Topic: Experimenting with Jalapeno firmware... - page 12. (Read 62632 times)

yes, but take the 1.2.6+ firmware due to XLINK

   You buy this open h/w and more use than Avr Dragon jtag http://www.seeedstudio.com/depot/bus-blaster-v3-p-1415.html?cPath=63_69


ps: Just restocks few days ago

Can we program the long boards (Little single / Single) with the Avalon DRAGON successfully?

Anyone have a dragon to sell me on the cheap? 40-45 bucks shipped

let me know

Danny

I have read that on BFL forum, but can't provide a link, also i don't remember exactly if the problem was with the 'rest of the chips' or the 'rest of the engines'.
You should look in the source of the firmware, not in the schematics

cgminer already does fan control as needed

Nothing is wasted. They already deliver over spec.

Interesting. I'm reading the first time about this. Where do you have this information from (link) ?

So this would mean that about ~6% hashing power of every ASIC is wasted, if multiple of them are on one PCB in a chain.
However, if a multiple ASICs per PCB design is handling the single ASIC's separately (not changed), it would have an advantage of about +6% hashing power over the other design, right?

I know that on the BFL PCB is also a FPGA or CPLD. I guess the chip-routing is going through this device and can be changed. Maybe with a another logic and an firmware update, the Engines 0 could be used. I have to take a look into the schematics.

engine 0 not working is a design limitation. If you have a single chip in the device you can use engine 0, but if you enable engine 0 on the first chip from a chain - you loose the rest of the chips

Does anyone know why it looks from all user-reports that engine 0 doesn't work?
In my 16-Asic Singles all engines 0 are disabled. But why?
Regarding the ASIC Datasheet from Butterflylabs all ASIC's should have 16 engines.
Also a look on to the ASIC Mask picture is clearly showing 16 (nearly) identical engines on the chip.

It's hard to believe that on all NON-Grad A chips engine 0 is not working. It would make much more sense if a more or less random engine doesn't work.

Does anyone know more about?

fans are always @ max at the jalapeno model. engine 0 wont work
using nonworking engines is also bad, you cant overclock as high due to the voltage drop.

Thanks Ashitank.

Thanks for pointing that out.

Another tip, if you want to hardcode all engines to be running (regardless it's functional or not) change __TOTAL_DIAGNOSTICS_RUN from 10 to 0. It will skip all tests on the engines. Beware of the high HW error rate.

Next thing I wanna try is to change DO_NOT_USE_ENGINE_ZERO and overclocking it to 350mhz on the frequency and see what will happen. Hopefully it won't brick 

Last but not least to max out the fan speed (FAN_CONTROL_BYTE_REMAIN_FULL_SPEED) if the above is successful.

"java API stats"

And it lists all the pools after the devices, so yes the last device (i.e. where [ID] != POOL* which come after the devices)

Very Nicely done Elitepork , Great Post 

Additional information.

How to test for the max speed your jallies can do.
1) I set the __ASIC_FREQUENCY_ACTUAL_INDEX to 1, comment out __RUN_HEAVY_DIAGNOSTICS_ON_EACH_ENGINE, and flash it to the jally. Boot up the jally
2) use the cgminer "java API stats" and look at the last miner (that should be last one due to hotplug)
3) you will see the numbers of engines that are functional. (for mine, it's 15 + 14)
4) Set __ASIC_FREQUENCY_ACTUAL_INDEX to 9 and flash it to the jally. Boot up the jally
5) use the cgminer "java API stat" and look at the last miner (that should be last one due to hotplug). If it is the same number as step 3) you got the max speed running with the most engine functional. If not, go to step 4 with __ASIC_FREQUENCY_ACTUAL_INDEX reduced by 1.

Hopes this help for others.

Did a photo shoot with the jallies over the long weekend over here. In this photo shoot, I have used this setup to flash the firmware while mining and it does not need to remove or change anything. Just that once you finish your flashing, put it back to the original shape...

These are the tools I used. There's a missing USB hub cable that used to connect the AVR to the laptop...


This is my jally. Working @ 5.5GH/s before the flash.


Removed the top cover using the hex screwdriver.


Then remove the fan.


Removed the 4 washers and "pillars" from the 4 long screws.


Now, removed the base and the 4 screws.


Flip it over and you will see this.


Unscrewed the 2 screws using the same hex screwdriver and you should see this


Now, put the plate on top and the heatsink below. Screw them back. Note that the screws should be loosely tight (turn until you feel a small resistance) and do not screw too tight.


Now, put back the base and 3 screws (less the screw that is on the fan side of the board)


Plug in the fan cable.


Then plug in the cable (can be 10 pin, can be 40 pins. I searched around in my home and I only managed to find the cable used for my floppy drive so I used that). You should be able to plug both in.


Now, put the fan back in. Using the 3 screws, make sure it is directly above. Do a manual spin (meaning use your hand/fingers to spin the fan) to make sure the blades are not hitting anything.


Now, turn back the jallies on and connect the AVR dragon to your PC/laptop.


Viola, now you can flash and mine without much hassle. Noted that after flashing, you might need to hard reboot your jally by turning its power off and back on.

After flashing, 7.8GH/s on one, 8.0 on the other.

get ck's -> http://ck.kolivas.org/apps/cgminer/BitForce_SC/
ive build myself a similiar one, which uses all engines responding correctly, if you want i can supply you with the .hex/.elf, you can also choose frequency of course

You may try ck firmware. I think it tries to use more engines every time you turn it on.

hey all, just wanted to say thank you!!! to the OP! This stuff is great, lol.

So I have two jally's and both have been upgraded to the 1.2.5 Firmware. I started with the image provided, but in the end compiled by own version of the firmware. However, the first questions I have is that to matter the index setting, 7,8,or 9 the hash rate does not change on the miners. Not sure why the changes I make are not uploading to the Jally. I know I have chosen the correct image when compiled to load up the jally, but something just is not right.

Also, I believe the chips I have a "lower" quality chips, as one is maxing out aroung 7.8 and the other around 7.2 with a higher error rate than the other, here is the

GetInfo result for Jally 1:

And here is Jally 2 GetInfo:

Any idea on how to get all the engines running on the first Jally so it can match the others performance? I see in the other precompiled images, there was something done to perform less checks on the procs, but not sure on all of it, just the index. Anyone point me in the right direction please?

Thanks in advance and thanks for everyones hard work!!!

I'm sorry, I cant help you with that, if you have that problem take it with Josh, not with me.