Stock S3+ voltage

[Jarredm]

So I recently aquired a couple used S3+ miners.  One of them I have had running for about 5 days and a couple of times one of the hash boards just stopped responding.  The interface showed all o's for its asics, but the temp dropped to room temp and the hash rate of the unit was cut in half.  The first reboot showed that hash board missing completely, but after fully removing power and starting back up, both boards came online and were hashing at full speed.  

After a little playing I found that it reboots a little more reliably at 218.75M frequency which is stock for the s3 whereas the s3+ stock is at 225.  When I first got this miner running, the previous owner had it running at 231.25 and had the voltage setting at 0785.  So my main question is.  What is the stock from factory voltage setting for the S3+?  Secondly, any tips on what steps to take while troubleshooting this issue?  The strange thing is that the fan speed is still being reported to the controller after the board stops hashing.

[1715492737]

1715492737

[adaseb]

On most S3 you can't actually adjust the voltage. Open up the unit and measure with a multimeter and it will be 0.78V for each.

Changing in the software, won't change it physically.

[RichBC]

As adaseb says the voltage is not adjustable, the setting is a left over from another miner. The S3+ is usually a very well behaved Miner and can often be run at a higer than standard frequency quite successfully, however not unheard of that the frequency needs to be dropped a bit. Is it 100% reliable at the lower frequency?


Rich

[adaseb]

What type of power supply are you using?

[Jarredm]

Hello all and thanks for the replys.

I have seen it posted before that the s3+ voltage setting in the webgui didn't do anything.  However, I've also seen contradictory information.  Most notably here: https://bitcointalk.org/index.php?topic=883197.0  So I guess I don't know what to believe.


Interesting that you ask about the power-supply abaseb.  I had not thought about this until you said it.  I have three of these miners but two power supplys.  Both PSUs are rated for 750W.  One is an XFX brand, the other is Seasonic.  The miner in question here is the one that is drawing power on one blade from Seasonic PSU and the other blade from the XFX.  Based on how I have them arranged, the board that seems to be having the trouble is the one attached to the XFX.

As for testing the voltage, can you point me to an image of what contact points should be used to test the voltage?  Lastly, based on your response, would 0.780 V (0780) be the stock setting?

[RichBC]

Hello all and thanks for the replys.

I have seen it posted before that the s3+ voltage setting in the webgui didn't do anything.  However, I've also seen contradictory information.  Most notably here: https://bitcointalk.org/index.php?topic=883197.0  So I guess I don't know what to believe.

Believe me, the S3 does not have software adjustable voltage. There is a version of the C1 hashboard that is adjustable, but I do not think it was ever fitted to the S3? However you can prove it for yourself by measuring, see below.

As for testing the voltage, can you point me to an image of what contact points should be used to test the voltage?  Lastly, based on your response, would 0.780 V (0780) be the stock setting?

Easiest place to mesure the core voltage is across one of the 8 Large 1500uF decoupling capacitors on the Hash board.



Not sure that there is a standard voltage, but you should see about 0.73V. You can then play with the software adjustment and confirm that the voltage does not change. 


Rich

[pinhead666]

Propably psu related problem. They can cause weird behavior sometimes.

[pekatete]

I agree, this is probably a PSU related problem.
The GUI voltage setting does not adjust the voltage at the board level (as / and there is no circuitry for that), rather at the chip level, and it is NOT a relic of other machines.
The stock / default setting for voltage is 0725 (or 0000 aka autovoltage), but the voltage setting will not magically cure an inadequately powered machine. Also remember, the voltage setting should be set according to the frequency per datasheet.

[RichBC]

I agree, this is probably a PSU related problem.
The GUI voltage setting does not adjust the voltage at the board level (as / and there is no circuitry for that), rather at the chip level,

Would you care to explain how it's possible to adjust at the Chip level but not the Board level? Does not make sense to me.

Rich

[pekatete]

I agree, this is probably a PSU related problem.
The GUI voltage setting does not adjust the voltage at the board level (as / and there is no circuitry for that), rather at the chip level,

Would you care to explain how it's possible to adjust at the Chip level but not the Board level? Does not make sense to me.

Rich

Look at the cgminer code for S3 (or S2 for that matter) and the datasheet for the chip. Even better, do away with your multimeter and run the rig with different voltage settings and you'll definitely see the change in hash-rate and / or HW.

[RichBC]

I agree, this is probably a PSU related problem.
The GUI voltage setting does not adjust the voltage at the board level (as / and there is no circuitry for that), rather at the chip level,

Would you care to explain how it's possible to adjust at the Chip level but not the Board level? Does not make sense to me.

Rich

Look at the cgminer code for S3 (or S2 for that matter) and the datasheet for the chip. Even better, do away with your multimeter and run the rig with different voltage settings and you'll definitely see the change in hash-rate and / or HW.

I have run with the full rage of voltage settings on an S3 & S3+ with no noticeable difference. What we are agreed on is that the voltage applied to the Core of the chip is not adjustable / measurable at the board level. What I think you are suggesting is that internal to the BM1382 chip the core voltage is adjustable?

Frankly this is not possible and if by some miracle it was then this would be evident by a change to the Power consumption / efficiency which I have also never seen reported?


Rich

[pekatete]

I have run with the full rage of voltage settings on an S3 & S3+ with no noticeable difference. What we are agreed on is that the voltage applied to the Core of the chip is not adjustable / measurable at the board level. What I think you are suggesting is that internal to the BM1382 chip the core voltage is adjustable?

Frankly this is not possible and if by some miracle it was then this would be evident by a change to the Power consumption / efficiency which I have also never seen reported?


Rich

You are lying through your teeth if you say you have run the full range of voltage settings without a noticeable difference to the hash and / or HW, an absolute lie! No difference to your multimeter readings I'd also argue is impossible, and remember to run the different vltages at the respective recomended frequencies in the datasheet too.

I take it you have looked at the cgminer code, and also looked at the datasheets(s) from bitmain to come to your evidenced conclusions. But to be absolutely & evidently certain that I am right, you'll have to de-assemble the S3 controller chip firmware (I can get you the raw file if you can't get it).

Suffice to say, the voltage setting in the UI is neither there by mistake nor a relic from another rig, if only that it was actually shipped at a latter date than the original S3 firmware.

[RichBC]

I have run with the full rage of voltage settings on an S3 & S3+ with no noticeable difference. What we are agreed on is that the voltage applied to the Core of the chip is not adjustable / measurable at the board level. What I think you are suggesting is that internal to the BM1382 chip the core voltage is adjustable?

Frankly this is not possible and if by some miracle it was then this would be evident by a change to the Power consumption / efficiency which I have also never seen reported?


Rich

You are lying through your teeth if you say you have run the full range of voltage settings without a noticeable difference to the hash and / or HW, an absolute lie! No difference to your multimeter readings I'd also argue is impossible, and remember to run the different vltages at the respective recomended frequencies in the datasheet too.

I take it you have looked at the cgminer code, and also looked at the datasheets(s) from bitmain to come to your evidenced conclusions. But to be absolutely & evidently certain that I am right, you'll have to de-assemble the S3 controller chip firmware (I can get you the raw file if you can't get it).

Suffice to say, the voltage setting in the UI is neither there by mistake nor a relic from another rig, if only that it was actually shipped at a latter date than the original S3 firmware.

If I say I have tried a range of voltages that is what I have done, why on earth would I lie?

Just to give some history I really wanted this to work and when I first bought an S3 & S3+ I read through your thread several times as I was keen to Undervolt to improve efficiency. I then tried all the versions of firmware I could find and several voltage settings. I was not able to see a change in the Core Voltage or at the wall power at any given frequency when varying the voltage and therefore concluded that the voltage was not changing.

In your thread all I can find is anecdotal information about miner improvements to hash rate / stability, nothing about improved efficiency and then several people also reporting that they could see no measurable improvement.

So at that point I gave up on attempting to adjust from within the software and added a Digital Pot to each stage controlled by an Arduino which gave the desired significant improvement to the efficiency.

The code in cgminer just writes to a Bitmain interface and beyond that we do not know what it does? In the case of the S3 I still think that is nothing at all.

So returning to this thread and the OP then please try adjusting the software voltage and report on what you find?

Rich

[pekatete]

^^^
Honestly, you were looking to under-clock / overlock and that has nothing at all to do with whether the voltage setting will offer you a solution. Saying that, it does not mean the voltage setting is useless, and I can without doubt state that it definitely affects the hash and / or HW depending on your frequency setting.

And no, I still do not believe you run the different voltage settings to establish whether it had an effect on hashing / HW, maybe to look for your under / over volting, thus you are lying through your teeth! You can try it now, do a 10 minute run on any frequency (stock or above) with 2 different voltage settings (excluding 0000).

The OP's issue is most likely due to the way he's powering the rig in conjunction with an auto voltage setting. I'd say set the voltage to the recomended datasheet setting for the frequency he's running, then if the board "falls off" again, then it is without doubt a power issue.

[RichBC]

^^^
Honestly, you were looking to under-clock / overlock and that has nothing at all to do with whether the voltage setting will offer you a solution. Saying that, it does not mean the voltage setting is useless, and I can without doubt state that it definitely affects the hash and / or HW depending on your frequency setting.

And no, I still do not believe you run the different voltage settings to establish whether it had an effect on hashing / HW, maybe to look for your under / over volting, thus you are lying through your teeth! You can try it now, do a 10 minute run on any frequency (stock or above) with 2 different voltage settings (excluding 0000).

The OP's issue is most likely due to the way he's powering the rig in conjunction with an auto voltage setting. I'd say set the voltage to the recomended datasheet setting for the frequency he's running, then if the board "falls off" again, then it is without doubt a power issue.

I am sorry but adjusting the Core Voltage has everything to do with the Frequency you can operate the chip at, the HW error rate & the efficiency that can be achieved. This as you have pointed out is clearly covered in the data sheet, and when I had added the digital pots was confirmed and backed up by measurements.

Also, with respect, the community on this Forum is very knowledgeable and quick to pick up on anything that will improve the efficiency of a Miner  and other than your thread there is no other mention of the Voltage setting actually working, which I find strange?

I have no intention of trying voltage setting from software again as I at the time spent many hours trying to get it to work. I concluded that either it does not work or that there are a only few S3 that it does work for and that mine was not one of those.

Please stop calling me a liar, which I can assure you I am not....

Rich

[pekatete]

You are a liar because you crystallised your argument based on your search for an under-volt efficiency, not on whether the voltage setting can improve the efficiency of the rig depending on which frequency you are running, and that is where you went wrong. That you adamantly peddle that as fact in direct contradiction with even the datasheet, makes you an even bigger liar.

But we digress, OP's issue, like he stated, is about how they are powering the problem rig. If you want to split hairs as to which diodes you looked at and what the readings of your multimeter were when you were looking to undervolt, then I'll leave you to it.

[RichBC]

You are a liar because you crystallised your argument based on your search for an under-volt efficiency, not on whether the voltage setting can improve the efficiency of the rig depending on which frequency you are running, and that is where you went wrong. That you adamantly peddle that as fact in direct contradiction with even the datasheet, makes you an even bigger liar.

But we digress, OP's issue, like he stated, is about how they are powering the problem rig. If you want to split hairs as to which diodes you looked at and what the readings of your multimeter were when you were looking to undervolt, then I'll leave you to it.

The OP raised voltage setting, all adseb and I pointed out was that is does not actually change the core voltage. If the S3 was not well past it's sell by date I would persist in the discussion, however as it's now academic I will call it a Day on the S3 Voltage setting discussion.

If anyone else want's to chip in with their experiences of software voltage setting on an S3 then please do so. I for one have had my fill of being called a liar in every other paragraph...


Rich

[pekatete]

You are a liar because you crystallised your argument based on your search for an under-volt efficiency, not on whether the voltage setting can improve the efficiency of the rig depending on which frequency you are running, and that is where you went wrong. That you adamantly peddle that as fact in direct contradiction with even the datasheet, makes you an even bigger liar.

But we digress, OP's issue, like he stated, is about how they are powering the problem rig. If you want to split hairs as to which diodes you looked at and what the readings of your multimeter were when you were looking to undervolt, then I'll leave you to it.

The OP raised voltage setting, all adseb and I pointed out was that is does not actually change the core voltage. If the S3 was not well past it's sell by date I would persist in the discussion, however as it's now academic I will call it a Day on the S3 Voltage setting discussion.

If anyone else want's to chip in with their experiences of software voltage setting on an S3 then please do so. I for one have had my fill of being called a liar in every other paragraph...


Rich

If you do not want to be called a liar, do not lie, simple.
The datasheet is quite clear on the recomended voltage setting for different frequencies, so persisting with your lies that the voltage setting has no use, while saying in the next breath that you do not know what the chip actually does with the settings from cgminer, simply cements your assertions as pure lies (and which liar wouldn't have their fill if they were found out?).

The voltage setting, like I stated, will not resurect an inadequately powered S3, so the OP has to ensure they are powering their rig adequately before even being concerned about the voltage setting. Matching the frequency they are running at with the recomended voltage setting (or a touch higer) will definitely reduce HW and / or improve hash speed.

EDIT:
Here's one example, properly powered with a Dell N750P-S0 server PSU with 4 custom made 16AWG 6 pin PCI-e cables, that's been running for a couple of weeks with the voltage setting matched to that recomended for the frequency in the datasheet.



I usually let these run for months on end without restarting save for when the PSU's fans die (common with these PSU's running in a shed in the garden) and the PSU shuts down.