Topic: Antminer S9 Volt Rocket Ship firmware mod! ~Time to Max Out Your Bitmain PSU~ - page 7. (Read 32694 times)
October 25, 2018, 04:02:22 PM
A few people have reported here https://bitcointalk.org/index.php?topic=5055138.60 about a decaying hash rate after some period of time with the LPM patch from Bitmain so I've decided to do a 12 hour test. Not sure when I'll be here to read the measurements but I will try to do at least a 3 hour, 6 hour, and 12 hour. So far 2 hours in I'm reading 13.607 th/s average at 1,177 watts at the PSU, assuming close to 1,273 watts at the wall.
October 25, 2018, 03:15:54 PM
By the way the last official firmware for S9 gives minus 200 watt with the same (default) speed!
You mean autotune?
October 25, 2018, 02:48:33 PM
By the way the last official firmware for S9 gives minus 200 watt with the same (default) speed!
October 25, 2018, 12:46:54 PM
I did mention that in the above post. The test was relevant to me so I could confirm that LPM had no effect on the software (which is something I had already assumed and wanted to verify). If anyone wants to argue that the 10 minute test is irrelevant I would agree with them.
October 25, 2018, 12:31:23 PM
Same 10 minute limited test. LPM makes no difference in BMMINER880 firmware, it would probably show closer numbers with a longer test.
[...]
[...]
Asic boost isn't enabled within rocketships bmminer.. So your lpm tests are completely irrelevant
October 25, 2018, 08:13:47 AM
Same 10 minute limited test. LPM makes no difference in BMMINER880 firmware, it would probably show closer numbers with a longer test.
Code:
FIRMWARE LPM FREQ VOLTS AMPS WATTS GH/S RT GH/S AVG W/TH
BMMINER 880 NO LPM 550 12.16 91.2 1,109 11775 11530 96.2
BMMINER 880 LPM 550 12.16 91.9 1,118 11764 11815 94.6
BMMINER 880 LPM 625 12.15 102.5 1,245 13226 13001 95.8
BMMINER 880 LPM 750 12.15 124 1,507 14184 14370 104.8
BMMINER 880 NO LPM 750 12.15 123.6 1,502 14140 14176 105.9
BMMINER 880 NO LPM 600 12.15 100.2 1,217 12729 12892 94.4
BMMINER 880 NO LPM 550 12.16 91.2 1,109 11775 11530 96.2
BMMINER 880 LPM 550 12.16 91.9 1,118 11764 11815 94.6
BMMINER 880 LPM 625 12.15 102.5 1,245 13226 13001 95.8
BMMINER 880 LPM 750 12.15 124 1,507 14184 14370 104.8
BMMINER 880 NO LPM 750 12.15 123.6 1,502 14140 14176 105.9
BMMINER 880 NO LPM 600 12.15 100.2 1,217 12729 12892 94.4
October 25, 2018, 05:11:42 AM
Thats good!
But did we find out what happens, if you replace bmminer again with taserz version?
I saw a drop in Watts but also in hash. I only upgraded without replacing the bmminer again.
But did we find out what happens, if you replace bmminer again with taserz version?
I saw a drop in Watts but also in hash. I only upgraded without replacing the bmminer again.
October 25, 2018, 01:41:23 AM
I have limited results as I only have access to a few machines while away from my data center. I was doing power measurements at the power supply tonight on 1 machine and only letting it hash for 10 minutes before taking readings. Machine is an April 2017 delivery Bitmain S9 13.5 Th/s, power supply is a Bitmain APW3+-12-1600-A3, also delivered during that time. Voltmeter and Ammeter are both Fluke. Again I want to stress that this is only 1 machine and very limited testing since I can't wait all day to take readings. This power supply is about 92.5% efficient at 100 amps according to this post https://bitcointalksearch.org/topic/reference-power-efficiency-for-apw3-apw3-and-apw7-5045057. Numbers are right around a 12.5% power savings and an overall efficiency improvement of 14.5%. I've only updated 3 machines so far and have have hundreds to go as well. Hoping to save about $2,500 in power costs per month with this firmware.
Nov 17 Firmware 12.15 vdc 110.6 amps 1,344 watts 13.226 th/s avg 101.6 w/th at the wall estimated ~1,453 watts
LPM Patch and checked 12.16 vdc 96.7 amps 1,176 watts 13.554 th/s avg 86.8 w/th at the wall estimated ~1,271 watts
LPM Patch and unchecked 12.15 vdc 110.5 amps 1,343 watts 13.424 th/s avg 100.0 w/th at the wall estimated ~1,452 watts
Nov 17 Firmware 12.15 vdc 110.6 amps 1,344 watts 13.226 th/s avg 101.6 w/th at the wall estimated ~1,453 watts
LPM Patch and checked 12.16 vdc 96.7 amps 1,176 watts 13.554 th/s avg 86.8 w/th at the wall estimated ~1,271 watts
LPM Patch and unchecked 12.15 vdc 110.5 amps 1,343 watts 13.424 th/s avg 100.0 w/th at the wall estimated ~1,452 watts
October 24, 2018, 10:21:27 PM
Some say it is not significant tuning the chips individually. I dont think so and I would like to try it. As House Xylander said, there is an interface, that checks the chips and shows them all coloured from green, to yellow, to red. Then you can pitch up the frequency for all green chips or each one by yourself etc.
I think you can decrese the HW errors by letting the poor chips hash slower and the good ones faster. That should increse the effectivity.
But unfortunately the only firmware I have found so far that supports this, has a dev fee included. So if someone has one without dev fee, that would be great. although now it would be interesting to get a firmware, that also supports asicboost.
I think you can decrese the HW errors by letting the poor chips hash slower and the good ones faster. That should increse the effectivity.
But unfortunately the only firmware I have found so far that supports this, has a dev fee included. So if someone has one without dev fee, that would be great. although now it would be interesting to get a firmware, that also supports asicboost.
It's not worth using it the dev fee is like 6% and the gain is little to none.
October 24, 2018, 07:56:37 PM
Some say it is not significant tuning the chips individually. I dont think so and I would like to try it. As House Xylander said, there is an interface, that checks the chips and shows them all coloured from green, to yellow, to red. Then you can pitch up the frequency for all green chips or each one by yourself etc.
I think you can decrese the HW errors by letting the poor chips hash slower and the good ones faster. That should increse the effectivity.
But unfortunately the only firmware I have found so far that supports this, has a dev fee included. So if someone has one without dev fee, that would be great. although now it would be interesting to get a firmware, that also supports asicboost.
I think you can decrese the HW errors by letting the poor chips hash slower and the good ones faster. That should increse the effectivity.
But unfortunately the only firmware I have found so far that supports this, has a dev fee included. So if someone has one without dev fee, that would be great. although now it would be interesting to get a firmware, that also supports asicboost.
October 24, 2018, 07:47:21 PM
I missed the fine details of that post, the part about "each chip" individually. Wouldn't that be a pain to do and monitor? Other than finite control, what benefit would that offer? I'm not trying to be argumentative, I'm legitimately asking if there is a benefit to this, cause if there is I might be interested.
I ran it through the Auto-Tune first and found the range that most of the chips showed green on the manual screen. After that, it was just a few chips that would show "red" or "crimson" (if the voltage and Freq where to high for the chip I would drop it 25-50m). When all the chips showed green they would typically have a value of 63-78. Sometimes during a hash a few would turn yellow and then back to green.
Just a hunch but I'm guessing that the red and crimson chip where the ones causing the most HW errors when setting it over 637m. That is why I listed the range 625-675m
October 24, 2018, 07:39:15 PM
Yes, I was looking for such thing! Is it really without dev fee? Can you tell me where to get it? Thanks in advance
If memory serves it was shared on these forums, I try looking when I get home and have a full PC at hand.
October 24, 2018, 12:04:45 PM
I missed the fine details of that post, the part about "each chip" individually. Wouldn't that be a pain to do and monitor? Other than finite control, what benefit would that offer? I'm not trying to be argumentative, I'm legitimately asking if there is a benefit to this, cause if there is I might be interested.
October 24, 2018, 12:02:22 PM
I could be mistaken, but doesn't Tazers mod (this forum) allow you to adjust frequency with no dev fee? I'm pretty sure Tazer is doing this simply because he is a cool dude (or chick?) and to stuff it in Bitmains face.
Yes it does, but not by each chip individually. Do you have a no dev fee firmware that can modify the chips individually?
October 24, 2018, 10:20:05 AM
Code:
NONCE2_AND_JOBID_STORE_ADDRESS
#define JOB_ID_OFFSET (0x0/sizeof(int))
#define HEADER_VERSION_OFFSET (0x4/sizeof(int))
#define NONCE2_L_OFFSET (0x8/sizeof(int))
#define NONCE2_H_OFFSET (0xc/sizeof(int))
#define JOB_ID_OFFSET (0x0/sizeof(int))
#define HEADER_VERSION_OFFSET (0x4/sizeof(int))
#define NONCE2_L_OFFSET (0x8/sizeof(int))
#define NONCE2_H_OFFSET (0xc/sizeof(int))
This is in btmc5 driver and is the masks the boost uses if you can figure out the usage
October 24, 2018, 08:54:55 AM
Yes, I was looking for such thing! Is it really without dev fee? Can you tell me where to get it? Thanks in advance
I could be mistaken, but doesn't Tazers mod (this forum) allow you to adjust frequency with no dev fee? I'm pretty sure Tazer is doing this simply because he is a cool dude (or chick?) and to stuff it in Bitmains face.
October 24, 2018, 08:37:33 AM
EDIT: I should say though, you do save ~10-13% in electricity but i've noticed about a 3-5% drop in hash rate as well.
Not trying to question you or "call you out", but you haven't measured the post LPM power usage at the wall? Please share what the source might be of the 1250 watts and the 3-5% drop in Hash rate claim.
I am slowly converting S9's over to LPM and I've actually noticed a slight increase in "avg Hashrate" over 24 hours and above. I've not had the opportunity to measure power at the wall yet, but if what you say is true, "10-13% power savings and my stable to slightly higher, avg Hash Rate....over 100's of S9's, this is significant.
October 24, 2018, 08:24:42 AM
I am unsure of the name of the firmware but it is the one that allows you to adjust the freq(?) on each chip.
Yes, I was looking for such thing! Is it really without dev fee? Can you tell me where to get it? Thanks in advance
October 24, 2018, 07:38:49 AM
whats that?
I am unsure of the name of the firmware but it is the one that allows you to adjust the freq(?) on each chip.
October 24, 2018, 06:58:56 AM
Jump to: