Pages:
Author

Topic: BITMAIN Antminer S3 support and OverClocking thread - page 42. (Read 158144 times)

hero member
Activity: 630
Merit: 500
legendary
Activity: 4256
Merit: 8551
'The right to privacy matters'
legendary
Activity: 1456
Merit: 1000
Weird thing is, the OC'ed one is doing the best. Obviously shares accepted, but hw errors are lowest,  and discarded are the lowest as well.

The temps on all machines are right at 40c

http://imgur.com/01NWkWj

They can preform very well if OC'ed right. (assuming you have a unit that likes it some of first batches do not).  But with OCing I've been very happy hottest 2 day's in a long time happened to be during my testing and the one that OC'ed well was 39 degrees and very good poolside.
full member
Activity: 168
Merit: 100
Weird thing is, the OC'ed one is doing the best. Obviously shares accepted, but hw errors are lowest,  and discarded are the lowest as well.

The temps on all machines are right at 40c

http://imgur.com/01NWkWj
sr. member
Activity: 294
Merit: 251
What's the highest frequency I can test without connecting 4 PCIE?  Right now I have 218.75 default frequency .

Thanks

Two cables @ 225, but my cables are hot. 600b bronze power supply, if that makes different.

 I can hold the cables in my hand, but i'm not going to push them further.

Best I'm getting is ~2 gh/s per frequency, i.e., 225 = 450 gh/s

OK thanks. 

I'm getting with 218.75 freq : 434 GH/s and 445 GH/s
full member
Activity: 168
Merit: 100
What's the highest frequency I can test without connecting 4 PCIE?  Right now I have 218.75 default frequency .

Thanks

Two cables @ 225, but my cables are hot. 600b bronze power supply, if that makes different.

 I can hold the cables in my hand, but i'm not going to push them further.

Best I'm getting is ~2 gh/s per frequency +4gh/s

I've got 3 s3's and this is after moving cables around on the 6pins and ssh overclocking 1 so far.

1 @ 225 = 454 gh/s
2 @ 218 = 440 gh/s


sr. member
Activity: 294
Merit: 251
What's the highest frequency I can test without connecting 4 PCIE?  Right now I have 218.75 default frequency .

Thanks
legendary
Activity: 1456
Merit: 1000
Well batch one here.  One runs at 420-425 overclocking makes no difference.  The other one i am able to get with 238 freq 440-460.  Seems to go down after a few hours.  Didn't try 250 freq as I am happy with the 440+ with temps under 40.  It's very very stable almost 2 days at overclocked speeds on 2nd unit no problems.

Very happy with my unit's wish first one overclocked, but with them going to refund/coupon for speed difference I can't complain at all.
full member
Activity: 168
Merit: 100
After changing the # to where I wanted them, I had to hit esc key and then :wq. I'm a win7 user, and this worked for me.
hero member
Activity: 857
Merit: 1000
Anger is a gift.
When you guys are re-applying thermal paste after cleaning with alchohol are you applying a thin layer with a piece of plastic/razor so the entire chip is covered or do you just put like a 1/2 pea sized drop in the center of each chip and let the heatsink smoosh it around?

Technically, the 2nd procedure you mentioned is correct.  Applying it w/ plastic or razor has the potential to create micro air bubbles in the paste and hinder its ability to transfer heat away from the chip.

I would test a small amount on one chip, and see what the coverage is like.

Once you have roughly the ideal amount, simply repeat the process down the line...

A little dab will usually suffice...

Yeah a little dab is all you need. This is not like a CPU where you can put a 1/2 of a pea worth and let the heat sink spread it out. Just a little dab on the chip and do not smooth it out. Worked well for me.
newbie
Activity: 5
Merit: 0
When you guys are re-applying thermal paste after cleaning with alchohol are you applying a thin layer with a piece of plastic/razor so the entire chip is covered or do you just put like a 1/2 pea sized drop in the center of each chip and let the heatsink smoosh it around?

Technically, the 2nd procedure you mentioned is correct.  Applying it w/ plastic or razor has the potential to create micro air bubbles in the paste and hinder its ability to transfer heat away from the chip.

I would test a small amount on one chip, and see what the coverage is like.

Once you have roughly the ideal amount, simply repeat the process down the line...

A little dab will usually suffice...
newbie
Activity: 30
Merit: 0
of 4 units:

a) 231.25MHZ = 452Gh
b) 218.75MHz = 429GH (one chip is shown as -  though, so presumably im losing about 12GH)
c) 250MHz = 502.5GH
d) 218.75MHz = 428GH (one chip is shown as -  though, so presumably im losing about 12GH)

ive tried rebooting the two systems with the dudding chips but to no avail. has anyone had and resolved this issue? (with the S1 a reboot always seemed to solve it)

overall im not terribly impressed with the quality of the components bitmain used. hopefully its a batch-specific problem though and wont occur again.

I had a unit that would X chips at any clock about 218.75. Had to take it apart and reapply thermal paste. No X chips after that. Been running for 2 days on stock clock with no issues.

When you guys are re-applying thermal paste after cleaning with alchohol are you applying a thin layer with a piece of plastic/razor so the entire chip is covered or do you just put like a 1/2 pea sized drop in the center of each chip and let the heatsink smoosh it around?
hero member
Activity: 857
Merit: 1000
Anger is a gift.
of 4 units:

a) 231.25MHZ = 452Gh
b) 218.75MHz = 429GH (one chip is shown as -  though, so presumably im losing about 12GH)
c) 250MHz = 502.5GH
d) 218.75MHz = 428GH (one chip is shown as -  though, so presumably im losing about 12GH)

ive tried rebooting the two systems with the dudding chips but to no avail. has anyone had and resolved this issue? (with the S1 a reboot always seemed to solve it)

overall im not terribly impressed with the quality of the components bitmain used. hopefully its a batch-specific problem though and wont occur again.

I had a unit that would X chips at any clock about 218.75. Had to take it apart and reapply thermal paste. No X chips after that. Been running for 2 days on stock clock with no issues.
legendary
Activity: 2128
Merit: 1005
ASIC Wannabe
of 4 units:

a) 231.25MHZ = 452Gh
b) 218.75MHz = 429GH (one chip is shown as -  though, so presumably im losing about 12GH)
c) 250MHz = 502.5GH
d) 218.75MHz = 428GH (one chip is shown as -  though, so presumably im losing about 12GH)

ive tried rebooting the two systems with the dudding chips but to no avail. has anyone had and resolved this issue? (with the S1 a reboot always seemed to solve it)

overall im not terribly impressed with the quality of the components bitmain used. hopefully its a batch-specific problem though and wont occur again.
hero member
Activity: 840
Merit: 1000
Quick power #'s, measured with my best unit.. and the only one that will hit 500GH.

Corsair TX850 80plus bronze PSU - single S3:
218 mhz -- 450 GH --  346 watt at the wall
225 mhz -- 480 GH --  370 watt at the wall
237 mhz -- 500 GH --  394 watt at the wall


I was a bit curious myself and tested one of my batch one units, but I'm getting different power numbers at the wall. See below:

218 mhz -- 440 GH --  360 watt at the wall
225 mhz -- 387 GH --  348 watt at the wall
237 mhz -- 408 GH --  360 watt at the wall
250 mgh -- 424 GH --  345 watt at the wall

I'm using a Corsair HX650 for each unit and testing using a Rosewill RHSP-13001. I've only tested one S3 thus far but there appears to be a power issue, or it's just not liking anything outside of stock frequency. I haven't flashed the firmware yet but am more concerned that I'm not pushing the same power to the unit at those higher frequencies. Any thoughts?


So in the apology post that Bitmain posted, ( https://bitcointalksearch.org/topic/m.7832797 ) they effectively indicated a bad batch of DC-DC and people weren't going to get the speeds they wanted.. hence the refund/credit.

I suspect they've set all S3's to 218 out of the box, and knowing that most of them won't be able to run over that point. I have (1) of (10) that can run over the stock.. the rest just start X'ing out ASIC's and never go any better.. they're actually faster are 218 than at 225/237/250/etc.

(I know a couple of -really- good board-level guys who could swap-out DC-DC supplies if I had any idea what to replace the faulty ones with)

As far as I know, there are no pin to pin replacement for the 53355 with same or higher amp in the same voltage values.
I already checked for overvolting.
Best bet would be to replace it with another batch of 53355, 3$ each, 50$ for each S3, and not sure it will solve anything.
for low cost fix, I'd try a slight over voltage by replacing the divider resistor. Going up to 0.8v could help.
sr. member
Activity: 467
Merit: 250
Quick power #'s, measured with my best unit.. and the only one that will hit 500GH.

Corsair TX850 80plus bronze PSU - single S3:
218 mhz -- 450 GH --  346 watt at the wall
225 mhz -- 480 GH --  370 watt at the wall
237 mhz -- 500 GH --  394 watt at the wall


I was a bit curious myself and tested one of my batch one units, but I'm getting different power numbers at the wall. See below:

218 mhz -- 440 GH --  360 watt at the wall
225 mhz -- 387 GH --  348 watt at the wall
237 mhz -- 408 GH --  360 watt at the wall
250 mgh -- 424 GH --  345 watt at the wall

I'm using a Corsair HX650 for each unit and testing using a Rosewill RHSP-13001. I've only tested one S3 thus far but there appears to be a power issue, or it's just not liking anything outside of stock frequency. I haven't flashed the firmware yet but am more concerned that I'm not pushing the same power to the unit at those higher frequencies. Any thoughts?


So in the apology post that Bitmain posted, ( https://bitcointalksearch.org/topic/m.7832797 ) they effectively indicated a bad batch of DC-DC and people weren't going to get the speeds they wanted.. hence the refund/credit.

I suspect they've set all S3's to 218 out of the box, and knowing that most of them won't be able to run over that point. I have (1) of (10) that can run over the stock.. the rest just start X'ing out ASIC's and never go any better.. they're actually faster are 218 than at 225/237/250/etc.

(I know a couple of -really- good board-level guys who could swap-out DC-DC supplies if I had any idea what to replace the faulty ones with)



sr. member
Activity: 448
Merit: 250
newbie
Activity: 22
Merit: 0
Quick power #'s, measured with my best unit.. and the only one that will hit 500GH.

Corsair TX850 80plus bronze PSU - single S3:
218 mhz -- 450 GH --  346 watt at the wall
225 mhz -- 480 GH --  370 watt at the wall
237 mhz -- 500 GH --  394 watt at the wall


I was a bit curious myself and tested one of my batch one units, but I'm getting different power numbers at the wall. See below:

218 mhz -- 440 GH --  360 watt at the wall
225 mhz -- 387 GH --  348 watt at the wall
237 mhz -- 408 GH --  360 watt at the wall
250 mgh -- 424 GH --  345 watt at the wall

I'm using a Corsair HX650 for each unit and testing using a Rosewill RHSP-13001. I've only tested one S3 thus far but there appears to be a power issue, or it's just not liking anything outside of stock frequency. I haven't flashed the firmware yet but am more concerned that I'm not pushing the same power to the unit at those higher frequencies. Any thoughts?
sr. member
Activity: 467
Merit: 250

generated a table to work with, based on the S1/S2 tables:

Quote
...package 'cgminer'

config 'asic-freq' 'default'

        #option 'freq_value'    '1486'  #262.5M
        #option 'chip_freq'     '262.5'
        #option 'timeout'       '15'

        #option 'freq_value'    '1386'  #250M
        #option 'chip_freq'     '250'
        #option 'timeout'       '16'

        #option 'freq_value'    '1306'  #243.75M
        #option 'chip_freq'     '243.75'
        #option 'timeout'       '17'

        #option 'freq_value'    '1286'  #237.5M
        #option 'chip_freq'     '237.5'
        #option 'timeout'       '17'

        #option 'freq_value'    '1206'  #231.25M   
        #option 'chip_freq'     '231.5'             
        #option 'timeout'       '18'                                                         
                                                                     
        #option 'freq_value'    '1185'  #225M   
        #option 'chip_freq'     '225'           
        #option 'timeout'       '18'           
                                               
        option 'freq_value'    '1106'  #218.75M
        option 'chip_freq'     '218.75'       
        option 'timeout'       '18'           
                                               
        #option 'freq_value'    '1086'  #212.5M
        #option 'chip_freq'     '212.5'       
        #option 'timeout'       '18'         
                                             
        #option 'freq_value'    '0782'  #200M
        #option 'chip_freq'     '200'       
        #option 'timeout'       '20'

hero member
Activity: 630
Merit: 500
i did my own calculations and here are the specs for 243.75M:

        option 'freq_value'    '1306'  #243.75M
        option 'chip_freq'     '243.75'
        option 'timeout'       '17'
Pages:
Jump to: