[ANN] AEON: Scalable, private, mobile-friendly cryptocurrency

[myagui]

Thanks for the feedback jwinterm. I know that most up-to-date ccminer forks are built with static libraries (so separate dll's are not required), but the cryptonight base from which cryptonight-lite was made, was a much older fork and generally not maintained, so I doubt it was configured in such fancy ways (not that I know enough to really make a call either way).
This helps though, as missing (or conflicting) dll's might also be causing the issue for SRBOOTH.

The issue for SRBOOTH was outdated drivers, just confirmed over PM and so all is well for him now  
I'm not sure now, but I think I had built with cuda 7.5, which would explain why a recent driver is required. Hmm... maybe I mentioned something around the time? Old farts have crappy memory  
 
For anyone interested in trying out my most up-to-date build, here's one with some personal tweaks and both cuda 6.5 & 7.5 flavors. I cannot for the life of me recall what compute levels I had included on this build, but at minimum, it should be compute 5.0 and upwards. Might also do better at regular cryptonight mining, for whatever that's worth...
ccminer-cryptonight-lite-beta-3

Edited to add: Note that most up-to-date build and personal tweaks don't necessarily make it any faster!    This version might do better on some cards, worse on others (for all I know, might do worse on all of them, unfortunately I cannot afford the testing time right now to go back and compare, all of my GPUs are tied up).

And here's a bunch of dll's, should have everything that might possibly be necessary to run any current x86 ccminer (not x64), as well as it most definitely has some extra (needless) stuff in there... It's my go to copy/paste set from many ccminer adventures over time  
ccminer-dll-pack

I'm exchanging PMs with SRBOOTH to see if we can figure out why the ccminer binaries that I posted are not working for him.
@jwinterm, if I read correctly, you also had some trouble on Windows. In case you used the binaries that I had posted, do you recall what was the error on your end?

@smooth, awesome job on the pruning release! I've been happy solo mining, hitting at least one block per day, sometimes two. Need all the power we can have against nefarious Moooooooooo.    I'll see if I can build the pruning release and start testing that as well tonight.

When I try to launch the binary, I get a "MSCVR120.DLL is missing from your computer" error box popup, but I'm pretty sure I have all the microsoft redistributables installed, and regular ccminer works ok. It works fine when built from source on Ubuntu computer though.

[1713965748]

1713965748

[1713965748]

1713965748

[SRBOOTH]

I'm exchanging PMs with SRBOOTH to see if we can figure out why the ccminer binaries that I posted are not working for him.
@jwinterm, if I read correctly, you also had some trouble on Windows. In case you used the binaries that I had posted, do you recall what was the error on your end?

@smooth, awesome job on the pruning release! I've been happy solo mining, hitting at least one block per day, sometimes two. Need all the power we can have against nefarious Moooooooooo.    I'll see if I can build the pruning release and start testing that as well tonight.

The relevant file causing the problem was my cudart32_65.dll was not up to date for the 6.5 build even though I had installed cuda 7. Found this out after updating my driver. Then copy and pasted the updated cudart32_65.dll to the .bat file on another machine where the driver was not yet updated and ccminer worked fine without going through the update. Kinda good to know in the case of a driver rollback or other testing. So it looks like , at least in this case, a matching dll would would work for others who have not or do not want to fully update. This was related to the 750ti btw.

[myagui]

Here are freshly compiled (and unofficial) AEON light-node binaries for testing on Windows:
I built these on Win-10, Boost 1.59, VStudio 2015, and look to be working fine on my Win-8 box as well.
Aeon-Win-x64-light-node-beta

Instructions starting up a light-node are here:
Pruning (aka Light Full Node) test release

And the FAQ on pruning just upthread:
Pruning FAQ

So for now, I'll be running a light-node instead of the archive one, and solo mining on the daemon as before. I'll most likely run a full node in the long term as I have the resources available, but anyhow I figured that testing light-node is more useful right now.

@SRBOOTH: thanks for the feedback 

Cheers!

[jwinterm]

Thanks for the feedback jwinterm. I know that most up-to-date ccminer forks are built with static libraries (so separate dll's are not required), but the cryptonight base from which .....

Working now myagui. Turns out I have cuda5.5, and both versions work, so I guess it was just the missing dll. I'm getting much lower hashrate on that windows computer though, around 250 h/s per 750ti, compared to about 425 h/s per card on ubuntu machine. So, don't know if that's cause I need to update cuda, or just cause windows sucks, but, anyway, it's working now

[SRBOOTH]

Thanks for the feedback jwinterm. I know that most up-to-date ccminer forks are built with static libraries (so separate dll's are not required), but the cryptonight base from which .....

Working now myagui. Turns out I have cuda5.5, and both versions work, so I guess it was just the missing dll. I'm getting much lower hashrate on that windows computer though, around 250 h/s per 750ti, compared to about 425 h/s per card on ubuntu machine. So, don't know if that's cause I need to update cuda, or just cause windows sucks, but, anyway, it's working now

I have similar deficiencies on certain comps running my 750ti's. Not absolutely positive but I believe it boils down to the architecture of the machine and/or bios. Bios update risk not worth the reward. ...I know this btw because I have swapped 3 different ti's between about 5 different machines and then played musical chairs with them and the deficiencies persisted among the mobo/processor/bios architectures and not the variations in cards or software. ....not likely your fault

[smooth]

Can somebody give me a brief summary of what I need to do to set up a pool. I've done it a few times before (for XMR) but I'm not really up on which versions of the code to use for AEON, any patches needed, etc.

Post if you feel it is of general interest, otherwise PM.

Thanks!

[SRBOOTH]

Can somebody give me a brief summary of what I need to do to set up a pool. I've done it a few times before (for XMR) but I'm not really up on which versions of the code to use for AEON, any patches needed, etc.

Post if you feel it is of general interest, otherwise PM.

Thanks!

I would love this info as well, I have a Dell R410 server machine that is pretty much idle and up to date.

[Arux]

Hi! I'm back at home. Today i will be busy with my day job (to explain my vacation at my office colleagues  ) but shortly i could active here.

[Arux]

Can somebody give me a brief summary of what I need to do to set up a pool. I've done it a few times before (for XMR) but I'm not really up on which versions of the code to use for AEON, any patches needed, etc.

Post if you feel it is of general interest, otherwise PM.

Thanks!

An important point: node v0.12 don't work. You must use node v0.10 else the compil fails.
With node v0.10 compiling and running the pool is ok. (Type node -v to check your node version)

[Arux]

There is a bug in the payment module of my pool. One miner is affected now. He's not paid since several days now. I'm aware of the problem and i will investigate ASAP
The coin are safe (but in the wallet pool, not in his personal wallet) His pending balance show correct information (the pool must paid to him this balance). My apologies for the delay.

[edit] 01/09/15, solved. there's no more pending balance above 0.5 . Payment module is Ok.

[myagui]

Thanks for the feedback jwinterm. I know that most up-to-date ccminer forks are built with static libraries (so separate dll's are not required), but the cryptonight base from which .....

Working now myagui. Turns out I have cuda5.5, and both versions work, so I guess it was just the missing dll. I'm getting much lower hashrate on that windows computer though, around 250 h/s per 750ti, compared to about 425 h/s per card on ubuntu machine. So, don't know if that's cause I need to update cuda, or just cause windows sucks, but, anyway, it's working now

I have similar deficiencies on certain comps running my 750ti's. Not absolutely positive but I believe it boils down to the architecture of the machine and/or bios. Bios update risk not worth the reward. ...I know this btw because I have swapped 3 different ti's between about 5 different machines and then played musical chairs with them and the deficiencies persisted among the mobo/processor/bios architectures and not the variations in cards or software. ....not likely your fault

Thank you both for the inputs.

I would bet on something specific to that Windows setup as fault for the low speed, if not the dreaded "operator error"  since that Windows hashrate sounds about right for regular cryptonight (as in missing the -A switch). Linux does generally better on the same hardware, that's for sure, and more so with any algorithm that is memory hard, where the Windows (WDDM) driver overhead really bogs down the performance.

Well, could also be that your Windows launch settings (threads & blocks) were not optimal. While on Linux, sub-optimal launch settings go partly unnoticed due to the comparative hashrate boost from Linux vs Windows. On Windows, less-than-perfect settings will cause much more of a performance hit.

I don't have the proper launch settings for the 750TI myself though, don't have those cards anymore. I do have reports of cryptonight-light performance in the region of 500 h/s on these cards, even on Windows. Note that both cryptonight & cryptonight-light allow for massive overclocks on the 750 TI, seeing as the GPU is hardly stressed at all.

[smooth]

I don't have the proper launch settings for the 750TI myself though, don't have those cards anymore. I do have reports of cryptonight-light performance in the region of 500 h/s on these cards

Well look, the algorithm has half the iterations of regular cryptonight so at a bare minimum the hash rate should be close to 2x.

Anything less than that indicates a severe tuning problem.

[myagui]

Correct smooth. But what I've seen in my limited experience here (as I don't have the ideal GPU models), is that without specific tuning for cryptonight-light, the performance will not automagically double, using the same launch settings as for regular cryptonight. I had observed ccminer hashing just marginally faster, if selecting cryptonight-light vs cryptonight on the same baseline settings (at least on Windows).

With proper launch settings, yes, the performance should be roughly double that of regular cryptonight. With ccminer, the ideal cryptonight-light performance, is probably found by using a launch string that uses double the number of threads as for regular cryptonight - in order to take advantage of the smaller workload.
As I understand, the ccminer thread synchronization and memory access behaves very differently in Linux, causing Linux miners to see most of the cryptonight-light gains even without tuning their launch parameters. In any case, proper tuning will provide the best results.

[GingerAle]

I didn't run it for long, but

-l 32x30

got me 520 on the 750 ti, and 578 h/s on that PNY OC version of the 750 ti (the one with 1220 clocks).

I also used nvidia 331 drivers. The driver version can affect things.

edited to add: this was on ubuntu 14.043

[myagui]

Thanks GingerAle,
That is exactly the performance range I would expect, like Smooth was saying, roughly double the performance of regular cryptonight. You can definitely get it to run faster with overclocking, as cryptonight & cryptonight-light are more tolerant to overclocking than most other algorithms. Anyhow, looking really good, cheers!

Also a +1 about drivers, some versions are faster, others are slower. Even across different systems, it's not always the same driver that performs best (cards on risers vs cards directly on slots, paired with fast CPU vs paired with slow CPU, etc).

I didn't run it for long, but

-l 32x30

got me 520 on the 750 ti, and 578 h/s on that PNY OC version of the 750 ti (the one with 1220 clocks).

I also used nvidia 331 drivers. The driver version can affect things.

[Blawpaw]

I really like this one. Heard about Aeon coin mining being enabled for android devices.

Any update regarding this topic?

[smooth]

I really like this one. Heard about Aeon coin mining being enabled for android devices.
Any update regarding this topic?

Not yet demonstrated, but the code should compile on Android if someone wants to try it.

[sulfurtank]

I think that's a shame since you would expect a minimum of functionality as an incentive for someone to set-up a full node for this coin.

[SRBOOTH]

Apologies for this TLDR question: Do I need to run the client on each comp where I want solo (gpu)? I currently am running a full windows version. I can't get the gpu's to connect to the wallet in the same network. Port 11181?? I solo mine 1 wallet (not a cryptonote coin)  from 3 different comps in the same network all the time. I have never gpu solo'd a cryptonote. Could someone point me to a good set of instructions please??

I want to set this up in anticipation of running the pruned version when the windows compilation comes out
Thanks in advance!

[myagui]

@SRBOOTH:

There's no external GPU solo miners for any cryptonote as far as I know. Most definitely not for nvidia, since ccminer will only work with pools. If you really want to solo mine with nvidias (or AMDs AFAIK), you'll need to setup a private pool. That should be an interesting/fun project btw.   Happy Mining!

Edit: on the pruned version compiled for windows, I've posted (unofficial) binaries just yesterday:
https://bitcointalk.org/index.php?topic=641696.msg12285871#msg12285871

Official ones should not take long I suppose, since Arux is back from all his summer partying  

Edit ^2: the light-node (pruned) working well for me so far. I have made a couple of transfers without any issues, I have solo mined a block, and have gone through the cycle seeing a misbehaving peer get blocked. Good stuff!