Bitcoin Forum
December 03, 2016, 03:41:38 PM *
News: Latest stable version of Bitcoin Core: 0.13.1  [Torrent].
 
   Home   Help Search Donate Login Register  
Pages: « 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 [23] 24 »  All
  Print  
Author Topic: FPGA development board "Lancelot" - accept bitsteam developer's orders.  (Read 95913 times)
loshia
Legendary
*
Offline Offline

Activity: 1372


View Profile
November 11, 2012, 11:55:06 AM
 #441

kano,

It may sound stupid but i am little bit confused of values i see. As you noted with new bitstream and timing 2:70 Mhs reporting will not be correct. To make things easy, assuming that cgminer shows  correct values for my gpu as follows:

MHS av   MHS 3s   Accepted   Hardware Errors Utility   
 391.03   391.16   6,129                    0       5.55/m

For my best lancelot board values are as follows:
MHS av   MHS 3s   Accepted   Hardware Errors        Utility   
506.35   662.20   6,651             131                  6.02

What about calculating Lancelot performance in following way
Lancelot Utility/GPU Utility*GPU MHS av
 6.02/5.55*391.03 = 424 Mhs

What about HW errors - are they taken into account when calculating utility. In my case for Lancelot they are 1.97% (131/6,651*100). Shall i do something like:
424/100*98.03=415.64

Thanks

Please help the Led Boy aka Bicknellski to make us a nice Christmas led tree and pay WASP membership fee here:
https://bitcointalk.org/index.php?topic=643999.msg7191563#msg7191563
And remember Bicknellski is not collecting money from community;D
1480779698
Hero Member
*
Offline Offline

Posts: 1480779698

View Profile Personal Message (Offline)

Ignore
1480779698
Reply with quote  #2

1480779698
Report to moderator
1480779698
Hero Member
*
Offline Offline

Posts: 1480779698

View Profile Personal Message (Offline)

Ignore
1480779698
Reply with quote  #2

1480779698
Report to moderator
1480779698
Hero Member
*
Offline Offline

Posts: 1480779698

View Profile Personal Message (Offline)

Ignore
1480779698
Reply with quote  #2

1480779698
Report to moderator
Advertised sites are not endorsed by the Bitcoin Forum. They may be unsafe, untrustworthy, or illegal in your jurisdiction. Advertise here.
1480779698
Hero Member
*
Offline Offline

Posts: 1480779698

View Profile Personal Message (Offline)

Ignore
1480779698
Reply with quote  #2

1480779698
Report to moderator
1480779698
Hero Member
*
Offline Offline

Posts: 1480779698

View Profile Personal Message (Offline)

Ignore
1480779698
Reply with quote  #2

1480779698
Report to moderator
1480779698
Hero Member
*
Offline Offline

Posts: 1480779698

View Profile Personal Message (Offline)

Ignore
1480779698
Reply with quote  #2

1480779698
Report to moderator
kano
Legendary
*
Offline Offline

Activity: 1918


Linux since 1997 RedHat 4


View Profile
November 11, 2012, 12:51:45 PM
 #442

kano,

It may sound stupid but i am little bit confused of values i see. As you noted with new bitstream and timing 2:70 Mhs reporting will not be correct. To make things easy, assuming that cgminer shows  correct values for my gpu as follows:

MHS av   MHS 3s   Accepted   Hardware Errors Utility   
 391.03   391.16   6,129                    0       5.55/m

For my best lancelot board values are as follows:
MHS av   MHS 3s   Accepted   Hardware Errors        Utility   
506.35   662.20   6,651             131                  6.02

What about calculating Lancelot performance in following way
Lancelot Utility/GPU Utility*GPU MHS av
 6.02/5.55*391.03 = 424 Mhs

What about HW errors - are they taken into account when calculating utility. In my case for Lancelot they are 1.97% (131/6,651*100). Shall i do something like:
424/100*98.03=415.64

Thanks

MHS of course will be wrong:
Every time work is aborted (no nonce found before timeout, or an LP occurs) it has to determine how many hashes were done.
The Hs value is used to calculate that
Every time it finds a share ... it knows how many hashes were done since the share value tells that.

However, the number of shares Accepted, Rejected, Stale, HW, U, pretty much everything else is correct.

Yes you can estimate your hash rate from U - but you'd have to run for a few days to ensure it's close.
Even after a few hours it can (rarely) be out by 10% (which is a lot)

anyway yep Hashes/s is simply (2^32) * U/60 (for 1 diff shares)

U is only accepted shares.

A, R, SS and HW are all independent.

To work out the HW % = HW / (A + R + SS + HW)

Pool: https://kano.is BTC: 1KanoiBupPiZfkwqB7rfLXAzPnoTshAVmb
CKPool and CGMiner developer, IRC FreeNode #ckpool and #cgminer kanoi
Help keep Bitcoin secure by mining on pools with Stratum, the best protocol to mine Bitcoins with ASIC hardware
loshia
Legendary
*
Offline Offline

Activity: 1372


View Profile
November 11, 2012, 01:08:21 PM
 #443

thank you Kano!

I got the U formula.


What about if pool is not 1 diff shares let us say it is dynamic? If it is static let us say 2 diff shares all is easy - 2^33 right?
Best

Please help the Led Boy aka Bicknellski to make us a nice Christmas led tree and pay WASP membership fee here:
https://bitcointalk.org/index.php?topic=643999.msg7191563#msg7191563
And remember Bicknellski is not collecting money from community;D
kano
Legendary
*
Offline Offline

Activity: 1918


Linux since 1997 RedHat 4


View Profile
November 11, 2012, 01:45:34 PM
 #444

thank you Kano!

I got the U formula.


What about if pool is not 1 diff shares let us say it is dynamic? If it is static let us say 2 diff shares all is easy - 2^33 right?
Best
U = 60 * accepted shares / elapsed time
However, I also added the difficulty versions of A + R + SS into the API in cgminer
(they are also printed in the summary when you exit)

So to get the correct U based on difficulty it's:
 60 * "Difficulty Accepted" / "Elapsed"
from the API 'summary'

(however note, HW is not difficulty based, it's 1diff share based)

Pool: https://kano.is BTC: 1KanoiBupPiZfkwqB7rfLXAzPnoTshAVmb
CKPool and CGMiner developer, IRC FreeNode #ckpool and #cgminer kanoi
Help keep Bitcoin secure by mining on pools with Stratum, the best protocol to mine Bitcoins with ASIC hardware
loshia
Legendary
*
Offline Offline

Activity: 1372


View Profile
November 11, 2012, 02:24:35 PM
 #445

Thank You!

Please help the Led Boy aka Bicknellski to make us a nice Christmas led tree and pay WASP membership fee here:
https://bitcointalk.org/index.php?topic=643999.msg7191563#msg7191563
And remember Bicknellski is not collecting money from community;D
loshia
Legendary
*
Offline Offline

Activity: 1372


View Profile
November 13, 2012, 07:35:49 PM
 #446

Kano,

I have tried --icarus-timing 2:70. I was running it for two days. I have saved the results.

Today i changed the resistors on a couple of my boards. With 2:70 i see a lot of times two yellow leds (on ones with VCC core increased -changed resistors) - The sign on top of the leds is IDLE. When i run cgminer with icarus-timing long No yellow leds - these broads (changed resistors) are blinking like the rest

Can you comment that pls.
Thank you

I can compare the results after two days or so...


Please help the Led Boy aka Bicknellski to make us a nice Christmas led tree and pay WASP membership fee here:
https://bitcointalk.org/index.php?topic=643999.msg7191563#msg7191563
And remember Bicknellski is not collecting money from community;D
kano
Legendary
*
Offline Offline

Activity: 1918


Linux since 1997 RedHat 4


View Profile
November 14, 2012, 02:25:50 AM
 #447

Kano,

I have tried --icarus-timing 2:70. I was running it for two days. I have saved the results.

Today i changed the resistors on a couple of my boards. With 2:70 i see a lot of times two yellow leds (on ones with VCC core increased -changed resistors) - The sign on top of the leds is IDLE. When i run cgminer with icarus-timing long No yellow leds - these broads (changed resistors) are blinking like the rest

Can you comment that pls.
Thank you

I can compare the results after two days or so...

What does the API say for 'stats' about the ICA devices? (both for the ones running 'long' and the ones running '2:70')
(There are 16 Icarus specific stats in there)

You of course have to enable the API with --api-listen and can then request API 'stats' from the computer running cgminer at 127.0.0.1 ... or read the API-README

Pool: https://kano.is BTC: 1KanoiBupPiZfkwqB7rfLXAzPnoTshAVmb
CKPool and CGMiner developer, IRC FreeNode #ckpool and #cgminer kanoi
Help keep Bitcoin secure by mining on pools with Stratum, the best protocol to mine Bitcoins with ASIC hardware
loshia
Legendary
*
Offline Offline

Activity: 1372


View Profile
November 14, 2012, 08:18:05 AM
 #448

Kano,
Unfortunately i did not saved api-stats with timing 2:70. I am posting results with timing long. If they are not enough i can run some boards with 2:70 and some with long and post back the results tomorrow


Full nonce for moded boards with U ranging between 6.3-6.5 is:
4.413108
4.329242
4.273607
4.347153

HS:

0.000000000999920
0.000000000981032
0.000000000965784
0.000000000984132


Full nonce for rest is with U ranging between 5.5 -5.7 is:
4.989389
4.936110
4.984062
4.896753

HS:
0.000000001134453
0.000000001115143
0.000000001132101
0.000000001110573

Is that info enough?

Best

Please help the Led Boy aka Bicknellski to make us a nice Christmas led tree and pay WASP membership fee here:
https://bitcointalk.org/index.php?topic=643999.msg7191563#msg7191563
And remember Bicknellski is not collecting money from community;D
kano
Legendary
*
Offline Offline

Activity: 1918


Linux since 1997 RedHat 4


View Profile
November 14, 2012, 10:53:15 PM
 #449

That's suggesting they are running at around 1GH/s so I guess there must be a problem (since the U isn't 1GH/s more like ~460MH/s)
Maybe the new bitstream divides the work up differently?

All 16 values for a device would be more useful ...

Anyway - if those numbers were correct then it would be something like: 1:40

Also, the Icarus code defaults to 2 devices (which you can change with --icarus-options) that divide the nonce-range in half (however the division 'method' based on the number of devices is fixed) - so I'm not sure exactly what the bitstream is doing different to give those numbers you are getting.

... however ... if the bitstream is adjusting the clock ... then yes timing long/short would not give correct results.

long/short timing assumes the clock is fixed - if it isn't it wont work without knowing the clock for each result (and me rewriting the timing code to handle that)

Try 1:40 and see what happens ...

Pool: https://kano.is BTC: 1KanoiBupPiZfkwqB7rfLXAzPnoTshAVmb
CKPool and CGMiner developer, IRC FreeNode #ckpool and #cgminer kanoi
Help keep Bitcoin secure by mining on pools with Stratum, the best protocol to mine Bitcoins with ASIC hardware
loshia
Legendary
*
Offline Offline

Activity: 1372


View Profile
November 15, 2012, 08:56:16 AM
 #450

Thanks Kano!
I am trying it and i will post the results
Best

Please help the Led Boy aka Bicknellski to make us a nice Christmas led tree and pay WASP membership fee here:
https://bitcointalk.org/index.php?topic=643999.msg7191563#msg7191563
And remember Bicknellski is not collecting money from community;D
loshia
Legendary
*
Offline Offline

Activity: 1372


View Profile
November 15, 2012, 02:33:16 PM
 #451

Kano,

Here are results with timing=long for 11 hours

Please note ICA0-ICA3 and ICA26-ICA29 are with changed resistors for sure. There are at lest 3-4 boards from the rest with changed resistors, but unfortunately i can not remember their numbers

Best

0GPU041672971221363.5930851.4000340.000000
1ICA041672123362.5107470.0482820.000002434.413108100.0000000009999200.1184841600427423166446050.008396160119longtrue11520022
2ICA141672124732.9614990.2144840.000002424.329242100.0000000009810320.1157431600424267736245410.00869616062longtrue11520022
3ICA241672126495.3150561.4768990.000002414.273607100.0000000009657840.1255971600428807018247070.00918016071longtrue11520022
4ICA341672125693.5022630.2624020.000002424.347153100.0000000009841320.1203401600429266287046060.008492160129longtrue11520022
5ICA441672110043.5664850.9466760.000002484.958089100.0000000011260980.1215351600427829935241150.007690160115longtrue11520022
6ICA541672108933.6354740.9450210.000002484.978525100.0000000011251780.1459211600426154570841120.007901160110longtrue11520022
7ICA641672109463.7116910.9985590.000002484.961865100.0000000011247890.1309351600427243334239750.007576160137longtrue11520022
8ICA741672109282.6618770.1212100.000002484.967016100.0000000011259690.1310151600429195698840960.00734516097longtrue11520022
9ICA841672109172.8412050.1197030.000002484.989389100.0000000011344530.1169521600427728159640720.00813516071longtrue11520022
10ICA941672109715.7188721.2405760.000000484.936110100.0000000011151430.1466091600429074699440030.0074421606longtrue11520022
11ICA1041672107524.5436381.3848640.000002484.984062100.0000000011321010.1217271600428198437239650.008019160130longtrue11520022
12ICA1141672110592.7433260.1193300.000001474.896753100.0000000011105730.1268801600426667058840250.00750816029longtrue11520022
13ICA1241672106154.5283841.2330070.000000495.045424100.0000000011304850.1900261600427846679239310.00759316094longtrue11520022
14ICA1341672107484.6139701.3523240.000002495.037712100.0000000011338360.1679241600428403990640160.00742616018longtrue11520022
15ICA1441672109002.7335900.1173770.000002495.037367100.0000000011498800.0986681600427780356840730.00779816066longtrue11520022
16ICA1541672109532.7183610.1199650.000002484.923011100.0000000011113310.1498821600426098408839910.007511160151longtrue11520022
17ICA1641672109502.5370140.1418950.000002484.981943100.0000000011370810.0982181600427535908440450.00827516043longtrue11520022
18ICA1741672123503.7266260.6644590.000002424.392341100.0000000009903620.1387701600425125986044970.00867116020longtrue11520022
19ICA1841672122502.9753540.1443060.000002434.405313100.0000000009961960.1266821600428820840844920.00866916011longtrue11520022
20ICA1941672106502.7219820.1446930.000002495.093104100.0000000011550940.1320131600429133208438300.007291160152longtrue11520022
21ICA2041672109132.6053710.1742660.000002484.966502100.0000000011184620.1627461600427724555239740.007577160138longtrue11520022
22ICA2141672107642.7266430.1476680.000001495.004606100.0000000011162130.2105081600427739587240420.00793716045longtrue11520022
23ICA2241672123752.6269030.1458370.000001424.360790100.0000000009914220.1026631600428856970244610.008659160140longtrue11520022
24ICA2341672107103.0866800.1449960.000002495.093802100.0000000011601900.1108241600429131565039850.007460160143longtrue11520022
25ICA2441672109222.7517430.1426710.000002484.997611100.0000000011273090.1558571600424687487239660.007127160126longtrue11520022
26ICA2541672108643.0416890.1411760.000002495.016164100.0000000011335260.1477071600425546535440460.00803816050longtrue11520022
27ICA2641672123123.0440150.3249070.000002434.447452100.0000000010155950.0855041600428400144845190.00844416040longtrue11520022
28ICA2741672124723.7492970.9244760.000002424.376254100.0000000009894040.1267961600428485033444520.008082160135longtrue11520022
29ICA2841672124062.9591080.1415560.000002434.419099100.0000000009995690.1259851600429242088646490.0090711609longtrue11520022
30ICA2941672126463.2335190.6637270.000001424.364028100.0000000009883830.1189561600426320887847910.008888160151longtrue11520022

Please help the Led Boy aka Bicknellski to make us a nice Christmas led tree and pay WASP membership fee here:
https://bitcointalk.org/index.php?topic=643999.msg7191563#msg7191563
And remember Bicknellski is not collecting money from community;D
luffy
Hero Member
*****
Offline Offline

Activity: 606



View Profile
November 17, 2012, 09:59:59 PM
 #452

i think i may have damaged the board while trying to solder the parallel resistors. now the Vin of every fpga
is stuck to 0.59v no matter how many times i try to solder the original smt registors back. the white leds are off,
all the other leds are on. i hope it is a matter of bad soldering and nothing else  Cry
loshia
Legendary
*
Offline Offline

Activity: 1372


View Profile
November 17, 2012, 10:04:56 PM
 #453

luffy,

Same happened to me also. Although original resistors look OK they were damaged. I suggest to find some 0603 Resistors and replace them. All will be fine. Another approach will be to unsloder them and cheek them with multimeter and  you will see that they are damaged (no resistance). Or you can measure the resistance between two ends of the resistors soldered on PCB. if they are damaged the resistance will be 10 KOhms if they are ok it should be about 5 Kohms

I was trying parallel soldering also. It turned out to be a bad idea because of easy "blowing out"
original ones. Better approach is just to replace them

Please help the Led Boy aka Bicknellski to make us a nice Christmas led tree and pay WASP membership fee here:
https://bitcointalk.org/index.php?topic=643999.msg7191563#msg7191563
And remember Bicknellski is not collecting money from community;D
hardcore-fs
Full Member
***
Offline Offline

Activity: 196


View Profile WWW
November 18, 2012, 12:27:51 AM
 #454

DON'T try to continually solder resistors to your PCB.....

Get some VERY THIN FLEXIBLE linking wire, then solder that to the PCB, THEN do your testing by soldering the resistors to the ends of the wire.
That way you only solder your VALUABLE PCB once or twice, but you can solder and abuse your shitty connection wire multiple times.

BTC:1PCTzvkZUFuUF7DA6aMEVjBUUp35wN5JtF
libertybuck
Full Member
***
Offline Offline

Activity: 137


View Profile
November 18, 2012, 12:41:43 AM
 #455

DON'T try to continually solder resistors to your PCB.....

Get some VERY THIN FLEXIBLE linking wire, then solder that to the PCB, THEN do your testing by soldering the resistors to the ends of the wire.
That way you only solder your VALUABLE PCB once or twice, but you can solder and abuse your shitty connection wire multiple times.


Good idea.

luffy
Hero Member
*****
Offline Offline

Activity: 606



View Profile
November 18, 2012, 06:18:01 AM
 #456

thanks guys. i measure them around 3Kohm each. i wonder if i have damaged any neighbour resistor!
loshia
Legendary
*
Offline Offline

Activity: 1372


View Profile
November 18, 2012, 06:56:49 AM
 #457

Probably you did. I can not be 100% sure but my core voltage was 0.5 like yours when they were 0 kohm each...being 3 kohm each it means that you should get very high VCC core (my oppinion) voltage or something else around is damaged which shall be the case

Try the wire thing as suggested. As stated PCB is easy to damage. If you happen to fix at least one of the cores you will able to measure correct core voltage

Please help the Led Boy aka Bicknellski to make us a nice Christmas led tree and pay WASP membership fee here:
https://bitcointalk.org/index.php?topic=643999.msg7191563#msg7191563
And remember Bicknellski is not collecting money from community;D
fpf
Newbie
*
Offline Offline

Activity: 20


View Profile
November 20, 2012, 04:52:12 PM
 #458

Dear N.G.Zhang,

Will the source code for that bitstream be published?
How does the FPGA determine the max. clock rate / hash rate?
Over the I2C temperature sensor that is available on the lancelot boards but not on the icarus boards?
And what does that mean in case of icarus boards.
What functions do the 4 dip switches have now? (Played a bit with them but it seems they lost their old functions and do some weird stuff now)

Regards

FPF


hardcore-fs
Full Member
***
Offline Offline

Activity: 196


View Profile WWW
November 20, 2012, 11:40:29 PM
 #459

Probably you did. I can not be 100% sure but my core voltage was 0.5 like yours when they were 0 kohm each...being 3 kohm each it means that you should get very high VCC core (my oppinion) voltage or something else around is damaged which shall be the case

Try the wire thing as suggested. As stated PCB is easy to damage. If you happen to fix at least one of the cores you will able to measure correct core voltage

There are always better solutions (routing/VHDL/Verilog/partitioning optimizations)

Really no one should be playing with the core voltage, UNLESS they have enabled both in-chip/external temp  and external  current  measurement.

Problem is that such things that allow you to fiddle with the core voltage vary significantly from batch to batch & the age of the Chip.

You will find that current and internal die temp start to show signs that you are getting close to the danger zone, whilst you still have time to back them off.
Certainly making mods to a small group of FPGA then retroactively applying the 'fix' to masses of production out in the field, really is only going to end one way, plus such damage is really going to F*** any chances of returning the units for trade in.

The question should be:
Is the mod going to make more bit coins between now and trade in, than the trade in value of the device?

If not then leave it the F***  ALONE!!!



BTC:1PCTzvkZUFuUF7DA6aMEVjBUUp35wN5JtF
burnin
Full Member
***
Offline Offline

Activity: 183

Mining HW-Dev


View Profile
November 22, 2012, 08:56:54 PM
 #460

Dear N.G.Zhang,

Will the source code for that bitstream be published?
How does the FPGA determine the max. clock rate / hash rate?
Over the I2C temperature sensor that is available on the lancelot boards but not on the icarus boards?
And what does that mean in case of icarus boards.
What functions do the 4 dip switches have now? (Played a bit with them but it seems they lost their old functions and do some weird stuff now)

Regards

FPF

From what he said before i don't think so.
Would be awesome though.
The Source won't be the only key, i think he did a lot of manual routing stuff.

donations accepted: BTC: 18tWnLb3y2E6B9utxDZE4bmu8DasSNBnqX
Register here for Project News: http://eepurl.com/zNH7v
Pages: « 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 [23] 24 »  All
  Print  
 
Jump to:  

Sponsored by , a Bitcoin-accepting VPN.
Powered by MySQL Powered by PHP Powered by SMF 1.1.19 | SMF © 2006-2009, Simple Machines Valid XHTML 1.0! Valid CSS!