markm
Legendary
Offline
Activity: 2940
Merit: 1090
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
April 20, 2013, 06:47:16 AM |
|
I have a couple of 5870's that give me 400 to 407 MHash doing SHA256 mining, but mucking around with all the suggested settings for litecoin the best I have been able to get them to do for scrypt mining is about 250 to 260 Khash. So to me GPUs are not looking real good for litecoin mining, certainly not the one Khash scrypt per 1 Mhash SHA256 one sees often as a rule of thumb estimate.
Maybe though there is more to getting the 5870 to work right than the thread about litecoin mining has mentioned?
Basically I am figuring I might as well use my GPUs for SHA256-based merged mining, mostly for the altcoins, and look toward FPGA for scrypt coins.
-MarkM-
|
|
|
|
razorfishsl
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
April 20, 2013, 07:05:47 AM |
|
I have my own Scrypt code for Xilinx FPGA and a pluggable rack system, that takes 10 boards, I had to mux them as 8+2 hot spares.(yep sometimes they drop in & out of service randomly)
Unfortunately... Performance is shite...... insofar as comparison to high-end CPU or GPUs. Who knows if I can get an improvement but it is going to be very hard to beat the GPU thrughput Vrs cost.
Interesting. Thanks for the detailed post, especially the last part where you share your results. This is good news as far as I am concerned, the whole point of Litecoin using Scrypt was so it would be difficult for specialised hardware to have a massive performance edge. K some VERY VERY preliminary verifiable measurements. (verifiable from my side). Iv'e had code for some time that takes about 1ms to do ONE round (no laughing at the back there), which would make it about a 1kh/s. This is actual code on an FPGA, yes it has not been optimized yet and yes it is a SINGLE core. I had not done any work on it, purely because I was working on a 'crypto job' for someone else.... then there was the work on getting the Bitcoin FPGA code operating at a higher level, currently thats clocking about 350MH/s on a xilinx, with bitcoins at a stupid level, and the fact that BFL & TOM both shafted me on deliveries of ASIC kit it is about priorities.(beer money) Since I recently took a slagging down on IRC about what some called 'stock images' that I : "had modified so they could not be google searched" , there were requests for a special 'proof' message which I have included specifically for the IRC doubters. ![](https://ip.bitcointalk.org/?u=https%3A%2F%2Fi.imgur.com%2FUHiTc3g.jpg&t=665&c=fYlzp8nVP9yynQ) This is not a 'get rich quick scheme' but rather a pure research task, possibly the first images of an FPGA product running litecoin. Crypto currencies are of interest to me, but I really will be surprised if people can break into high double figures with a single core on an FPGA.
|
|
|
|
peacefulmind
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
April 20, 2013, 08:57:57 AM |
|
Looking forward to hear more on this project.
Any idea on prototype ETA? - some already have talked about various prototypes in the works?
The key is the transparency imho - is there a REAL product, and are they consistent in quality? Does it verifiably beat forthcoming 8750 and 8790 in measurements of performance and power usage?
|
"I think you are to hung up on this notion about 'pre-mining' being a No-No." - from journeys into the dark depths of the alt coin forum....
|
|
|
razorfishsl
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
April 20, 2013, 11:14:57 PM |
|
NO.... I cannot see FPGA beating 8750 or 8790
But the real issue is no one is sharing any Litecoin FPGA figures, considering all these 'prototypes' floating about ...... It is the 'Tom'/ BFL fuckfest all over again..........
Personally I have great difficulty believing some of the figures that are being bandied about for litecoin FPGA. I've see statements such as
"Spartan6 XC6SLX150 fits only 8 scratchpads. If BRAMs are not used for bitcoin computations, it is possible to implement LTC mining for XC6SLX150 at about 50 - 100 kh/s per chip with about 80% of slices free."
Quite frankly I find this very difficult to believe.. Because I'm seeing figures of about 3k flipflops & 5k =LUTS for a SCRYPT engine.
Xc6SLX150 is specd at 23,000 slices each slice contains 4 LUTS & 8 FF so ~23,000*4 =~98k LUTS
So just taking into account the LUTS: 8 cores = 8*5k =~40k LUTS, which in no way leaves "about 80% of slices free", and NO way am I seeing anything near 100Kh/s.
To have ~80% free you would need to implement 8 engines in ~20k LUTS ,~2.5K LUTS per engine. Sorry but for a SCRYPT engine I have to call bullshit...
Then suddenly I see this sort of thing in the same discussion. "multiple smaller DRAM chips working in parallel will do best job... Allowing about 500 mega-transfers for low-cost / mid-cost fpga, that is 500 giga-bits per second or 60 gigabytes per second. Overall cost of DRAM will be about 150 EUR- and of FPGA to handle that about 300 EUR-. If works in fully-pipelined manner it would give about 500 kh/s mining performance for litecoin application."
So ABRACADABRA... we have a 5 fold increase in performance, and not a SINGLE analysis on actual cycle times any place to be seen.
|
|
|
|
phk
Newbie
Offline
Activity: 28
Merit: 0
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
April 20, 2013, 11:41:47 PM |
|
Personally I have great difficulty believing some of the figures that are being bandied about for litecoin FPGA.
Can you share a link to the threads with your bandied about numbers? I haven't seen anybody say anything that outrageous here.
|
|
|
|
kelpy
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
April 21, 2013, 12:01:57 AM |
|
Interested
|
Bitrated user: kelpy.
|
|
|
3ham3
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
April 21, 2013, 12:38:18 AM |
|
I hope we will be seeing some updates soon.
|
|
|
|
Trommie
Newbie
Offline
Activity: 25
Merit: 0
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
April 21, 2013, 12:48:56 AM |
|
K some VERY VERY preliminary verifiable measurements. (verifiable from my side).
Iv'e had code for some time that takes about 1ms to do ONE round (no laughing at the back there), which would make it about a 1kh/s. This is actual code on an FPGA, yes it has not been optimized yet and yes it is a SINGLE core.
.
.
.
This is not a 'get rich quick scheme' but rather a pure research task, possibly the first images of an FPGA product running litecoin. Crypto currencies are of interest to me, but I really will be surprised if people can break into high double figures with a single core on an FPGA.
Interesting stuff. Thinking about having a play around with this myself. Is there any opensource Scrypt VHDL out there? Or should I just roll my own with the SHA1 and mod the Salsa20 that is up on opencores site? I'm not thinking about mining or anything just to get some throughput data/energy costs etc and also work out what the problems are people will face who want todo it properly. If all else fails I'm thinking of running it on top of a NIOS II implementation. Completely pointless but got to be worth a shot just for the laughs.
|
|
|
|
Operatr (OP)
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
April 21, 2013, 12:57:57 AM |
|
Looking forward to hear more on this project.
Any idea on prototype ETA? - some already have talked about various prototypes in the works?
The key is the transparency imho - is there a REAL product, and are they consistent in quality? Does it verifiably beat forthcoming 8750 and 8790 in measurements of performance and power usage?
The team is just now being cemented while I get our home base in order, so no ETA yet but hopefully soon. This is pretty much day 1 ![Smiley](https://bitcointalk.org/Smileys/default/smiley.gif) And I agree, which is why BlockBurner will be a community driven company. Much like Local Motors (the community engineered Ralley Fighter car), the idea is to crowd source a design for a production device and utilizing open source solutions whenever possible. My position in it is overall project management and business logistics, organizing funding campaigns, design, and handling final production and distribution in house. Everything will be made transparent as to company operations and finances. Bitcoin is a community built protocol, I think businesses around it should be built the same way. If this is the future of currency, then we must create the future of business to support it. The community spirit is the most pervasive element of the protocol. The response I am seeing here is evidence of that. This is why I have no wish to simply fish for people willing to submit their money to a pre-order and hope for the best like other companies. A properly crowd-funded campaign is the better way to go for a bootstrap industry, so that is the approach I am going to take. BlockBurner has nothing to hide. Funds will only be delivered when a known point is reached that should suffice to fund that stage of development. This keeps expectations in check, and clarity to the process. Unlike BFL there will be constant updates and publicly accessible accounting so you know where your money is going. There will be a way to opt out as well if you change your mind, though refunds can only be done so long as those funds are not invested toward the goal. Until goals are reached, BlockBurner will take possession of nothing, safe in escrow until then and not before. The dev team will be compensated for their good work at project completion, in a way that is decided as fair, likely a reserved front row stake in a Batch 1 device. Nothing is wrong with crowd funded efforts, but these companies are doing it wrong. I want to fly right by the community with transparency so I never see "is BlockBurner a scam?" pop up here someday. There is no trust being developed by these companies by being dodgy, not a good way for a business to start out with a cloud of doubt over it. The response here is awesome guys, I have been working on getting things set up on my end non stop, I will do my best not to disappoint you. All devs interested in this project are welcome to PM me, I am getting things organized and we will break things up into separate focus groups soon and start talking about the logistics. To the devs I have talked to so far, thank you again for your interest, you will be hearing from me soon as I get the BlockBurner site forums up over the next day or so. I am going to also try an integrate a project management system. New logo ![Cool](https://bitcointalk.org/Smileys/default/cool.gif) ![](https://ip.bitcointalk.org/?u=http%3A%2F%2Fblockburner.nailedmedia.com%2Ffiles%2F2013%2F04%2Fblockburner-paragram-logo.png&t=665&c=nlXxqjEwp2e-lg) Operatr
|
|
|
|
phk
Newbie
Offline
Activity: 28
Merit: 0
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
April 21, 2013, 01:06:46 AM |
|
If all else fails I'm thinking of running it on top of a NIOS II implementation. Completely pointless but got to be worth a shot just for the laughs.
You could certainly start with a pure software approach and then gradually optimize it with custom instructions. You wouldn't ever get the speed of a pure RTL design, but it's a good (rewarding) exercise. With the reduced resource, utilization you could end up with the fastest implementation on the BeMicro.
|
|
|
|
mrSprinkles
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
April 21, 2013, 01:08:28 AM |
|
I have a couple of 5870's that give me 400 to 407 MHash doing SHA256 mining, but mucking around with all the suggested settings for litecoin the best I have been able to get them to do for scrypt mining is about 250 to 260 Khash. So to me GPUs are not looking real good for litecoin mining, certainly not the one Khash scrypt per 1 Mhash SHA256 one sees often as a rule of thumb estimate.
Maybe though there is more to getting the 5870 to work right than the thread about litecoin mining has mentioned?
Basically I am figuring I might as well use my GPUs for SHA256-based merged mining, mostly for the altcoins, and look toward FPGA for scrypt coins.
-MarkM-
I'm getting 380 - 400Kh/s with each of my 5850s. Using 12.8 drivers, 12.8 APP SDK. The only thing I can think of is that you may be using a low intensity associated with SHA256 such as 11. With litecoin mining, you should be using an intensity of 18, and move up to 19 and 20 if your card doesn't crash at these. Actually you should be getting roughly 10% more Kh/s than Mh/s with scrypt.
|
|
|
|
lame.duck
Legendary
Offline
Activity: 1270
Merit: 1000
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
April 21, 2013, 09:43:32 AM |
|
Personally I have great difficulty believing some of the figures that are being bandied about for litecoin FPGA.
Can you share a link to the threads with your bandied about numbers? I haven't seen anybody say anything that outrageous here. These numbers are from bitfury somewhere in the forum here. While its not so easy to verify the numbers i think he is a quite skilled HW-Designer so i would careful with 'ABRACADABRA'. http://bitfury.org/
|
|
|
|
|
tacotime
Legendary
Offline
Activity: 1484
Merit: 1005
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
April 21, 2013, 03:23:24 PM |
|
It's weird that he would say Hmm ... maybe even Litecoin chip would be done before Bitcoin chip, as it seems to be much simpler and uses well-known techniques - less competition :-) The implementation of scrypt for Litecoin requires you to do the equivalent number of SHA256 hashes to the number of execution of the salsa for loop. If the Litecoin implementation of scrypt includes SHA256 (the algo bitcoin uses) and memory intensive salsa stream cipher rounds, why would it be easier than just implementing SHA256?
|
XMR: 44GBHzv6ZyQdJkjqZje6KLZ3xSyN1hBSFAnLP6EAqJtCRVzMzZmeXTC2AHKDS9aEDTRKmo6a6o9r9j86pYfhCWDkKjbtcns
|
|
|
phk
Newbie
Offline
Activity: 28
Merit: 0
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
April 22, 2013, 12:46:08 AM |
|
why would it be easier than just implementing SHA256?
I think that post is just (old) off-the-cuff remarks and has multiple errors (not fully fleshed out ideas).
|
|
|
|
razorfishsl
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
April 23, 2013, 01:22:22 AM Last edit: April 23, 2013, 01:37:01 AM by razorfishsl |
|
Personally I have great difficulty believing some of the figures that are being bandied about for litecoin FPGA.
Can you share a link to the threads with your bandied about numbers? I haven't seen anybody say anything that outrageous here. learn to use google...(that magic thing you can Copy& paste sections of text into, to track down stuff) https://bitcointalk.org/index.php?topic=98535.5;wap2But specifically it is the 80% free I take exception to. Anyway I finally received some 50A TI power modules, so I can run Bitcoin & Litecoin research boards together.
|
|
|
|
razorfishsl
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
April 23, 2013, 01:27:35 AM |
|
K some VERY VERY preliminary verifiable measurements. (verifiable from my side).
Iv'e had code for some time that takes about 1ms to do ONE round (no laughing at the back there), which would make it about a 1kh/s. This is actual code on an FPGA, yes it has not been optimized yet and yes it is a SINGLE core.
.
.
.
This is not a 'get rich quick scheme' but rather a pure research task, possibly the first images of an FPGA product running litecoin. Crypto currencies are of interest to me, but I really will be surprised if people can break into high double figures with a single core on an FPGA.
Interesting stuff. Thinking about having a play around with this myself. Is there any opensource Scrypt VHDL out there? Or should I just roll my own with the SHA1 and mod the Salsa20 that is up on opencores site? I'm not thinking about mining or anything just to get some throughput data/energy costs etc and also work out what the problems are people will face who want todo it properly. If all else fails I'm thinking of running it on top of a NIOS II implementation. Completely pointless but got to be worth a shot just for the laughs. You will need to roll your own, the implementations are not too good.
|
|
|
|
Operatr (OP)
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
April 25, 2013, 03:56:23 PM |
|
Update:
The dev team is now getting on to our new project home with Atlassian, discussions are already flowing for hardware considerations to develop the ideal FPGA for Scrypt and an initial Git has been posted. We will be breaking up into specific teams in due time as everyone gets settled in. The project page comes complete with a Git repository as well for, as the intent is to release the client software as an open source package.
Meanwhile I myself am getting the home website up and running, with community forums.
More to come soon!
Operatr
|
|
|
|
VJain
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
April 25, 2013, 05:07:21 PM |
|
Definitely interested! If things look like they'd have a good ROI term, would look at maybe 50-100BTC~ of pre-orders (if you have a pre-order newsletter list, please append me to it ![Smiley](https://bitcointalk.org/Smileys/default/smiley.gif) )
|
Making Apps and Websites for people. I charge reasonable rates ($30-40/hour in BTC).
|
|
|
Operatr (OP)
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
April 25, 2013, 07:19:30 PM |
|
Definitely interested! If things look like they'd have a good ROI term, would look at maybe 50-100BTC~ of pre-orders (if you have a pre-order newsletter list, please append me to it ![Smiley](https://bitcointalk.org/Smileys/default/smiley.gif) ) That is definitely a goal. This machine is to be specialized for Scrypt unlike other FPGA miners available now, which perform better for SHA256. While they can mine Scrypt coins as well they are not as productive as they could be. Though the power savings alone over running several GPU's may be enough ![Smiley](https://bitcointalk.org/Smileys/default/smiley.gif) Along with BTC, I think as a Litecoin mining device there will be an option for Litecoin payments too, and regular USD to appeal to those who want to mine coins instead of buying them (naturally they would have no coins yet). As part of the web building I will be doing through the rest of the week, bi-weekly newsletter will be incorporated ![Cool](https://bitcointalk.org/Smileys/default/cool.gif) We'll have Facebook, Twitter and some other things as well coming up to keep everyone updated in a variety of ways. Operatr
|
|
|
|
|