Algorithmically placed FPGA miner: 255MH/s/chip, supports all known boards

[eldentyrell]

Glasswalker mentioned in the enterpoint thread that the bitstream he is working on doesn't require a DCM

Link?

I assume you mean he's using a PLL instead of a DCM.  I would find it pretty hard to believe he was using neither.

Ok, quick update. Unfortunately my last build once again (sigh...) had a bug in the new LVDS clocking.

Ah, differential signaling.  Okay, yes, I can see how that would let you get a ~200mhz clock into the fabric without using a DCM or PLL.  Unfortunately it means that any clock calibration has to be done at the clock-generator (I have no idea what they're using for that).

[eldentyrell]

Unfortunately the added fixes to adjust DCM locking and relocking provided no benefit.

You mean that it did not re-lock?

That's really strange; the DCM shouldn't be able to tell the difference between when it is reset on startup (when you say it IS locking) and the reset being sent to it after losing a lock (when you say it is NOT locking).

Could you please post the log file?

[eldentyrell]

If ChrisP releases his USB driver code and somebody gives me SSH access to a Linux machine with an Enterpoint board attached to it I can take over the debugging/troubleshooting from here on out.  Apparently he has gotten the board-specific communication path from the host to the chip working reliably, and that's the only part I am stubborn about not doing.  There are also quite a lot of tricks for working around this DCM weirdness, but the ones we haven't tried yet require rebuilding the bitstream (I can do that).

[eldentyrell]

TML-1.04 is released.  No new bitstreams but a major improvement in efficiency in the code that fetches new work.  You should now see actual hashrates much closer to the expected hashrate.


10.Aug.2012  Release v1.04
             add monitors for time spent on various network tasks
             performance/efficiency improvements in getwork
             add check_dcm_lock, freeze_dcm, reset_dcm_if_lock_lost
             fix leaf-clock-mux settings for rph

[eldentyrell]

If you're running a "g-series" bitstream (i.e. its name starts with the letter "g"), please try moving up to one of the h-series.

If this causes problems for you, or results in noticeably reduced performance, please contact me.

Thanks.

[Keninishna]

If ChrisP releases his USB driver code and somebody gives me SSH access to a Linux machine with an Enterpoint board attached to it I can take over the debugging/troubleshooting from here on out.  Apparently he has gotten the board-specific communication path from the host to the chip working reliably, and that's the only part I am stubborn about not doing.  There are also quite a lot of tricks for working around this DCM weirdness, but the ones we haven't tried yet require rebuilding the bitstream (I can do that).

Has Chrisp contacted you yet? or you him? The best bitstream now is getting 200 mh/s chip so a nice 55 mh/s boost could make some good btc for my 100 chips.

[eldentyrell]

Has Chrisp contacted you yet? or you him? The best bitstream now is getting 200 mh/s chip so a nice 55 mh/s boost could make some good btc for my 100 chips.

Yes, he has contacted me and has given me his code.  Unfortunately this is just about the worst possible time in terms of me having free time to integrate it -- I have zero availability until a week from today.

In return for his code I'm sending him 25% of the commissions earned from enterpoint board users.  This is a standing offer for anybody else out there thinking of implementing support for any other boards' proprietary interfaces.  But your code has to be good -- the offer doesn't apply to steaming piles.  It must be written in Java, no exceptions.  If you need to use any third-party libraries you must check with me first (before you spend a lot of time on this).  JNI is very strongly discouraged; using a java.io.Process to talk to some native-code utility (like urjtag) is strongly preferred.

[eldentyrell]

TML-1.05 is out.  Two new bitstreams "jensen" and "jaskowski".  Upgrading is very strongly encouraged.  More details when I am no longer under the gun.

[cablepair]

Has Chrisp contacted you yet? or you him? The best bitstream now is getting 200 mh/s chip so a nice 55 mh/s boost could make some good btc for my 100 chips.

Yes, he has contacted me and has given me his code.  Unfortunately this is just about the worst possible time in terms of me having free time to integrate it -- I have zero availability until a week from today.

In return for his code I'm sending him 25% of the commissions earned from enterpoint board users.  This is a standing offer for anybody else out there thinking of implementing support for any other boards' proprietary interfaces.  But your code has to be good -- the offer doesn't apply to steaming piles.  It must be written in Java, no exceptions.  If you need to use any third-party libraries you must check with me first (before you spend a lot of time on this).  JNI is very strongly discouraged; using a java.io.Process to talk to some native-code utility (like urjtag) is strongly preferred.

Hello

I would like to start a relationship with your organization and mine, to start moving forward with making the ModMiner Quad able to mine with your bitstream. Ive got some very talented programmers on my team, that could help you out - and develop some cool stuff and make a whole new (and large) group of customers able to mine with your tricone bitstream. A collaberation between the two of us could benefit both of our projects, I will contact you via email, hopefully we can get together with you and my team and put together a game plan.

Thanks Eldentyrell

-Tom

[nelisky]

Hey,

Just got myself 10 mins to try this out, seems to be working fine except I don't see much of a difference in actual share count per period when compared to same voltage/cooling and ztex firmware.

Still, there is one small issue with the pool URL parser that I'd like to note (has probably been mentioned before as it is obvious, but I couldn't find anything), and that's the usage of @ in the username (deepbit at least does this), so "http://username@mail.com:password@pit.deepbit.com:8332" will not be a valid pool url (at least on tml-1.5) because it parses "mail.com:password@pit..." as server address.

[kakobrekla]

Hey,

Just got myself 10 mins to try this out, seems to be working fine except I don't see much of a difference in actual share count per period when compared to same voltage/cooling and ztex firmware.

Well, the latest batch of bitstreams (J series) is a tad slower than H series.

ET: Any new bitstreams to try out yet?

[Keninishna]

Anything new on enterpoint boards? I can give you a remote connection to a board to test with if you need one.

[purelithium]

Hello

I would like to start a relationship with your organization and mine, to start moving forward with making the ModMiner Quad able to mine with your bitstream. Ive got some very talented programmers on my team, that could help you out - and develop some cool stuff and make a whole new (and large) group of customers able to mine with your tricone bitstream. A collaberation between the two of us could benefit both of our projects, I will contact you via email, hopefully we can get together with you and my team and put together a game plan.

Thanks Eldentyrell

-Tom

Any word on this collaboration?

[ShadesOfMarble]

Also, any news on compatibility with Enterpoint-Hardware? Last post sounded promising, but now it got silent.
At least it's no longer that urgent because 200+ MH/s bitstreams are now available... But I would like to try TML nonetheless.

[eldentyrell]

I have chrisp's code, so if somebody gives me remote access to a linux box with an Enterpoint board attached to it, I will integrate his code and make it part of the next release.

For secure communications, you can always get my current PGP key from namecoin:


  namecoind name_scan p/eldentyrell 1

[eldentyrell]

TML-1.10 has been released.

This includes a new bitstream, kreisel, which is a major performance improvement over all previous bitstreams.  I will post exact numbers after letting my mine run overnight.

[ShadesOfMarble]

I have chrisp's code, so if somebody gives me remote access to a linux box with an Enterpoint board attached to it, I will integrate his code and make it part of the next release.

Is Windows ok, too?

[Dexter770221]

TML-1.10 has been released.

This includes a new bitstream, kreisel, which is a major performance improvement over all previous bitstreams.  I will post exact numbers after letting my mine run overnight.

Please also post how to force urjtag to work. I'm trying few days now and no luck to download SVN 2027 or apply patch to it. I know, dowload linux or Cygwin and compile yourself, well I cannot. I can comunicate with Amontec JTAGkey and thats all what achieved so far. No luck with xilinx platform cable USB
I have Lancelot, that is compatible with Icarus. As I can see no one have confirmed that your bitstream is working with it. I belive it's that complicated urjtag thing. You're loosing alot of hashpower becuse of that.

[eldentyrell]

I have chrisp's code, so if somebody gives me remote access to a linux box with an Enterpoint board attached to it, I will integrate his code and make it part of the next release.

Is Windows ok, too?

Maybe as a last resort.  I got an offer of access to an Ubuntu machine, so it looks like I'm set for now.

[kakobrekla]

Hrm im suffering from a strange issue ... One day (even before 1.10) I got this message and I can't get around it.

Code:

 java -Xbootclasspath/a:ZtexBTCMiner-120221.jar -jar tml-1.10.jar ztex:0 host

_________________________________________________________________________
Tricone Mining Logic, host software v1.10

   IF YOU EXPERIENCE HIGH ERROR RATES: try running just one ring at a
   time (e.g. use 'ztex:0:0.0' on command line instead of 'ztex:0').
   If each ring works error free on its own, but you get errors when
   running all three, it means your power supply is sagging.

             adding work source host

------------------------------------------------------------------------------
Here is a list of the devices that Ztex's Java code claims it can see:
0: bus=002  device=14 (`014')  ID=221a:100
   Manufacturer="ZTEX"  Product="btcminer for ZTEX FPGA Modules"    SerialNumber="04A346E225"
   productID=10.13.1.1  fwVer=0  ifVer=1
If your board is not on the list, please contact him and tell him to fix his code.
------------------------------------------------------------------------------

[0              ] found new board ztex:0
[ztex:0         ] found new chip
[ztex:0         ] programming FPGA with bitstream kreisel
             starting long poll thread
[ztex:0         ]   done programming FPGA
H:0 X:0 E:0 T:10s   |  H:0 E:0 A:0 R:0 T:10s 

java.io.IOException: you forgot to upload the bitstream to the FPGA
        at com.triconemining.board.Ztex$ZtexChip.write(Ztex.java:206)
        at com.triconemining.board.Ztex$ZtexChip.scan(Ztex.java:191)
        at com.triconemining.board.MiningChip.read(MiningChip.java:52)
        at com.triconemining.miner.ChipWrapper.checkMagicNumber(ChipWrapper.java:294)
        at com.triconemining.miner.ChipWrapper.<init>(ChipWrapper.java:54)
        at com.triconemining.miner.BoardWrapper.getChip(BoardWrapper.java:49)
        at com.triconemining.miner.BoardWrapper.run_(BoardWrapper.java:101)
        at com.triconemining.miner.BoardWrapper.run(BoardWrapper.java:73)
        at java.lang.Thread.run(Thread.java:722)

Its not the board (been runing ztex at 240+/1.4 fine). Won't run TML again no matter what.