Bitcoin Forum

the default is 282, and hashrate is  60G.

Does anyone tried to set it to 300?

yes. i've been running that way for a while now. i think xiangfu said it should not be a problem.


would be nice to know if we can go past 300 though.

This is not a video card or CPU you can replace by overnighting one from Newegg if you fuck it up.

Good point. I was curious about running it at 300 (not a huge "overclock" over 282) but I don't want to chance it. I'd also think that power consumption will naturally go up and that poor Antec 650 PSU is pretty much almost overloaded. If I had watercooling blocks for the Avalon units and a beefier PSU then I'd feel more comfortable upping the clock rate.

What hash rate are you getting at 300MHz?

Another thing is I have one avalon hashrate at 65G, another only 55G.

Both 282 mhz.

Someone else mentioned on a different thread (can't remember which one) that if you set the clock to 300 the hashrate bounces all over the place from 60 to 80 GH/s. Can anyone else confirm this wild fluctuation in hashing rate and what would be the average hashrate for an Avalon clocked at 300?

I believe if you can provide a efficient cooling. overclock at this level is acceptable.

I've been running at 300MHZ for a couple of days (in an air conditioned room). It seems to give around 71GH.

I tried mine at 300 for about 20 minutes.  The hash rate and the shares submitted didn't change.  So I put it back to 282.  My unit was in a cold room, 0 C.

The thing is running so well that I am loath to mess with it.  I have a better spot to put it but I don't want to turn it off.

Did you restart the cgminer process?  I'm definitely getting more hashes at the pool end.

It was only a 20 minute attempt.  Yes I did restart cgminer.  The thing is running so well that I am loath to mess with it.  I have a better spot to put it but I don't want to turn it off.

How about your hashrate at 282Mhz?

it's very strange 1 of them is only 55, while another is 65.

https://i.imgur.com/dm7cImp.png

and, and ,how about the different of utility?

282 --> 65  (590W, 220v)
300 --> 71 (640W, 220v)
i am running it stable at 300 without problems although it is a bit loudy :)
i am not sure about PSU, i haven't opened it,  is it really 650W?  :o

thanks.
I just tried with 1 avalon, it's 66G -> 72G .
looks stable. the temp is 42 -> 44
and fan 1260 -> 1960

another one unlucky.  55G -> 56G only.

lol I'm in the same situation. I think a good investment would be to put it on a UPS. I got my eyes on a 1500VA unit - should be more than enough juice to keep it mining during brief power outages.

Hmm extra 50 watts for 300. I guess that will for sure overwhelm the Antec 650 watt PSU if one runs at 120v. I don't think I will try this until I slap on a 850 watt or 7500 watt Gold or Platinum rated PSU.

Are you running the latest firmware? And this happens no matter what clock rate the chips are running at? What do the temps say? Are the fans spinning at above 1000 rpm and can you feel the air flowing?

both the latest firmware of 0325. it seemd over clock doesn't help.
the temp3 are both 44. ( I have turn on air condition and set to 21 degree room temp.)

fans are 2200 rpm. air pressure are good. and I have additional cooling fans outside.

Everything are the same of both machine :  same pool, same firmware, in the same room.

Hmm, odd. Everything seems ok. One thing tthat _might_ help is to make sure you do not have additional cooling fans outside the case blowing in. Leave the cooling setup as it came. This was a recommendation by ngzhang and I believe it is also mentioned in the official wiki.

You could also try opening the thing up and disconnecting and make sure everything is connected solidly and reconnect it if not.

Looks like some chips are not working. an Avalon has 240 chips so thats 250mhz for each chip.
So i would say 20 chips are not working, weird that its exactly 20, maybe a part of a is not working.
A slot has 8 times 10 chips so maybe 2 sets of 10 not getting power? whats your power usage?

Don't confuse input wattage to output wattage. The PSU is rated at 650 watts output. When the input wattage is 640, the output more like 500-550 watts (depending on efficiency), so that's fine for a decent PSU.

avalon on 120v USA, pull 658W at 300 and 615w at 282

they do not run any hotter.

yes i think you are right.

i need a screw tomorrow ....

but i'm afraid of mess things up :) maybe the second avalon will say good bye to me after tomorrow.


ahh ..Yes!!
I found the PSU are not the same!
maybe I'll change the psu first!!

really?

that's the first i've heard of that. ???

Anyone have confirmed 220/240v power consumption numbers? Difference between 110/120v? I'm trying to quantify the power savings vs 220/240v infrastructure costs for 2 Avalons(7 modules). Perhaps I should start a new thread.

I can test it later.

ok, guys, let me tell you what I did today.

First, I open the case and check every connector. I also find out a TT 850W PSU from my storage room ( aha , that's good to find it) , preparing to replace the origin PSU.

this is what I see when open the case:
https://i.imgur.com/4Ztl0kX.jpg

It's a antec PSU. but there is no label on it so I can't tell the model . I guess the label must be on the other side.
https://i.imgur.com/4ztvar6.jpg

So I tried to check every connector . And yes, I found one loose!
https://i.imgur.com/L6VTZS8.jpg

After replug the loosed line, turn it on, every thing back to normal : 65G at 282Mhz.
Sorry @SolarSilver I didn't pull the PSU out to see the label. I think of put it back to work quickly.

It's hard to put the case cover back than open it.
I have to use some paper to help . like this:
https://i.imgur.com/f5WCLLW.jpg

And, @solarsilver, I didn't open another working avalon. But I did take a picture of that PSU from outside:
https://i.imgur.com/rrRAZd7.jpg

Everything is good. I fixed the 55G avalon to 65G .And I found some other person in this forum also meet the line loose problem.
Hope this thread can be helpful to others.

And now I run 2 of my avalon at 300Mhz, it shows 71G in the gui.

The 2nd PSU looks like a Corsair Gaming Series GS700
700W, 58A Singlerail

300Mhz = safe, but result may vary.

You guys are so cool, you could have say Avalon is 70G ,because it really did. But you only promise 60G. So you give every one 10G extra.

Cheers!

Hmm yes indeed results may vary, but I think play it safe and keep it a 282. 300 may be necessary when competition for hashing power gets really tough, but for now I'm thankful I got 67 GH/s which is something I would have never dreamed of or thought I could achieve back in the GPU mining days.

I think when you overclock it the power usage will go up much more, after all its overclocking.

the trick is to leave the sides with 4 circles still screwed in but loose and slide it to fit closed then use needle nose pliers on the sides with 3 to keep them from falling back in case before you tighten them one by one.

I'm glad to see you found the loose plug, nobody is going to lose sleep over the power supply label ;-)

You can run the systems at 300, they work fine here (because temperature is low due to extended winter in our hemisphere), don't forget they do consume 50 Watts extra if you push them from 282 to 300

is this how it came?  no panel to cover the psu?

I have a new firmware we ( ckolivas,xiangfu ) just built.

newest cgminer 2.10.5a, much more stable. no more restarts.

PM for a download link if you wish to try.

I tried this this morning.  Hash went from 68,000 -> 71,000.  It ran for about 2 hours and then stopped.  I had to call my wife an teach her how to restart the computer over the phone.  We set it back to 282 MHz and it has been running for 5 hours.  In 4 days it has never stopped when set to 282 Mhz.  I think I will leave it there.

thank you for this.

is the cgminer 2.10.5a solely for Avalon use?

I have been running mine at 300MHz for a while now.  I have not noticed any ill effects.  The fans seem to run a little faster.

I believe ngzhang said that running at 300MHz will increase power consumption, and will raise the HW error percent slightly.

it's proportional, fan is dynamic according to temp.

I did some overclocking with GPUs and as far as i read the fear of people that simple overclocking can dramatically reduce the lifetime of a chip isnt correct. The biggest problem for chips is when you set the voltage higher. The electron migration (or something called similar) will make the chip go down faster this way. Thats a fast way to crash a chip.
The second way, but far less dangerous is high temperature. At least GPUs shut down when they are too hot. Asics dont do this so one should be cautious with the temperature. But a higher temperature is shortening the life of a chip but not very much.
The third is to overclock the frequency. It doesnt hurt in its own. The only danger is the high temperature. So if one manages to keep the temperature at the same level while overclocking it shouldnt hurt the chip in way that its noticeable.
If you want to make your chip live long and have better power then overclock and undervolt it. That will give higher speed, lower temperature and the overall live of the chip will be extended.

Im not a pro so if im wrong let me know.

I have a feeling the reason it stopped was because the power supply could not handle it.

Maybe someone can correct me if I'm wrong but I do believe the Avalon ASICS will shutdown if they get too hot. Now I doubt that each chip has a thermal sensor in it but at the very least they might have a thermal sensor somewhere on the ASIC PCBs or somewhere around. It would be pretty silly if they went through all the trouble of designing bitcoin mining ASICS and built no thermal protection in the system whatsoever. After all these are money printing machines, no?

Well the fans adjust there speed so i guess there are sensors on there yeah.

True, but the sensors could be just ambient air temp sensors - meaning they just sense the temp in a given area, and not exactly on or near the ASIC die itself.

I'm surprised they're running the power supplies so close to the rated capacity. I wouldn't feel comfortable running anything less than an 850w one.

I wouldn't be surprised if you got a better hashrate with more power headroom either.

correct, there's an hard-control on shutdown on when it overheats.

so its secure, and it has its own sesnsors in it. I love it

Well I took the plunge and clocked it at 300 a few minutes ago. Hashrate went up to 74G for a bit..average is 71G. Not sure for how long it will remain stable at this clock rate or if the PSU can handle it. If the Antec unit is just 650 running it at 300 mhz is probably running the PSU at 100% load. I really need to get a 850w PSU.

I can confirm this is the case. The Hashrate bounces all over the place, from 54GH/s to 80+.

Average is about 72~74. Consider though that the average is usually also all over the place.

I looked into this last week. Though the 1500VA (by APC) are only rated for 6 minutes operation with a standalone Avalon unit.

To be honest, you'd need more juice than that just to keep it going while the power goes out. Average power outage is about 2 hours over here. 1500VA is definitely not a lot.

In my case it appears the hashrate does not bounce around. It's usually staying in the 68 to 74 GH/s range. My Average is 71. I'm running 03-25 firmware. Ambient temps are between 15 to 17 C.

So far I'm happy with the overclock but my main concern now is how much longer can the PSU hold under this heavy stress. But I already said many times on this thread and others that I plan to swap out that overworked antec 650 unit.

True, but higher VA units are not so easy to setup in a residential type of situation. I don't want to wire 30 amp (or heavier) circuits for a UPS. The electrical work alone will be expensive, and so will the UPS itself ($5000 to $10,000). There are other factors too that make it a challenge to engineer a adequate 100% uptime solution for these units in a RESIDENTIAL situation. A 1500VA unit is enough to ride the usual power flickers that happen. Here in my area we have reliable power and the worst that typically happens is brownouts. One solution would be to keep another miner at a family member that you trust. This way if you have an extended outtage at your home location you at least aren't losing out as much in daily earnings.

Ideally you'd want to host these in a data center, but that's a whole other discussion. Would you trust someone else with your miner?

how are u monitoring your avalons?  via ssh over the network?

I got an Avalon that initially appeared to be a lemon. cgminer would not start through the web interface. It could only be started from an SSH term using a start - ctrl-C - start - ctrl-C ... sequence until the units were coaxed into hashing properly. Otherwise I'd get "FPGA contoller mess up" "Avalon initialized" messages looping over and over. Once started, hashing at 282 MHz was a little unstable with the hash rate bouncing around a little.

After reading some of the comments in the other thread, I replaced the supplied Corsair GS700 PSU with a Fortress 750W Platinum I had lying around. This solved the cgminer  starting problem and the unsteady hash rate. In fact, when I switched the clock to 300 MHz it now hashes perfectly at a steady 71.5 Gh/s with the somewhat upgraded PSU.

You might want to have look at your power supply if you are having these problems.

Both. I typically just keep an eye on the cgminer status page via http, and also if I notice that cgminer has been running for a short time I hop on over to ssh and see if the actual system rebooted. Typically it's just cgminer acting up and having to be restarted.

Very nice. I just finally pulled the trigger on a AX860 (80 Plus Platinum rated) so hopefully that will work out ok. Just a word of caution to those of you who are thinking of replacing the PSU, watch out and make sure the length of the PSU is not over 7 inches. A decent 800+ watt unit should be around 6 to 6.3".

however someone who posted pics, i think libertybuck, is running his psu's outside the avalon just fine and has measured temps actually lower with this configuration.

xiangu updated the wiki today. there's now a list of power supplies

https://en.bitcoin.it/wiki/Avalon#Comfirmed_working_PSU_list


Also a public test firmware is available here. included cgminer 2.11.4 ( there is still a bug with this one where sometimes 0 mhash happens. Is/Will be fixed soon. )

http://downloads.qi-hardware.com/people/xiangfu/avalon/next-testing/openwrt-ar71xx-generic-tl-wr703n-v1-squashfs-factory-20130408.bin

is this the same as it was PM?

no.

one in PM is based off of 2.10.5a. PM one has working restart.

Newest testing firmware 20130408 has an issue when you first start or when you change pools, cgminer doesn't restart. It hangs and sits for several minutes until watchdog kills process. Then it works ok.

20130408 is very nice when running though, and is based off of 2.11.4 and a new avalon driver. It greatly reduces rejects and HW errors.

ok thx then I will wait ;)

No. I did not posted it. Should be someone else.

Aseras, Please let us know when this issue is fixed.

Will do. Ckolivas is working on it with xiangfu. My avs are limping now, my Internet is out, running on 3G until they fix it. Sucks. Probably have to drag my avalons back to work.

A new [Next] firmware has been released. Anyone brave enough to give it a test and see if all the issues have been ironed out?

http://downloads.qi-hardware.com/people/xiangfu/avalon/next-testing/openwrt-ar71xx-generic-tl-wr703n-v1-squashfs-factory-20130410.bin

I was going to say the same thing. I'm curious to know what has been changed/fixed/updated in this latest firmware. A "CHANGELOG" file in the testing folder would be really nice.

I'm testing it on two units right now.  So far it doesn't hang after I update the cgminer config in the gui and it updates faster.

is it save Keep settings?

new build fixes the delay when starting. It is also cgminer up to 2.11.4 code. It fixes many bugs. It greatly reduces rejects and HW errors.

safe to keep settings.


If you want to know what was changed, simply look at the git

https://github.com/BitSyncom/cgminer/tree/avalon

https://github.com/BitSyncom/cgminer-openwrt-packages

flashed to 0410 firmware just now.

the process is smooth, and the cgminer got upgrade. worked ok at 300Mhz for 10 minutes.

Still have some HW ( maybe cause I'm working at 300Mhz).
Still many rejects with slush's pool ( but all other pool is ok. ozco, btcguild)

Before flash, the hashrate average is 70-71 G each. now one is 67G and one is 71G.

Will keep monitoring and report later.

the new flash and cgminer reimplement the miner worker threads.

https://github.com/BitSyncom/cgminer/commit/6c5c6e99a32315af9ee345830a10fd770fa615b8

it makes the work numbers different. It is much more consistent now. Before work would return kinda at random, you you'd see swings between high values and lower ones.

Aseras, question for you. What percentage of HW errors would you consider to be "OK" while running at 300 MHz. I get close to 8% at the moment. With 03-25 I got pretty much the same. Admittedly it has been only 1hr and 20 min since I flashed to 04-10 so maybe I need a longer sampling period?

I got 1045 accept, 48 HW for 1 hour 14 min for one.
and 557 accept, 35 HW for 32 min for another.

A handful per hour, not more than few hundred per day. 1 hw error for several thousand diff1 shares.

The reason for the hw errors is there is a bug in Fpga controller, it cannot be fixed in software ( according to xiangfu ) but trying to find a work around. The controller won't reset on end of work and stalls apparently no matter what you do.

Ckolivas is going to rewrite the USB driver and code next week.  if all goes well well have much better control soon.

Looking forward to having the HW issue resolved..

Interesting, there is a FPGA chip on that data controller  :)

godd to know that the top guys are trying to debug avalon :P

Bold text are the numbers that seem odd to me. Discarded 5278?!  Hmm. HW 1315. It's mainly the discarded that I'm worried about. Could this be due to the pool or the avalon firmware?

I'm running 4-10.

Discards are pool or protocol. In stratum/GBT you make up work as you go, if the block changes, the work is discarded. If it's stale and submitted or duplicated work, its rejected. HW errors are just that, the hardware hung doing something and had to be reset.

as long as your work is mostly accepted shares, you are good. If you see a very high % ( >10 % ) of rejected/stale or HW errors then something is wrong.

Thanks for the info Aseras! :)

So far I think my stats look good.