Topic: [HOWTO] flash your jalapeno to 8+ ghs - page 14. (Read 85144 times)
I successfully flashed my Jally as well. I noticed that the voltage reading on CGMINER is floating around 3.95-4.0v. Is this normal?
Got my AVR Dragon last night and successfully flashed my "hot mosfet" Jalapeno with the heatpipe heatsink.
Speed went 5.2 -> 7.750 GH/s 49% increase
Power consumption 32 -> 45W 41% increase
Temperature 29 -> 34C
Utility @ diff 4 18.24 -> 26.98 44% increase
HW Errors 0 -> 142 in 19 hours with 30415 diff4 shared accepted, 0.11% error rate
Speed went 5.2 -> 7.750 GH/s 49% increase
Power consumption 32 -> 45W 41% increase
Temperature 29 -> 34C
Utility @ diff 4 18.24 -> 26.98 44% increase
HW Errors 0 -> 142 in 19 hours with 30415 diff4 shared accepted, 0.11% error rate
Well done, not sure if heatpipe necessarily signifies what the MOSFETs are, but keep a very good eye on them.
Meaning there may have been a different number of old single sinks and hot MOSFET boards to use.
What is ND?
Also, it seems cheapest direct from atmel.
Also, it seems cheapest direct from atmel.
You must put diff at 7
Got my AVR Dragon last night and successfully flashed my "hot mosfet" Jalapeno with the heatpipe heatsink.
Speed went 5.2 -> 7.750 GH/s 49% increase
Power consumption 32 -> 45W 41% increase
Temperature 29 -> 34C
Utility @ diff 4 18.24 -> 26.98 44% increase
HW Errors 0 -> 142 in 19 hours with 30415 diff4 shared accepted, 0.11% error rate
Very happy with the results
Might try tweaking with the speed and binning code later
Speed went 5.2 -> 7.750 GH/s 49% increase
Power consumption 32 -> 45W 41% increase
Temperature 29 -> 34C
Utility @ diff 4 18.24 -> 26.98 44% increase
HW Errors 0 -> 142 in 19 hours with 30415 diff4 shared accepted, 0.11% error rate
Very happy with the results
Might try tweaking with the speed and binning code later
Very nice work.
after this flashing process that seems to be simple enough, does the amount of errors increase?
just to add, I tried setting
#define __ASIC_FREQUENCY_ACTUAL_INDEX 8
in std_defs.h
and flashed the 7GH Jalapeno (with binned chips) and was able to get 14 engines @309MHz and 15 engines @267MHz
http://imgur.com/peKactI
however the error count was in the 1000s and showing at 5%
anyhow, went back to the BFL released hex file
#define __ASIC_FREQUENCY_ACTUAL_INDEX 8
in std_defs.h
and flashed the 7GH Jalapeno (with binned chips) and was able to get 14 engines @309MHz and 15 engines @267MHz
http://imgur.com/peKactI
however the error count was in the 1000s and showing at 5%
anyhow, went back to the BFL released hex file
Here is precompiled firmware that will try to enable all the engines for you if you can't get up to the ~8.2GH that should be doable:
http://ck.kolivas.org/apps/cgminer/BitForce_SC/BitForce_SC_1.2.5ck.elf.zip
(Unzip it and load the file BitForce_SC.elf into the "Flash(128KB)" box after the step where you click on "Memories" as kano posted above.)
It's just like the old unlocked shaders thing with 6950s, some of them will just run like 6970s anyway. Usual warnings apply, but I am using this firmware on my own, so it's not a precompiled brick.
http://ck.kolivas.org/apps/cgminer/BitForce_SC/BitForce_SC_1.2.5ck.elf.zip
(Unzip it and load the file BitForce_SC.elf into the "Flash(128KB)" box after the step where you click on "Memories" as kano posted above.)
It's just like the old unlocked shaders thing with 6950s, some of them will just run like 6970s anyway. Usual warnings apply, but I am using this firmware on my own, so it's not a precompiled brick.
thanks ckolivas.
Is the only change
Code:
#elif (__HEAVY_DIAGNOSTICS_MODERATE_3_NONCES)
if (iReadBackNonceCount < 4)
{
// This engine is DEAD!
bWasAnyEngineDecommissioned = TRUE;
DECOMMISSION_PROCESSOR(cDiagChip, cDiagEngine);
}
#endif
from
Code:
if (iReadBackNonceCount < 4)
Code:
if (iReadBackNonceCount < 2)
or did you tweak other things in there as well?
Just got my AVR dragon, connected it to to AVR studio and did an upgrade for the Dragon. All good so far. Then erased my Jalapeno, also good. Then tried to flash it using Kano's settings...ERROR. Something about not being able to execute a command. I had a mini-heartattack . Then i tried reading info from the Jalapeno atmel chip. Turned out i have a different model chip on my Jalapeno then Kano. Mine was a AT32UC3A1256 instead of a AT32UC3A1128. After i selected the correct chip in AVR studio it flashed perfectly.
Went from 5.2Gh/s to 7.5Gh/s with same error count!
I managed to break 2 blades from the BFL fan (cheap quality plastic) while doing this. Be careful with removing it or powering it while not mounted on the heatsink.
Went from 5.2Gh/s to 7.5Gh/s with same error count!
I managed to break 2 blades from the BFL fan (cheap quality plastic) while doing this. Be careful with removing it or powering it while not mounted on the heatsink.
Ordered my programmer. Can't wait for it to arrive
My friend and I have units to flash! I'm curious how mine will go. It mines at 6.15GH/s all day long.
My friend and I have units to flash! I'm curious how mine will go. It mines at 6.15GH/s all day long.
Just to give you another datapoint, i got my jala to turn on 30 engines by cooling it and retrying. I agree that the checks seem a bit arbitrary if i can restart it 5 times and get 30, 28, 28, 29, 29 engines on 5 consecutive restarts.
For now, im good with getting 29 engines most of the time.
For now, im good with getting 29 engines most of the time.
Then it won't decommission any engines regardless. My firmware was slightly more conservative and made sure the engine returned at least one valid nonce.
Avr studio wont work with simple jtag and if you are going to build simple jtag watch out might damage your Jal or short.
ps: you will need program s/w for jtag.
ps: you will need program s/w for jtag.
The firmware tests the engine (and chip if you have multiple chips) performance at startup and adjusts the speed accordingly based on how "correct" the answers are from each engine. It is not unusual for the existing Jals out there to be somewhere between 27 and 30 units when run at the higher speeds. 30 is the most that can be activated on 2+ chip units. When run at their lowest speed as 5GH units, most of the engines seem to work making all 30 start at 189MHz ~5.7GH. I posted firmware which relaxes the test dramatically making it far more likely to activate all 30 engines at speed setting "7" which is 274MHz, so 30 engines at 274 ~8.2 GH. I can't comment on the binned versus unbinned policy for chips by BFL but it looks like earlier units were unbinned making them potentially full speed, while later ones were chosen to reach the guaranteed speed as a minimum.
Unfortunately some of the hardware errors are currently counted towards the hashrate so you cannot simply look at the hashrate and determine whether you're creating all good work. I'm planning on rectifying that next cgminer version.
Can you be so kind to please post the code to the modified firmware, since I want to test my Jally at speeds "9" and using all engines. Unfortunately some of the hardware errors are currently counted towards the hashrate so you cannot simply look at the hashrate and determine whether you're creating all good work. I'm planning on rectifying that next cgminer version.
Just comment out these lines in ASIC_Engine.c
Code:
if (iReadBackNonceCount < 4)
{
// This engine is DEAD!
bWasAnyEngineDecommissioned = TRUE;
DECOMMISSION_PROCESSOR(cDiagChip, cDiagEngine);
}
btw. i dont think you need a dragon to programm a simple avr.
just build a serial cable and you are ready to go...
avrs are very simple interface and if you have old motherboard with serial port, thats all you need.
or you can buy some simple usb mini avr programmers for some euros only.
they should also work.
i post you later a simple shematic for the avr cable.
its realy easy cable with no electeonics needed (some resistors would be good but not needed).
i programmed many flasher boxes for phones with that already.
this boxes had atmega 8 and atmega 128 chips on board.
is this possible? i will really like to know since i was gonna order my dragon this week
Soooo excited... I cant wait to rape my jallly then have it rape the difficulty
I'm going to spitball and say that big single sinks with the supplied fan are going to be doing something like 0.3 c/watt 30 ish above ambient. I'm guessing that fan does about 40 CFM. I am guessing again that if you upgrade the fan to something loud and grunty you could get the cooling system performance to something closer to 0.2 C/W... this would give you approx 50W cooling headroom, which I think would be approx 25% overclock. Provided i) you get lucky with the chips in terms of heat and power consumption ii) you get lucky with the MOSFETs in terms of heat and power consumption.
Anyway, I do believe we'll see at least one or two adventurous souls managing to get 80Gh/s out of a big single, but how long they will last at that output is open to conjecture.
The little singles should be much easier, but depending on how the yield falls, might have higher clocked chips with less active engines in the first place. We're seeing jallies with 15 engine chips but has that just been to get orders out the door?
Anyway, I do believe we'll see at least one or two adventurous souls managing to get 80Gh/s out of a big single, but how long they will last at that output is open to conjecture.
The little singles should be much easier, but depending on how the yield falls, might have higher clocked chips with less active engines in the first place. We're seeing jallies with 15 engine chips but has that just been to get orders out the door?
Cool guide. Got one for us Single owners? OH that's right only about 10 people have their Singles at this point so what's the point?
Don't rub it in 
I meant to say Got One for Us Single Preorder Owners
BFL being awarded the Most Professional Company ever award just decides to ship Jalapeno's and completely disregarded their other product line
Do they get lifetime warranty?
Maybe not after you mod the firmware and push it hard. They certainly would not be required to do so. I imagine when these break because of overclocking they will have a pretty similar characteristic. So if you are doing this, know what you are doing, don't go crazy, and you MUST upgrade the cooling. This includes all parts further stressed by more heat, not just the ASIC chips themselves.
The later devices, 25-60GH/s I'd firstly be surprised if they could be clocked much faster, and secondly be worried that it would kill it very quickly without major work on the hardware cooling.
Obviously they will already have the 1.2.5 or later firmware so this simple reflash wont make any difference, but messing with the firmware to clock it up higher ... I can see ASIC death and destruction on the horizon with the bigger devices
i think at 9
you will have a lot of errors
or not???
you will have a lot of errors
or not???
Jump to: