[ANN] Spondoolies-Tech - carrier grade, data center ready mining rigs

[zvisha]

Please write to me and remo, i would like to investigate. Did others here get degradation or hw error increase with some specific version?
My test miners show up to 3% difference between self reported and pool. Do you see more?

zvisha has been helping me out the last day or so with troubleshooting and getting some temporarily-lost stability back to the miner, however, i am still having some slight issues:

1) fans at 70% used to be fine but as of today sounded terrible. ive soundproofed somewhat and reduced to 60%, but its still a similar volume to what an unmuffled 60% would have sounded like last week. Without a dB meter i could be crazy though - getting this in a DC is top priority. the unit also has very slight vibrations that go away if i put a few pounds of pressure on the middle back edge of the lid.

2) The unit seems to always have a 10% or higher (right now, its about 16%!) error rate, even after hours of operation, temperatures of 26/77C, and most of the chips at 77C with a few at 113C.  Miner says 1.41Th while the pool says 1.25-1.3TH.  What is the accepted/error ratio for others?

the product is excellent, but it needs to be operated in a datacenter or somewhere that the fans running at 90-100% wont burst eardrums. I would really love to see the SP10 fit in an SP30 case to better dissipate the heat and use bigger/quieter fans.  Most DC racks cant handle anywhere near the 40-50kW draw that the SP10 and SP30 models enable

I was also always getting 1.3TH consistently, then put the new firmware on. Now I'm down to 1.2TH. So I'm going the other way. I'm still on a 110V circuit and get switched over to a 208V soon so hopefully I'll see the nice 1.45TH that I see in all the Spondoolie video's.

[psahx]

Kobi's reply:
It’s hard to tell as the mechanical details of the holes and the screws are missing.
It seems to me that the “One Space (1U) Rack Mount Bracket (Ear)MCM Part #: 555-13214  |  Penn Elcom Part #: R1206/1U” won’t fit while the “Cisco STK-RACKMOUNT-1RU Rackmount Brackets” seems to fit.
Please verify the screws are #6-32 x 3/8”.
In addition, those mounting brackets are not adjustable so please pay attention to the SP10 depth which is 475mm and the rack rails distance

Thanks a lot. I guess, I will just give it a try with Cisco, does not cost that much. Will let you know.

[zvisha]

Ok, Con knows of the issue, it will be fixed soon.
For now use previous version for p2p.

Did the 1.4.2 firmware update do something to disable mining on p2pool?  I have updated, and now my SP10 will not connect to ANY p2pool node.  I can connect to Eligius, BTCGuild, GHash, etc.  My local p2pool node?  No.  Any other p2pool node?  No.  Before anyone asks... yes, my local node is up.  I can confirm this because my 2 S1s are happily hashing away on it.  I have also pointed those two to another p2pool node and they hashed there as well.

Thanks for the help.

[klondike_bar]

WORDS

Then go ahead, put it in a DC. Can't run that at home unless you have a way of not hearing it.

I actually run it in the backroom of a small business - but the other day it became so noisy (seemingly out of the blue that morning) at 70% fans they asked me to shut it down remotely until i could come in later the day and correct the noise issues.

If all goes well the unit will be in a bitcoin colocation by the end of the month with $0.05/kWh power

[zvisha]

The "Spondoolies hero of the day" medal goes to user Klondike_bar, for discovering a problem of high error rate on his machine. We released a fix (1.4.3) that should bring the HW error rate down from ~8% to 2%.

Please update ASAP and let me know if the fix works.

[guytechie]

Please write to me and remo, i would like to investigate. Did others here get degradation or hw error increase with some specific version?
My test miners show up to 3% difference between self reported and pool. Do you see more?

I just upgraded all 3 SP10s to 1.4.2 and they are now much slower at 1.2TH.  They were getting 1.3TH before the firmware upgrade (1.3.49).

As I checked for the previous version number I noticed you just promoted 1.4.3 from test version to stable.  What is the changelog between all 3 versions?

[zvisha]

Please write to me and remo, i would like to investigate. Did others here get degradation or hw error increase with some specific version?
My test miners show up to 3% difference between self reported and pool. Do you see more?

I just upgraded all 3 SP10s to 1.4.2 and they are now much slower at 1.2TH.  They were getting 1.3TH before the firmware upgrade (1.3.49).

As I checked for the previous version number I noticed you just promoted 1.4.3 from test version to stable.  What is the changelog between all 3 versions?

Please try 1.4.3 and let me know how it behaves as self-reported rate, rate in pool and error rate.
Also please send me your Webpages with DCRs, ASICs shots and front page of the version you have now (I guess 1.4.2) and 1.4.3.

Note that at start miners always work slower, and then rise their rate till they find optimal point.

[merv77]

The "Spondoolies hero of the day" medal goes to user Klondike_bar, for discovering a problem of high error rate on his machine. We released a fix (1.4.3) that should bring the HW error rate down from ~8% to 2%.

Please update ASAP and let me know if the fix works.

my HW errors on firmware version 1.4.2 settled down to about 5%

I will try 1.4.3 and report back tomorrow with results.

thanks zvisha and klondike_bar

BTW, what did klondike_bar discover?

[zvisha]

The "Spondoolies hero of the day" medal goes to user Klondike_bar, for discovering a problem of high error rate on his machine. We released a fix (1.4.3) that should bring the HW error rate down from ~8% to 2%.

Please update ASAP and let me know if the fix works.

my HW errors on firmware version 1.4.2 settled down to about 5%

I will try 1.4.3 and report back tomorrow with results.

thanks zvisha and klondike_bar

BTW, what did klondike_bar discover?

His machine had exceptionally high error rate of over 10%. I have error rate of about 3-4%. He let me use his machine to understand the situation, and I found out that the fix for his machine improves HW errors in all my machines too. It is the way I use BIST to find optimal point, the BIST I did was too weak for many ASICs.
 

[murdof]

Hey Zvi,

One miner is working at :
26C/72C (in/out) - 227V - AC2DC power=1256[1265], temp=0, leading zeroes=41
[H:HW:1446Gh/Th:1446Gh/W:1446Gh,W:945,L:24,A:192,ER:1446,EP:1222,MMtmp:26]

The other is working at:
27C/77C (in/out) - 222V - AC2DC power=1254[1265], temp=0, leading zeroes=40
[H:HW:1397Gh/Th:1398Gh/W:1397Gh,W:944,L:24,A:192,ER:1397,EP:1222,MMtmp:27]

Just wondering how come the second miner never breaks that 1.42TH and is around 1.4TH all the time - while the first is stuck at 1.46TH....
I know it is within the 10% tolerance - just wondering...

Both miners overclocked according to your post.

[bobsag3]

I have not had any problems with p2p on any firmware on any of my sp10s.

[guytechie]

I just upgraded all 3 SP10s to 1.4.2 and they are now much slower at 1.2TH.  They were getting 1.3TH before the firmware upgrade (1.3.49).

As I checked for the previous version number I noticed you just promoted 1.4.3 from test version to stable.  What is the changelog between all 3 versions?

Please try 1.4.3 and let me know how it behaves as self-reported rate, rate in pool and error rate.
Also please send me your Webpages with DCRs, ASICs shots and front page of the version you have now (I guess 1.4.2) and 1.4.3.

Note that at start miners always work slower, and then rise their rate till they find optimal point.

Nevermind the changelog.  I read in your previous post.  I can't wait to test the 110v limit lift.  The cables I am using are 14 gauge, so they should be ok.

I can't do screenshots yet because I'm at work.  Well, I can, but I can't... (can't have anything too obvious on my screen to show I'm not working, lol).

I was able to remote in to check the hashrate and the miners report a slower rates than before by about 10 to 30 GH.  One is 10 GH less, another is 20 GH less, and the 3rd is 30 GH less.  However, BTC Guild reports FASTER rates than before.  I don't know what they were individually, but normally all 3 would give me around 3.8 TH, but now it's 4 TH.

Maybe there were more HW errors in 1.3.49 than in 1.4.3?  Who knows.  I honestly didn't pay attention to the HW errors.

[raskul]

pools report hashrates using packet data sent from your miner, so the hashrate showing at your pool is likely to be showing on a slight delay, and usually not exact. Best always to go by the stats showing on the UI / Cgminer.

[jonnybravo0311]

Ok, Con knows of the issue, it will be fixed soon.
For now use previous version for p2p.

Did the 1.4.2 firmware update do something to disable mining on p2pool?  I have updated, and now my SP10 will not connect to ANY p2pool node.  I can connect to Eligius, BTCGuild, GHash, etc.  My local p2pool node?  No.  Any other p2pool node?  No.  Before anyone asks... yes, my local node is up.  I can confirm this because my 2 S1s are happily hashing away on it.  I have also pointed those two to another p2pool node and they hashed there as well.

Thanks for the help.

Thanks Zvisha.  I tried 1.4.3, and it also does not see any p2pool node.  I'll go back to 1.3.49 until a fix can be made.

[Biodom]

can we have a status update on June batch please? thanks

Didn't you hear the news, they sold all remaining Sp-10 to Taras [LNU].

Biodom, I have an order for 2 SP10 from June batch, i guess you were just too late. that's how it is mucker  
a s-t-a-t-u-s update - you perceive this to mean something other than a s-t-a-t-u-s update.

Don't act so butt-hurt about it.

hey, i am hurt about it because i communicated with them on saturday night and was given no indication that I have to do it in BTC that very moment and when I was about to go to the bank on Monday am and was checking the forum-boom, they had sold it.

"Callum decided to call his father-in-law the "Exorcist" because every time he came to visit he made the spirits disappear"

[zvisha]

I just upgraded all 3 SP10s to 1.4.2 and they are now much slower at 1.2TH.  They were getting 1.3TH before the firmware upgrade (1.3.49).

As I checked for the previous version number I noticed you just promoted 1.4.3 from test version to stable.  What is the changelog between all 3 versions?

Please try 1.4.3 and let me know how it behaves as self-reported rate, rate in pool and error rate.
Also please send me your Webpages with DCRs, ASICs shots and front page of the version you have now (I guess 1.4.2) and 1.4.3.

Note that at start miners always work slower, and then rise their rate till they find optimal point.

Nevermind the changelog.  I read in your previous post.  I can't wait to test the 110v limit lift.  The cables I am using are 14 gauge, so they should be ok.

I can't do screenshots yet because I'm at work.  Well, I can, but I can't... (can't have anything too obvious on my screen to show I'm not working, lol).

I was able to remote in to check the hashrate and the miners report a slower rates than before by about 10 to 30 GH.  One is 10 GH less, another is 20 GH less, and the 3rd is 30 GH less.  However, BTC Guild reports FASTER rates than before.  I don't know what they were individually, but normally all 3 would give me around 3.8 TH, but now it's 4 TH.

Maybe there were more HW errors in 1.3.49 than in 1.4.3?  Who knows.  I honestly didn't pay attention to the HW errors.

This is exactly what I would expect. Great news, we just gained ~5%!

[zvisha]

Hey Zvi,

One miner is working at :
26C/72C (in/out) - 227V - AC2DC power=1256[1265], temp=0, leading zeroes=41
[H:HW:1446Gh/Th:1446Gh/W:1446Gh,W:945,L:24,A:192,ER:1446,EP:1222,MMtmp:26]

The other is working at:
27C/77C (in/out) - 222V - AC2DC power=1254[1265], temp=0, leading zeroes=40
[H:HW:1397Gh/Th:1398Gh/W:1397Gh,W:944,L:24,A:192,ER:1397,EP:1222,MMtmp:27]

Just wondering how come the second miner never breaks that 1.42TH and is around 1.4TH all the time - while the first is stuck at 1.46TH....
I know it is within the 10% tolerance - just wondering...

Both miners overclocked according to your post.

It depends on the HW you got. Our first batch was our best and with low variance between machines, it gets to 1.5 easy with the over-clocking (and even 1.57 in iceland temperatures ) and 1.35 for 110V. The second batch was much more diverse and very temperature dependant, ranging from 1.36 to 1.48 for 220V and 1.25 to 1.33 for 110V.

[zvisha]

Ok, Con knows of the issue, it will be fixed soon.
For now use previous version for p2p.

Did the 1.4.2 firmware update do something to disable mining on p2pool?  I have updated, and now my SP10 will not connect to ANY p2pool node.  I can connect to Eligius, BTCGuild, GHash, etc.  My local p2pool node?  No.  Any other p2pool node?  No.  Before anyone asks... yes, my local node is up.  I can confirm this because my 2 S1s are happily hashing away on it.  I have also pointed those two to another p2pool node and they hashed there as well.

Thanks for the help.

Thanks Zvisha.  I tried 1.4.3, and it also does not see any p2pool node.  I'll go back to 1.3.49 until a fix can be made.

BTW, do you know how much bandwidth P2P node takes in cgminer for SP10? What hash rate it provides and what is the leading zeroes?

[Guy Corem]

can we have a status update on June batch please? thanks

Didn't you hear the news, they sold all remaining Sp-10 to Taras [LNU].

Biodom, I have an order for 2 SP10 from June batch, i guess you were just too late. that's how it is mucker  
a s-t-a-t-u-s update - you perceive this to mean something other than a s-t-a-t-u-s update.

Don't act so butt-hurt about it.

hey, i am hurt about it because i communicated with them on saturday night and was given no indication that I have to do it in BTC that very moment and when I was about to go to the bank on Monday am and was checking the forum-boom, they had sold it.

"Callum decided to call his father-in-law the "Exorcist" because every time he came to visit he made the spirits disappear"

Biodom, the sold-out was very fast due to several large orders we got. We didn't expect it either. On Saturday night we had solid stock.

Regards,
Guy

[guytechie]

pools report hashrates using packet data sent from your miner, so the hashrate showing at your pool is likely to be showing on a slight delay, and usually not exact. Best always to go by the stats showing on the UI / Cgminer.

The stats in the UI are definitely lower than before.  The pool - you're right, there's variance because it's based on data packets sent.  It's now at 3.9 TH.  Still better than before though, which ranged from 3.7 to 4 TH (super lows and super highs).

Right now my super lows and highs are 3.8 TH and 4.1 TH.  You can never go by those numbers of course.  Avg seems to be about 100 GH more on the pool (or about 33 GH faster each miner when you divide by 3).