luffy
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
November 07, 2012, 07:57:48 AM |
|
@kano: many thanks, i will try it @Dexter770221: can you please confirm the registors' location given by rgzen? thanks
|
|
|
|
Dexter770221
Legendary
Offline
Activity: 1029
Merit: 1000
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
November 07, 2012, 08:13:02 AM |
|
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
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
November 07, 2012, 09:18:27 AM |
|
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
Activity: 1610
Merit: 1000
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
November 07, 2012, 10:07:12 AM |
|
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.htmCalculation showed about 90-100K +1 for the question where to measure VCC out core voltage
|
|
|
|
Dexter770221
Legendary
Offline
Activity: 1029
Merit: 1000
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
November 07, 2012, 10:37:06 AM |
|
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
Activity: 1610
Merit: 1000
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
November 07, 2012, 11:02:39 AM |
|
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:)
|
|
|
|
Dexter770221
Legendary
Offline
Activity: 1029
Merit: 1000
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
November 07, 2012, 11:55:23 AM |
|
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
Activity: 1610
Merit: 1000
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
November 07, 2012, 12:09:31 PM |
|
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...
|
|
|
|
rgzen
Member
![*](https://bitcointalk.org/Themes/custom1/images/star.gif)
Offline
Activity: 93
Merit: 10
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
November 08, 2012, 09:33:19 PM |
|
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
Activity: 1029
Merit: 1000
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
November 08, 2012, 09:50:46 PM |
|
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
![*](https://bitcointalk.org/Themes/custom1/images/star.gif)
Offline
Activity: 93
Merit: 10
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
November 08, 2012, 10:15:47 PM |
|
hey can you tell me with the info I have posted how much is my hardware error rate? ![Huh](https://bitcointalk.org/Smileys/default/huh.gif) It is possible yo be 0%? thanks
|
|
|
|
rgzen
Member
![*](https://bitcointalk.org/Themes/custom1/images/star.gif)
Offline
Activity: 93
Merit: 10
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
November 08, 2012, 10:19:20 PM |
|
or it can be R/(A+R)? in this case around 1.2%
|
|
|
|
kano
Legendary
Offline
Activity: 4592
Merit: 1851
Linux since 1997 RedHat 4
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
November 08, 2012, 10:59:41 PM |
|
hey can you tell me with the info I have posted how much is my hardware error rate? ![Huh](https://bitcointalk.org/Smileys/default/huh.gif) 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](https://bitcointalk.org/Smileys/default/tongue.gif)
|
|
|
|
rgzen
Member
![*](https://bitcointalk.org/Themes/custom1/images/star.gif)
Offline
Activity: 93
Merit: 10
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
November 09, 2012, 08:44:46 AM |
|
ok, I have upgrade my software.
thanks
|
|
|
|
luffy
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
November 11, 2012, 07:39:10 AM |
|
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
Activity: 4592
Merit: 1851
Linux since 1997 RedHat 4
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
November 11, 2012, 11:19:16 AM |
|
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)
|
|
|
|
loshia
Legendary
Offline
Activity: 1610
Merit: 1000
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
November 11, 2012, 11:55:06 AM |
|
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
|
|
|
|
kano
Legendary
Offline
Activity: 4592
Merit: 1851
Linux since 1997 RedHat 4
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
November 11, 2012, 12:51:45 PM |
|
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)
|
|
|
|
loshia
Legendary
Offline
Activity: 1610
Merit: 1000
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
November 11, 2012, 01:08:21 PM Last edit: November 11, 2012, 01:23:32 PM by loshia |
|
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
|
|
|
|
kano
Legendary
Offline
Activity: 4592
Merit: 1851
Linux since 1997 RedHat 4
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
November 11, 2012, 01:45:34 PM |
|
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)
|
|
|
|
|