Avalon A7 announced

[Kuulani]

WiFi does work with the controller I just got from Canaan.io it appears to be based on a Pi3-B it is disabled by default. That must be what they mean by "No-wifi mode". There is a wifi section on the very bottom of the main overview page and its grayed out saying its disabled. I clicked enable and it asked me for an ssid. I hit the scan on that box the Pi gave me a list to chose from. I picked mine and it rebooted itself an took off mining. When I get off work i'll look what version Canaan has on the PI, as it came with a micro SD card with the firmware already on it.

[PeaMine]

Review on the 721 and 741.
So far zero issues, and no reboots required since I first powered them on.
I have owned the 721 for about 4 months and the 741s for 3 months.
Controller works great as well, no issues.
Heat management seems much better and even than my S5, s7, s9.
So far it seems they're running at 0 (default) volt setting, but doing about 10% more hashrate than advertised.
Their first wave of summer heat will be the true test.
No failed chips or boards and no hash rate drop.

[NotFuzzyWarm]

Same here with mine but I must add that they are twitchy if power blips long enough for the PSU to drop out but not long enough for the RiPi to notice. Other than that - behaving perfectly

Both 741's I have had that happen. Yes they came right back to life but even after 8hrs only ramped up to about 5-6THs vs the normal 8.5ish I usually see. Once I was able to get to them, cycled the PSU's, waited for life, used the GUI to soft boot the miners again (always a good idea after a hard boot) and all was well again. Ramped right back up to target in just over an hour.

Oddly enough, the 721's attached to the same RiPi had no problem.

[philipma1957]

my three 721's are just about perfect  on kano.is

[NotFuzzyWarm]

my three 721's are just about perfect  on kano.is

Ja. Now if only someone can explain why Awesome Miner reports the Avalon's hash rate like this...

That pic is with just the 3x 721's.  For a reference, that band in the middle are most of my s9's pushing 12-14THs each
After adding the 741 to same node now spikes around 60-80THs are frequent. Impressive but....  

The average hash rate is correct as per what kano reports. Just looked at kano and the 3x 721' along with 1x 741 I have on a single Ripi is reporting 27.67THs.

Sooooo...... wtf is with the spikes? Patrike (Awesome Miner author) says that is what the API is reporting when queried. Most odd but - hey, KanoPool loves it so okely-dokely....

[Kuulani]

I have been running a pair of A741's since Feb. 8. And aside from a forced reboot ( power outage for 39hrs after a huge windstorm a couple weeks ago) they have been running just fine.  Couple of things I noticed tho maby others can comment on.

1. Mine don't have the voltage offset option like the one Phil has in his screenshots. It's just not there.

2. They start off slow and take a few hours to "ramp up to speed" but from what I have seen from others here that's normal for the A7's.

3. My Pair seam to run a bit above spec. but they also draw a bit more power to. All I have to go on for power measurement is a kill-a-watt pro so that could be off a bit as Phil has one that's a bit hot as well. I'm getting 1405w at the wall with a evga 1600g2 with an avg hash rate of 8.358th on 1 machine and 1389w at the wall also with an evga 1600g2 with an avg hash rate of 8.261th on the 2nd machine. Just a bit over the 1150w @ the wall with 7.3th spec posted by Canaan.

[PeaMine]

I have been running a pair of A741's since Feb. 8. And aside from a forced reboot ( power outage for 39hrs after a huge windstorm a couple weeks ago) they have been running just fine.  Couple of things I noticed tho maby others can comment on.

1. Mine don't have the voltage offset option like the one Phil has in his screenshots. It's just not there.

2. They start off slow and take a few hours to "ramp up to speed" but from what I have seen from others here that's normal for the A7's.

3. My Pair seam to run a bit above spec. but they also draw a bit more power to. All I have to go on for power measurement is a kill-a-watt pro so that could be off a bit as Phil has one that's a bit hot as well. I'm getting 1405w at the wall with a evga 1600g2 with an avg hash rate of 8.358th on 1 machine and 1389w at the wall also with an evga 1600g2 with an avg hash rate of 8.261th on the 2nd machine. Just a bit over the 1150w @ the wall with 7.3th spec posted by Canaan.

I haven't measured my watt usage yet, also using 1600watt.  I am also getting around 8~TH though, so perhaps I might need to use -1 or -2 to give my PSU some extra overhead?
Though I think Phil said his power usage didn't change with those settings, not sure.
So far no issues though under 31C ambient as well.
I do have one 741 that is restarting every 15 minutes or so but pool hashrate doesn't seem to notice, but I need to update the MM for them all any way.

[fanatic26]

To get real power figures you should use an amp clamp and do the math. In my experience none of those meters are ever accurate.


Also the 741 is listed at 1150w +15% so the actual power draw of the unit is ~1320 watts with a very efficient PSU. So you really arent over what they claim by very much.

[IntroVert]

...

1. Mine don't have the voltage offset option like the one Phil has in his screenshots. It's just not there.

...

If you upgrade your controller software to 12/29/2016 version (OpenWrt driver version 50019) for your RPi, you'll get the voltage offset setting.

https://canaan.io/2016/12/30/new-avalonminer-controller-avalonminer-721-mm-software-released/

[StevenMosher]

I have been running a pair of A741's since Feb. 8. And aside from a forced reboot ( power outage for 39hrs after a huge windstorm a couple weeks ago) they have been running just fine.  Couple of things I noticed tho maby others can comment on.

1. Mine don't have the voltage offset option like the one Phil has in his screenshots. It's just not there.

2. They start off slow and take a few hours to "ramp up to speed" but from what I have seen from others here that's normal for the A7's.

3. My Pair seam to run a bit above spec. but they also draw a bit more power to. All I have to go on for power measurement is a kill-a-watt pro so that could be off a bit as Phil has one that's a bit hot as well. I'm getting 1405w at the wall with a evga 1600g2 with an avg hash rate of 8.358th on 1 machine and 1389w at the wall also with an evga 1600g2 with an avg hash rate of 8.261th on the 2nd machine. Just a bit over the 1150w @ the wall with 7.3th spec posted by Canaan.

 
 Hi,

   First post here and I work for canaan.  There is one thing we have not adequately explained to people. The spec is 7.3 TH/s.  What that means is the device should never produce less than that. They very often will perform "above spec".

The reason for this is deep in the design of the chips. Rather than "bin sort" chips that could run at different frequencies,  every chip
dynamically adjusts according to its real time performance. We dynamically validate the chip output ( no hash errors ) and adjust the speeds according to the current operating environment. Each chip is pushed to run as fast as it can ( with no errors) and if there are output errors, the speed is throttled back. In no case should it fall below 7.3. So its a minimum spec. The tuning is actually more fine grain than I describe here, and as far as I know our decision to actually valididate that the output is error free is unique. Not all hash rate is created equal.

That's why we call it RTH/s  Reliable Tera hash. We are dynamically verifying the output of every chip at its current clock rate. If it can go faster, we up its clock.  Fixed frequency design based on bin sorted chips can't do this.

This makes it hard to win spec sheet wars

[NotFuzzyWarm]

Glad to see someone from Canaan here. Welcome StevenMosher!

Query: Is the error control process and speed throttling why the miners report hash rate to monitor programs like we see in post #345? (saving eyeball room by not reposting pic so soon). The plot is 1hr long.

I talked to the author of Awesome Miner and he looked over what the api reports and said that graph is accurate and is what cgminer reports when queried by Awesome every 15 seconds. Don't get me wrong - the graph may not be pretty but the total average hash rate seen by kano is correct so this is not a problem, just a puzzle. I just wonder what the Canaan does to make their GUI graph so even.

Love your miners and so far have 3x 721's and 2x 741's

[StevenMosher]

Glad to see someone from Canaan here. Welcome StevenMosher!

Query: Is the error control process and speed throttling why the miners report hash rate to monitor programs like we see in post #345? (saving eyeball room by not reposting pic so soon). The plot is 1hr long.

I talked to the author of Awesome Miner and he looked over what the api reports and said that graph is accurate and is what cgminer reports when queried by Awesome every 15 seconds. Don't get me wrong - the graph may not be pretty but the total average hash rate seen by kano is correct so this is not a problem, just a puzzle. I just wonder what the Canaan does to make their GUI graph so even.

Love your miners and so far have 3x 721's and 2x 741's

 Ah it's good to be back in the ASIC world  (old guys might remember me from 3D graphics)  The chips function pretty much as I described. As for the GUI, it depends what variable you are looking at-- I will get a crash course later this week,  but in general the values on the real time graphs would be  moving averages with different time windows ( as I understand it ). Consequently, they would be smooth. 

[PeaMine]

Wanted to mentioned I updated the MM on the 741 and noticed no more self reboots so far which is good from the miners and more stable hash rate.
What all was changed, or was it just from the power cycling perhaps?
Hardware errors seem to be up a bit, but getting 7.8TH after invalids poolside with volt 0, so I'm not really going to complain haha.
I also took the 721 off of the controller, though I didn't have anything chained to it.  They get 6.5 TH Poolside +1 volt.

[in2tactics]

Glad to see someone from Canaan here. Welcome StevenMosher!

Query: Is the error control process and speed throttling why the miners report hash rate to monitor programs like we see in post #345? (saving eyeball room by not reposting pic so soon). The plot is 1hr long.

I talked to the author of Awesome Miner and he looked over what the api reports and said that graph is accurate and is what cgminer reports when queried by Awesome every 15 seconds. Don't get me wrong - the graph may not be pretty but the total average hash rate seen by kano is correct so this is not a problem, just a puzzle. I just wonder what the Canaan does to make their GUI graph so even.

Love your miners and so far have 3x 721's and 2x 741's

 Ah it's good to be back in the ASIC world  (old guys might remember me from 3D graphics)  The chips function pretty much as I described. As for the GUI, it depends what variable you are looking at-- I will get a crash course later this week,  but in general the values on the real time graphs would be  moving averages with different time windows ( as I understand it ). Consequently, they would be smooth. 

The "Creative Labs" Steven Mosher?

[StevenMosher]

Glad to see someone from Canaan here. Welcome StevenMosher!

Query: Is the error control process and speed throttling why the miners report hash rate to monitor programs like we see in post #345? (saving eyeball room by not reposting pic so soon). The plot is 1hr long.

I talked to the author of Awesome Miner and he looked over what the api reports and said that graph is accurate and is what cgminer reports when queried by Awesome every 15 seconds. Don't get me wrong - the graph may not be pretty but the total average hash rate seen by kano is correct so this is not a problem, just a puzzle. I just wonder what the Canaan does to make their GUI graph so even.

Love your miners and so far have 3x 721's and 2x 741's

 Ah it's good to be back in the ASIC world  (old guys might remember me from 3D graphics)  The chips function pretty much as I described. As for the GUI, it depends what variable you are looking at-- I will get a crash course later this week,  but in general the values on the real time graphs would be  moving averages with different time windows ( as I understand it ). Consequently, they would be smooth. 

The "Creative Labs" Steven Mosher?

 Yup. After Creative I went to do Linux cell phone,  then spent a bunch of years doing Open source  ( R stats) packages. back to hardware.

[philipma1957]

I still have some speakers from creative labs  small world.


back to mining

I have 3 of the avalon 721's nice gear

[in2tactics]

Glad to see someone from Canaan here. Welcome StevenMosher!

Query: Is the error control process and speed throttling why the miners report hash rate to monitor programs like we see in post #345? (saving eyeball room by not reposting pic so soon). The plot is 1hr long.

I talked to the author of Awesome Miner and he looked over what the api reports and said that graph is accurate and is what cgminer reports when queried by Awesome every 15 seconds. Don't get me wrong - the graph may not be pretty but the total average hash rate seen by kano is correct so this is not a problem, just a puzzle. I just wonder what the Canaan does to make their GUI graph so even.

Love your miners and so far have 3x 721's and 2x 741's

 Ah it's good to be back in the ASIC world  (old guys might remember me from 3D graphics)  The chips function pretty much as I described. As for the GUI, it depends what variable you are looking at-- I will get a crash course later this week,  but in general the values on the real time graphs would be  moving averages with different time windows ( as I understand it ). Consequently, they would be smooth. 

The "Creative Labs" Steven Mosher?

 Yup. After Creative I went to do Linux cell phone,  then spent a bunch of years doing Open source  ( R stats) packages. back to hardware.

Well, I think it is pretty awesome to see you around again. I was always a fan of Creative Labs products and extensively used their MP3 players alongside their other products.

Anyway, it would be really cool to see Canaan Creative build and sell another Avalon based usb miner. I would love to see how well the A3212 performs in a device of that type.

[NotFuzzyWarm]

Arrg. The 741 I have at work just got twitchy and has dropped to 6.4THs for the past hour vs its normal 8.2ish. So, I'll update the firmware tomorrow and see if takes care of this.

Any reason to NOT try using the even more recent 20170310 release vs the Dec 29 one?

[philipma1957]

Glad to see someone from Canaan here. Welcome StevenMosher!

Query: Is the error control process and speed throttling why the miners report hash rate to monitor programs like we see in post #345? (saving eyeball room by not reposting pic so soon). The plot is 1hr long.

I talked to the author of Awesome Miner and he looked over what the api reports and said that graph is accurate and is what cgminer reports when queried by Awesome every 15 seconds. Don't get me wrong - the graph may not be pretty but the total average hash rate seen by kano is correct so this is not a problem, just a puzzle. I just wonder what the Canaan does to make their GUI graph so even.

Love your miners and so far have 3x 721's and 2x 741's

 Ah it's good to be back in the ASIC world  (old guys might remember me from 3D graphics)  The chips function pretty much as I described. As for the GUI, it depends what variable you are looking at-- I will get a crash course later this week,  but in general the values on the real time graphs would be  moving averages with different time windows ( as I understand it ). Consequently, they would be smooth. 

The "Creative Labs" Steven Mosher?

 Yup. After Creative I went to do Linux cell phone,  then spent a bunch of years doing Open source  ( R stats) packages. back to hardware.

Well, I think it is pretty awesome to see you around again. I was always a fan of Creative Labs products and extensively used their MP3 players alongside their other products.

Anyway, it would be really cool to see Canaan Creative build and sell another Avalon based usb miner. I would love to see how well the A3212 performs in a device of that type.

yeah it would work nicely on my stud hub  A power usb stick miner has many people looking to see one for sale

[Naokia980]

It is time to back mining again. Now i need to find my old equipment