Topic: Ultra Under-overclock image for A2 Innosilicon by Emdje - V5.0 - page 22. (Read 79786 times)

Image is posted, see first post for the link.
And let me know if it works for the guys/girls with the A2MINI Terminator

In about half an hour the pc will finish compressing an image for the A2MINI Terminator (based on the assumption of 10 chips and 54 cores per chip). Then I will upload it and post the link.

Cgminer does start on my A2, but the log shows some weirdness (character wise).
Miningrig rentals showed normal hashing.
Testing will be at you own risk. Make sure you have adequate cooling and a stable power source.
When someone test the image, please report either the successful or failed attempt. And please post logs of that attempt if possible.

Good luck I hope it works  Grin
And if it does, feel free to send me some btc: 1FiHCFcP4ovbV89nGhYEMWXSZsWnGUASSe

Page 25 - Table 6 Pin description. But who knows what does it mean... 0.93V is the target.

where does it say you can play with pin settings?
 

This is what Innosilicon released: http://www.usbminers.nl/Innosilicon_A2_PG_v120140424.pdf

I'd kill for a schematic...

I have no problem with soldering and have some basic knowledge of how such electronics works. I will start doing irresponsible things after I will not have decent income from this. I had an impression that voltage can be adjusted without changing the resistor by simply adjusting the pin settings as manual A2 states. Good luck on you experiments and keep us informed.

I've got an a2 mini from some months back and I just bought a piece of shit 'working' a2box from minereu.

anyway, my mini runs at 0.83v, the a2box cards run at ~0.865v.  as I've got two dead cards fro the a2box already, it's now a hybrid. If you want to chnage the volkatge, you need to change a smd resistor. my mini cards have one marked 04C that is 10.7k, my box cards have an 05C (11k) in the same place. It's voltage setting divider from a 2v reference. The actual voltage will be slightly lower than calculated using resistor voltage, so there's some current drain on it. Higher value resistor for more volts, lower for less volts. give it  a month and I'll be undervolting mine. There are 4 voltage regulator controllers per blade. if you aren't really good at smd soldering, I'd not recommend you even try this. I'm thinking that as you need to ask, you should probably just forget it.

I ran emdjes image for a day but then it barfed, it seems that I have a THIRD blade from the A2 box going flaky, in 3 days.
I tried reburning the image but it still not playing. I get this at the top of the screen:

[2014-08-29 03:40:46] Started cgminer 3.9.0 [2014-08-29 03:40:46] Run Reset=0 [2014-08-29 03:40:46] SPI Speed 4000 kHz [2014-08-29 03:40:46] ST MCU - Disable [2014-08-29 03:40:46] GPIO CS is OFF [2014-08-29 03:40:47] Failure(cs0)(20): missing ACK for cmd 0x04 [2014-08-29 03:40:47] ACK(cs0) timeout:cmd_RESET_BCAST-1.1678s [2014-08-29 03:40:47] No any A1 board


minereu scripta image that's been running the mini for months is still ok, it sometimes see a blade has stopped working

Re Emjde's question about the new minis

I'd be amazed if it was new silicon. There are now 10 chips per blade. My scripta firmware thinks it has 54 cores per chip, the datasheet I found suggests 27.


I would love to know how to disable voltage regulators, they have a shutdown input, but I can't work out what the chip is or find a schematic. I've got one blade that eats double it's power ration (290W at the wall), I'm hoping that by shutting a core or four down, I can get some use from it.

Gimme money! I promise to do something completely irresponsible with it: 151vBLqdvQFCGHZB86ybq5RQXVy9VuaLxQ

You are welcome. Regarding your questions, both a no 
My A2's are running nicely so I am not turning them off to take the blade apart to look how to adjust voltage. And I can't find any decent high quality pictures of them. Maybe, when you take yours apart you can make good photo's from the front and back and post them?

Thanks for nice description, will try that. By the way, did you find any way how to increase voltage that goes to the chip? And any success with new cgminer binary?

The blades can be taken apart with a Philips screwdriver. Make a rudimentary picture of the blade on a piece of paper. Place the screws on the picture where you took them from, so that everything goes back where it came from. Make a mental note of how fixed text screws were so you don't crush the chips when you place it back.
Apply thermal grease on the chip. Not to much! Put grease on the entire heatsinc and place the silicone thermal pad on it.  Now you can screw it back on. Do the same on the other side between the heatsinc and the pad. (on that side I believe you can pit grease right on the board but not shure about that). I gave pit small heatsincs on the voltage regulators as well.

Hope this helps.

Im sometimes getting overheat warnings like BA13 cs:3 A:8 -C:432 (54-54-54-54-54-54-54-54) | (5s):15.13M (avg):14.28Mh/s | A:205920 R:311 HW:13234 WU:53.9/m | T:52C*Hot*(25-25-29-30-29-29-38-52)

Seems like some chips are not covered with thermal glue as emdje mentioned. What is the process of unbundling the blade and heat sink?

Hard to say, depends on if you have a first or second generation A2 terminator. If you have one of the newer versions you might get another 10 MH/s or so (they seem more optimized). If you have one of the older versions you I believe you are pretty much at your max, maybe gain just a few MH/s. What are you running it on now?

my A2 is running at 94MH right now, whats the speed i can get to with you latest ISO ?

good suggestions... i noticed A2 chips are not heat friendly, they do tend to perform better in cooler environment.. unlike KNC Jupiter back in the days the higher the temp the better they used to run..

I did not notice anything different in the amount of course (sometimes I have all of them sometimes I 'miss' one or two).
I did notice a slightly lower temperature, the cooler it runs the better the OCing.

Did you gain anything by reapplying TIM? Less dead cores, cooler temps, better OCing etc?

When I took my blades apart I noticed that one of the chips had not much, if any, cooling paste on it. The Silicone pad was scorched because of the heat. When the blades heat up the amount of work that is sent automatically reduces. I don't know if an permanently overheated core are recognized or are regarded as dead. Might be worth to have a look. take it apart, clean the thermal grease of, apply new good thermal grease between the chips and the silicone pad. Apply grease between the silicone pad and the heat-sync as well. Off course not to much, the less the better, but enough to make contact. It might work it might not.....

emdje,

i believe same as before: username: pi password: innosilicon

btw, i loaded up your image and they indeed rock! both miners are hashing stably.. batch1 doesn't take higher values due to dead cores so i could go as far as 1220.. however batch2 can go up to 96mh easily... good job indeed!

Because this is the version that seems to work best for me. 'Don't change a winning team', they say.
I will look at your suggestions though  

bclcjunkie: What is the login and password for this one: https://www.dropbox.com/s/zz8uj8xbt8tkqi4/ltc-a2-2014-05-21.img.gz