Avalon ASIC users thread - page 66. | Bitcointalksearch.org

GandalfG

sr. member

Activity: 259

Merit: 250

Dig your freedom

Quote from: AdamKD on October 17, 2013, 04:52:39 PM

Quote from: GandalfG on October 17, 2013, 04:07:00 PM

Quote from: AdamKD on October 17, 2013, 09:41:27 AM

If you have issues with hardware error rate add some solder around the buffers. That fixes the HW error rate which is *usually* related to the buffers on the individual hash boards.

There isn't buffers in original Avalon design. May you mean 74LVC2G34 on the signaling bus ? (Config_P/N ReportP/N)
But I never see bad solder on this parts.

Those. There's 3 of them (U13, U14, and U15) per hash board. My friend and I've seen hundreds (if not thousands) of them under a solder rework microscope. They all looked good from almost every angle. Touching them up all of them usually fixed a HW error problem (Once we narrowed it down to the right hash board). I'm not going to guarantee that it will fix a HW error problem. I'm just going to state that there's a good probability that it will fix a HW error problem.

Thx for advice. I reflow all my boards and have 1.6% HW at 430MHz.
But observe interesting thing, HW percentage are depends from length and quality ribbon cable. Maybe in conjunction with solder quality on buffers ?

ste0024

sr. member

Activity: 462

Merit: 250

Quote from: natbyte on October 17, 2013, 03:50:50 PM

Buffers? Care to share a picture of where you mean, thanks.

+100

AdamKD

member

Activity: 98

Merit: 10

Quote from: GandalfG on October 17, 2013, 04:07:00 PM

Quote from: AdamKD on October 17, 2013, 09:41:27 AM

If you have issues with hardware error rate add some solder around the buffers. That fixes the HW error rate which is *usually* related to the buffers on the individual hash boards.

There isn't buffers in original Avalon design. May you mean 74LVC2G34 on the signaling bus ? (Config_P/N ReportP/N)
But I never see bad solder on this parts.

Those. There's 3 of them (U13, U14, and U15) per hash board. My friend and I've seen hundreds (if not thousands) of them under a solder rework microscope. They all looked good from almost every angle. Touching them up all of them usually fixed a HW error problem (Once we narrowed it down to the right hash board). I'm not going to guarantee that it will fix a HW error problem. I'm just going to state that there's a good probability that it will fix a HW error problem.

GandalfG

sr. member

Activity: 259

Merit: 250

Dig your freedom

Quote from: AdamKD on October 17, 2013, 09:41:27 AM

If you have issues with hardware error rate add some solder around the buffers. That fixes the HW error rate which is *usually* related to the buffers on the individual hash boards.

There isn't buffers in original Avalon design. May you mean 74LVC2G34 on the signaling bus ? (Config_P/N ReportP/N)
But I never see bad solder on this parts.

natbyte

member

Activity: 70

Merit: 10

Buffers? Care to share a picture of where you mean, thanks.

AdamKD

member

Activity: 98

Merit: 10

If you have issues with hardware error rate add some solder around the buffers. That fixes the HW error rate which is *usually* related to the buffers on the individual hash boards.

driksson

sr. member

Activity: 294

Merit: 250

Quote from: smoothrunnings on October 17, 2013, 07:41:14 AM

Quote from: natbyte on October 17, 2013, 04:09:13 AM

Nope sorry for the confusion, I was just giving some details on my hardware setting, each Avalon is slightly different, I have also replaced the TIM with paste on each module to allow it to run cooler allowing me to run at 360 Mhz stable. From my testing the hashing indeed remains constant, however H/W increase over time e.g. over 48 hours without restart, after running for long periods I see 6 - 8% which is just wasted electricity, so a quick cgminer restart solves all this

My advice for your setup is to have cgminer restart every x amount of hours keeping H/W errors low, use whatever Mhz setting keeps your units hashing high as possible and stable. Hope this makes sense.

Sorry for budding in here, but are you saying the HW on the status page is the errors? And when you mention keeping it low, what's the magic number?

Thanks

what i realised, was that my units was running (b2 3u) at 83-85gh for 3-4 days on eligius. then all of a sudden, they drop to 75-70gh.. for no real reason, hashrate seems same on machines, but hw rate increases. Basically, if you see 83gh on eligius or in your pool always, you don't have our problem. if you see it dropping, add reset script line.

smoothrunnings

hero member

Activity: 630

Merit: 501

Quote from: natbyte on October 17, 2013, 04:09:13 AM

Nope sorry for the confusion, I was just giving some details on my hardware setting, each Avalon is slightly different, I have also replaced the TIM with paste on each module to allow it to run cooler allowing me to run at 360 Mhz stable. From my testing the hashing indeed remains constant, however H/W increase over time e.g. over 48 hours without restart, after running for long periods I see 6 - 8% which is just wasted electricity, so a quick cgminer restart solves all this

My advice for your setup is to have cgminer restart every x amount of hours keeping H/W errors low, use whatever Mhz setting keeps your units hashing high as possible and stable. Hope this makes sense.

Sorry for budding in here, but are you saying the HW on the status page is the errors? And when you mention keeping it low, what's the magic number?

Thanks

driksson

sr. member

Activity: 294

Merit: 250

Quote from: natbyte on October 17, 2013, 04:09:13 AM

Nope sorry for the confusion, I was just giving some details on my hardware setting, each Avalon is slightly different, I have also replaced the TIM with paste on each module to allow it to run cooler allowing me to run at 360 Mhz stable. From my testing the hashing indeed remains constant, however H/W increase over time e.g. over 48 hours without restart, after running for long periods I see 6 - 8% which is just wasted electricity, so a quick cgminer restart solves all this

My advice for your setup is to have cgminer restart every x amount of hours keeping H/W errors low, use whatever Mhz setting keeps your units hashing high as possible and stable. Hope this makes sense.

Thanks yes i understood that. super. I mean, the line you gave, maybe ill change it to every 20 hours, as it looks quite stable and high for even 70 hours.. but keeping a restart on a daily basis should be fine.
As for 360, maybe i will try 350, since they are in 20dgr ac room, but no hardware mods at all.

natbyte

member

Activity: 70

Merit: 10

Nope sorry for the confusion, I was just giving some details on my hardware setting, each Avalon is slightly different, I have also replaced the TIM with paste on each module to allow it to run cooler allowing me to run at 360 Mhz stable. From my testing the hashing indeed remains constant, however H/W increase over time e.g. over 48 hours without restart, after running for long periods I see 6 - 8% which is just wasted electricity, so a quick cgminer restart solves all this

My advice for your setup is to have cgminer restart every x amount of hours keeping H/W errors low, use whatever Mhz setting keeps your units hashing high as possible and stable. Hope this makes sense.

driksson

sr. member

Activity: 294

Merit: 250

Quote from: natbyte on October 17, 2013, 02:36:51 AM

I have tried the latest official and reverted back to the last CK firmware version as it was more stable for me, I run at 360 Mhz with H/W at about 1.4%

However after running cgminer for days the H/W error rate would always creep up, so now I just run a cgminer restart in scheduled task, below will restart cgminer every 8 hours. Keeps my H/W rate low.

0 */8 * * * /etc/init.d/cgminer restart

so you are suggesting to set fixed 360 instead of 340? and add the restart each 8 hours?
I assume that its H/W that goes up for me aswell, as hashingpower seems to remain, and all boards are hashing at next to identical numbers.

natbyte

member

Activity: 70

Merit: 10

I have tried the latest official and reverted back to the last CK firmware version as it was more stable for me, I run at 360 Mhz with H/W at about 1.4%

However after running cgminer for days the H/W error rate would always creep up, so now I just run a cgminer restart in scheduled task, below will restart cgminer every 8 hours. Keeps my H/W rate low.

0 */8 * * * /etc/init.d/cgminer restart

driksson

sr. member

Activity: 294

Merit: 250

Hey, i realised that my units are much stabler, on -avalon-auto and chip freq 340 fixed. (without --avalon-freq 340-375, if i had this on, it defaulted on much less than 340 after a few days).

however, now it went down to a default again to around 330 after 4 solid days.
Anything i can do? or must i put a cron job to restart every 24 hours to keep it solid?
(i'm running latest ckolivas, have not tried the latest official yet)

smoothrunnings

hero member

Activity: 630

Merit: 501

Does anyone know where I can get an extra thermal sensor for my B1? I noticed in the B2 there are two thermal sensors connected to module one and module 3 (of course there is on on one of the boards making it 3 total), this might explain why the B1 shows -1 one Temp2 because it only comes with one, well two technically thermal sensors, one connected to module 1 and the other on one of the boards. I would like to know if anyone has attempted to get another thermal sensor for their B1?

Thanks,

smoothrunnings

hero member

Activity: 630

Merit: 501

Can the WiFi be setup so that my Avalon connects to my AccessPoint instead of hosting its own? If so how do I set it up to do this and use WPA2 encryption?

I had a hook and the options aren't really straight forward.

Thanks

Update:

I got it to connect but I can't access the avalon from http once I remove the ethernet cord from the machine over wifi. Sad

Tigggger

legendary

Activity: 1098

Merit: 1000

Quote from: smoothrunnings on October 16, 2013, 03:14:04 PM

Thanks for the info.

Do you know if the units run cooler with their lids off than having it on? I have a 18" fan blowing air at my B1 and B2 units.

No problem.

They are designed to stay closed, and I'd leave it that way. Nzhang (sp?) said a lot was done on the airflow and pressures.

PS Solved my remote problem, I had set it up correctly the first time the problem was with chrome, when I looked at the error it was "ERR_UNSAFE_PORT", quick google search showed that adding --explicitly-allowed-ports=xxx to the end of it's shortcut was all that was required to get it going again

GandalfG

sr. member

Activity: 259

Merit: 250

Dig your freedom

Quote from: smoothrunnings on October 16, 2013, 03:14:04 PM

Quote from: Tigggger on October 16, 2013, 03:03:23 PM

Quote from: smoothrunnings on October 16, 2013, 02:59:11 PM

what should I set the --auto-cutoff on my 4 module?

I had it set to 60 when I had 3 modules.

Batch #1 & #2 (50 temp, 70 cutoff)
Batch #3 (70 temp, 90 cutoff)

Thanks for the info.

Do you know if the units run cooler with their lids off than having it on? I have a 18" fan blowing air at my B1 and B2 units.

I make many test and most important thing are inlet air temp. 18 or lower.
Next. Best result I get when lid cover 75% case. About 10cm from rear. Its help for better exhaust hot air.

smoothrunnings

hero member

Activity: 630

Merit: 501

Quote from: Tigggger on October 16, 2013, 03:03:23 PM

Quote from: smoothrunnings on October 16, 2013, 02:59:11 PM

what should I set the --auto-cutoff on my 4 module?

I had it set to 60 when I had 3 modules.

Batch #1 & #2 (50 temp, 70 cutoff)
Batch #3 (70 temp, 90 cutoff)

Thanks for the info.

Do you know if the units run cooler with their lids off than having it on? I have a 18" fan blowing air at my B1 and B2 units.

Tigggger

legendary

Activity: 1098

Merit: 1000

Quote from: smoothrunnings on October 16, 2013, 02:59:11 PM

what should I set the --auto-cutoff on my 4 module?

I had it set to 60 when I had 3 modules.

Batch #1 & #2 (50 temp, 70 cutoff)
Batch #3 (70 temp, 90 cutoff)

smoothrunnings

hero member

Activity: 630

Merit: 501

Quote from: GandalfG on October 16, 2013, 02:10:10 PM

Remains of thermal paste is not a problem. I doubt Bitsyncom use conductive in this case.
BTW. How much You pay for module ?

I bought them from Yifu, 3.8BTC each. Shipping was good, got them in 3 business days from China.

Topic: Avalon ASIC users thread - page 66. (Read 438582 times)