Topic: [XPM]unofficial jhPrimeminer thread - page 33. (Read 180246 times)

v4 v5 v6
Which version is best? Thanks.

I may not be able to answer all your questions, but I can do this one. Val/h is "share value per hour". And Share Values are exactly what they sound like, the value of each found and submitted share (6ch = 0.1, 7ch = 1, etc)

mumus, is there an IRC chatroom on freenode that people in this thread and use for discussion?  

I'm thinking of starting a mining hardware comparison page or something for primecoin.

With my A10-4600m, using the following settings for V6 AVX: -s 200000 -d 16 -t 4
I am managing between 25000-19000 PPS with 9.1 Val/h.. whatever that means.

The strange thing about the performance of my A10, is that i must disable Primorial Multiplier Auto Tunning, or it goes from its start up value of 67 (which is good for this cpu, as far as I can tell), down to 19.. which lowers PPS and Val/h.

Also, the startup tuning seems odd.  At parts of it, i have much higher PPS, but after awhile, it slows down... and its final value always seems to be slower.  Is there anyway we could get flags to disable the auto tuning, and manually program the L1cache value?  Also, another flag to disable Primorial Multiplier Auto Tunning on startup.  It simply doesn't work for my CPU.

I'm not sure what Val/h is, but my Core2 Duo downstairs is getting about the same, with much lower PPS.

If those two accounts on ypool do belong to you mumus, I am interested to hear what hardware you are mining with.

v6 AVX has been a champ for me on this machine.  I've gone from the 1900 PPS i could muster out of HP9, to this 20000 PPS setup, and I somehow managed to mine not 1, but two blocks lastnight!

Yes, we can only tell after numerous days of testing and probably just luck. Started running hp9 and found a block after 8 hours although I have been running hp9 in the past for over 2-3 days without finding anything. I'll probably stick to a mixture of solo and pool.

Thanks.

As far as I see the official 0.4 version should contain most of the optimizations and even more. (See the changes on https://github.com/jh000/jhPrimeminer)
I haven't tried to run it for longer period (I'm running tests for my ideas) but it definitely worth a try.

mumus, just wondering, will you make an optimized version of 0.4 as well? I am doubting whether to use yours or the newer 0.4.

Just for the record,official 0.4 was out a few days ago. Changelog:

Just to be correct, I didn't developed the miner. I just put some pieces together to be more efficient.

You know that this is pure luck. I was in the both sides, and I still I prefer using the pool.
I'm nut sure that the jhPrimeminer is faster or better than mikalh's HP9 primecoin client. I never compared them and it's hard to compare without running with enough computer power and over a longer period of time.

Hi Mumus,

Is there any chance you can have an option to solo mine?

Over a 5 day period my desktop i7 found 3 blocks (2 on v5, 1 on v6) for the pool and I only got 14 XPM for my efforts 

It would be really good if I could solo mine with your miner.

Hi Mumus,

I hope you really mean "no hard feelings" as it definitely wasn't my intention to offend.  You have spent a lot of time and effort in to developing this miner, and myself and everyone else really appreciate it.  As soon as I have some XPM in my wallet I'll make a donation for your efforts.

I'm a suspicious person by nature though, lol.

You're absolutely right though, a wireshark trace or compiling the binaries from the source (after checking the code of course) would resolve any doubts.

I for one, although suspicious, am also lazy so am using the compiled binaries that you posted! 

Thanks for your understanding and open mind.

Cheers,

Beanhead.

  No hard feelings. I can understand your thinking and feelings.
I wish I could have one of those accounts.
But getting these result is no such a big deal. Using 50-100 instances of VPS can make this easily, when somebody has the guts to invest. Just check the other forum ("Primecoin High Performance"), some peoples are using enormous computing power to do solo mining. Actually these users with big hashing power just helps the pool to have less variance,  getting more  consistent income and less dependent on luck.
If you want to solve your concerns that the binary somehow submit shares to my account you may ask someone who trust to compile the code and publish it. jh00 can be one of them. I think everybody trusts him. Also with a sniffer or connection monitor (tcpview) can be easily discovered if the binaries behaves strangely. BTW, the idea is not that stupid, could be easily implemented, but I'm sure some of you already checked the binaries and would be discovered already. Anyway who don't trust the binaries used the code to compile for themselves.

that's exactly what I thought... it's a proprietary protocol and a proprietary miner, open sourced, yes, but we have no idea if the compiled version is the open source one... that discrepancy in the share count is eye catching.

anyone running v6 with a larger than default sieve? Figured id try some different settings,trying 2M for awhile everything else default. just started pool minining recently im running an i5 2410M (laptop) 2 core 4 thread 2.6ghz

My laptop has found 2 blocks with V6

Looks like V6 is performing better on my quad core, have yet to get meaningful comparison between my 2 old dual cores.

Thank you for the nice comments, I meant no ill will to Mumus or anyone else by my comments, I was theorising and you obviously missed my point.  There is no escaping that those stats do stand out.

The following is how sad and stupid I am,

How do we know that the compiled version that we have all downloaded contains the same source code as what has been posted on github?  Unless someone has taken the time to disassemble it? Then we don't.

So in theory, and I am not suggest this is the case, there could be some code in the miner we are all running that submits the odd share here or there to an account other than the one specified when you run it.  The odd share here or there times a couple of thousand miners? You do the maths.

This is how it is potentially possible that we are mining for him. What a ridiculously stupid idea though.

Again I was theorising around the possibilities of that vast power advantage those two users appear to have over everyone else.  It may not even be Mumus, and in which case the theory above is completely null and void.

Again I would like to apologise if I have caused offense.  I am no troll and that was not my intention.

How effing stupid are you?  HOW are we mining for him???  In a quick look, he's found 9 out of last 30 blocks.  1.41-1.52 average per block x 30 blocks = 42.3-45.6 XPM.  If he'd been mining solo that'd be well over 90!  Sure looks to me like HE's mining for US.  (Prob shouldn't tell him, he'll quit the pool and then where are we?)

So what if he's running a VPS farm with 400 CPUs?  I know I am close to break-even, have earned 97XPM through the pool and my stats say I've found 9 blocks.  So I would have been bettter off if I'd been solo, but I'd not have had a constant income.  It's a trade off.

Yes, but no one else in the pool is getting anywhere near those stats.  That is a serious amount of horse power to get that kind of figure.  The closest is about 11000, but those two combined are 32000!  Looks out of place in the list of other miners.  Maybe I'm just being paranoid. Maybe there is a GPU version out there? Maybe he owns a server farm? Maybe we are all mining for him? Maybe  it isn't even him? Maybe I should shut up lol.

Remember, shares count for 40 blocks...