sidehack (OP)
Legendary
Offline
Activity: 3374
Merit: 1859
Curmudgeonly hardware guy
|
|
June 13, 2016, 11:55:00 PM |
|
Also, I've settled on terms for the group buy so that's live now.
|
|
|
|
adaseb
Legendary
Offline
Activity: 3878
Merit: 1733
|
|
June 14, 2016, 06:19:51 AM |
|
This reminds me of the S4 I once had. I got it really cheap and it was crazy loud, and I had to have it in my room that I sleep in.
I remember I managed to make it almost silent at 1.3TH/s running at around 0.55W/GHS and I swapped in some S3 fans and played around with the fan controller settings. You could actually change the voltage in the control panel. No need to pencil mod.
The loudest part was actually that stock server PSU. Sure miss that miner.
Its amazing how with an undervolt and underclock you can turn the loudest miner into something manageable.
|
|
|
|
sidehack (OP)
Legendary
Offline
Activity: 3374
Merit: 1859
Curmudgeonly hardware guy
|
|
June 14, 2016, 12:24:26 PM |
|
I got a mostly-working S4 for free end of last year and gave it the same treatment to heat my apartment over the winter. Well, I didn't replace the fans but they did run a lot slower.
|
|
|
|
sidehack (OP)
Legendary
Offline
Activity: 3374
Merit: 1859
Curmudgeonly hardware guy
|
|
June 14, 2016, 04:28:28 PM |
|
Inbetween things today, I'll be getting some operating data from different voltages and clock setpoints. I've got a stack of pregenerated hex files for different hot/cold settings and when I put up the analysis numbers I'll also post a ZIP file with all of 'em in it within a few days.
Right now it looks like 750MHz might be just a bit high for the 720/690 profile ( .03% errors) for 3.38TH from 900W at 61C, but that's still pretty decent. I think the cold-running is about the same setpoint as a stock full S7.
|
|
|
|
Sweminer777
|
|
June 16, 2016, 12:34:13 PM |
|
Inbetween things today, I'll be getting some operating data from different voltages and clock setpoints. I've got a stack of pregenerated hex files for different hot/cold settings and when I put up the analysis numbers I'll also post a ZIP file with all of 'em in it within a few days.
Right now it looks like 750MHz might be just a bit high for the 720/690 profile ( .03% errors) for 3.38TH from 900W at 61C, but that's still pretty decent. I think the cold-running is about the same setpoint as a stock full S7.
this 2 boards s7 lookssimilar to Batch6 boards. I was looking around, took some photos too. I can overclock abit if iset up the voltmetere to around 12.120Vis good from PSU else it gets alot of errors. mine hashs 4050, but is doing around 4200-4400 wich is fine considering that b6 is not overclockable. ill post some pictures soonof my boards så you can take a look on the similarity. regards.
|
|
|
|
RichBC
|
|
June 17, 2016, 10:00:04 AM |
|
I have got hold of a Hash Board from a Batch 8 135 Chip S7. Have it running on the Bench with an S5 Controller Board loaded with the S7 Firmware. I have been playing with the various different firmware versions and noticed something interesting with the latest single fan release. S7firmware-201605231558-700M-1fan-6000.tar.gz MD5:7A38217ED4B93D662F4F0EBB85006694 Immediately after power up the Pic programs the Digital pot, with my board this is with 14 which sets the string Voltage to 10.667V. However with the latest single Fan firmware I detected on the Analyser a second programming of the Pot about 37 seconds after booting. This was also to 14. Looking at the Kernel Log and also monitoring on the Serial Interface of the BBB I see the following. [ 36.708381] clear FPGA nonce buffer.. [ 37.014489] btm_tx_conf.. [ 37.017098] Set asic frequency {100}.. [ 37.020850] bitmain_set_pic_voltage: n = 14.. [ 37.025232] set_pic_voltage cmd_buf[0]: 0xab, cmd_buf[1]: 0xb0, cmd_buf[2]: 0xe, cmd_buf[3]: 0xd4.. [ 37.034503] send BC data:..0x0000: 0x03 0x00 0x00 0x1a 0xab 0xb0 0x0e 0xd4 .. [ 37.056664] send BC data:..0x0000: 0x03 0x04 0x00 0x1a 0xab 0xb0 0x0e 0xd4 . So it looks like Bitmain might be using the the TX connection to the Pic to pass a Digital Pot setting after booting. Looking on the Analyser there is a Burst of data on the TX, which is also echo'd 100uS later on RA4 which is connected to the 2 Pin P2 connector on the board edge. Then 4.8mS later the Pot is programmed with 14. Unfortunately as my system programs 14 at power up and then 14 again I cannot tell if the Pic is just repeating it's inbuilt value or passing the new value. I do not know if there will be a file that can be SSH into with this 14 Value in or if it's embedded in some Bitmain code. I cannot find a file, but if anyone knows their way around the firmware perhaps they could have a look? Rich
|
|
|
|
sidehack (OP)
Legendary
Offline
Activity: 3374
Merit: 1859
Curmudgeonly hardware guy
|
|
June 17, 2016, 12:13:52 PM |
|
Not unexpected. What's the voltage setting given in the cgminer.conf for that one?
Today's a short day, and also a fairly full one, so I'm not sure if I'll have time to play with this info yet. But I'll work on it.
|
|
|
|
RichBC
|
|
June 17, 2016, 12:34:19 PM |
|
cgminer.conf shows "bitmain-voltage" : "0706" which I suspect is not used, but could try changing it?
Rich
|
|
|
|
sidehack (OP)
Legendary
Offline
Activity: 3374
Merit: 1859
Curmudgeonly hardware guy
|
|
June 17, 2016, 12:43:00 PM |
|
Given that's the same number I've seen everywhere else, and have seen several different numbers written by the PIC, I doubt it's what we're looking for. But I could be wrong.
Also, when you give the values written by the PIC here and in the S7 repair thread, are you giving hex or decimal?
|
|
|
|
RichBC
|
|
June 17, 2016, 12:47:31 PM |
|
Yes I have seen that number before, tried changing it but 0x14 still written. The written value is in Hex, the 14 from the kernel log I had at first assumed to be decimal but must also be hex. Not sure what I have put in the repair log, whatever it is could be either?
Rich
|
|
|
|
sidehack (OP)
Legendary
Offline
Activity: 3374
Merit: 1859
Curmudgeonly hardware guy
|
|
June 17, 2016, 06:28:25 PM |
|
Say, possibly dumb question. Where's the kernel log on these?
|
|
|
|
alh
Legendary
Offline
Activity: 1846
Merit: 1052
|
|
June 17, 2016, 06:31:01 PM |
|
Say, possibly dumb question. Where's the kernel log on these?
On most Linux based systems, /var/log/messages is it. That would have Kernel and possibly other messages. I expect cgminer to place it's logs elsewhere, though I don't really know.
|
|
|
|
sidehack (OP)
Legendary
Offline
Activity: 3374
Merit: 1859
Curmudgeonly hardware guy
|
|
June 17, 2016, 06:33:00 PM |
|
Not sure if I'm doing something wrong, but I'm getting nothing out of /var/log is why I'm asking.
EDIT - perhaps at some point I should remember it's viewable in the webconfig.
|
|
|
|
HagssFIN
Legendary
Offline
Activity: 2464
Merit: 1710
Electrical engineer. Mining since 2014.
|
|
June 17, 2016, 06:45:22 PM |
|
.. EDIT - perhaps at some point I should remember it's viewable in the webconfig.
Yeah. My S7 (B9) shows nothing in the kernel log when I'm viewing it with Chrome, but with Microsoft Edge it works ok.
|
|
|
|
sidehack (OP)
Legendary
Offline
Activity: 3374
Merit: 1859
Curmudgeonly hardware guy
|
|
June 17, 2016, 06:50:34 PM |
|
So I just put that firmware on a stock S7. I didn't probe the I2C lines but measuring the voltage says it should have been about 0x10 (hex 16). After putting that firmware on, now I'm seeing a voltage corresponding to 0x0E (hex 14) so I reckon it actually is doing the update.
|
|
|
|
RichBC
|
|
June 17, 2016, 07:22:14 PM |
|
I get the kernel log in under System, seems to show initially but then of often goes blank. So good news that it seems to take the Parameter. Just need to find where it is stored so we can change it? Here is the full Data sent from RA4 which is connected to P2 on the board edge. 12:06:10.177 73 0E FF FF FF FF FF FF s.ÿÿÿÿÿÿ 12:07:16.080 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................................................. 12:07:16.096 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ............... 12:07:16.137 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 01 00 00 00 00 00 00 00 00 00 00 00 00 ................................................................ 12:07:16.178 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 02 00 00 00 00 00 00 00 00 00 00 00 00 ................................................................ 12:07:16.217 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 03 00 00 00 00 00 00 00 00 00 00 00 00 ................................................................ 12:07:16.257 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 04 00 00 00 00 00 00 00 00 00 00 00 00 ................................................................ 12:07:16.273 FE þ 12:07:16.289 24 24 24 $$$ 12:07:16.305 24 24 $$ 12:07:16.321 24 24 24 $$$ 12:07:16.337 24 24 $$ 12:07:16.353 24 $ 12:07:38.305 84 00 00 11 „... 12:07:39.678 AB B0 0E D4 «°.Ô 12:07:39.694 14 33 The first line is sent soon after Power up then the others at the times shown. They are obviously filtering the TX in some way, can only assume it is there to help debug or who knows? Rich
|
|
|
|
sidehack (OP)
Legendary
Offline
Activity: 3374
Merit: 1859
Curmudgeonly hardware guy
|
|
June 17, 2016, 07:29:37 PM Last edit: June 17, 2016, 07:56:07 PM by sidehack |
|
Right. I'll spend a bit of time trying to track down the command to TX that info. Not a lot, since I still got stuff to do and it's BURGER NIGHT.
Also thinking about it, that cmd_buf looked familiar. It's in driver-bmsc in what currently exists of bitmain's cgminer fork code, and in Icarus in U3 code. So, it's probably tied into cgminer itself and not in a script somewhere. But where is it stashed in a config? Hm...
|
|
|
|
sidehack (OP)
Legendary
Offline
Activity: 3374
Merit: 1859
Curmudgeonly hardware guy
|
|
June 17, 2016, 09:12:36 PM Last edit: June 22, 2016, 11:51:22 PM by sidehack |
|
Decided to summarize some of the mass of data down there in a handy graph. Note the bars represent a "maximum stable GH" for each voltage level; there is none for 720mV because it was still good at 750MHz and I didn't want to take it any higher for fear of blowing the regulator. The operating frequency can be derived from the hashrate by dividing by 4.5 So here's a not quite complete but pretty extensive chart of measured performance from my hacked S7LN. Columns: mmm/nnn - hot starting voltage/cold running voltage xxx - operating frequency Calc GH - Calculated GH of two 45-chip boards at running frequency VAC, IAC - voltage and current measurements in cold running WAC - Watts AC, power measure during cold running (calculated value) Calc J/GH - Calculated J/GH based on Calc GH and WAC HW% - percentage of HW errors after 30 minutes runtime GH - 30min - Measured average GH after 30 minutes runtime Hash Eff - Measured GH as a percentage of Calc GH Actual J/GH - calculated from WAC and 30min measured GH 720/720 | Calc GH | VAC | IAC | WAC | Calc J/GH | HW% | GH - 30min | Hash Eff | Actual J/GH | 750 | 3375 | 116 | 8.8 | 1021 | 0.302 | 0.000 | 3383 | 1.00 | 0.302 | 725 | 3262.5 | 115 | 8.6 | 989 | 0.303 | 0.000 | 3250 | 1.00 | 0.304 | 700 | 3150 | 115 | 8.25 | 949 | 0.301 | 0.000 | 3165 | 1.00 | 0.300 | 675 | 3037.5 | 115 | 7.9 | 909 | 0.299 | 0.000 | 3035 | 1.00 | 0.299 | 650 | 2925 | 115 | 7.6 | 874 | 0.299 | 0.000 | 2889 | 0.99 | 0.303 | 710/710 | | | | | | | | | | | 775 | 3487.5 | 115 | 8.85 | 1018 | 0.292 | 0.008 | 3453 | 0.99 | 0.295 | 750 | 3375 | 115 | 8.5 | 978 | 0.290 | 0.001 | 3326 | 0.99 | 0.294 | 725 | 3262.5 | 116 | 8.1 | 940 | 0.288 | 0.000 | 3256 | 1.00 | 0.289 | 700 | 3150 | 117 | 7.8 | 913 | 0.290 | 0.000 | 3109 | 0.99 | 0.294 | 675 | 3037.5 | 117 | 7.45 | 872 | 0.287 | 0.000 | 3060 | 1.01 | 0.285 | 650 | 2925 | 117 | 7.2 | 842 | 0.288 | 0.000 | 2935 | 1.00 | 0.287 | 700/700 | | | | | | | | | | | 800 | 3600 | 116 | 8.5 | 986 | 0.274 | 0.104 | 3365 | 0.93 | 0.293 | 775 | 3487.5 | 116 | 8.3 | 963 | 0.276 | 0.033 | 3436 | 0.99 | 0.280 | 750 | 3375 | 116 | 8.05 | 934 | 0.277 | 0.011 | 3368 | 1.00 | 0.277 | 725 | 3262.5 | 116 | 7.7 | 893 | 0.274 | 0.001 | 3221 | 0.99 | 0.277 | 700 | 3150 | 116 | 7.45 | 864 | 0.274 | 0.000 | 3164 | 1.00 | 0.273 | 675 | 3037.5 | 116 | 7.2 | 835 | 0.275 | 0.000 | 3015 | 0.99 | 0.277 | 650 | 2925 | 117 | 6.9 | 807 | 0.276 | 0.000 | 2940 | 1.01 | 0.275 | 720/690 | | | | | | | | | | | 750 | 3375 | 117 | 7.70 | 901 | 0.267 | 0.032 | 3293 | 0.98 | 0.274 | 725 | 3262.5 | 116 | 7.60 | 882 | 0.270 | 0.007 | 3243 | 0.99 | 0.272 | 700 | 3150 | 116 | 7.25 | 841 | 0.267 | 0.001 | 3128 | 0.99 | 0.269 | 675 | 3037.5 | 116 | 7 | 812 | 0.267 | 0.000 | 3069 | 1.01 | 0.265 | 650 | 2925 | 117 | 6.6 | 772 | 0.264 | 0.000 | 2949 | 1.01 | 0.262 | 625 | 2812.5 | 118 | 6.35 | 749 | 0.266 | 0.000 | 2809 | 1.00 | 0.267 | 710/680 | | | | | | | | | | | 775 | 3487.5 | 117 | 7.60 | 889 | 0.255 | 0.332 | 2799 | 0.80 | 0.318 | 750 | 3375 | 115 | 7.70 | 886 | 0.262 | 0.090 | 3170 | 0.94 | 0.279 | 725 | 3262.5 | 115 | 7.40 | 851 | 0.261 | 0.024 | 3200 | 0.98 | 0.266 | 700 | 3150 | 117 | 6.90 | 807 | 0.256 | 0.010 | 3135 | 1.00 | 0.258 | 675 | 3037.5 | 117 | 6.60 | 772 | 0.254 | 0.001 | 3034 | 1.00 | 0.255 | 650 | 2925 | 117 | 6.35 | 743 | 0.254 | 0.000 | 2880 | 0.98 | 0.258 | 625 | 2812.5 | 117 | 6.10 | 714 | 0.254 | 0.000 | 2806 | 1.00 | 0.254 | 600 | 2700 | 117 | 5.80 | 679 | 0.251 | 0.000 | 2698 | 1.00 | 0.252 | (approx stock) | 700/670 | | | | | | | | | | | 750 | 3375 | 116 | 7.25 | 841 | 0.249 | 0.429 | 2593 | 0.77 | 0.324 | 725 | 3262.5 | 116 | 7.00 | 812 | 0.249 | 0.126 | 3058 | 0.94 | 0.266 | 700 | 3150 | 116 | 6.80 | 789 | 0.250 | 0.037 | 3079 | 0.98 | 0.256 | 675 | 3037.5 | 117 | 6.50 | 761 | 0.250 | 0.008 | 3030 | 1.00 | 0.251 | 650 | 2925 | 117 | 6.20 | 725 | 0.248 | 0.001 | 2919 | 1.00 | 0.249 | 690/660 | | | | | | | | | | | 750 | 3375 | 116 | 7.00 | 812 | 0.241 | 0.503 | 2465 | 0.73 | 0.329 | 725 | 3262.5 | 116 | 6.80 | 789 | 0.242 | 0.239 | 2815 | 0.86 | 0.280 | 700 | 3150 | 116 | 6.50 | 754 | 0.239 | 0.107 | 2915 | 0.93 | 0.259 | 675 | 3037.5 | 116 | 6.30 | 731 | 0.241 | 0.031 | 2946 | 0.97 | 0.248 | 650 | 2925 | 116 | 6.05 | 702 | 0.240 | 0.007 | 2909 | 0.99 | 0.241 | 680/650 | | | | | | | | | | | 700 | 3150 | 118 | 6.20 | 732 | 0.232 | 0.270 | 2646 | 0.84 | 0.276 | 675 | 3037.5 | 118 | 6.00 | 708 | 0.233 | 0.096 | 2848 | 0.94 | 0.249 | 650 | 2925 | 118 | 5.75 | 679 | 0.232 | 0.032 | 2867 | 0.98 | 0.237 | 625 | 2812.5 | 118 | 5.60 | 661 | 0.235 | 0.005 | 2820 | 1.00 | 0.234 | 600 | 2700 | 118 | 5.30 | 625 | 0.232 | 0.000 | 2681 | 0.99 | 0.233 | 670/640 | | | | | | | | | | | 650 | 2925 | 117 | 5.60 | 655 | 0.224 | 0.096 | 2756 | 0.94 | 0.238 | 625 | 2812.5 | 118 | 5.30 | 625 | 0.222 | 0.027 | 2765 | 0.98 | 0.226 | 600 | 2700 | 118 | 5.20 | 614 | 0.227 | 0.005 | 2703 | 1.00 | 0.227 | 575 | 2587.5 | 117 | 5.00 | 585 | 0.226 | 0.001 | 2559 | 0.99 | 0.229 | 550 | 2475 | 118 | 4.70 | 555 | 0.224 | 0.000 | 2434 | 0.98 | 0.228 | 660/630 | | | | | | | | | | | 625 | 2812.5 | 118 | 5.10 | 602 | 0.214 | 0.125 | 2606 | 0.93 | 0.231 | 600 | 2700 | 118 | 4.95 | 584 | 0.216 | 0.028 | 2646 | 0.98 | 0.221 | 575 | 2587.5 | 117 | 4.80 | 562 | 0.217 | 0.005 | 2574 | 0.99 | 0.218 | 550 | 2475 | 118 | 4.55 | 537 | 0.217 | 0.000 | 2503 | 1.01 | 0.215 | 525 | 2362.5 | 118 | 4.30 | 507 | 0.215 | 0.000 | 2365 | 1.00 | 0.215 | 650/620 | | | | | | | | | | | 625 | 2812.5 | 117 | 5.00 | 585 | 0.208 | 0.343 | 2235 | 0.79 | 0.262 | 600 | 2700 | 118 | 4.75 | 561 | 0.208 | 0.140 | 2462 | 0.91 | 0.228 | 575 | 2587.5 | 118 | 4.50 | 531 | 0.205 | 0.041 | 2508 | 0.97 | 0.212 | 550 | 2475 | 118 | 4.30 | 507 | 0.205 | 0.009 | 2467 | 1.00 | 0.206 | 525 | 2362.5 | 118 | 4.10 | 484 | 0.205 | 0.001 | 2365 | 1.00 | 0.205 | 500 | 2250 | 118 | 3.90 | 460 | 0.205 | 0.000 | 2240 | 1.00 | 0.205 | 475 | 2137.5 | 118 | 3.70 | 437 | 0.204 | 0.000 | 2111 | 0.99 | 0.207 | 450 | 2025 | 118 | 3.50 | 413 | 0.204 | 0.000 | 2017 | 1.00 | 0.205 | 425 | 1912.5 | 118 | 3.30 | 389 | 0.204 | 0.000 | 1913 | 1.00 | 0.204 | 400 | 1800 | 118 | 3.10 | 366 | 0.203 | 0.000 | 1799 | 1.00 | 0.203 | 640/610 | (unstable) | | | | | | | | | | 575 | 2587.5 | 118 | 4.40 | 519 | 0.201 | 0.082 | 2419 | 0.93 | 0.215 | 550 | 2475 | 118 | 4.20 | 496 | 0.200 | 0.025 | 2437 | 0.98 | 0.203 | 525 | 2362.5 | 118 | 4.00 | 472 | 0.200 | 0.004 | 2344 | 0.99 | 0.201 | 500 | 2250 | 118 | 3.80 | 448 | 0.199 | 0.001 | 2253 | 1.00 | 0.199 | 475 | 2137.5 | 117 | 3.60 | 421 | 0.197 | 0.000 | 2131 | 1.00 | 0.198 | 450 | 2025 | 118 | 3.35 | 395 | 0.195 | 0.000 | 2007 | 0.99 | 0.197 | 425 | 1912.5 | 118 | 3.20 | 378 | 0.197 | 0.000 | 1912 | 1.00 | 0.197 | 400 | 1800 | 119 | 3.00 | 357 | 0.198 | 0.000 | 1784 | 0.99 | 0.200 | 630/600 | (unstable) | | | | | | | | | | 500 | 2250 | 119 | 3.60 | 428 | 0.190 | 0.008 | 2232 | 0.99 | 0.192 | 475 | 2137.5 | 118 | 3.50 | 413 | 0.193 | 0.003 | 2114 | 0.99 | 0.195 | 450 | 2025 | 118 | 3.30 | 389 | 0.192 | 0.002 | 2023 | 1.00 | 0.192 | 425 | 1912.5 | 118 | 3.10 | 366 | 0.191 | 0.001 | 1913 | 1.00 | 0.191 | 400 | 1800 | 118 | 2.90 | 342 | 0.190 | 0.001 | 1800 | 1.00 | 0.190 |
|
|
|
|
IITravel01
|
|
June 18, 2016, 04:38:00 PM |
|
That chart is really informative. Looks like the sweet spot at 650/620 is 500M to 550M. I suspect 525M or 537.5M (should bring 2.4+TH/s at 495w, the only thing that shows bad for 550M is the HW errors creeping up) might be best.
|
|
|
|
sidehack (OP)
Legendary
Offline
Activity: 3374
Merit: 1859
Curmudgeonly hardware guy
|
|
June 18, 2016, 04:49:42 PM |
|
Yeah, 620 is a good setpoint. Assuming mine is a representative sample, the group-buy miners will be programmed with 650/620 unless otherwise specified, and shipped set at 450MHz for 2TH at a shade over 400W. That does give the owner still a lot of room to play, including getting close to stock speed (since 550MHz versus 600MHz only costs about 225GH) but saves close to 200W. All those numbers are given with free-running fans, by the way. Turning fans down will save a shade of power on fans, but as chip heat increases the chip-level power requirement increases for a given voltage and frequency. It's a tradeoff.
Once the group-buy batch miners start arriving about Tuesday I'll be able to get measures from a bigger sample set and compare.
|
|
|
|
|