Cointerra Hardware Support **Unofficial

[trixter]

I have apparently confirmed that the DFU firmware file is infact just an ARM binary.  It is basically a memory dump so its not packaged as an a.out or elf or anything that way.  

I am still trying to work on this when I get some time.  I am planning on getting a loan from BTCjam (references greatly needed ) to expand my hardware holdings which in turn will allow me to free up one miner for experimentation with the firmware.  I have a few ideas that might be able to increase the speed of the cointerrible miners should certain things pan out (I dont know enough about the ASIC to know yet).  If I can increase my hardware holdings then I should have less time spent chasing the carrot to pay bills that I can dedicate some serious time to delving into the firmware and seeing if it is possible to improve performance and maybe fix some bugs.  

The first steps will be to map out the firmware.  There appears to be 108 functions that I would have to go through one at a time and figure out exactly what each one does and how it does it.  They may not all be functions, sometimes compilers make what look like functions out of blocks of code during the optimization process.  Once mapped I plan on rewriting functions one at a time in C and publishing on github under a OSI approved FOSS license (unless cointerrible is going to sue me since that might violate some copyright law somewhere given the fact  that I am doing the disassembly and spec writing and coding of the new firmware - see Compaq in reversing the IBM PC BIOS).  The DMCA clearly does not apply as there are no "locks" to be broken and reverse engineering under limited circumstances is allowed.  Technically this would help cointerrible anyway as they could use any of my published code as they see fit, but sometimes companies dislike users doing things they did not approve of (see Apple).  

It may be required that someone else code from the spec who has never seen the firmware but we can cross that bridge if cointerrible decides to issue a takedown notice.

EDIT I was wrong!!  Say it aint so.  Upon further reading the DFU file is wrapped using Ti's dfuwrap tool as part of Tivaware.  This means that to properly analyze it one must unwrap it.  I am not able to find a utility to do that so I have to read up on how its done and write a tool to undo it.  Fortunately its well documented so I should be able to figure it out.

[trixter]

I was able to unwrap the DFU header.  It flashes to address 0x00000000 so a complete rewrite.  I also noticed that the payload grew by 10kB between 0.4.5 (0.7.6) and 0.5.29 (0.7.46? something not yet published I have 3 versions).   I may not yet have it perfect, I had a problem validating the DFU suffix meaning its either not there or is in a format other than I expected it to be in.

This means that I can start looking at the code.  Hopefully the rest is more straightforward.

EDIT:
I started on this and have most of it "done".  There is the Ti library that I have to weed out, and because the padding bits are known standards for the arm (well unknown to me as this is my first arm reverse engineering project) I discovered it is thumb so if anyone else starts set IDA T=1 or whatever tools you use set it to use thumb. 

By done I mean that I have marked data sections, code, disassembled the code, know what some of the functions are, but that is the easier part.  Disassembly is rarely perfect, and converting it to equivalent C code is troublesome in most instances.  So the real work will be the next step which is extremely time consuming.  I honestly do not think I will have any time for the next week maybe more.  If anyone else feels they are qualified to actually do something meaningful with what I have done I can provide the work I have done so far and maybe get something going.  PM me if you are interested, have time and can help

[marvinmartian]

I was able to unwrap the DFU header.  It flashes to address 0x00000000 so a complete rewrite.  I also noticed that the payload grew by 10kB between 0.4.5 (0.7.6) and 0.5.29 (0.7.46? something not yet published I have 3 versions).   I may not yet have it perfect, I had a problem validating the DFU suffix meaning its either not there or is in a format other than I expected it to be in.

This means that I can start looking at the code.  Hopefully the rest is more straightforward.

You know all their firmware was on Github not long ago.  It's quite possible you may be able to find it in a net archive someplace.  If you did, it would save you a ton of trouble.

I feel silly for not grabbing it myself.

[trixter]

I did not know that, thanks.

So far all I have found is a monitor for it, and web pages.  Not the firmware that gets loaded onto the mining boards.  I have tried various different search terms, do you know what the project was called?

I even searched for some strings in the CTA firmware to see if google can find anything, it couldnt.  I tried several strings, just the string and nothing else.  Are you certain it was the CTA (miner board) firmware and not something else that was published?  I cant believe anything got deleted from the internet and I cant believe that google cant find it if its there and available publicly.

[marvinmartian]

I'm pretty sure it was the code for the CTAs, yes.  It may have even been linked on a thread on BTC forum.  I can't recall but perhaps someone here can find the link.  May have been another forum.  I'll keep looking too.

[thevinci]

My piece of shit cointerra died today. I am going to take a sledgehammer and smash it. Then I am going to douse it with gasoline and burn it.

[trixter]

My piece of shit cointerra died today. I am going to take a sledgehammer and smash it. Then I am going to douse it with gasoline and burn it.

A friends half died yesterday.  It seems they had the 30 day warranty (originally 3 months) because they knew there were problems that gave them a short life.  Is it a total death or is one CTA board still good?  I may be interested in the working board.

[marvinmartian]

My piece of shit cointerra died today. I am going to take a sledgehammer and smash it. Then I am going to douse it with gasoline and burn it.

I suspect it can be fixed, even if partially.  If not, you can sell it on eBay either as one unit or part it out.  EG., the power supplies sell for almost $300 ea.

[marvinmartian]

My piece of shit cointerra died today. I am going to take a sledgehammer and smash it. Then I am going to douse it with gasoline and burn it.

A friends half died yesterday.  It seems they had the 30 day warranty (originally 3 months) because they knew there were problems that gave them a short life.  Is it a total death or is one CTA board still good?  I may be interested in the working board.

Try disconnecting the "black and white" cable that goes to the PSU.  Google around (and on this site) for pics and other info.  It's the most common cure for the "one dead board" syndrome.

[Mattster28]

Well, the day has come where it costs more money to run my Terraminer than it makes. Time to shut her down. 
Its been fun. GL all.

[derel1cte]

Well, the day has come where it costs more money to run my Terraminer than it makes. Time to shut her down. 
Its been fun. GL all.

I think you need to check you math.
1.6Ths @ Diff of 23B is still 0.034BTC per day, about $16.50
2200W costs $9.00 per dat at 0.17 $/kwh.

You're still making $7.50 a day.

[Mattster28]

I think you need to check you math.
1.6Ths @ Diff of 23B is still 0.034BTC per day, about $16.50
2200W costs $9.00 per dat at 0.17 $/kwh.

You're still making $7.50 a day.

How can you come on here and tell me to check my math? Do you know where I'm located and how much I pay for electricity? I didn't think so.

[derel1cte]

How can you come on here and tell me to check my math? Do you know where I'm located and how much I pay for electricity? I didn't think so.

Ok... so what are you paying for electricity?

Mining is barely profitable for me here in the states paying $0.17 per kwh. If you are paying more than that for power IDK why you'd even attempt to mine. Unless you are paying more than $0.32 you are still breaking even. Factor in speculation that the $/BTC increases, the TerraMiner is still profitable for now.

[Mattster28]

Ok... so what are you paying for electricity?

Mining is barely profitable for me here in the states paying $0.17 per kwh. If you are paying more than that for power IDK why you'd even attempt to mine. Unless you are paying more than $0.32 you are still breaking even. Factor in speculation that the $/BTC increases, the TerraMiner is still profitable for now.

Do you really need to know the whole story? Ok well listen. When I started out mining my electric rates were $0.19 per kw. After I added the Terraminers in March my electric usage went up considerably and every month my electric company threatened to raise my rates. Well they finally did it last month. They raised it to $0.34 kw. Happy now?

[derel1cte]

Do you really need to know the whole story? Ok well listen. When I started out mining my electric rates were $0.19 per kw. After I added the Terraminers in March my electric usage went up considerably and every month my electric company threatened to raise my rates. Well they finally did it last month. They raised it to $0.34 kw. Happy now?

Yeah, actually. I wasn't trying to attack you. It just seemed strange to declare a piece of hardware unprofitable when it is really only a unique (and very shitty) situation that caused it to no longer be viable. Sorry to hear about your power company. Hopefully you can find some co-location or sell it to someone in an area where power is more affordable.

[Mattster28]

Yeah, actually. I wasn't trying to attack you. It just seemed strange to declare a piece of hardware unprofitable when it is really only a unique (and very shitty) situation that caused it to no longer be viable. Sorry to hear about your power company. Hopefully you can find some co-location or sell it to someone in an area where power is more affordable.

It's all good. And yes it is a shitty situation.

[wls]

I've tried to re-thermal past one of my units, just where one of the groups of four are bad on both boards. The result was no change in temperatures, which leads me  to believe my coolant must be low or the cooling blocks are clogged. How do you add more coolant to the system? Is their a plug to top off the system or do I have to cut and patch a hose to do it?

Thanks!
Posted from Bitcointa.lk - #ZS4Ze8HI51nhwAMK

[trixter]

I've tried to re-thermal past one of my units, just where one of the groups of four are bad on both boards. The result was no change in temperatures, which leads me  to believe my coolant must be low or the cooling blocks are clogged. How do you add more coolant to the system? Is their a plug to top off the system or do I have to cut and patch a hose to do it?

Thanks!
Posted from Bitcointa.lk - #ZS4Ze8HI51nhwAMK

According to cointerrible you never ever need to because magic.

Its a sealed system so you are going to have to yank off one of the hoses and hope that you can seal it back up when you reattach it.  I have never done this as I have not appeared to require it.  Others have reported leaks so hopefully they can chime in.

[keeek]

My „Early January Batch“ Terraminer IV stopped hashing on one board after 5 months constant hashing. About every 4 weeks I cleaned the grid at the front side of the miner. These cleanings have been the only shutdowns of the miner. I kept the fingers crossed at the reboot, but it did not help. After the last cleaning only one board started hashing again. You could see both boards on the status page, but just one was hashing. At the faulty board pump speed was zero. I thought about just leaving it as it is, but now I have a new hosting possibility at much cheaper power costs... so I decided to try to change the pumps!

First of all... It worked! Both boards are hashing at full speed again (about 1.6 Th/s):

You can not buy the Terraminer water cooling system at a electronic store. They put two pumps on one giant radiator. So I bought two Corsair H60 instead (about 60 Euro each). Each H60 has a pump, a fan and a radiator.

To place these two radiators inside the miner, I removed one of the fans in the back. After removing this fan, there is enough space for two new radiators + fans. The next step is removing the old cooling system. Then I removed the old paste from the 2x 4 ASICs. The new pump came with heat transfer paste on the lower side of the pump. I thought that this is fine and installed the two H60s. This was a bad idea, because there is a frame around the 4 ASIC chips. This frame is slightly higher than the ASICs. Result: The ASICs had no contact to the pump and shutdown immediately because of heat. At least they did not get damaged.

The second try: I removed the H60s again. Then I removed the heat transfer paste from the pumps. Now I put Coollaboratory Liquid Ultra on the ASICs. One package lasts for 8 ASICs. After installing the new water cooling system I looked out for a place for the old fan from the back. It is not a perfect installation, but now it is mounted at the front side.


Beautiful silver ASICs:






Back side with one of the new radiators:




The new cooling system installed:












... and finally - the old fan at the front and the miner hashing again:

[trixter]

keeek: thanks for the pics. 

More  generically for other problems ... To aid in diagnosing these problems like pump failures there is a cgminer log file you can access with ssh which will contain error messages, for an example of the previous post error it will be
"Pump bad" (which will contain numbers letting you know which board and pump).

There are many errors that get reported that can help a bit, or perhaps just muddy the waters.

the cgminer log is in /var/log/cgminer.log (symlink to /Angstrom/Cointerra/logs/cgminer.log).  This might help you diagnose what exactly is going on with the unit so you know what corrective action to take.

The web interface may give pointers (such as was pointed out by the prior post) however the log file can give you more information when the web interface fails to provide anything meaningful.