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

[redphlegm]

Is it already too difficult to get some results?

I'm mining with an i7 and get ~800-1000 pps. But didn't find anything in the last 20h 

I think it might be already too difficult. I'm running on OS X 10.8.4, and I set up an Ubuntu OS in VMWare Fusion. It's running around 600-700 ops, has gone through almost 80,000 blocks in 24 hours, and there's nothing produced so far. I'll let it keep running for a few days. I understand that luck is a major factor, so maybe lady luck will visit me.

Why not run native? you could probably get better PPS than that. and the code compiles fine native OSX.

I've tried it. For some reason running it in a VM on Ubuntu gives 2-3x the performance. No idea why.

[1714916806]

1714916806

[1714916806]

1714916806

[1714916806]

1714916806

[Trillium]

So it seems like a few people cant get sievesize to 4M or more.

I am not experienced enough with the code/theory to explain why this might be happening, maybe mikaelh can work it out of we post our infos.

I have it crashing on both machines that I mine with when I set it to 4M.

As another member pointed out, it seems possible to mine with 3.6M but going to 3.7M will make it crash every time.

My machines are fairly different:

Main rig:
Core2Quad Q9550 (2.83 GHz)
4 GB DDR3 @ 667 MHz
Asus P5Q3 Deluxe Wifi@n version
6870+5970+5750
Vista 64 fully updated

Repurposed Dell office PC:
Core2Duo E8400 (3 GHz)
4 GB DDR2 @ 400 MHz
Dell 0Y958C (intel?) motherboard
PCI 8 MB Ati Rage graphics card + onboard
Win7 64 fully updated

both now running hp7 of course.

Looking further, it seems that the limit is around 3642400... but it is not a fixed thing. For example sometimes it will work and othertimes it wont, but it seems to center around this value for me. If it doesn't crash immediately then it seems to mine happily. It crashes as soon as a network connection is established and mining begins, if it's going to crash.

[giblet]

Complete scrub to this, should I be running the primecoin-qt or primecoind?  Running the primecoin-qt, my laptop has an Ivy Bridge i5 and NVS 5400m and is getting about 1600 prime/sec, desktop has an Athlon X2 6000+ and gets about the same, does this sound about right?  How long does it take to get XPM?

[gigawatt]

Complete scrub to this, should I be running the primecoin-qt or primecoind?

If you're running it and want visual interaction, you want to run primecoin-qt.
If you want to control it via command line or are using a headless system, primecoind.

You should get the same performance with each.

[giblet]

Complete scrub to this, should I be running the primecoin-qt or primecoind?

If you're running it and want visual interaction, you want to run primecoin-qt.
If you want to control it via command line or are using a headless system, primecoind.

You should get the same performance with each.

Awesome, thanks!

[OnlyC]

I get this everytime I start up primecoind --daemon and execute  watch -n 5 'primecoind getmininginfo & primecoind listaccounts & primecoind getconnectioncount'

1st problem (during launch)
error: couldn't connect to server
error: couldn't connect to server
error: couldn't connect to server

then I have to go back, and execute primecoind --daemon again until I see this: Error: Cannot obtain a lock on data directory /root/.primecoin. Primecoin is probably already running.

about 10-20 seconds later, the stats show up, and its connected. Do you think this problem could be affecting my possibility to mine blocks? about 2 days ago, I mined 2 blocks in 1 hour.. then nothing the rest of the day, and i am curious if this error has something to do with it? My other server doesn't have this problem, and loads after the first execution of primecoind. Any ideas?

I am using HP7 with directory of /primecoin/src when executing primecoind.

2nd problem (closing down)
Primecoin server stopping
root@# primecoind stop
Primecoin server stopping
root@# primecoind stop
Primecoin server stopping

I have to use killall -9 primecoind (at twice) to get it to stop. Any ideas to get it back on the right track? 1 server executes launch and stop commands fine, one server doesn't.

1st:
primecoind take 100% CPU usage so that make VPS extreme slow, just let it watch, then will be ok. Sleep 5; after --daemon if you want.

2st: Normally it'll stop after 3-5sec, if you can't stop it, reboot the VPS.

[torbank]

I get this everytime I start up primecoind --daemon and execute  watch -n 5 'primecoind getmininginfo & primecoind listaccounts & primecoind getconnectioncount'

1st problem (during launch)
error: couldn't connect to server
error: couldn't connect to server
error: couldn't connect to server

then I have to go back, and execute primecoind --daemon again until I see this: Error: Cannot obtain a lock on data directory /root/.primecoin. Primecoin is probably already running.

It takes some time to start the primecoind --daemon, so the first round of primecoind requests are not able to connect to the server.

[blackboy]

I'm about to start renting dedicated servers by the dozen. I think that I found n4ru's special place...

Supposing a GPU miner comes up next week, how could I monetize the servers?

[notme]

I'm about to start renting dedicated servers by the dozen. I think that I found n4ru's special place...

Supposing a GPU miner comes up next week, how could I monetize the servers?

use capybara or a similar framework to script watching videos for btc:
http://www.videobtc.com/?ref=0184e4f464

[torbank]

Anyone have any luck getting this to run on the Xeon Phi?

http://www.intel.com/content/www/us/en/processors/xeon/xeon-phi-detail.html

[Tuck Fheman]

I am definitely using the 64 bit version as I had previously.

My prime rate was 2400 with v5 then I extracted v7 and ran it, only gets 1200-1300 now. It only took me 1 minute to extract the new version and change the target I run, this is not comparing between days.

I was trying to get 4M sievesize with this shortcut command to start it:
C:\Windows\System32\cmd.exe /c start "runlow" /low C:\primecoin-0.1.1-hp7-winx64\primecoin-qt.exe -gen -setgenerate true -sievesize=4000000


except... it crashes on 4M. So I have it running on 2M right now:

ditto to all the above for my i7 quad. much slower results, less blocks on testnet with v7 vs any other version.  mainnet yielded 40% less PPS & CPM on avg.

on testnet v7 would find consecutive blocks almost immediately, then nothing for a few minutes, then 1 BPM or less thereafter during all test using all settings I've tried on other versions and new ones.  Setting to default settings on testnet gave fastest results, but blocks were still much less than any other version tested during same time frame.

FWIW, yes my settings are set correctly and always have been.

UPDATE : Well, I have to take back the above now. I lowered my proc to 4 and now have higher PPS/CPM(?) than previously and it spikes to the highest values I've seen on mainnet w/other builds. Then again, it could all be a matter of when I'm fetching the info. It fluctuates a lot (as I'm sure you know) so it's hard to say if there were higher values at times I didn't check. I can only report what I see.

[Tuck Fheman]

So it seems like a few people cant get sievesize to 4M or more.

v3 has worked well above 6M sievesize for me. it's the only version that works on mainnet above 2.2M for my i7 quad.

it's also the last version I found blocks with on mainnet. 

v6 was fastest (most BPMs), but I can't get it to run on mainnet with the settings I had on testnet (6144000).

[maco]

So it seems like a few people cant get sievesize to 4M or more.

v3 has worked well above 6M sievesize for me. it's the only version that works on mainnet above 2.2M for my i7 quad.

it's also the last version I found blocks with on mainnet. 

v6 was fastest (most BPMs), but I can't get it to run on mainnet with the settings I had on testnet (6144000).

What did you set 6M at? I think 61440000 did good for me, but that was only on the testnet and never did good when it came down to the nuts and bolts of mining, haha.

[Tuck Fheman]

What did you set 6M at? I think 61440000 did good for me, but that was only on the testnet and never did good when it came down to the nuts and bolts of mining, haha.

HP3 had best results (on testnet) at 6553600. It outperformed 6M & 6144000 on HP3. However, I never found a block on mainnet using those settings and it yields low PPS on mainnet.

Right now I'm using v7 at 2M/8% and getting good PPS (good=not as good as others, but acceptable) for this difficulty (compared to other settings/builds).  This is with an older i7 860 quad @ 3.22 on Win7/64.

Even though I'm seeing lower PPS w/v7 I'm sticking with it for Sunny's fix.

[Tamis]

People starting to mine should know that by now block finds are rare even on i7 3770k @ 4.4 !
I used to find a block a day and I have found none during the last 72h...

[arnuschky]

Can someone confirm that -hp7 works? Didn't find a block since 24hrs.

Can someone of the "big players" could confirm that -hp7 works fine for them?

[paulthetafy]

Can someone confirm that -hp7 works? Didn't find a block since 24hrs.

Can someone of the "big players" could confirm that -hp7 works fine for them?

Yes hp7 works fine.

[masterOfDisaster]

Can someone confirm that -hp7 works? Didn't find a block since 24hrs.

Can someone of the "big players" could confirm that -hp7 works fine for them?

I might not be a big player but I can nevertheless confirm that hp7 works fine - if you mean in terms of solving blocks with it. I can confirm that for both mikaelh's 64 bit binary for linux and the 64 bit binary for windows.

[rahlquist]

Getting a dependency error on my fresh cent 6.4 install, any thoughts? I dont see a source for this lib for this version of Cent.

[rahlquist@corgi 64]$ ./primecoin-qt
./primecoin-qt: error while loading shared libraries: libminiupnpc.so.9: cannot open shared object file: No such file or directory

[Trillium]

Anyone have any luck getting this to run on the Xeon Phi?

http://www.intel.com/content/www/us/en/processors/xeon/xeon-phi-detail.html

Phi are coprocessors, and probably require specific coding to utilize them. On the other hand, intel might've designed them to be virtualized by certain OS so that they can seamlessly be used as 'normal' CPU by any processes, but I doubt it.

From your link:

Ease of programming Intel® architecture

Applications can support both Intel® Xeon® processors and Intel® Xeon Phi™ coprocessors which use common languages, models, and tools, including:

    Familiar and standard development tools—no need for learning new languages or tools
    Intel® tools like Intel® Cluster Studio XE, as well as many tools from other vendors
    Popular approaches such as Message Passing Interface, OpenMP, Fortran’s DO CONCURRENT*, Intel® Threading Building Blocks (Intel® TBB) , and Intel® Cilk™ Plus

It sounds like it could be done, but someone would have to sit down with them and develop the code, then do tests.