Official Open Source FPGA Bitcoin Miner (Last Update: April 14th, 2013) - page 32.

TheSeven

hero member

Activity: 504

Merit: 500

FPGA Mining LLC

Quote from: OrphanedGland on July 02, 2011, 11:53:01 AM

440MH/s is the raw underlying hash rate for your configuration, but you need to consider delays associated with presenting the golden nonce, getting new data from the pool etc. You would never be able to achieve 440MH/s in practice.

Actually making that run at 100% efficiency is fairly trivial, and my Xilinx design is already doing it. Just upload the next work when the old one is like 80% processed, and while you upload it, the FPGA will continue to old one, and still report shares if it found them during the upload.

440MH/s? What kind of badass FPGA is that?

makomk

hero member

Activity: 686

Merit: 564

Quote from: themike5000 on July 02, 2011, 06:10:05 PM

My linear tech chips are running very hot, especially the one circled. Two of them are too hot to touch for longer than a second or two. Should I try to cool it off?

The hottest one is a Linear LTM4601V.

Hmmmm. The data sheet for that chip suggests that maximum safe case temperature is 100 degrees centrigrade and that it should probably be able to operate without forced cooling or a heatsink across the full rated amperage range at typical room temperatures, so I don't know... check the temperature and see, or just point a fan across it?

themike5000

member

Activity: 99

Merit: 10

My linear tech chips are running very hot, especially the one circled. Two of them are too hot to touch for longer than a second or two. Should I try to cool it off?

The hottest one is a Linear LTM4601V.

lame.duck

legendary

Activity: 1270

Merit: 1000

Quote from: OrphanedGland on July 02, 2011, 11:53:01 AM

440MH/s is the raw underlying hash rate for your configuration, but you need to consider delays associated with presenting the golden nonce, getting new data from the pool etc. You would never be able to achieve 440MH/s in practice.

Well, there would be some sort of FIFO @ the input that holds the next workload. Since with high hashing speed the performance loss will rise there could be some demand for such a solution.

comboy

sr. member

Activity: 247

Merit: 252

Quote from: themike5000 on July 02, 2011, 06:25:14 AM

Quote from: comboy on July 02, 2011, 05:09:18 AM

To me, themike5000's output looks good, I have very similar one.

I was just wondering, how can I list devices name to know what to put into mine.tcl? I see "hardware_name" in Quartus so that's not a problem, but I don't know where can I check "device_name". The first part in the device name seems to be models that Quartus are asking me too choose between when plugging the device, but what is this 0x020F70DD? I thought maybe checksum but it doesn't seem to fit. Is there any way I can list all devices in a format that I'll be able to use in this param?

Run the programfpga.bat file. It will find any attached devices and list them in that format.

I'm on linux, but yes, I found what I needed there, thanks.

OrphanedGland

member

Activity: 70

Merit: 10

440MH/s is the raw underlying hash rate for your configuration, but you need to consider delays associated with presenting the golden nonce, getting new data from the pool etc. You would never be able to achieve 440MH/s in practice.

lame.duck

legendary

Activity: 1270

Merit: 1000

yes, on each core.

themike5000

member

Activity: 99

Merit: 10

Quote from: lame.duck on July 02, 2011, 11:20:45 AM

Quote from: themike5000 on July 02, 2011, 10:46:44 AM

Any way to find a more accurate hashrate output than relying on bitcoins.lc to calculate for me? Am I running at 150 or 220? I'm confused...

You are running @220MHz and with a lully unrolled design you should have a hashrate of 220 MHash/s

220MHash/s on each core though, right?

lame.duck

legendary

Activity: 1270

Merit: 1000

Quote from: themike5000 on July 02, 2011, 10:46:44 AM

Any way to find a more accurate hashrate output than relying on bitcoins.lc to calculate for me? Am I running at 150 or 220? I'm confused...

You are running @220MHz and with a lully unrolled design you should have a hashrate of 220 MHash/s

themike5000

member

Activity: 99

Merit: 10

Quote from: LazarusLong on July 02, 2011, 06:42:19 AM

"result: false" looks like the share was not accepted.
Accepted shares normally should have a result "true" in the JSON response

It seems like its working. I have it working for bitcoins.lc because it tells me the hashrate of the workers (any other way to do this in the software?). It leveled out a around 300Mhash/second. That seems odd because its running on two cores and should be at 220 clock rate (see picture)

Any way to find a more accurate hashrate output than relying on bitcoins.lc to calculate for me? Am I running at 150 or 220? I'm confused...

LazarusLong

newbie

Activity: 16

Merit: 0

"result: false" looks like the share was not accepted.
Accepted shares normally should have a result "true" in the JSON response

themike5000

member

Activity: 99

Merit: 10

Quote from: comboy on July 02, 2011, 05:09:18 AM

To me, themike5000's output looks good, I have very similar one.

I was just wondering, how can I list devices name to know what to put into mine.tcl? I see "hardware_name" in Quartus so that's not a problem, but I don't know where can I check "device_name". The first part in the device name seems to be models that Quartus are asking me too choose between when plugging the device, but what is this 0x020F70DD? I thought maybe checksum but it doesn't seem to fit. Is there any way I can list all devices in a format that I'll be able to use in this param?

Run the programfpga.bat file. It will find any attached devices and list them in that format.

I'm going to try to run it for a bit longer and see if slush recognizes my output as work.

lame.duck

legendary

Activity: 1270

Merit: 1000

If i remember correctly, pro programming script has some provisions to detect what Jtag-adapter an which device on each adaper are connected. Eventually i added some output lines or just commented them out (don't remember exactly). But you can also use the scheme give in the mine.tcl script (even i failed to do so manually) when copying it did work.
Eventually we could trying to find a canonical naming scheme for the bitstreams so the software could select from the jtag-id.

comboy

sr. member

Activity: 247

Merit: 252

To me, themike5000's output looks good, I have very similar one.

I was just wondering, how can I list devices name to know what to put into mine.tcl? I see "hardware_name" in Quartus so that's not a problem, but I don't know where can I check "device_name". The first part in the device name seems to be models that Quartus are asking me too choose between when plugging the device, but what is this 0x020F70DD? I thought maybe checksum but it doesn't seem to fit. Is there any way I can list all devices in a format that I'll be able to use in this param?

TheSeven

hero member

Activity: 504

Merit: 500

FPGA Mining LLC

Sounds like either the FPGA design has a bug that makes it calculate garbage, or the nonce offset value on the client side doesn't fit.

themike5000

member

Activity: 99

Merit: 10

I adapted the "program-fpga-board.bat" program to load "OrphanGland's" .sof file for the stratix 4 boards onto my fpga (my chip is the 230, his is the 530). Then I changed the minc.tcl file to point towards my device. I run the program and this is what spits out:

What am I supposed to see?
(I put in my slush credentials into the file as well)

TheSeven

hero member

Activity: 504

Merit: 500

FPGA Mining LLC

Quote from: comboy on June 30, 2011, 11:14:53 AM

Quote from: lame.duck on June 30, 2011, 10:14:51 AM

Quote from: comboy on June 30, 2011, 09:26:08 AM

I've been trying to run this miner from repo under linux, but I have problems with TclUrl. I did not have it, so I downloaded it and compiled, but now I get:

ERROR: Can't load library: libTclCurl7.19.6.so. The operating system reports the following error: libTclCurl7.19.6.so: wrong ELF class: ELFCLASS64

any hints on how to solve it?

Sounds to me as you try to use a 64bit lib on a 32 bit linux. For me, your aproach works fine.

That was a good hint. I still don't understand it though. I'm on 64bit, I've compiled it and also tried debian 64bit binary, in both cases getting mentioned error. After your post I gave some 32bit binary a try (on my 64bit system), and it worked.

Sounds like you're running 32bit Tcl on a 64bit system and tried to load a 64bit module, which won't work.

comboy

sr. member

Activity: 247

Merit: 252

Quote from: lame.duck on June 30, 2011, 10:14:51 AM

Quote from: comboy on June 30, 2011, 09:26:08 AM

I've been trying to run this miner from repo under linux, but I have problems with TclUrl. I did not have it, so I downloaded it and compiled, but now I get:

ERROR: Can't load library: libTclCurl7.19.6.so. The operating system reports the following error: libTclCurl7.19.6.so: wrong ELF class: ELFCLASS64

any hints on how to solve it?

Sounds to me as you try to use a 64bit lib on a 32 bit linux. For me, your aproach works fine.

That was a good hint. I still don't understand it though. I'm on 64bit, I've compiled it and also tried debian 64bit binary, in both cases getting mentioned error. After your post I gave some 32bit binary a try (on my 64bit system), and it worked. And so the adventure begins

It definitely needs some leds blinking Cool

I was also wondering, do these chips have any built-in temperature sensor?

lame.duck

legendary

Activity: 1270

Merit: 1000

Quote from: comboy on June 30, 2011, 09:26:08 AM

I've been trying to run this miner from repo under linux, but I have problems with TclUrl. I did not have it, so I downloaded it and compiled, but now I get:

ERROR: Can't load library: libTclCurl7.19.6.so. The operating system reports the following error: libTclCurl7.19.6.so: wrong ELF class: ELFCLASS64

any hints on how to solve it?

Sounds to me as you try to use a 64bit lib on a 32 bit linux. For me, your aproach works fine.

comboy

sr. member

Activity: 247

Merit: 252

I've been trying to run this miner from repo under linux, but I have problems with TclUrl. I did not have it, so I downloaded it and compiled, but now I get:

ERROR: Can't load library: libTclCurl7.19.6.so. The operating system reports the following error: libTclCurl7.19.6.so: wrong ELF class: ELFCLASS64

any hints on how to solve it?

Topic: Official Open Source FPGA Bitcoin Miner (Last Update: April 14th, 2013) - page 32. (Read 432950 times)