Power requirement is about 9.5W (=8.5W*sqr(1.27/1.23)*200/192)
Solid. Still above 20MH/W. Sure beats the ~2.5MH/W I get with GPUs now.
Topic: BTCMiner - Open Source Bitcoin Miner for ZTEX FPGA Boards, 215 MH/s on LX150 - page 28. (Read 161727 times)
November 08, 2011, 05:13:39 PM
Solid. Still above 20MH/W. Sure beats the ~2.5MH/W I get with GPUs now.
November 08, 2011, 05:10:38 PM
Would like to know the amp draw as well. If it is the same it would be 8.78 watts at 200MHz
Pls remove my private data.
Power requirement is about 9.5W (=8.5W*sqr(1.27/1.23)*200/192)
November 08, 2011, 04:38:15 PM
Is there any chance we can sweet talk you into dropping a 1.26v resistor onto one to see the results? =)
I'll try this out next week. I tested it with a prototype. Here are the results:
At 1.23V core voltage (default, R12=1.3kΩ, R13=2.4kΩ): 192 MHz @ 0.0% error rate
At 1.27V core voltage (R12=1.3kΩ, R13=2.2kΩ): 200 MHz @ 0.0% error rate
Those who increase the core voltage (by replacing R12 and/or R13) have to do this on their own risk. The absolute maximum voltage according to the specs is 1.32V. But voltage overshoot has to be taken into account. These overshoots occur at load removal (FPGA reset) and are caused by the energy stored in the coil. The overshoot voltage at a nominal voltage of 1.27V is larger than 1.32V.
On the other hand, the limit of 1.32V is calculated by 1.2V + 10%, i.e. this value is (at least a little bit) arbitrary. IMHO 1.27V is still quite save, but I cannot guarantee for it.
Nice to know. You didn't by any chance get power draw (amps) did you?
November 08, 2011, 04:30:23 PM
Is there any chance we can sweet talk you into dropping a 1.26v resistor onto one to see the results? =)
I'll try this out next week. I tested it with a prototype. Here are the results:
At 1.23V core voltage (default, R12=1.3kΩ, R13=2.4kΩ): 192 MHz @ 0.0% error rate
At 1.27V core voltage (R12=1.3kΩ, R13=2.2kΩ): 200 MHz @ 0.0% error rate
Those who increase the core voltage (by replacing R12 and/or R13) have to do this on their own risk. The absolute maximum voltage according to the specs is 1.32V. But voltage overshoot has to be taken into account. These overshoots occur at load removal (FPGA reset) and are caused by the energy stored in the coil. The overshoot voltage at a nominal voltage of 1.27V is larger than 1.32V.
On the other hand, the limit of 1.32V is calculated by 1.2V + 10%, i.e. this value is (at least a little bit) arbitrary. IMHO 1.27V is still quite save, but I cannot guarantee for it.
November 07, 2011, 09:52:42 AM
I got a question about that. Does GPL protect your IP from a "covert distro". Meaning someone take your code, tweaks it to get 3% more performance and then sells FGPA already loaded with it. They still have to release the source code of their improvements right?
Bitstream cannot be stored permanently in the FPGA, i.e. Bitstream is uploaded by the software during the start-up.
But besides of that, source code has also be released if firmware or bitstream is 'hidden in hardware'.
Good to know. Would hate to see a competitor gain (beyond what GPL allows) from your awesome code. Hopefully someone does improve upon that in the future but any improvements should be shared w/ all FPGA capable hardware.
November 07, 2011, 09:07:34 AM
Thank you so much again for the quick response. Yea, minimal cost for sure. Do you know what the CFM, voltage and amperage are for those fans?
6.97 CFM
7 to 13.2V
0.06A @ 12V
Ahh, very friendly numbers. Thanks again. =)
November 07, 2011, 03:59:35 AM
I got a question about that. Does GPL protect your IP from a "covert distro". Meaning someone take your code, tweaks it to get 3% more performance and then sells FGPA already loaded with it. They still have to release the source code of their improvements right?
Bitstream cannot be stored permanently in the FPGA, i.e. Bitstream is uploaded by the software during the start-up.
But besides of that, source code has also be released if firmware or bitstream is 'hidden in hardware'.
November 07, 2011, 03:48:25 AM
Thank you so much again for the quick response. Yea, minimal cost for sure. Do you know what the CFM, voltage and amperage are for those fans?
6.97 CFM
7 to 13.2V
0.06A @ 12V
November 07, 2011, 03:43:34 AM
I bought your 1.15d card and the power supply module on the internet a couple of months ago. I was trying to get my own bitcoin miner working, but had some problems since my hardware background isn't quite good enough. (I write linux software drivers for Android for a living, so I have good low-level software experience but not much experience with Verilog.) However, I downloaded your design and it is working great. I'm getting about 150-160 MH/s out of the Spartan6 LX150.
The obvious problem is that it gets too hot. When I first power it on it can do about 190 MHz, but as it heats up it starts dropping frequency until it gets down to 168 MHz. The passive heatsink you supply just isn't enough for this application. It's super-hot to the touch. Can you recommend a fansink that will work well with the 1.15d board to keep it cooler?
The obvious problem is that it gets too hot. When I first power it on it can do about 190 MHz, but as it heats up it starts dropping frequency until it gets down to 168 MHz. The passive heatsink you supply just isn't enough for this application. It's super-hot to the touch. Can you recommend a fansink that will work well with the 1.15d board to keep it cooler?
1.15d module has just a 25mm heat sink. As stated on the BTCMiner page this variant requires some airflow, e.g. from a fan positioned near the FPGA Board or using Active cooling upgrade for USB-FPGA Module 1.15d.
November 06, 2011, 10:43:19 PM
ztex,
I bought your 1.15d card and the power supply module on the internet a couple of months ago. I was trying to get my own bitcoin miner working, but had some problems since my hardware background isn't quite good enough. (I write linux software drivers for Android for a living, so I have good low-level software experience but not much experience with Verilog.) However, I downloaded your design and it is working great. I'm getting about 150-160 MH/s out of the Spartan6 LX150.
The obvious problem is that it gets too hot. When I first power it on it can do about 190 MHz, but as it heats up it starts dropping frequency until it gets down to 168 MHz. The passive heatsink you supply just isn't enough for this application. It's super-hot to the touch. Can you recommend a fansink that will work well with the 1.15d board to keep it cooler?
Thanks.
I bought your 1.15d card and the power supply module on the internet a couple of months ago. I was trying to get my own bitcoin miner working, but had some problems since my hardware background isn't quite good enough. (I write linux software drivers for Android for a living, so I have good low-level software experience but not much experience with Verilog.) However, I downloaded your design and it is working great. I'm getting about 150-160 MH/s out of the Spartan6 LX150.
The obvious problem is that it gets too hot. When I first power it on it can do about 190 MHz, but as it heats up it starts dropping frequency until it gets down to 168 MHz. The passive heatsink you supply just isn't enough for this application. It's super-hot to the touch. Can you recommend a fansink that will work well with the 1.15d board to keep it cooler?
Thanks.
November 06, 2011, 05:16:32 PM
I mean: "Would you mind if somebody using your arithmetic on other venders development board? certainly follow the GNU license. "
It's Open Source, i.e. everyone is allowed to port this software / core to other boards. The only limitation is that is has to be published under GPL, if it is distributed in some way.
I got a question about that. Does GPL protect your IP from a "covert distro". Meaning someone take your code, tweaks it to get 3% more performance and then sells FGPA already loaded with it. They still have to release the source code of their improvements right?
November 06, 2011, 05:14:21 PM
Would there be any cost savings to order them with just the heatsink and no fan? Or is the cost so negligible it doesn't really matter? Thanks again for your replies, m8
The fan belongs to the heat sink and is installed on it. AFAIR heat sink with fan costs $4-$5, i.e. there is not much potential for saving costs.
Thank you so much again for the quick response. Yea, minimal cost for sure. Do you know what the CFM, voltage and amperage are for those fans?
November 06, 2011, 04:59:03 PM
Would there be any cost savings to order them with just the heatsink and no fan? Or is the cost so negligible it doesn't really matter? Thanks again for your replies, m8
The fan belongs to the heat sink and is installed on it. AFAIR heat sink with fan costs $4-$5, i.e. there is not much potential for saving costs.
November 06, 2011, 04:51:56 PM
I mean: "Would you mind if somebody using your arithmetic on other venders development board? certainly follow the GNU license. "
It's Open Source, i.e. everyone is allowed to port this software / core to other boards. The only limitation is that is has to be published under GPL, if it is distributed in some way.
November 06, 2011, 04:37:10 PM
Is there any chance we can sweet talk you into dropping a 1.26v resistor onto one to see the results? =)
I'll try this out next week. Btw, I assume the heatsinks are not inculded.? Where did you get yours and are they cost effective if buying 100+ of them?
A heat sink with fan is included, see http://www.ztex.de/usb-fpga-1/usb-fpga-1.15x.e.html#hsAhh, my poor comprehension skills. Some of the verbage threw me off I believe. 'belongs to the content of delivery', it makes sense now. Should possibly read, "is included content with delivery of fpga"
Would there be any cost savings to order them with just the heatsink and no fan? Or is the cost so negligible it doesn't really matter? Thanks again for your replies, m8
Cheers,
Derek
November 06, 2011, 04:07:52 PM
Would you mind using your arithmetic on other venders development board? certainly follow the GNU license.
Why should I do this?
I mean: "Would you mind if somebody using your arithmetic on other venders development board? certainly follow the GNU license. "
November 06, 2011, 04:04:45 PM
Is there any chance we can sweet talk you into dropping a 1.26v resistor onto one to see the results? =)
I'll try this out next week. Quote
Btw, I assume the heatsinks are not inculded.? Where did you get yours and are they cost effective if buying 100+ of them?
A heat sink with fan is included, see http://www.ztex.de/usb-fpga-1/usb-fpga-1.15x.e.html#hs
November 06, 2011, 03:57:48 PM
Whould you mind using your arithmetic on other venders development board? certainly follow the GNU license.
Why should I do this?
November 06, 2011, 01:21:06 PM
Very awesome on the 190 achievment. Is that primarily a chip limitation or is it possible to get any higher with good cooling and highly 'clean' power? I.e., 88% PSUs, voltage regulation, and commercial grade ventilation?
It's a limitation of the bitstream (i.e the implemented design) and the FPGA.
'Clean power' does not help since core voltage quality is defined by the on board regulators, not he external voltage.
Standard cooling (with heatsink and fan) is more than sufficient.
The only thing that may help is to 'overvolt' the FPGA from (e.g. from 1.23V to 1.26V) by replacing one resistor.
Thank you very much for the response. Is there any chance we can sweet talk you into dropping a 1.26v resistor onto one to see the results? =)
Btw, I assume the heatsinks are not inculded.? Where did you get yours and are they cost effective if buying 100+ of them?
November 06, 2011, 10:16:33 AM
Very awesome on the 190 achievment. Is that primarily a chip limitation or is it possible to get any higher with good cooling and highly 'clean' power? I.e., 88% PSUs, voltage regulation, and commercial grade ventilation?
It's a limitation of the bitstream (i.e the implemented design) and the FPGA.
'Clean power' does not help since core voltage quality is defined by the on board regulators, not he external voltage.
Standard cooling (with heatsink and fan) is more than sufficient.
The only thing that may help is to 'overvolt' the FPGA from (e.g. from 1.23V to 1.26V) by replacing one resistor.
Whould you mind using your arithmetic on other venders development board? certainly follow the GNU license.
Jump to: