note, I pushed more fixes to Github. I think I finally fixed the CCheckQueueControl assertion errors with this commit:
https://github.com/mikaelh2/primecoin/commit/e1784f808908e5ac5c17125091ab3666e8e8c591