Avalon ASIC users thread

[dogie]

I cracked it! Hitting 120GH on a 3 module now

[1714821922]

1714821922

[1714821922]

1714821922

[A Meteorite]

I cracked it! Hitting 120GH on a 3 module now

Surely you're pulling our leg. 450 is the max clock rate which puts a 3 module somewhere around 100GH. If not, do spill your secret.

[kano]

...
Thanks for the answer. I guess, it comes down to MHSAV right? Doesnt that take everything into account and reflect pretty much what is valid and going to the pool?

I'd actually suggested DiffAccepted, since I'm not sure what gets counted in MHSav on the avalon.

To determine your Accepted MH/s it's:
   DiffAccepted * (2^32) / Elapsed
That shows the MH/s the pool is paying you so ignores anything that doesn't count.

[dogie]

I cracked it! Hitting 120GH on a 3 module now

Surely you're pulling our leg. 450 is the max clock rate which puts a 3 module somewhere around 100GH. If not, do spill your secret.

Spent all of yesterday going through the firmware code, there is a trick people are missing. Uses some information from the github as well.

[kano]

I cracked it! Hitting 120GH on a 3 module now

http://img.ctrlv.in/img/51d9f1d64f595.png

Surely you're pulling our leg. 450 is the max clock rate which puts a 3 module somewhere around 100GH. If not, do spill your secret.

Spent all of yesterday going through the firmware code, there is a trick people are missing. Uses some information from the github as well.

Lulz - a few things:
1) The pool numbers are not accurate
2) It's easy to get them high when the share difficulty is high - just look at the pool around the time you find a share ...
3) Being the one who wrote all the Icarus timing code I know quite well how to get icarus to fake the numbers ...

[SolarSilver]

Spent all of yesterday going through the firmware code, there is a trick people are missing. Uses some information from the github as well.

Care to post of *what* you modified? Not just a printf I hope...

And can you show a screenshot of the cgminer status page so we can judge from the numbers?

[Sitarow]

I cracked it! Hitting 120GH on a 3 module now

Surely you're pulling our leg. 450 is the max clock rate which puts a 3 module somewhere around 100GH. If not, do spill your secret.

Spent all of yesterday going through the firmware code, there is a trick people are missing. Uses some information from the github as well.

Post something like this please

Accepted Shares after 5 days.
8257580 / 7209.39 =1145.39 accepted shares a min or 68723.539 accepted shares per hour.

[silverston]

hello.
 looking for a replacement power supply.please advise a good unit for stable overclocking and stable work

[SolarSilver]

hello.
 looking for a replacement power supply.please advise a good unit for stable overclocking and stable work

I used a SilverStone Strider Gold 1200W

http://www.silverstonetek.com/product.php?pid=284

it fits right into the Batch #2 machines, it does not fit into the Batch #1 machines

In the Batch #1 I used an OCZ ZX 1250 as it's shorter

http://ocz.com/consumer/psu/zx-series-850w-1250w-power-supply

I basically used them because I cannibalize them from my old GPU systems, for future repairs I'm also looking for a Platinum 1200 power supply, that fits into the case

Any suggestions?

[elasticband]

I cracked it! Hitting 120GH on a 3 module now

I hope you upgraded the PSU before doing this on units you plan on selling

[demkd]

silverston
I can recommend Corsair AX860i platinum. Not expensive.
But If you want best available PSU than get Seasonic 860 platinum.
If you remove stock PSU from the case (to prevent overheating) than any 750W PSU is good. (at 354 clock 3 module avalon drain only 630w from PSU (735-740w from wall for gold rated PSU))

[silverston]

silverston
I can recommend Corsair AX860i platinum. Not expensive.
But If you want best available PSU than get Seasonic 860 platinum.
If you remove stock PSU from the case (to prevent overheating) than any 750W PSU is good. (at 354 clock 3 module avalon drain only 630w from PSU (735-740w from wall for gold rated PSU))

So if I have inside  750W? there is no reason to change? a model is  unknown ... something in Chinese written on it.
 is stable at 300MHz, 325 goes to idle mode and orange light blinking. the temperature is not more than 48 degrees. it is the power supply or something else? I connected directly to PC because my TP-link internal router not working

[elasticband]

just kill cgmiiner in the process tab,(if you see anything about max children, kill it also) kill it 2 times and then goto startup and then RESTART cgminer

[-ck]

just checked my freq after 15h of continuous uptime.  

i've got settings of --avalon -auto --avalon -temp 40 and i have 325 set in default box above.  according to con the latter shouldn't matter.

problem is that hashrate is averaging only about 70 Mh/sec.  i thought it was supposed to go as high as 80-85?

edit:  and freq are btwn 285-305

It's going to really struggle to keep it at 40 degrees. That's probably why.

well, if that's the case then i'm going to try this:  --avalon-auto --avalon-freq 350-375 --avalon-temp 40

i don't care about killing the fans.  they can be replaced.  i want to get the freq and resulting HR up (w/o killing the chips of course  )

If you don't care about killing the fans then what you want is to set the fanspeed high, not set the target temperature low because if it can't keep the temperature low it will stop the avalon from clocking higher. Remove the avalon-temp argument and add --avalon-fan 80 to keep it at 80%. I don't think there's much advantage to going above 80% fanspeed due to it reversing the flow of the third fan.

[-ck]

Just as an aside, I am currently touring through Europe for the next 6 weeks.

Care to elaborate which countries?

Just Greece from here on Net access is not so great so far where I am...

[loshia]

Just as an aside, I am currently touring through Europe for the next 6 weeks.

Care to elaborate which countries?

Just Greece from here on Net access is not so great so far where I am...

Have a nice holiday CON!

Something to share:
My batch 1 unit makes 83+GH steady autoclock reaching 355 Mhz stable err <2%
My batch 2 unit makes 80+ steady autoclock reaching 343  Mhz stable  err <2%

I have tried batch 2 unit clackoed at 350 auto-clock disabled - same performance  80+ with err about 5%

I am sure it is due to HW itslef. I am wondering what may be different. I will play with FPGA USB plugs(jumpers) when i have time

Can anyone share batch 2 results - auto clock ?

 

[goxed]

Batch 2 autoclock results:

ambient 80 - 85 F, 340 - 344MHz
ambient 75 - 80F, 348 - 353MHz

[goxed]

I cracked it! Hitting 120GH on a 3 module now

Surely you're pulling our leg. 450 is the max clock rate which puts a 3 module somewhere around 100GH. If not, do spill your secret.

Spent all of yesterday going through the firmware code, there is a trick people are missing. Uses some information from the github as well.

I rechecked the schematics, I doubt it's got something to do with voltage mod, since the voltage regulator is a fixed one.
 That leaves coding trickery somewhere, unless I have grossly misunderstood something somewhere.

[Elokane]

There was a bug on BTCGuild yesterday. My 3 units also showed 350GH/S at the peak.
I'd post the picture, but I'm lazy.

[demkd]

So if I have inside  750W? there is no reason to change? a model is  unknown ... something in Chinese written on it.
 is stable at 300MHz, 325 goes to idle mode and orange light blinking. the temperature is not more than 48 degrees. it is the power supply or something else? I connected directly to PC because my TP-link internal router not working

Move out PSU from case.
I run my avalon with external PSU at 336 with 750 stock PSU (Enermax ENM750EWT-S RG) and it never restarts.
Tested @354 but noise level unacceptable, 750w PSU easily handle moderately overclocked units.
if problem still occurs than check you "cgminer api log"
It must be something like that (my unit @336 2 days w/o restart):
   [no_matching_work] => 0
   [match_work_count1] => 20069
   [match_work_count2] => 20320
   [match_work_count3] => 20197
   [match_work_count4] => 19944
   [match_work_count5] => 20103
   [match_work_count6] => 19962
   [match_work_count7] => 19976
   [match_work_count8] => 20163
   [match_work_count9] => 20064
   [match_work_count10] => 19961
   [match_work_count11] => 20056
   [match_work_count12] => 20235
   [match_work_count13] => 20173
   [match_work_count14] => 20481
   [match_work_count15] => 20215
   [match_work_count16] => 20042
   [match_work_count17] => 20044
   [match_work_count18] => 19992
   [match_work_count19] => 20040
   [match_work_count20] => 20398
   [match_work_count21] => 19627
   [match_work_count22] => 19978
   [match_work_count23] => 20354
   [match_work_count24] => 20186
All counters have almost same values. (~5% diffs) if not you have defective modules or some module overheated (usually second module w/o temp sensor ).

And you can try to reset your router, may be it's not dead.
https://bitcointalk.org/index.php?topic=140539.msg2672023#msg2672023