Post
Topic
Board Mining (Altcoins)
Re: Swedish ASIC miner company kncminer.com
by
CYPER
on 23/12/2013, 16:55:29 UTC

There will be a tuning suite for Nov, but not this side of Jan, as everything there is winding down. There's just some ongoing Neptune work. There will also be the means to vary SPI clock frequency. I saw it in the original tuning suite as a feature, but that hadn't been developed into the recent release. It was purely part of a template I saw.



Can you please elaborate on what that option does as I believe many of us are just changing values without knowing what they actually do  Grin
The voltage settings are clear, but the SPI voltage and frequency not so much. I know they are related to the communication between each board and the controller board, so do they increase the signal to noise ratio or am I talking rubbish  Grin

By changing the voltage you are altering the power consumption and the noise to signal ratio (communication error rate - although reported by CGMiner as hardware errors they are not actual hardware errors so don't worry)

The less voltage the higher noise to signal ratio, the higher the probability of hardware errors.

The greater the voltage the less noise to signal ratio, the smaller the probability of hardware errors.

The higher the frequency, again the higher noise to signal ratio, which is not what you want.

If the SPI frequency is too low then there is not enough bandwidth to collect all the good nonces found. So you want to find an equilibrium where by SPI frequency is high enough not to miss any of the nonces found, but low enough to retain a healthy noise to signal ratio and thus minimise hardware errors.

Also, my bad, the future firmware update is not about SPI frequency, it's about configuring the PLL settings, which is directly correlated with the hashing speed as you are directly altering the clock speed of the die.

The SPI is not the hashing speed, but the speed of the communication between the dies and the controller board.

Hope that clears things up bud, and Merry Christmas, or if you're not into all that happy holidays all the same. Wink

Thanks for the info Smiley
I have one other concern in relation to changing certain hex values, that increase some other values, that lead to increased hashrate  Grin which bothers me lately: according to one KNC engineer, which I met in person  Grin anything above 50Amps per VRM is considered not safe for long term usage. Or 200Amps per board. He wasn't very clear about it. Yet the 0.99 Tune firmware has the option to increase the current per VRM up to 64Amps, which brings the question: if over 50Amps is considered not safe, why then we have 64Amps at our disposal?
What is considered safe? Judging by the latest firmware I would assume 64Amps to be safe or at least any logical, non-technical person would believe that as it is included in the Advanced tab options.

Happy holidays and don't drink too much  Tongue We need you back here in the new year  Cool


The GE modules are rated for 40 Amps, but can safely be clocked above that. 64 Amps is the max where we feel comfortable clocking the VRMs to, and strongly advise against considering anything higher. Obviously pushing things too far reduces operational life expectancy, but if it's been included in an official firmware then be confident that we are comfortable releasing it as such.

Not sure where you can take the Ericssons. That said i was out of the office for a couple of weeks touring with a Nov Jupiter, and then back briefly for a week before heading home for Christmas tonight. I know precisely who you are talking about and will bend his ear when we next see eachother, he's not in today, they celebrate Christmas earlier here so i'll pick his brains when we next catch up, unless you catch him first. Wink

I am even more confused now  Grin
He literally scared me and I lowered my Amps below 52 and you say that up to 64 should be OK as KNC feels comfortable with that.
I like your position better  Cool Please verify it with him and come up with a uniform position  Roll Eyes

In regards to November Jupiters: they are hashing like beasts, so no point trying to change anything there.
But just as a comparison: the total current for a November board is around 200Amps (8x 25Amps), so maybe there is some correlation between that and the proposed max safe limit of 50Amps (200 in total) for an October board? But then again that 8x 25Amps might just be fairly reserved and not close to the true safe maximum.
I wish we can get an official response, just for the sake of people who decide to set all to max voltage (64Amps) and have no clue that it might be dangerous.