Avalon ASIC users thread

[silverston]

Just watch the freq carefully when using --avalon-auto. New firmware is much more aggressive in overclocking.

After 1.5 hrs on a B3 4M unit with stock PSU, the HW error rate started climbing rapidly while running @360 MHz (6%+ error rate over the runtime, up from 1.7% minutes earlier). Clock speed was not ramping to down to compensate.

Now using --avalon-freq 345-355 to see if that prevents a runaway.

And before anyone says it, yes that unit is getting a PSU swap later this weekend.

we'll see. my power supply already replaced.

[mdbssm]

Just watch the freq carefully when using --avalon-auto. New firmware is much more aggressive in overclocking.

After 1.5 hrs on a B3 4M unit with stock PSU, the HW error rate started climbing rapidly while running @360 MHz (6%+ error rate over the runtime, up from 1.7% minutes earlier). Clock speed was not ramping to down to compensate.

Now using --avalon-freq 345-355 to see if that prevents a runaway.

And before anyone says it, yes that unit is getting a PSU swap later this weekend.

we'll see. my power supply already replaced.

My other B3 unit with an external 1200w PSU is behaving fine.

[tarui]

anyone tried replacing the thermal paste between the chips and the heatsink?

or adding thermal paste between the heatsink and the casing (where the heatsinks are mounted onto the case) ?

I am 

So what did u do? Replace thermal paste or put thermal paste between hs and case?
What kind of thermal paste did use? Non conductive?

[tarui]

Just watch the freq carefully when using --avalon-auto. New firmware is much more aggressive in overclocking.

After 1.5 hrs on a B3 4M unit with stock PSU, the HW error rate started climbing rapidly while running @360 MHz (6%+ error rate over the runtime, up from 1.7% minutes earlier). Clock speed was not ramping to down to compensate.

Now using --avalon-freq 345-355 to see if that prevents a runaway.

And before anyone says it, yes that unit is getting a PSU swap later this weekend.

What's your hashrate?

[silverston]

anyone tried replacing the thermal paste between the chips and the heatsink?

or adding thermal paste between the heatsink and the casing (where the heatsinks are mounted onto the case) ?

I am 

So what did u do? Replace thermal paste or put thermal paste between hs and case?
What kind of thermal paste did use? Non conductive?

replaced the thermal pad between the board and the heatsink on the thermal paste. used Arctic Ceramic and MX-2

[silverston]

What's your hashrate?

after 35 min is 86Gh 

[mdbssm]

What's your hashrate?

Batch3 4module 1200wPSU, 110.8 GH/s, 352-353 MHz, 0.3% reject rate, 1.75% HW errors
Batch3 4module stock PSU, 111.6 GH/s, 355 MHz, 0.2% reject rate, 0.94% HW errors

The new FW fixed a high reject rate problem on the first unit. Was running 2-3% rejects on the 0703 firmware.

[CoinHoarder]

anyone tried replacing the thermal paste between the chips and the heatsink?

or adding thermal paste between the heatsink and the casing (where the heatsinks are mounted onto the case) ?

I am 

There are a couple guys I've seen do this. Not sure of a link tho, you could prob. use search. It did improve chip temperatures.

[silverston]

anyone tried replacing the thermal paste between the chips and the heatsink?

or adding thermal paste between the heatsink and the casing (where the heatsinks are mounted onto the case) ?

I am 

There are a couple guys I've seen do this. Not sure of a link tho, you could prob. use search. It did improve chip temperatures.

it did improve more stable overclocking

[tarui]

I'm a bit concerned with the thermal paste used. The manufacturer seems to use so much of it that it overflows thru the holes and on to the circuit board itself and contacting the chip and other components.

Unless they use non conductive thermal paste.

[ProfMac]

Changed options 1 hour ago:

--avalon-options 115200:24:10:d:355 --avalon-auto --avalon-freq 256-375 --avalon-temp 50

My match_work_count shows more spread:
min: 3001
max: 3288
0.5 * (3288-3001) / (3288+3001) --> 0.002_28

When the uptime is several days, these numbers become very close.

My freq dropped to 351 MHz.
The orange light is still blinking.
Temperatures are now 22, 46, 47°C
Fans are 2760, 3720, 3720 and seem to vary a lot, and the temperature wanders around also.

   [timeout] => 35


I show 3387 accepted, 4 rejects, 83372 diff1shares at 1h 11m 18s

[tarui]

What's your hashrate?

Batch3 4module 1200wPSU, 110.8 GH/s, 352-353 MHz, 0.3% reject rate, 1.75% HW errors
Batch3 4module stock PSU, 111.6 GH/s, 355 MHz, 0.2% reject rate, 0.94% HW errors

The new FW fixed a high reject rate problem on the first unit. Was running 2-3% rejects on the 0703 firmware.

Mine looks worst. Batch 1 4 module

Hashrate swings wildly so it's hard to judge.

1250w psu and around 0.56% rejected.
Cant remember how to read the hw error rate though.


The fact that I'm in a tropical country doesn't help much with thr temps

[mdbssm]

Cant remember how to read the hw error rate though.

HW error rate in % = HW / (diff1shares + HW) x 100

Or better yet, apply the patch:
https://bitcointalk.org/index.php?topic=254331.0

[whizter]

I'm getting very good results so far with the new 20130818 firmware. Orange LED is blinking but I don't care.

Today I discovered a potential problem with my Fractal Newton 1000W PSU, the PSU Fan doesn't work at all, it just stands still. I don't know how the Fan Control is implemented in this PSU, but I can't imagine the Fan not running at all @ 680W load. I already contacted Fractal about this, but I don't expect an answer before monday. So I'll just let it run like this and see what happens. Cooling should be enough with those powerful Avalon fans.

[kano]

Cant remember how to read the hw error rate though.

HW error rate in % = HW / (diff1shares + HW) x 100

Or better yet, apply the patch:
https://bitcointalk.org/index.php?topic=254331.0

Except that the patch is wrong
Use current cgminer and it reports the % figures directly, and correctly, in the API

... and in older versions of cgminer, through the API, they are called: "Hardware Errors" and "Diff1 Work" to calculate them

[zoro]

excellent work ckolivas with new FW 130818. after 4 hours, utility is increased due to more accepted shares!

[mdbssm]

Cant remember how to read the hw error rate though.

HW error rate in % = HW / (diff1shares + HW) x 100

Or better yet, apply the patch:
https://bitcointalk.org/index.php?topic=254331.0

Except that the patch is wrong
Use current cgminer and it reports the % figures directly, and correctly, in the API

... but not in the cgminer status page of the avalon itself, and not in the miner.php generate page.

The 0817 avalon firmware reports that it is using cgminer 3.3.4.

[kano]

Cant remember how to read the hw error rate though.

HW error rate in % = HW / (diff1shares + HW) x 100

Or better yet, apply the patch:
https://bitcointalk.org/index.php?topic=254331.0

Except that the patch is wrong
Use current cgminer and it reports the % figures directly, and correctly, in the API

... but not in the cgminer status page of the avalon itself, and not in the miner.php generate page.

The 0817 avalon firmware reports that it is using cgminer 3.3.4.

Well it's in 3.3.4 in the master git - I put it there before it was released
I don't have an avalon - so I can only guess someone missed something or the avalon does something different to my miner.php ...

[-ck]

The avalon web interface does not report the API devs command back, so you don't see the hw error% reported anywhere unless you're logged into the avalon some other way, using the devs command. The calculated HW error percentage is wrong in that lua modification by BenTuras as kano said.

[GenTarkin]

Yeah the 0818 firmware is treating me well so far. I can now hold 360mhz @ less than 2% HW Errors. This was not possible before, 360mhz would be 3+ % HW error rate on previous firmware.
So my MHSav is constantly 84GH+ ... Im wondering what exactly was changed?

Also, the results are during the hotter part during the day, so at night I imagine it will be even better.