Pages:
Author

Topic: Raspberry Coins - Cyclone V - 28nm based FPGA miner - page 2. (Read 8320 times)

full member
Activity: 196
Merit: 100
Good info and good thread - we could probably throw the Raspberry Coins CM-1 miner into the bunch of boards that are running the open-source code (might need to make some minor adjustments for the Raspberry Pi - SPI interface).

... So what makes Litecoin mining interesting is :
  - scrypt runs the SHA256 in a loop (1024 times)
  - uses about 128K of memory

... the end result is that scrypt rates are  slower (in the range of 1000+ KHash/sec) while bitcoin (using straight SHA256) is in the (1000+) MHash/sec areas.   As a result, the difficulty level should be lower for scrypt.  In addition, the Cyclone V appears to have more than enough memory to run scrypt without going external.  This makes mining Litecoin (and running scrypt) appear promising to me.

Well, you're going to have some competition (as a little googling would have shown) http://www.litecoinfpga.com/ (yes funny pic, but very serious dev guy, click on "you can ask here" for the forum).
newbie
Activity: 16
Merit: 0
NO! You are making big mistake. Scrypt runs 4096 iterations of salsa20_8 for calculation of ONE hash.
newbie
Activity: 33
Merit: 0
The second part of my reply is FLEXIBILITY.  These devices are NOT purpose built-ASIC's.  We are investigating utilizing the 12.4 Mb built-in memory (1,220 M10K blocks) per FPGA (24.8 Mb, 2,440 M10K blocks total) for use in Litecoin mining.  A purpose-built, Bitcoin mining ASIC can never be re-targeted for other uses.  These FPGA's on the other hand are MUCH more flexible.

Question for the FPGA experts. Is this really enough memory for scrypt mining. Its on-die and wide bandwidth, so beats external memory, but what performance could be expected? I understand scrypt uses a 100kB lookup table, so does 24.8Mb (3MByte) equate to ~30   threads, or not? How will it compare with GPU?

I don't think anyone can accurately answer that without testing first. Just like ASIC's, tests are showing up different than theories.

Yeah, but there are guys posting on the fpgaminer thread https://bitcointalksearch.org/topic/official-open-source-fpga-bitcoin-miner-last-update-april-14th-2013-9047 who are using the big Xilinx Kintex-7 chips , I just wondered if they had had any insight to the feasability of scrypt with only this much onboard ram.

Whoa, I didn't even know about that thread. The last I heard was from the people looking to crowd-source ASIC's/FPGA's. Thanks for the heads-up!

Good info and good thread - we could probably throw the Raspberry Coins CM-1 miner into the bunch of boards that are running the open-source code (might need to make some minor adjustments for the Raspberry Pi - SPI interface).

... So what makes Litecoin mining interesting is :
  - scrypt runs the SHA256 in a loop (1024 times)
  - uses about 128K of memory

... the end result is that scrypt rates are  slower (in the range of 1000+ KHash/sec) while bitcoin (using straight SHA256) is in the (1000+) MHash/sec areas.   As a result, the difficulty level should be lower for scrypt.  In addition, the Cyclone V appears to have more than enough memory to run scrypt without going external.  This makes mining Litecoin (and running scrypt) appear promising to me.
legendary
Activity: 1036
Merit: 1001
/dev/null
Tony, sorry for this kind of question, but you really think, that your miner (if I will consider lead time) will meet ROI? I know about difficulty raise but still..

one thing is cool idea, open source HW and super cool design, but other is costs. I was really interested about this project from the start, but you simply miss a train.

I somebody will buy this device, it is just for support an idea and for your great work, but definitely not for make a profit.

good luck

newbie
Activity: 33
Merit: 0
You know CM-1 already exists, right? Cairnsmore-1 ?

Cairnsmore-1 is based on Four (4) Xilinx Spartan 6 - LX150's (with a total of about 600K LC).  The LX150 is 65nm technology and when fully pipelined and overclocked runs up to 200 MHz.  (This is in the 800 MHash/sec area) It requires an external computer to drive it via the USB port.  The Cairnsmore-1 requires USB hubs and USB cables to connect and control, and obtain status.  It is from a totally different company.

The Raspberry Coins - Coin Miner 1 (CM-1) is based on Two (2) Altera Cyclone V-A9's (with a total of 602K LC).  The Cyclone V A9 is 28 nm technology.  When fully pipelined, the -8 (slow) speed grade, has restricted fmax of 240 MHz (around 960 MHash/sec area), and an unrestricted fmax in the range of 380+ MHz (which is potentially in the 1500+ MHash/sec area).  The CM-1 is connected to the (included) Raspberry Pi, and is capable of standalone (does not require another computer) for control or status.  The Raspberry Coins - CM1 units may be connected together using a simple ethernet hub/switch.  It is from Raspberry Coins.

Perhaps we should not have chosen the CM-1 as the name of our product.  I did not realize that somebody else had used those particular letters and numerals.  Oh well, we'll probably have to change the name to the full name (Coin Miner 1) vs the shortened name (CM-1).  (We always called our product the CM-1 internally).
sr. member
Activity: 401
Merit: 250
You know CM-1 already exists, right? Cairnsmore-1 ?
legendary
Activity: 1988
Merit: 1007
The second part of my reply is FLEXIBILITY.  These devices are NOT purpose built-ASIC's.  We are investigating utilizing the 12.4 Mb built-in memory (1,220 M10K blocks) per FPGA (24.8 Mb, 2,440 M10K blocks total) for use in Litecoin mining.  A purpose-built, Bitcoin mining ASIC can never be re-targeted for other uses.  These FPGA's on the other hand are MUCH more flexible.

Question for the FPGA experts. Is this really enough memory for scrypt mining. Its on-die and wide bandwidth, so beats external memory, but what performance could be expected? I understand scrypt uses a 100kB lookup table, so does 24.8Mb (3MByte) equate to ~30   threads, or not? How will it compare with GPU?

I don't think anyone can accurately answer that without testing first. Just like ASIC's, tests are showing up different than theories.

Yeah, but there are guys posting on the fpgaminer thread https://bitcointalksearch.org/topic/official-open-source-fpga-bitcoin-miner-last-update-april-14th-2013-9047 who are using the big Xilinx Kintex-7 chips , I just wondered if they had had any insight to the feasability of scrypt with only this much onboard ram.

Whoa, I didn't even know about that thread. The last I heard was from the people looking to crowd-source ASIC's/FPGA's. Thanks for the heads-up!
full member
Activity: 196
Merit: 100
The second part of my reply is FLEXIBILITY.  These devices are NOT purpose built-ASIC's.  We are investigating utilizing the 12.4 Mb built-in memory (1,220 M10K blocks) per FPGA (24.8 Mb, 2,440 M10K blocks total) for use in Litecoin mining.  A purpose-built, Bitcoin mining ASIC can never be re-targeted for other uses.  These FPGA's on the other hand are MUCH more flexible.

Question for the FPGA experts. Is this really enough memory for scrypt mining. Its on-die and wide bandwidth, so beats external memory, but what performance could be expected? I understand scrypt uses a 100kB lookup table, so does 24.8Mb (3MByte) equate to ~30   threads, or not? How will it compare with GPU?

I don't think anyone can accurately answer that without testing first. Just like ASIC's, tests are showing up different than theories.

Yeah, but there are guys posting on the fpgaminer thread https://bitcointalksearch.org/topic/official-open-source-fpga-bitcoin-miner-last-update-april-14th-2013-9047 who are using the big Xilinx Kintex-7 chips , I just wondered if they had had any insight to the feasability of scrypt with only this much onboard ram.
newbie
Activity: 16
Merit: 0
One can very accurately answer the question by looking in the ltc mining code. There is a nice long loop, what makes FPGA miners useless.
legendary
Activity: 1988
Merit: 1007
The second part of my reply is FLEXIBILITY.  These devices are NOT purpose built-ASIC's.  We are investigating utilizing the 12.4 Mb built-in memory (1,220 M10K blocks) per FPGA (24.8 Mb, 2,440 M10K blocks total) for use in Litecoin mining.  A purpose-built, Bitcoin mining ASIC can never be re-targeted for other uses.  These FPGA's on the other hand are MUCH more flexible.

Question for the FPGA experts. Is this really enough memory for scrypt mining. Its on-die and wide bandwidth, so beats external memory, but what performance could be expected? I understand scrypt uses a 100kB lookup table, so does 24.8Mb (3MByte) equate to ~30   threads, or not? How will it compare with GPU?

I don't think anyone can accurately answer that without testing first. Just like ASIC's, tests are showing up different than theories.
hero member
Activity: 572
Merit: 500
I think you have missed the boat, bad timing bro!!
sr. member
Activity: 245
Merit: 250
so expensive.i really do not know who need this?
full member
Activity: 196
Merit: 100
The second part of my reply is FLEXIBILITY.  These devices are NOT purpose built-ASIC's.  We are investigating utilizing the 12.4 Mb built-in memory (1,220 M10K blocks) per FPGA (24.8 Mb, 2,440 M10K blocks total) for use in Litecoin mining.  A purpose-built, Bitcoin mining ASIC can never be re-targeted for other uses.  These FPGA's on the other hand are MUCH more flexible.

Question for the FPGA experts. Is this really enough memory for scrypt mining. Its on-die and wide bandwidth, so beats external memory, but what performance could be expected? I understand scrypt uses a 100kB lookup table, so does 24.8Mb (3MByte) equate to ~30   threads, or not? How will it compare with GPU?
newbie
Activity: 16
Merit: 0
~15 W for 1 GH/s btc hashing.
legendary
Activity: 1946
Merit: 1006
Bitcoin / Crypto mining Hardware.
How is it better than http://www.xilinx.com/products/boards-and-kits/EK-K7-KC705-G.htm
 running it @ 1GH/s or more. And it had potential to mine LTC as well.

Excellent question.

That is a single board for $1695 which contains a single Xilinx XC7K325T-2FFG900C (326K logic cells total) + lots of additional circuitry (ethernet, flash memory, DDR SDRAM, LCD display, lots of connectors, etc.)

The Raspberry Coins - CM-1 board contains two Cyclone V-A9 devices (301K logic cells each), for a total of 602K logic cells.  The interface is via Raspberry Pi (SPI port).  So, CM-1 board is significantly simpler, and you get much more logic cells per dollar.

Both the Kintex-7 and the Cyclone V are 28 nm technology.

Cool, thanks for the clarification. Makes sense. Altera seems much for value for money. Any idea about the power draw of the V-A9 FPGAs while hashing?
newbie
Activity: 33
Merit: 0
How is it better than http://www.xilinx.com/products/boards-and-kits/EK-K7-KC705-G.htm
 running it @ 1GH/s or more. And it had potential to mine LTC as well.

Excellent question.

That is a single board for $1695 which contains a single Xilinx XC7K325T-2FFG900C (326K logic cells total) + lots of additional circuitry (ethernet, flash memory, DDR SDRAM, LCD display, lots of connectors, etc.)

The Raspberry Coins - CM-1 board contains two Cyclone V-A9 devices (301K logic cells each), for a total of 602K logic cells.  The interface is via Raspberry Pi (SPI port).  So, CM-1 board is significantly simpler, and you get much more logic cells per dollar.

Both the Kintex-7 and the Cyclone V are 28 nm technology.
legendary
Activity: 1775
Merit: 1032
Value will be measured in sats
22K for 22-28ghs?

How much are avalons selling for? (if ever shipped)
How muych $$ would BFL take off you for the hope of get a 30ghs single?
How much are block ejaculator blades selling for?

*IF* any asic supplier were actually shipping anything consistently you have to say the prices are too high.
That said I wouldn't order at that price to get in the first 150 chips, but could be interested in batch 2 if prices come down.

Also I'd really like to buy boards without the RPi and run from the computer hardware I already have, and that computer can support far more than 4 boards.

stop deluding yourself. this guy is just another c*** looking to get rich off the BTC crazy by lining his own pockets and selling wildly overpriced equipment that will be worth nothing by the time he ships if he ever does...

USB ASIC miners are overpriced as it is already BUT at least they are a real tangible good that exists now. These Ris-poos are more expensive  $/Ghz and have not even been made yet. There is no point arguing "oh well if someone else can ship a ASIC in 6 wks time as these speeds then..." because we just don't know if this guy will by shipping be then too. Good idea man but come back when you have something to sell and the attitude might be more receptive.
sr. member
Activity: 298
Merit: 250
Play2Live pre-sale starts on January 25th
By the way, why the price is so high? If you are building 300 boards and have inside connection at Altera, single FPGA would cost ~180$, raspberry pi is cheap, enclosure makes another 20$ from some cheap labor country. PCB+passives+power makes another 50$. So your profit is like 500$ from single unit :O Plus mining profits from testing the produced units couple weeks before delivery.  
+1

I like it, but not at this price.
full member
Activity: 154
Merit: 100
Look, it's a cool idea, but it's far too expensive per Gh, even if it was available today. Given the timeline and the risks, this is near-hopeless - but I genuinely hope I'm wrong.

Do something interesting, prove it can run scrypt efficiently, then we can talk. Smiley
legendary
Activity: 1988
Merit: 1007
22K for 22-28ghs?

How much are avalons selling for? (if ever shipped)
How muych $$ would BFL take off you for the hope of get a 30ghs single?
How much are block ejaculator blades selling for?

*IF* any asic supplier were actually shipping anything consistently you have to say the prices are too high.
That said I wouldn't order at that price to get in the first 150 chips, but could be interested in batch 2 if prices come down.

Also I'd really like to buy boards without the RPi and run from the computer hardware I already have, and that computer can support far more than 4 boards.

$22,200 for 22-28 GH/s
$22,200 / 200 = 111 * .3 = 33.3 GH/s via USB miners

Not to mention you have it in-hand within a couple days. Not relying on a company that may or may not be able to achieve their goal.
Pages:
Jump to: