Topic: GekkoScience NewPac / Terminus R606 (BM1387) Official Support Thread - page 37. (Read 61923 times)

Sounds to me like he's mining the alt's with CPU cores.

1TH on power level 3? That's a pretty sweet unit you got.

my only question is - how are you mining XMR with a sha-256 miner? or am I misunderstanding?

I'm a new R606 owner and here are my anecdotal tests and observations.

I purchased my R606 from Eyeboot on Amazon and it came with a R606 wrapped in bubble wrap with a USB cable in a ziplock bag taped to the outside in a Amazon box.  No paperwork or how to get started guide or anything.  I found this place and read through hundreds of posts to get my bearings.

I powered my 606 with a spare PCIe cable from my altcoin mining rig and used a spare raspberry pi 3B to control it (my mining rig runs on hiveos)

With the 3B the best I could do was 800ish GBs running full power.  I read that the Pi4 does better so I bought one with 4GB ram a case with fan and all of the accessories.

With the Pi4 I got right around 1Tb on power setting 3 with it having no difficulty reaching 750 frequency (I read on here somewhere 750 is about the max safe frequency)

I tuned it back down to 550 frequency on power setting 4 (stock settings I think) while I waited for a thermometer to come in the mail.

With a digital thermometer mounted to ass end of the top heat sink it was time to experiment.

I spent a lot of time upgrading the Pi to a 64 bit kernel and installing XMrig to see what would happen.

With XMrig running all four cores I was back in the 800GBs range no matter the power setting.  I expected this.  What I didn't expect was that I couldn't get the two miners to work together to give me a happy medium.

After much experimentation I ended up with the R606 on level 5 and using 3 cores for Monero.

This seems to be the best setup with my hardware but the R606 will just barely put up 1TB and the cpu miner will find several shares over several minutes and then go flat for several hours before finding more shares again.  With all four cores mining I can get a consistent 110hs mining Monero but there isn't enough room for the R606.

Updating to the 64 bit kernel doesn't seem to have any effect on the 606.

In conclusion to this post I believe that a Pi4 is the best way to go considering all the trouble and headache that I see a lot of users go through and how painless it was for me to set up on the Pi.  The 64 bit kernel doesn't do anything and if you try to push the Pi to do anything else it just won't without fucking up everything else.

Right now I'm running about 1.01T on power setting 3 and a temp of 55.7C and I'm quite happy.

Adding them to the README.md (github) and the `--help` output would also be really swell.


--gekko-start-freq and --gekko-step-freq are flags I found elsewhere and have been working very well for my 2pacs; sadly, no effect on the compac. Hmm. Maybe that means the r606 patches have made it to master branch. I have been, until today, unaware there even was a feature-branch r606 and these two flags, at least, are working as of my last update and compile cycle.

I'll need to read up on tune-down. I'm not sure it's what I'm looking for. I was thinking something like:


where N is a setting
  0=disable autotune,
  1=max-hash (per freq setting),
  2=max electrical-efficiency (disregard freq statement, if any);

or something to that effect. Not that this is a big concern, really. I can work out the math from the power tables provided and use frequency values. Of course, it would help if the power table was expanded to include all sane (and some insane) freq values; or a W=vcore*freq formula.

I know its daunting looking through the 80 pages of posts. Maybe Sidehack/VH can update the first page with all the config settings cgminer now supports?

you may want to try --gekko-start-freq, I havent gone through the code on github to see if it works for terminus but it would be worth a quick try. like you said, letting it autotune is best, but this is the one your looking for: --gekko-tune-down

https://bitcointalksearch.org/topic/m.50506833

 [90-95]% is a healthy range, 95 is default, set to 0 to disable.
        1.  Adjusts frequency according to each miner's unique charateristics
        2.  To keep a hash performance above 95%.
        3.  Based on the idea that (95% of 275MHz) hashes better than (80% of 300MHz)

In preparation for picking up my newpacs and hub from 419mining, the following questions are important, because I'm too dumb to figure it out from the github repo:

Has the r606 branch been merged with master, or is it still recommended to build from r606 (if I want autotune and other new features)?

Once autotune has settled on a good frequency, can I then disable autotune for future restarts and set the tuned freq in cgminer.conf? Understanding that each I may lose optimal tuning on future restarts, but this would allow for rapid ramp-up to "good enough."

Is there a setting to cause autotune to choose best hash:power ratio, when that is less than max hash rate?

Thanks a bunch.

(eegads, weeding through 80 pages of posts is exhausting. Admittedly, after around page 35, I just skimmed for posts from sidehack and vh making sure I spotted all the software update notes.)

I'll probably fix it sometime. I tended to lean on my webserver for image hosting but the connection was fairly slow so I also used photobucket. The server's now set up in an actual data center with a solid connection so I should port everything over to it.

The pic's in that thread have been down for a really long time now.
Here's the reply I got from the other thread:

I googled it and found a picture. But it would be handy if the first post pics were back up.

Looks like the announcement thread pictures are not displaying. Maybe it's me on my end?

Just looking to refresh on the voltage settings and how they look displayed on the unit.

Got a couple of high end ones there then. I’ve just received my second/replacement and working on building it up to 1th (or close).

nice! i cant get mine to go over 650 mhz even at 650 mhz after a week its down to 350-400 and i have to restart it to get it back up to 650.

Both are running 725Mhz, one power level 5 and the other power level 6 for a total of 1.94THs.

What are your settings?

My two R606's are pulling 207 Watts, including the Power Supply and minus the Pi 4.

In theory, how many 606’s could you run off a 1600 watt 110v circuit?

No worries. It seems that chip 0 always wants to be one step below whatever freq I set it at. I've got another unit on the way which hopefully wont be too long. I may take a closer look at it once it arrives. Do the heatsinks come off at all? I haven't tried yet but didn't notice anything on the areas I could see.

EDIT, Update:
She's been running at 550m V5 for the past couple of days, tried 575m but didnt like it. It seems thats all the puff she has now. At least shes stable though 

No, 12 would be the 12th chip (indexed as chip 11) at the other end of the string. It would have been replaced before re-testing. I may be remembering a different unit but I believe that chip was having trouble breaking 400MHz.

0bit, I totally agree with os2sam. I was one of the people who was having trouble running newpacs on a Win10 pc. I helped get some data together for vh (who developed the “gekko” version of cgminer) to try and get things running stable on Windows machines but it was a PITA. I ended up getting a Pi4 (not a pi3) as it was the best for running them on. They are even tested on Linux machines by the manufacturer.

Performance will be BETTER and it will run BETTER on a linux based OS. The reason being, the asic boost feature of vh’s cgminer will run properly on a linux OS, on a windows machine it wont so you wont reach the potential peak hashrate. Also I found that it would constantly restart the sticks and I ended up running out of USB com ports so eventually they would all stop working on Win10.

But if you are insistent on cross compiling it then feel free to waste your time.

---

Update on my issues:

So ive been testing it over the past couple of days and it seems to run for almost 24hours on V5 at 600-650m. But seems to always cut out. I've switched it down to 575m V5 to try and run it at almost stock clocks to see how it goes. One thing I have noticed is that it is ALWAYS "chip 0" which doesnt like  the freq. sidehack, would this possibly be the "12 slow" that was marked on the pcb or is that the other end of the string?