Topic: Cairnsmore1 - Quad XC6SLX150 Board - page 95. (Read 286370 times)

As ebereon worked out - try changing SW6 1 (that, I guess, changes the baud rate?)

The real cgminer uses 115200 in the icarus code

Works fine in windows, I just wish I could figure out what Linux wants to communicate with it. I couldn't get mpbm working with it either. I still think I have a driver issue even though the ttyUSB0-4 are all showing up properly, and the drivers look good in lsmod.

thanks kano!

mpbm works also with the icarus miner, just change the bautrate to 57600 as kano has described. SW6 = 1234 ON

If it is actually working fine, then '--icarus-timing short' should also ensure it works out the correct timing for reporting the hash rate and for minimising waits between work

'waits' should be almost zero, but if the default values are wrong they could be sizeable depending on the FPGA performance

Also note (as it says in FPGA-README) '--icarus-timing short' needs the computer to be 'quiet', other than cgminer, to make sure it gets the numbers right

Once it has the right 2 numbers you can feed that into cgminer every other time you run it and not need to run the 'short' timing mode

Got it now running with standard cgminer 2.4.3 with SW6 1 = off, 234 on  (no --icarus-timing option)

But EP-cgminer wont work now...
and the other way around.

So at my board SW6 dip 1 changes the possibility of standard cgminer or the enterpoint one.

I use now the standard cgminer.

The only change in the "driver-cairnsmore.c" code (other than copying the icarus 2.3.4 code and changing the text everywhere to say cairnsmore instead of icarus) was to change the serial speed to 57600 instead of 115200
(and no idea why but they did that to the icarus driver also)

I have no idea if that even fixes anything - but the code will certainly report the MH/s wrong and if it doesn't run exactly the same MH/s speed as an icarus Rev3 it could have other issues (2.3.4 is quite old in the FPGA arena ...)

Again as I wrote in here many pages back, the whole point of the --icarus-timing option was to deal with devices with different speeds to the standard icarus Rev3 (but pretending to be an icarus Rev3)

I even provided that without ANY compensation at all (from anyone with any FPGA ) so all they needed to say was cairnsmore needs the Serial/USB to be 57600 (if that is even true?)

If it's not needed then the current cgminer 2.4.2/3 should be able to mine with the correct --icarus-timing options

(edit: and yes 2.3.4 doesn't have --icarus-timing)

There, fixed that for you.

How would I know with 2 4*5870 rigs running cgminer 2.4.3 custom 

So, who got this board running with Enterpoint-CGMiner? I'm trying to (Win7 64 / WinXP 32), but always get "Test failed ... get 00000000". Drivers are installed.

If I remember right all the dip switches are on(or maybe all off) except on SW3 and SW4 where the second bit in from the left (controller/cable end side) on both of those. All of the boards should have gone out with the correct settings already set. I don't have a physical board with me that would allow me to be more descriptive. Basically SW2/3/4/5 follow a polarity inverted Icarus setup. First and second bits from left are all that matter. I don't think the last 2 bits are used.

SW!/SW6 I think all switches should be set the side nearest the board edge. Some of the bits here control frequency and you can tell from the red LED flash rate if you get one of those. Others control JTAG functions.

I'll try and get a marked up picture to explain this on the website. It's just a job that needs doing.

I tested some switch positions of the 4 dip switch-banks, one for every fpga.

 _________
|              |
|              |
|   1      3  |
|              |
|   2      4  |
|_________|



Switch position shipping:
1 = 1234 -> on
2 = 1 -> off, 234 -> on
3 = 1234 -> on
4 = 1 -> off, 234 -> on
That was hashing @ ~50Mh


This setup mean's for me only fpga 1 and 3 is active. When EP-cgminer is working, the orange led's turn off and flashing red only on fpga 1 and 3. fpga 2 and 4 stay at orange led. Blue led turns on when a share is found. Orange led turns on when longpoll breaks actual work.(?)  

Attention! That's only the information from my own tests and i think it works like that.

To turn on the other two fpga's just switch the dip 1 -> on on fpga 2 and 4, then they will do something, but please turn the board off before you do it. I turned off the power AND the usb.
In this position the blue led will turn on, on fpga 1 and 2 (also 3 and 4) at the same time when a share is found.

I thought every fpga should do his own work and tested futher.

My switch positions now:
1 = 134 -> on, 2 -> off
2 = 1234 -> on
3 = 134 -> on, 2 -> off
4 = 1234 -> on
That is hashing @ ~100Mh

I really don't know if that make's sence, but my unit is now hashing @100Mh instead of only 50Mh. Every fpga turns on his blue led when a share is found.

Some technical information about the dip switches and led's would be nice  

I've heard of it yes.
I don't see why.
I'm not sure I understand.

• cgminer has backup pools in one of its multiple pool support modes (in fact it's the default).
• Bitstream upload, I don't know about.
• Switches MHz, IIRC I've seen users of cgminer discussing how it works, so it surely is included.

Has cgminer backup pools support was the main reason why I switched from DiabloMiner 10 months ago, I very much doubt you know cgminer enough to help me form my opinion if you cite this as an advantage for MPBM...

+1. And sorry for moaning yohan.

There are temperature sensors next to every FPGA and we can sense fan speed on every fan header. Those features are not enabled as yet but will come eventually.

the moaner says: No moan, just full ack, Enterpoint never "bfl"d us (taking money and then make us wait for the hardware). At BFL  you had to pay in front and wait longer than the time from idea to delivery of the product, than cairnsmore (okay, bfl hashed quickly at delivery, but THIS IS NO MOAN!, also a lotta time left)
The hardware product is built with foresight to multiple use cases (diff power, power distribution, stacking, up and down link instead of usb). Also I tried to heat up the board a bit by running it with the fan switched off. As soon as you start the fan again its almost "directly" at ambient temperature. Cooling wise its pretty perfect, the fan is a good choice.

Does the board have temperature sensors for the fpga? Or will you check via the error rate for overheating fpgas? (Thinking of a fan failure/slow down (Should be some time until that happens with arctic cooling fans))

And a comment to the bfl asic release: For my opinion its a little too sarcastic (jalapeno---usb powered coffee warmer...)

Ok CGminer (windows) is back and with source code.

As to BFL our consumer protection authorities always say if something looks too good to be true then it probably isn't. Technically those hashing rates could achieved in ASIC but I don't think the financial model stacks up. The technology they would have to use in my opinion would be very expensive. In my personal opinion it has the appearence of a ponzi scheme. Even if they somehow have a large enough market to make the financials work why would they hand away profit by pricing so low. They could offer x2 hashes per dollar and kill all of the competition off easily not x16 as they are. That does not stack up at all especially if a real VC is envolved. I know enough about them to be pretty certain on that.

Press releases without dates and real names smell like a very rotten fish.

BFL as far as I am aware don't have any timeline for delivery. It might be they are gathering orders and planning an ASIC development then based on holding peoples money to fund it. if that is what they are doing expect a very long delivery and maybe never a working product. It's not uncommon for ASICs to need one or more respins and that can be a very large amount of money and time.

As to doing an ASIC ourselves. It might be done in 6 months but that would be pushing it even by our standards. However I will say I don't think we have seen the best FPGAs can bring as yet. Not by a very long way. That's almost a BFL style statement and you can all ponder that one. Not to boast in any way but merely a statement what the team have done in Bitcoin, in the shape of Cairnsmore1, is extremely simple compared to what we do in size and complexity for the general HPC marketplace. Don't confuse that with being a poor product. I am exceptionally pleased with Cairnsmore1. It had a design target and we have mostly met that in the target timescale. I do think is that we probably have the most capable FPGA product in the Bitcoin market. None of you have not really seen what it is capable of doing as yet. I know several people have praised our product and I am glad at that response. It is always our target to please customers and the team have a large amount of pride in achieving that result. As an example very few people yet have the data for say the thermal performance and I can say it appears to have done everything that we aimed to do in the design. Particularly the rearward heat projection. That's not there by accident. The front heatsinking within the limits of the Spartan package is also working extreme well. The F12 fan can almost not be heard and absolutely blasts heat out of the board. I have personally taken a board without fan running at 80 deg to 30 deg in about 1 minute by turning on the fan. That's good measure of how good the fan/heatsink combo is.

Anyway one more piece of speculation for you. Anything that BFL can actually bring to the market we can almost certainly compete with with assuming we choose to do that. That's more a commercial decision that anything and if we had projects of better return we might do them in preference. I would say if I was running BFL I should be doing the same. That's one of the things that does not make sense in their offering. I don't need to say any more. You all have common sense I hope and you should use that and think through the difference between the way we run and BFL. I'm sure Enterpoint will still be trading in 10 years time which more that I can say for other companies.

Yohan

BFGMiner works better on FPGAs.

Spare me the self promotion.

What is the difference between cgminer and BFGMiner ?

Maybe officially endorsed by BFL ( in the Mini Rig screenshots ) 

No, that's BFGMiner.

I didn't see any bitching.  All I saw was a notice that Enterpoint had failed to comply with the license requirements for fair use and distribution of cgminer.  It wouldn't have mattered if they had done it with MPBM, then they would have been in violation of MPBM's license agreement instead of cgminer.

It was a mistake they made to release a compiled version without the source and license.  They pulled it and said they will comply with the license when it is re-released.

Hardly "Bitching".