Topic: Innosilicon releases A9 Zmaster 50ksol/s Equihash miner at 620W - page 13. (Read 45686 times)

Has anyone gotten confirmation of their september batch shipping? I'm waiting on 6 "free" models from one order and 15 "free" models from another order. Just curious if anyone else has gotten anything. It's kinda crucial that these things go on soon. What's interesting is the spikes of difficulty still. I think the difficulty will increase btu there is also more than one equihash algorithm. If all these asics coming out disperse across the different coins difficulty will only be marginally highly as opposed to everyone going to Zcash or Horizen. Just hope we don't get burned with these asics, I have confidence that if people disperse we will all be fine.

After 38-44h all machines restarted because the memory was full. For the first time actually one froze. It had the July FW. Otherwise all back to normal. I restarted all at the same time again to compare another round.

Hi All, Sorry i may off topic.  But i just want to find out whether anyone deal with https://cryptodrilling.com?  i am looking to get one but not sure they are legit.

Someone had their miner from September batch shipped already?

So after 36h the cache is over 90% (90-94% depending on unit). On the August and July FW. Its the same.

Yes, no problems at all.

Interestingly now some miners are at 99% memory usage. After 34h. But still hashing fine!

Everything becomes a D3 eventually.

I think anyone that missed Inno's $10k for 2 in late June, or $650 batch 2 Z9 mini (From buying $250 coupons online) has missed the boat.

i wonder how long until equihash ASICs become the next D3

Sidenote: Youve really tempted me to tear apart me computer to try my bad unit on a high quality power supply. I tried swapping and no effect on eithers hashrate. Ive been so baffled by my 28ksol Unit. Slight improvements through overclocking but nothing significant. Also, beautiful boards with nice solder. No weird heat spots on FLIR. Just seems like a bad bunch of chips. Unless maybe its some wierd power issue. My gut says bad silicon tho.

Do your units all hash well?

Interesting. Im not arguing just pondering. Increasing the VID would have a definite increase in power usage. So would the fans being set higher. I wonder if there is autotuning happening under the hood, even tho cgminer doesnt report it. And on a full speed unit like yours. The autotune backs off the VID.

Or maybe a net efficiency gain by running cooler is occurring and ive missed it since if been running overridden fan speeds the whole time.

After nearly 24h, I can tell that on the August FW I am getting 1% rejects, on the July FW I am getting 2% rejects. One could say its 1% less. Marketing guys would say its a 50% decrease of rejects.
Concerning power: I am running 2 units on one PSU (same models). I am getting 1322W - 1325W for 2x A9 with August fw. As I observed it, it dipped some times for a moment under 1000W. 2 units with the July FW gives me 1320-1324W. As I observed, the dips occured here also, but not that frequent and not that severe. I am using TP-Link Smart plugs.

For me they use the same power. I can not tell the difference. Maybe the lower temps are because of the fans.

Thanks for the info! 

Dont think of it as Inno providing anything. They didn't. Things are just thrown together to work. Its more like, how can I exploit Linux openness to suit my need. Quick search suggests this is nginx webserver for the web gui. Its not something Im familiar with but its likely possible todo all the cammands in the gui with direct http requests. Probably poorly documented im guessing. There's cgminer api commands that seem to have decent documentation.If it provides what you need. Dont know if this helps at all.



So.. Ramp up firmwares? Ive only noticed ramp up changes in the GUI, due to Inno messing around with hash/timeframe measurement. Never noticed much pool side changes. Has anyone noticed a definite pool side ramp increase?

Andartis, how accurate is your power consumption measurement? Im pretty sure performance+ will use more as it run the chips at a slightly higher voltage. Unless that voltage gets scaled back on some units? Doesnt on mine. Even the fan running faster should draw more power.

Some power #'s for anyone interested. Probably not precise and has an HS100 Smartplug on them too.

Crap 28ksol Unit 
** Seems to run 29ksol at higher vid but also set to noauto/PLL (Mhz) 1300-1350. I don't know that Pll adjust does anything after initial startup. noauto is necessary for VID change.
220 VID 835 Watts - no auto,  fan1 Max (200%?), fan2 Max (200%?) - 29k
205 VID 790 Watts - no auto,  fan1 Max (200%?), fan2 Max (200%?) - 29k
195 VID 774 Watts - no auto,  fan1 Max (200%?), fan2 Max (200%?) - 29k
175 VID   ?   Watts - no auto,  fan1 Max (200%?), fan2 Max (200%?) - Seem to run poorly again. 28k. Probably will retest.

50k sol Unit **No significant gains ever noticed!
175VID 666 Watts - Performance+ mode,  fan1 "100%", fan2 Max (200%?)
205 VID 770 Watts- no auto,  fan1 "100%", fan2 Max (200%?)
205 VID 747 Watts - no auto,  fan1 "100%", fan2 "100%"

Setting PLL too high makes the unit stop working. But setting 1300-1375 doesnt show significant gains on my 50k unit. Maybe clock is strictly managed after a time and it only starts up on set PLL. Never seen a device not work better at higher (stable) frequency. Seems like underclocking might be more of a thing on this unit unless inno changes something. kinda seems like they're working on something tho.

Anyway, no real point to be made here accept maybe that these things light up as you pump more voltage into them.

Side note: This is my personal Capacitor temperature scale
40°C - Enjoys life, not much to complain about
50°C - works for a living
60°C - Agitated, always yells at Nancy
70°C - Eye twitchy Stress
80°C - Thinks about suicide way to oftern
90°C - Dead man walking
I was concerned that inno sent me a $5k machine that ran at close to 70°C, unnecessarily. Feel like it was for low wattage marketing. obviously im overdoing it @ "100%" and ~"200%" with extension and cut wire blue, but meh. less statistical chance of an issue.

Last, i have some FLIR images if anyone wants to see. Didnt find anything with it except a strange 1K resistor from 12V to GND. Its a hard working little guy, continuously pumping 1/8 watt to ground. Light up on the FLIR. Was initially going to change it thinking it may be bad, until I realize that a 1k resistor hooked up to 12V WOULD run hot. Having a hard time guessing what its for. 12V Stabilization? seems excessive. Kinda want to take apart a good one and make sure it has the same part. lol. but that seems so unlikely it would be wrong part.

oh, any everyone has figured out how to change pools without any ramp up right? 

Same problem here -manual restart-, I can confirm that new firmware version doesn't fix the memory leak.

Best regards!

Nicehash at the moment.

which pool you use?

Due to maintenence, it came that I powered on all the miners at th esame time. So after 12 hours nearly everything is identical. No reboots or freezes, 12h uptime on all miners of the 6th July and 28th August firmware.
Cache is exactly the same on all. Power consumption on performance+ vs performance (July fw) are about the same.

Only fans spin a bit faster on the August fw (5-10%) and the temps are about 5° lower. Ramping up, as reported, takes very long on August fw.

The hashrate graphs look smoother on the August fw because it shows past 12h, the July fw shows only past 4h. So dont get distracted by the smoother graphs.

Thats it for the last 12h. I will see now which unit freezes or restarts first.

Does anybody know if there's an HTTP API that Innosilicon provides?

Somebody posted this snippet earlier in the thread from a log file:


Trying to add support for this miner to a project that I'm working on.  Getting the output from this URI would be amazing, assuming that's a valid API endpoint.

Anybody know?  Thanks in advance!

Very often when I get memory full issue, it does not restart the process automatically, and I must manually restart.

It seems not. But I dont know a sufficient way to tell, rather than looking at the uptime and the memory used.
In general I dont unterstand why it is such a big problem. It restarts the mining process automatically. I never had big problems, except the dip in hashrate. To keep the dip short, you should use the faster ramping up firmware. The new one seems not to be better.

Hashrates are the same. The performace settings are a joke. In every version I tested so far.

I'd be very curious to see if this fixes the memory leak issue. Interested to hear your results.