BITMAIN Antminer S3 support and OverClocking thread

[tyfo]

i thought the S3's shut down if they reached 80 degrees celcius ?
anyone tested this since i see alot of talk about fire hazards.

also could get a tellstick and some kind of nexa unit and turn off the power if the temperature inside the room reaches X degrees

If they shut down - what is the point of mining? Sure, the A/C will cool down the air eventually and ASICs will run again at some point. In this case the best option would be the chilled air flowing directly to the S3s and PSUs. Enclosing such heat makes it a hot air electric oven unless your A/C is more powerful than the S3s and PSUs combined (you can also think about throwing the PSUs outside the cold area if you don't care about cooling them).

[1713915287]

1713915287

[1713915287]

1713915287

[oskuro]

Question guys.

With 218.75 and 225 its necessary to use 4 PCIE? Or with just 2 PCIE its ok?? For 218.75 im sure i can just use 2 PCIE, but for 225 frequency im not that sure.

right now i have 4 PCIE for both, 716W total

[chadwickx16]

Question guys.

With 218.75 and 225 its necessary to use 4 PCIE? Or with just 2 PCIE its ok?? For 218.75 im sure i can just use 2 PCIE, but for 225 frequency im not that sure.

right now i have 4 PCIE for both, 716W total

I've heard you can do it with two cables, just make sure they aren't getting too hot. As is hot to the touch.

[allcoinminer]

Question guys.

With 218.75 and 225 its necessary to use 4 PCIE? Or with just 2 PCIE its ok?? For 218.75 im sure i can just use 2 PCIE, but for 225 frequency im not that sure.

right now i have 4 PCIE for both, 716W total

My recommendation is to use four cables for better stability.
But its not a necessity. If you are overclocking I now strongly recommend(unlike earlier) to use 4 power cables.

[oskuro]

Question guys.

With 218.75 and 225 its necessary to use 4 PCIE? Or with just 2 PCIE its ok?? For 218.75 im sure i can just use 2 PCIE, but for 225 frequency im not that sure.

right now i have 4 PCIE for both, 716W total

My recommendation is to use four cables for better stability.
But its not a necessity. If you are overclocking I now strongly recommend(unlike earlier) to use 4 power cables.

I wanted to use Corsair CX750M for both, but yeah i have one at 225 frequency, so thats OC and i suppose with just 2 PCIE wont work very well maybe

[orviwan]

Does anyone have any information about the JDH1238B fan used on the S3? Specifically Air Flow CFM and Noise Level dbA.

I'm looking to replace them with something quieter if possible.

Thanks

[kebabman]

Which two of the four PCI-E connectors do you need to plug power into? Or does it not matter?

[philipma1957]

Does anyone have any information about the JDH1238B fan used on the S3? Specifically Air Flow CFM and Noise Level dbA.

I'm looking to replace them with something quieter if possible.

Thanks

they move 90-105 cfm.  the high end 30 a piece deltas move 210 cfm.  so if you buy them and turn down the rpm they will be quieter.  wait for link,


these can be found for less.


http://www.frozencpu.com/products/9321/fan-581/Delta_120_x_38mm_Ultra_High-Speed_PWM_Fan_-_21038_CFM_QFR1212GHE-PWM.html?tl=c365s935b113&id=mPpv8As6


they are 4 pin and will run slower then 2000 rpm and still move more air then stock.

most other solutions are meh.  but someone may find a better one for you.

[TracerX]

Which two of the four PCI-E connectors do you need to plug power into? Or does it not matter?

The blades run lengthways, and you need to power each.  So one on the "left-side" and one on the "right-side", if you are looking at it from the front side with the ethernet connection.

[allcoinminer]

Which two of the four PCI-E connectors do you need to plug power into? Or does it not matter?

You can connect any of the two, one in each blade.
Its better and safe to have all the 4 cables connected to Ant S3.
I found improved stability when connected all the fours.
Make sure that you are not connecting a single blade from multiple PSUs.
Same time you can connect one PSU to one blade and Another PSU to the other blade
if you cannot have a single PSU which can power both the blades.

[la7eralus]

Work Input timing
Clock Freq.(M)   Hash rate(G)   Reg.value   Ideal delay(ms)   Recommenddelay(ms)
 100                   6.30                    0783           42.6                    38
 125                   7.88                    0983           34.0                    31
 150                   9.45                    0b83           28.4                    26
 175                  11.03                    0d83           24.3                    22
 200                  12.60                    0782           21.3                    19
 225                  14.18                    0882           18.9                    17
 250                  15.75                    0982           17.0                    15
 275                  17.33                    0a82           15.5                    14
 300                  18.90                    0b82           14.2                    13
 325                  20.48                    0c82           13.1                    12
 350                  22.05                    0d82           12.2                    11
 375                  23.63                    0e82           11.3                    10
 400                  25.20                    08f2           10.6                    10

Has anyone successfully broken the above down into smaller incremental steps?  Preferably 1/4 of their current spacing?  Or even 1/6th?  Thanks!

[v0n]

Both of my 1st batch S3's will overclock to 225 and beyond, at least for short time, but both hash slower at 225 and 250 than at 218.5. As in - both 5s and average GH/s figures are lower (barely reaching 400 at 225), whereas they will happily do stable throughput of 440 at 218.5.   

[la7eralus]

Clock Freq.(M)   Hash rate(G)      Reg.value      Ideal delay(ms)      Recommenddelay(ms)
 100                   6.30                    0783           42.6                    38
-106.25
-112.5
-118.75
 125                   7.88                    0983           34.0                    31
-131.25
-137.5
-143.75
 150                   9.45                    0b83           28.4                    26
-156.25
-162.5
-168.75
 175                  11.03                    0d83           24.3                    22
-181.25
-187.5
-193.75
 200                  12.60                    0782           21.3                    19
-206.25
-212.5
-218.75
 225                  14.18                    0882           18.9                    17
-231.25
-237.5
-243.75
 250                  15.75                    0982           17.0                    15
-256.25
-262.5
-268.75
 275                  17.33                    0a82           15.5                    14
-281.25
-287.5
-293.75
 300                  18.90                    0b82           14.2                    13
-306.25
-312.5
-318.75
 325                  20.48                    0c82           13.1                    12
-331.25
-337.5
-343.75
 350                  22.05                    0d82           12.2                    11
-356.25
-362.5
-368.75
 375                  23.63                    0e82           11.3                    10
-381.25
-387.5
-393.75
 400                  25.20                    08f2           10.6                    10

I was able to seperate the above frequencies into smaller incremental steps, but I have no idea how to calculate the "reg. value" or the "ideal delay" for them.  Anyone with a little more know-how than myself willing to help out?  I know a few other people in this thread were looking for something like this as well, and any help at all would be appreciated.  Thanks!

[JakeTri]

I was able to seperate the above frequencies into smaller incremental steps, but I have no idea how to calculate the "reg. value" or the "ideal delay" for them.  Anyone with a little more know-how than myself willing to help out?  I know a few other people in this thread were looking for something like this as well, and any help at all would be appreciated.  Thanks!

Adjusting the S1 formula based on S3 ASIC spec I was able to get following register values (and timeout). Use the settings below at your own risk!

Clock_Freq	Register value	Timeout
(chip_freq)	(freq_value)	(timeout)
100	0783	40
103.125	1007	39
106.25	0803	38
109.375	1107	37
112.5	0883	36
115.625	1207	35
118.75	0903	34
121.875	1307	33
125	0983	32
128.125	1407	31
131.25	0A03	30
134.375	1507	30
137.5	0A83	29
140.625	1607	28
143.75	0B03	28
146.875	1707	27
150	0B83	27
153.125	1807	26
156.25	0C03	26
159.375	1907	25
162.5	0C83	25
165.625	1A07	24
168.75	0D03	24
171.875	1B07	23
175	0D83	23
178.125	1C07	22
181.25	0E03	22
184.375	1D07	22
187.5	0E83	21
190.625	1E07	21
193.75	0F03	21
196.875	1F07	20
200	0782	20
206.25	1006	19
212.5	0802	19
218.75	1106	18
225	0882	18
231.25	1206	17
237.5	0902	17
243.75	1306	16
250	0982	16
256.25	1406	16
262.5	0A02	15
268.75	1506	15
275	0A82	15
281.25	1606	14
287.5	0B02	14
293.75	1706	14
300	0B82	13
306.25	1806	13
312.5	0C02	13
318.75	1906	13
325	0C82	12
331.25	1A06	12
337.5	0D02	12
343.75	1B06	12
350	0D82	11
356.25	1C06	11
362.5	0E02	11
368.75	1D06	11
375	0E82	11
381.25	1E06	10
387.5	0F02	10
393.75	1F06	10
400	0F82	10

[la7eralus]

I was able to seperate the above frequencies into smaller incremental steps, but I have no idea how to calculate the "reg. value" or the "ideal delay" for them.  Anyone with a little more know-how than myself willing to help out?  I know a few other people in this thread were looking for something like this as well, and any help at all would be appreciated.  Thanks!

Adjusting the S1 formula based on S3 ASIC spec I was able to get following register values (and timeout). Use the settings below at your own risk!

Clock_Freq	Register value	Timeout
(chip_freq)	(freq_value)	(timeout)
100	0783	40
103.125	1007	39
106.25	0803	38
109.375	1107	37
112.5	0883	36
115.625	1207	35
118.75	0903	34
121.875	1307	33
125	0983	32
128.125	1407	31
131.25	0A03	30
134.375	1507	30
137.5	0A83	29
140.625	1607	28
143.75	0B03	28
146.875	1707	27
150	0B83	27
153.125	1807	26
156.25	0C03	26
159.375	1907	25
162.5	0C83	25
165.625	1A07	24
168.75	0D03	24
171.875	1B07	23
175	0D83	23
178.125	1C07	22
181.25	0E03	22
184.375	1D07	22
187.5	0E83	21
190.625	1E07	21
193.75	0F03	21
196.875	1F07	20
200	0782	20
206.25	1006	19
212.5	0802	19
218.75	1106	18
225	0882	18
231.25	1206	17
237.5	0902	17
243.75	1306	16
250	0982	16
256.25	1406	16
262.5	0A02	15
268.75	1506	15
275	0A82	15
281.25	1606	14
287.5	0B02	14
293.75	1706	14
300	0B82	13
306.25	1806	13
312.5	0C02	13
318.75	1906	13
325	0C82	12
331.25	1A06	12
337.5	0D02	12
343.75	1B06	12
350	0D82	11
356.25	1C06	11
362.5	0E02	11
368.75	1D06	11
375	0E82	11
381.25	1E06	10
387.5	0F02	10
393.75	1F06	10
400	0F82	10

Awesome!  I'm testing the following now:

        option 'freq_value'    '1306'
        option 'chip_freq'     '243.75'
        option 'timeout'       '16'

One of my Ant's works fine at 250, however the other does not and had to be bumped down to the 237.5 to refrain from attaining X's.

This seems to be stable so far, but only time will tell as it took about an hour/2 for it to show X's the first time around.

Currently averaging 485-490 GH/s.

[MinersAnonymous]

Do you guys feel it is good to instantly remove the heatsinks and apply new thermal paste to the chips?

I am unsure of the quality paste used from manufacturer but I highly doubt it is premium due to production costs.

[la7eralus]

Do you guys feel it is good to instantly remove the heatsinks and apply new thermal paste to the chips?

I am unsure of the quality paste used from manufacturer but I highly doubt it is premium due to production costs.

I have been running both of my Ant's OC'd to 500 GH/s and 490 GH/s and both temperatures are well within normal ranges and all chips are stable and hashing with no X's.  One of them has been running a few hours as if you look up a post or two you'll see where we came up with the newer smaller incremental frequency steps, and that is the one running at 490 GH/s up from the 237.5 clock which was in the 470-480 GH/s range, while the other has been running for 72 hours at 500 GH/s.  

Highest temperature reached by any miner is 44C, while averaging 40-43 C in use, so I would say replacing the thermal compound is unnecessary.

That being said, if you do decide to go ahead and upgrade/replace the thermal compound, do not use any compound with Silver in it as it is conductive, and make absolutely sure that whatever brand you decide to go with DOES NOT CONDUCT ELECTRICITY - YOU WILL FRY YOUR ANTMINER!!!!!

[davinrover]

Hi Guys

I just got my antminer s3 and having big problems.

It's only mining at about 250ghs and gets around 75% hw error.

I am using a brand new Corsair CX500 and have also tried with my Corsair HX850.
I have changed the thermal paste on all the chips.
I have flashed it to latest firmware.
I have very stable internet connection.

Anyone have advice for me?

After 5min: http://i62.tinypic.com/2psln39.jpg
After 10min: http://i61.tinypic.com/28vaza.jpg
HW error: http://i58.tinypic.com/262qnmq.jpg

[MinersAnonymous]

Do you guys feel it is good to instantly remove the heatsinks and apply new thermal paste to the chips?

I am unsure of the quality paste used from manufacturer but I highly doubt it is premium due to production costs.

I have been running both of my Ant's OC'd to 500 GH/s and 490 GH/s and both temperatures are well within normal ranges and all chips are stable and hashing with no X's.  One of them has been running a few hours as if you look up a post or two you'll see where we came up with the newer smaller incremental frequency steps, and that is the one running at 490 GH/s up from the 237.5 clock which was in the 470-480 GH/s range, while the other has been running for 72 hours at 500 GH/s.  

Highest temperature reached by any miner is 44C, while averaging 40-43 C in use, so I would say replacing the thermal compound is unnecessary.

That being said, if you do decide to go ahead and upgrade/replace the thermal compound, do not use any compound with Silver in it as it is conductive, and make absolutely sure that whatever brand you decide to go with DOES NOT CONDUCT ELECTRICITY - YOU WILL FRY YOUR ANTMINER!!!!!

I see Yes my current have same temperature, from some pictures I have seen of heatsinks removed I noticed very bad application of the paste. I don't expect too much from the assembly line.

Silly question but I must ask, outside of trial and error how are you coming up with the newer reg codes you posted above? Should I completely ignore the first page and use the recent posting?

[la7eralus]

Do you guys feel it is good to instantly remove the heatsinks and apply new thermal paste to the chips?

I am unsure of the quality paste used from manufacturer but I highly doubt it is premium due to production costs.

I have been running both of my Ant's OC'd to 500 GH/s and 490 GH/s and both temperatures are well within normal ranges and all chips are stable and hashing with no X's.  One of them has been running a few hours as if you look up a post or two you'll see where we came up with the newer smaller incremental frequency steps, and that is the one running at 490 GH/s up from the 237.5 clock which was in the 470-480 GH/s range, while the other has been running for 72 hours at 500 GH/s.  

Highest temperature reached by any miner is 44C, while averaging 40-43 C in use, so I would say replacing the thermal compound is unnecessary.

That being said, if you do decide to go ahead and upgrade/replace the thermal compound, do not use any compound with Silver in it as it is conductive, and make absolutely sure that whatever brand you decide to go with DOES NOT CONDUCT ELECTRICITY - YOU WILL FRY YOUR ANTMINER!!!!!

I see Yes my current have same temperature, from some pictures I have seen of heatsinks removed I noticed very bad application of the paste. I don't expect too much from the assembly line.

Silly question but I must ask, outside of trial and error how are you coming up with the newer reg codes you posted above? Should I completely ignore the first page and use the recent posting?

I separated out the operating speed's into incremental steps and JakeTri calculated the appropriate Freq values and Timings.