Bitcoin Forum
March 19, 2024, 11:06:08 AM *
News: Latest Bitcoin Core release: 26.0 [Torrent]
 
   Home   Help Search Login Register More  
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 101814 times)
luffy
Hero Member
*****
Offline Offline

Activity: 607
Merit: 500



View Profile
November 07, 2012, 07:57:48 AM
 #421

@kano: many thanks, i will try it
@Dexter770221: can you please confirm the registors' location given by rgzen?
thanks
1710846368
Hero Member
*
Offline Offline

Posts: 1710846368

View Profile Personal Message (Offline)

Ignore
1710846368
Reply with quote  #2

1710846368
Report to moderator
1710846368
Hero Member
*
Offline Offline

Posts: 1710846368

View Profile Personal Message (Offline)

Ignore
1710846368
Reply with quote  #2

1710846368
Report to moderator
TalkImg was created especially for hosting images on bitcointalk.org: try it next time you want to post an image
Advertised sites are not endorsed by the Bitcoin Forum. They may be unsafe, untrustworthy, or illegal in your jurisdiction.
1710846368
Hero Member
*
Offline Offline

Posts: 1710846368

View Profile Personal Message (Offline)

Ignore
1710846368
Reply with quote  #2

1710846368
Report to moderator
Dexter770221
Legendary
*
Offline Offline

Activity: 1029
Merit: 1000


View Profile
November 07, 2012, 08:13:02 AM
 #422

Thanks Kano for the additional info about that timings. Propably "long" will report very close to your calculations. After 12 hours U:5.8 (from 5.7), so it's little improvment, however hardware errors dropped to ~3% (previously was ~5%).
Yes luffy, resistors showed by rgzen are correct one.

Under development Modular UPGRADEABLE Miner (MUM). Looking for investors.
Changing one PCB with screwdriver and you have brand new miner in hand... Plug&Play, scalable from one module to thousands.
luffy
Hero Member
*****
Offline Offline

Activity: 607
Merit: 500



View Profile
November 07, 2012, 09:18:27 AM
 #423

ok!
then what resistor can we solder in parallel with the current one in order to get 1.24v-1.28v?
and what are the pins where we measure the VCC voltage?
loshia
Legendary
*
Offline Offline

Activity: 1610
Merit: 1000


View Profile
November 07, 2012, 10:07:12 AM
 #424

ok!
then what resistor can we solder in parallel with the current one in order to get 1.24v-1.28v?
and what are the pins where we measure the VCC voltage?


http://www.1728.org/resistrs.htm
Calculation showed about 90-100K
+1 for the question where to measure VCC out 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
Dexter770221
Legendary
*
Offline Offline

Activity: 1029
Merit: 1000


View Profile
November 07, 2012, 10:37:06 AM
 #425

ngzhang mentioned that this resistor should be somwhere around 9k. R=(R1*R2)/(R1+R2)=(10*100)/(10+100)=1000/110=9.09k. 100k in parallel should be perfect.
When board is placed in front of you as is on rgzen pictures, 4 holes for measure voltage are placed on the right edge of the board. They are visible on bottom picture, near 270uF capacitor and coil.

Under development Modular UPGRADEABLE Miner (MUM). Looking for investors.
Changing one PCB with screwdriver and you have brand new miner in hand... Plug&Play, scalable from one module to thousands.
loshia
Legendary
*
Offline Offline

Activity: 1610
Merit: 1000


View Profile
November 07, 2012, 11:02:39 AM
 #426

ngzhang mentioned that this resistor should be somwhere around 9k. R=(R1*R2)/(R1+R2)=(10*100)/(10+100)=1000/110=9.09k. 100k in parallel should be perfect.
When board is placed in front of you as is on rgzen pictures, 4 holes for measure voltage are placed on the right edge of the board. They are visible on bottom picture, near 270uF capacitor and coil.
Hey,
thank you for the update! It seems that you know hardware stuff better than me (i am not expert at all). From what i have seen in pdf (core_power) schematic (i might be wrong of course) we have to measure a voltage VCCINT1V2_A and  VCCINT1V2_B. Is that true? If it is, are the wholes you are referring to equal to "VCCINT1V2_A and  VCCINT1V2_B". If yes, would it be very hard for you just to mark them on the picture, because it is very hard for me to find them out:)
Once again excuse me but as i said i am not an expert at all:)

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
Dexter770221
Legendary
*
Offline Offline

Activity: 1029
Merit: 1000


View Profile
November 07, 2012, 11:55:23 AM
 #427

http://i.minus.com/iOoHctgIXSDTB.JPG - picture from first post (courtesy of ngzhang).
At the edge of the board you can see 4 holes (right side, white squares around it) labeled GND, V.CORE1, V.CORE2, V.AUX. Those are test points for this voltages. V.AUX is 3,3V for IO.

Under development Modular UPGRADEABLE Miner (MUM). Looking for investors.
Changing one PCB with screwdriver and you have brand new miner in hand... Plug&Play, scalable from one module to thousands.
loshia
Legendary
*
Offline Offline

Activity: 1610
Merit: 1000


View Profile
November 07, 2012, 12:09:31 PM
 #428

http://i.minus.com/iOoHctgIXSDTB.JPG - picture from first post (courtesy of ngzhang).
At the edge of the board you can see 4 holes (right side, white squares around it) labeled GND, V.CORE1, V.CORE2, V.AUX. Those are test points for this voltages. V.AUX is 3,3V for IO.

Thanks!
I got it...

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
rgzen
Member
**
Offline Offline

Activity: 93
Merit: 10



View Profile
November 08, 2012, 09:33:19 PM
 #429

Sixteen hours past since the miner started it throws the following:

(5s):526.7 (avg):483.0 Mh/s | Q:1250  A:5336  R:64  HW:0  E:427%  U:5.7/m

According to my calcs it is more less 411 MH/s or around 8% improvement...
I repeat that I have the first version flashed with the V42 bitstream and the resistors changed and --icarus-timing 2:70 with cgminer 2.7.4.
So I am asking... What are your results with other configurations??

PD: And I have too curiosity of know what are exactly the differences between the first and the second version of the board.
Dexter770221
Legendary
*
Offline Offline

Activity: 1029
Merit: 1000


View Profile
November 08, 2012, 09:50:46 PM
 #430

After my mod (136k in parallel with 10k resistors, Vcore@1.22V) I'm getting U:5.8, error rate ~3%. It's not big improvement but always it's something. I think that ngzang just posted little modified bitstream with auto adjusted clock, version "handy placed small cores" is still in his possesion and he don't want to share with it...

Under development Modular UPGRADEABLE Miner (MUM). Looking for investors.
Changing one PCB with screwdriver and you have brand new miner in hand... Plug&Play, scalable from one module to thousands.
rgzen
Member
**
Offline Offline

Activity: 93
Merit: 10



View Profile
November 08, 2012, 10:15:47 PM
 #431

hey
can you tell me with the info I have posted how much is my hardware error rate?Huh
It is possible yo be 0%?
thanks
rgzen
Member
**
Offline Offline

Activity: 93
Merit: 10



View Profile
November 08, 2012, 10:19:20 PM
 #432

or it can be R/(A+R)? in this case around 1.2%
kano
Legendary
*
Offline Offline

Activity: 4438
Merit: 1794


Linux since 1997 RedHat 4


View Profile
November 08, 2012, 10:59:41 PM
 #433

hey
can you tell me with the info I have posted how much is my hardware error rate?Huh
It is possible yo be 0%?
thanks
cgminer before 2.7.6 doesn't report HW: for the Icarus driver.
So you won't known unless you are using 2.7.6 (where I added it) or later.
... preferably the latest version Tongue

Pool: https://kano.is - low 0.5% fee PPLNS 3 Days - Most reliable Solo with ONLY 0.5% fee   Bitcointalk thread: Forum
Discord support invite at https://kano.is/ Majority developer of the ckpool code - k for kano
The ONLY active original developer of cgminer. Original master git: https://github.com/kanoi/cgminer
rgzen
Member
**
Offline Offline

Activity: 93
Merit: 10



View Profile
November 09, 2012, 08:44:46 AM
 #434

ok, I have upgrade my software.

thanks
luffy
Hero Member
*****
Offline Offline

Activity: 607
Merit: 500



View Profile
November 11, 2012, 07:39:10 AM
 #435

i upgarded too. i noticed that even icarus board without any modifications has a few HW hits although none of my VGAs (7950,5870,5970) has any!
kano
Legendary
*
Offline Offline

Activity: 4438
Merit: 1794


Linux since 1997 RedHat 4


View Profile
November 11, 2012, 11:19:16 AM
 #436

i upgarded too. i noticed that even icarus board without any modifications has a few HW hits although none of my VGAs (7950,5870,5970) has any!
I get about 0.2% on my 2 original Icarus boards.

The GPU code always reported HW: errors, but the others didn't.
GPU's often get none - usually only get them when over heating of over clocking too far
(I'm not sure if I've ever had a HW error on one my 6950 GPUs in 16 months - the other one fan failed and got replaced so had a few when that happened)

Anyway, I changed it so all devices (except ztex) go through the same (new) HW check code
(after I realised that it wasn't doing that - which I thought it was in the FPGA code)

Pool: https://kano.is - low 0.5% fee PPLNS 3 Days - Most reliable Solo with ONLY 0.5% fee   Bitcointalk thread: Forum
Discord support invite at https://kano.is/ Majority developer of the ckpool code - k for kano
The ONLY active original developer of cgminer. Original master git: https://github.com/kanoi/cgminer
loshia
Legendary
*
Offline Offline

Activity: 1610
Merit: 1000


View Profile
November 11, 2012, 11:55:06 AM
 #437

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
kano
Legendary
*
Offline Offline

Activity: 4438
Merit: 1794


Linux since 1997 RedHat 4


View Profile
November 11, 2012, 12:51:45 PM
 #438

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 - low 0.5% fee PPLNS 3 Days - Most reliable Solo with ONLY 0.5% fee   Bitcointalk thread: Forum
Discord support invite at https://kano.is/ Majority developer of the ckpool code - k for kano
The ONLY active original developer of cgminer. Original master git: https://github.com/kanoi/cgminer
loshia
Legendary
*
Offline Offline

Activity: 1610
Merit: 1000


View Profile
November 11, 2012, 01:08:21 PM
Last edit: November 11, 2012, 01:23:32 PM by loshia
 #439

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: 4438
Merit: 1794


Linux since 1997 RedHat 4


View Profile
November 11, 2012, 01:45:34 PM
 #440

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 - low 0.5% fee PPLNS 3 Days - Most reliable Solo with ONLY 0.5% fee   Bitcointalk thread: Forum
Discord support invite at https://kano.is/ Majority developer of the ckpool code - k for kano
The ONLY active original developer of cgminer. Original master git: https://github.com/kanoi/cgminer
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:  

Powered by MySQL Powered by PHP Powered by SMF 1.1.19 | SMF © 2006-2009, Simple Machines Valid XHTML 1.0! Valid CSS!