Topic: FPGA development board "Icarus" - DisContinued/ important announcement - page 32. (Read 207279 times)
February 03, 2012, 05:02:36 PM
我想请问一下张博士，这个用人民币买多少钱一个？本人住的离北航不远，运费可以忽略不计的，或者自己去取也行啊？
February 03, 2012, 04:45:51 PM
let's discuss about the pool selection.
in brief, if you have a large ping, please use the queue version of the miner software, but it will cause a higher stale rate. about 1-2%. but the FPGA could run at a full speed.
but if your connection is unstable, it has already been found that both miner software sometimes get stuck.
the both simple and queue miner will not continue or fix. we will change to a more common use miner, like CGminer or some other stuff who want to support icarus. there are already some people working on this.
hmmmm, the FPGA simply act as a wire. pass the host data to the next one, and store-and-forward data to the host.
in brief, if you have a large ping, please use the queue version of the miner software, but it will cause a higher stale rate. about 1-2%. but the FPGA could run at a full speed.
but if your connection is unstable, it has already been found that both miner software sometimes get stuck.
the both simple and queue miner will not continue or fix. we will change to a more common use miner, like CGminer or some other stuff who want to support icarus. there are already some people working on this.
Quote
I'm sure it hasn't yet. My question was more a what if scenario aimed at understanding the dependence between the FPGAs. If one were to be damaged/abused how would the other one be affected. If one gets sick/dead does that render the other also inoperable? I'm not saying anything is wrong but it's good to understand how they relate to each other. From looking at the code it seems like the first dead would knock out the second, but not the other way around; unless I'm wrong about how the second communicates via the first.
hmmmm, the FPGA simply act as a wire. pass the host data to the next one, and store-and-forward data to the host.
I have changed to queue miner and now get 374.25 +/- 5.93 MHash/s
I also get "upstream result: False" about every day and the mining stops --- I throught it may be my pc -- so I am in process of setting up a new pc to run software/icarus
I will run some traceroutes and ping test and post results and respond to ecki
February 03, 2012, 06:06:05 AM
let's discuss about the pool selection.
in brief, if you have a large ping, please use the queue version of the miner software, but it will cause a higher stale rate. about 1-2%. but the FPGA could run at a full speed.
but if your connection is unstable, it has already been found that both miner software sometimes get stuck.
the both simple and queue miner will not continue or fix. we will change to a more common use miner, like CGminer or some other stuff who want to support icarus. there are already some people working on this.
hmmmm, the FPGA simply act as a wire. pass the host data to the next one, and store-and-forward data to the host.
in brief, if you have a large ping, please use the queue version of the miner software, but it will cause a higher stale rate. about 1-2%. but the FPGA could run at a full speed.
but if your connection is unstable, it has already been found that both miner software sometimes get stuck.
the both simple and queue miner will not continue or fix. we will change to a more common use miner, like CGminer or some other stuff who want to support icarus. there are already some people working on this.
Quote
I'm sure it hasn't yet. My question was more a what if scenario aimed at understanding the dependence between the FPGAs. If one were to be damaged/abused how would the other one be affected. If one gets sick/dead does that render the other also inoperable? I'm not saying anything is wrong but it's good to understand how they relate to each other. From looking at the code it seems like the first dead would knock out the second, but not the other way around; unless I'm wrong about how the second communicates via the first.
hmmmm, the FPGA simply act as a wire. pass the host data to the next one, and store-and-forward data to the host.
February 03, 2012, 05:13:33 AM
OF
Hello Hardpick,
361 ms from Australia to Thailand? This is very bad...
Which Internet Provider do you use? Telstra?
I have less than 200ms to my customers in Sydney from here (Thailand)... so it is the same distance!
Antwort von 203.37.115.162: Bytes=32 Zeit=189ms TTL=236
Antwort von 203.37.115.162: Bytes=32 Zeit=190ms TTL=236
Antwort von 203.37.115.162: Bytes=32 Zeit=188ms TTL=236
Antwort von 203.37.115.162: Bytes=32 Zeit=184ms TTL=236
Ping-Statistik für 203.37.115.162:
Pakete: Gesendet = 4, Empfangen = 4, Verloren = 0
(0% Verlust),
Ca. Zeitangaben in Millisek.:
Minimum = 184ms, Maximum = 190ms, Mittelwert = 187ms
Could you PM me your traceroute to miner.ecki.net please?
Note: NOT ecki.net, because this will be in Germany and not Thailand where the pool server is located!
The ping time is important, but the variance is also very important.
The difference between Minimum, Maximum and Average should be as little as possible!
But most important for mining are the number of packet drops!
So just ping the target 10000 times and have a look about the % of packet losses.
Ping is important, but it is not all...
If you have a high number of packet drops to your current pool, then choose another one and it will increase your current hash rate of 360 MH/s!
For more answers or information about network questions please PM me, this is off topic here in this thread (I assume)... even you are using an Icarus board.
Cu,
Ecki
Hello Hardpick,
how important is ping time?
I am getting 360 mh/s from 2nd batch icarus on deepbit which has a ping time of 410 ms
(my isp ping time is 11ms -- have adsl2 at 12.34 mbps/ 0.83 mbps -- I live in sydney australia)
I have tested other pools
...
ecki.net.-----361ms
...
I am getting 360 mh/s from 2nd batch icarus on deepbit which has a ping time of 410 ms
(my isp ping time is 11ms -- have adsl2 at 12.34 mbps/ 0.83 mbps -- I live in sydney australia)
I have tested other pools
...
ecki.net.-----361ms
...
361 ms from Australia to Thailand? This is very bad...
Which Internet Provider do you use? Telstra?
I have less than 200ms to my customers in Sydney from here (Thailand)... so it is the same distance!
Antwort von 203.37.115.162: Bytes=32 Zeit=189ms TTL=236
Antwort von 203.37.115.162: Bytes=32 Zeit=190ms TTL=236
Antwort von 203.37.115.162: Bytes=32 Zeit=188ms TTL=236
Antwort von 203.37.115.162: Bytes=32 Zeit=184ms TTL=236
Ping-Statistik für 203.37.115.162:
Pakete: Gesendet = 4, Empfangen = 4, Verloren = 0
(0% Verlust),
Ca. Zeitangaben in Millisek.:
Minimum = 184ms, Maximum = 190ms, Mittelwert = 187ms
Could you PM me your traceroute to miner.ecki.net please?
Note: NOT ecki.net, because this will be in Germany and not Thailand where the pool server is located!
The ping time is important, but the variance is also very important.
The difference between Minimum, Maximum and Average should be as little as possible!
But most important for mining are the number of packet drops!
So just ping the target 10000 times and have a look about the % of packet losses.
Ping is important, but it is not all...
If you have a high number of packet drops to your current pool, then choose another one and it will increase your current hash rate of 360 MH/s!
For more answers or information about network questions please PM me, this is off topic here in this thread (I assume)... even you are using an Icarus board.
Cu,
Ecki
February 03, 2012, 03:49:37 AM
Quote
this event may only happen when the FPGA is broken. and NEVER happened until now.
Just to clarify before someone panics, I'm sure ngzhang meant to say that "it has NEVER happened yet" or hasn't happened yet. Grammar Syntax Nazi on the attack
lolOK。
it has NEVER happened yet
February 03, 2012, 03:16:12 AM
how important is ping time?
I am getting 360 mh/s from 2nd batch icarus on deepbit which has a ping time of 410 ms
(my isp ping time is 11ms -- have adsl2 at 12.34 mbps/ 0.83 mbps -- I live in sydney australia)
I have tested other pools
50btc.com-----382ms
A1BitcoinPool.com-----359ms
asspennypool.com-----308ms
BitClockers.com-----233ms
bitcoin.cz-----387ms
bitcoins.lc-----388ms
bithasher.com-----277ms
bitparking.com-----247ms
bitpenny.com-----247ms
btcguild.com-----243ms
btcmine.com-----275ms
btcserv.net-----413ms
deepbit.net-----410ms
ecki.net.-----361ms
eclipsemc.com-----239ms
eligius.st-----412ms
mitminter.com-----395ms
mkalinin.ru-----347ms
mtred.com-----279ms
NMCBIT.COM-----271ms
ozco.in-----234ms
simplecoin.us-----217ms
Triplemining.com-----396ms
I am getting 360 mh/s from 2nd batch icarus on deepbit which has a ping time of 410 ms
(my isp ping time is 11ms -- have adsl2 at 12.34 mbps/ 0.83 mbps -- I live in sydney australia)
I have tested other pools
50btc.com-----382ms
A1BitcoinPool.com-----359ms
asspennypool.com-----308ms
BitClockers.com-----233ms
bitcoin.cz-----387ms
bitcoins.lc-----388ms
bithasher.com-----277ms
bitparking.com-----247ms
bitpenny.com-----247ms
btcguild.com-----243ms
btcmine.com-----275ms
btcserv.net-----413ms
deepbit.net-----410ms
ecki.net.-----361ms
eclipsemc.com-----239ms
eligius.st-----412ms
mitminter.com-----395ms
mkalinin.ru-----347ms
mtred.com-----279ms
NMCBIT.COM-----271ms
ozco.in-----234ms
simplecoin.us-----217ms
Triplemining.com-----396ms
February 03, 2012, 01:53:13 AM
Quote
this event may only happen when the FPGA is broken. and NEVER happened until now.
Just to clarify before someone panics, I'm sure ngzhang meant to say that "it has NEVER happened yet" or hasn't happened yet. Grammar Syntax Nazi on the attack
lolOK。
it has NEVER happened yet
February 03, 2012, 01:36:46 AM
Quote
this event may only happen when the FPGA is broken. and NEVER happened until now.
Just to clarify before someone panics, I'm sure ngzhang meant to say that "it has NEVER happened yet" or hasn't happened yet. Grammar Syntax Nazi on the attack
lol February 03, 2012, 12:17:18 AM
Hi there, I have some questions:
-In the first post, I can see this:
"$: 569 (1) / 469$ (multiple of 30)"
so 1 board is 569$ and 30 boards is 469$ each, is that correct? If 30 boards is 469$ each, I might be interested in buying 30 Icarus boards.
-Also, to use them, you only need the boards and a PC where you connect the boards with USB. So, in theory, the more there is USB ports, the more you can plug into that PC, correct?
-Does the PC have some requirements on their specs? If I use a basic Sempron 130 with 1 Gb of RAM, is it going to be ok? I also suppose that using Linux on that PC is ok?
-You ship to Canada? What's the shipping cost?
-Let's say that in theory, I have a motherboard with 8 USB ports and 4 PCI-Express slots. So, always in theory, I could plug 8 FGPA board and 4 GPU in that motherboard and they could all mine together, correct?
Thanks for your answers. This project seems to have potential
-In the first post, I can see this:
"$: 569 (1) / 469$ (multiple of 30)"
so 1 board is 569$ and 30 boards is 469$ each, is that correct? If 30 boards is 469$ each, I might be interested in buying 30 Icarus boards.
-Also, to use them, you only need the boards and a PC where you connect the boards with USB. So, in theory, the more there is USB ports, the more you can plug into that PC, correct?
-Does the PC have some requirements on their specs? If I use a basic Sempron 130 with 1 Gb of RAM, is it going to be ok? I also suppose that using Linux on that PC is ok?
-You ship to Canada? What's the shipping cost?
-Let's say that in theory, I have a motherboard with 8 USB ports and 4 PCI-Express slots. So, always in theory, I could plug 8 FGPA board and 4 GPU in that motherboard and they could all mine together, correct?
Thanks for your answers. This project seems to have potential
yes, if your amount is larger than 30, the unit price is 469$, if you do not need the adapter (recommend for bulk orders), the unit price is 464$. save you 150$ to get a good power.
there is nearly no requirements for the PC. even somebody is working on put the miner into an ARM based platform. a PC can run windowsXP or linux is ok.
about the shipment cost, if using EMS, a 15KG package cost is around 300$.
the USB port can expand with HUBs. so 100+ FPGA boards may operate on one PC.
Hi!
I´m from Brazil and I would like to buy the FPGA to mine. Can you send to Brazil?
Thank you!
I´m from Brazil and I would like to buy the FPGA to mine. Can you send to Brazil?
Thank you!
certainly , some Brazilian is already my customers.
Code:
[quote author=ngzhang link=topic=51371.msg727871#msg727871 date=1328204144]
present configuration is the FPGA will stop working when it find a valid nonce. but the other FPGA in chain will continue working (because it didn't know what is happened on the other FPGA).
this idea is to keep a simple event. so software miner should push a work to FPGA immediately when received a valid nonce.
Does this behaviour also mean that if the second FPGA goes dead/locks up the first one will continue working? I'd assume if the first one goes dead it prevents comm to the 2nd one but I'm curious about independence when the first is still reachable but the second doesn't respond any more?
i designed the logic inside the FPGA as a robust way. this event may only happen when the FPGA is broken. and NEVER happened until now.
February 02, 2012, 08:04:42 PM
Code:
[quote author=ngzhang link=topic=51371.msg727871#msg727871 date=1328204144]
present configuration is the FPGA will stop working when it find a valid nonce. but the other FPGA in chain will continue working (because it didn't know what is happened on the other FPGA).
this idea is to keep a simple event. so software miner should push a work to FPGA immediately when received a valid nonce.
Does this behaviour also mean that if the second FPGA goes dead/locks up the first one will continue working? I'd assume if the first one goes dead it prevents comm to the 2nd one but I'm curious about independence when the first is still reachable but the second doesn't respond any more?
February 02, 2012, 06:04:40 PM
Hi!
I´m from Brazil and I would like to buy the FPGA to mine. Can you send to Brazil?
Thank you!
I´m from Brazil and I would like to buy the FPGA to mine. Can you send to Brazil?
Thank you!
February 02, 2012, 05:54:10 PM
-Let's say that in theory, I have a motherboard with 8 USB ports and 4 PCI-Express slots. So, always in theory, I could plug 8 FGPA board and 4 GPU in that motherboard and they could all mine together, correct?
You can use USB hubs and additional USB controllers too. 
February 02, 2012, 05:24:23 PM
Hi there, I have some questions:
-In the first post, I can see this:
"$: 569 (1) / 469$ (multiple of 30)"
so 1 board is 569$ and 30 boards is 469$ each, is that correct? If 30 boards is 469$ each, I might be interested in buying 30 Icarus boards.
-Also, to use them, you only need the boards and a PC where you connect the boards with USB. So, in theory, the more there is USB ports, the more you can plug into that PC, correct?
-Does the PC have some requirements on their specs? If I use a basic Sempron 130 with 1 Gb of RAM, is it going to be ok? I also suppose that using Linux on that PC is ok?
-You ship to Canada? What's the shipping cost?
-Let's say that in theory, I have a motherboard with 8 USB ports and 4 PCI-Express slots. So, always in theory, I could plug 8 FGPA board and 4 GPU in that motherboard and they could all mine together, correct?
Thanks for your answers. This project seems to have potential
-In the first post, I can see this:
"$: 569 (1) / 469$ (multiple of 30)"
so 1 board is 569$ and 30 boards is 469$ each, is that correct? If 30 boards is 469$ each, I might be interested in buying 30 Icarus boards.
-Also, to use them, you only need the boards and a PC where you connect the boards with USB. So, in theory, the more there is USB ports, the more you can plug into that PC, correct?
-Does the PC have some requirements on their specs? If I use a basic Sempron 130 with 1 Gb of RAM, is it going to be ok? I also suppose that using Linux on that PC is ok?
-You ship to Canada? What's the shipping cost?
-Let's say that in theory, I have a motherboard with 8 USB ports and 4 PCI-Express slots. So, always in theory, I could plug 8 FGPA board and 4 GPU in that motherboard and they could all mine together, correct?
Thanks for your answers. This project seems to have potential
February 02, 2012, 02:22:24 PM
Another question about the communication specs: How do the boards behave after they found a solution? Do they keep working on the current work till they get new work, possibly resulting in 3 or more solutions per work unit?
present configuration is the FPGA will stop working when it find a valid nonce. but the other FPGA in chain will continue working (because it didn't know what is happened on the other FPGA).
this idea is to keep a simple event. so software miner should push a work to FPGA immediately when received a valid nonce.
ADD:
about the detection. there is a simple way to impalement, with out any RTL change.
just push a 512bits dummy work package to the COM port, fot example:
Code:
['data'] = "0000000120c8222d0497a7ab44a1a2c7bf39de941c9970b1dc7cdc400000079700000000e88aabe1f353238c668d8a4df9318e614c10c474f8cdf8bc5f6397b946c33d7c4e7242c31a098ea500000000000000800000000000000000000000000000000000000000000000000000000000000000000000000000000080020000"
['midstate'] = "33c5bf5751ec7f7e056443b5aee3800331432c83f404d9de38b94ecbf907b92d"
the boards will return "063c5e01" in a short time.
Hehe I found this in our miner source, too. Helped me a lot to get the communication working
.About the stop-searching-after-a-nonce-is-found thing, is there some specific reason for this behaviour or could this be changed? I think it would simplify miner developement if it kept searching, because currently a new getwork request has to be made after a nonce is found. Laggy internet connections make the FPGAs idle more and lead to lower hash rates (I think that's why my boards currently only make about 350 mhash each).
If the FPGAs kept searching, a simple getwork-thread could feed the FPGAs new work every about <11 seconds, minimizing the negative effects of a bad connection.
I think for now I'm going to implement some queueing mechanism as a workaround (couldn't flash a different bitstream anyways because I don't yet have a platform cable).
i know what you mean. and i just solve this problem by a work queue function.
my suggest is DO NOT CHANGE THE BITSTEAM, because a near 200M speed bitsteam is very hard to implement. or you can try to synthesize the code.
February 02, 2012, 02:04:12 PM
Another question about the communication specs: How do the boards behave after they found a solution? Do they keep working on the current work till they get new work, possibly resulting in 3 or more solutions per work unit?
present configuration is the FPGA will stop working when it find a valid nonce. but the other FPGA in chain will continue working (because it didn't know what is happened on the other FPGA).
this idea is to keep a simple event. so software miner should push a work to FPGA immediately when received a valid nonce.
ADD:
about the detection. there is a simple way to impalement, with out any RTL change.
just push a 512bits dummy work package to the COM port, fot example:
Code:
['data'] = "0000000120c8222d0497a7ab44a1a2c7bf39de941c9970b1dc7cdc400000079700000000e88aabe1f353238c668d8a4df9318e614c10c474f8cdf8bc5f6397b946c33d7c4e7242c31a098ea500000000000000800000000000000000000000000000000000000000000000000000000000000000000000000000000080020000"
['midstate'] = "33c5bf5751ec7f7e056443b5aee3800331432c83f404d9de38b94ecbf907b92d"
the boards will return "063c5e01" in a short time.
Hehe I found this in our miner source, too. Helped me a lot to get the communication working
.About the stop-searching-after-a-nonce-is-found thing, is there some specific reason for this behaviour or could this be changed? I think it would simplify miner developement if it kept searching, because currently a new getwork request has to be made after a nonce is found. Laggy internet connections make the FPGAs idle more and lead to lower hash rates (I think that's why my boards currently only make about 350 mhash each).
If the FPGAs kept searching, a simple getwork-thread could feed the FPGAs new work every about <11 seconds, minimizing the negative effects of a bad connection.
I think for now I'm going to implement some queueing mechanism as a workaround (couldn't flash a different bitstream anyways because I don't yet have a platform cable).
February 02, 2012, 12:35:44 PM
Another question about the communication specs: How do the boards behave after they found a solution? Do they keep working on the current work till they get new work, possibly resulting in 3 or more solutions per work unit?
present configuration is the FPGA will stop working when it find a valid nonce. but the other FPGA in chain will continue working (because it didn't know what is happened on the other FPGA).
this idea is to keep a simple event. so software miner should push a work to FPGA immediately when received a valid nonce.
ADD:
about the detection. there is a simple way to impalement, with out any RTL change.
just push a 512bits dummy work package to the COM port, fot example:
Code:
['data'] = "0000000120c8222d0497a7ab44a1a2c7bf39de941c9970b1dc7cdc400000079700000000e88aabe1f353238c668d8a4df9318e614c10c474f8cdf8bc5f6397b946c33d7c4e7242c31a098ea500000000000000800000000000000000000000000000000000000000000000000000000000000000000000000000000080020000"
['midstate'] = "33c5bf5751ec7f7e056443b5aee3800331432c83f404d9de38b94ecbf907b92d"
the boards will return "063c5e01" in a short time.
February 02, 2012, 12:27:20 PM
Another question about the communication specs: How do the boards behave after they found a solution? Do they keep working on the current work till they get new work, possibly resulting in 3 or more solutions per work unit?
February 01, 2012, 12:18:59 AM
hi, guys.
the holiday is finally over.
i'm going to have some meeting with my supplier and factory, and place some order for the 3rd batch.
it's sure there will be some BOM changes to increase power efficiency, enhanced cooling, and so on.
so later 2-3 days i will very busy. PM and email reply will much slower. please forgive me. confirm an accurate manufacture plan is the most important thing to do now.
regards everyone.
the holiday is finally over.
i'm going to have some meeting with my supplier and factory, and place some order for the 3rd batch.
it's sure there will be some BOM changes to increase power efficiency, enhanced cooling, and so on.
so later 2-3 days i will very busy. PM and email reply will much slower. please forgive me. confirm an accurate manufacture plan is the most important thing to do now.
regards everyone.
I thought the Chinese New Year holiday is 15 days long ?!?!
Wikipedia seems to say so: http://en.wikipedia.org/wiki/Chinese_new_year
So, if day 1 was Jan. 23rd, day 9 is Jan 31st, day 10 is Feb 1st and day 15 is Feb. 6th
yeah, 15 days long "in the country". or even 1 month long.
but in city, about 7-10 days long.
February 01, 2012, 12:03:03 AM
hi, guys.
the holiday is finally over.
i'm going to have some meeting with my supplier and factory, and place some order for the 3rd batch.
it's sure there will be some BOM changes to increase power efficiency, enhanced cooling, and so on.
so later 2-3 days i will very busy. PM and email reply will much slower. please forgive me. confirm an accurate manufacture plan is the most important thing to do now.
regards everyone.
the holiday is finally over.
i'm going to have some meeting with my supplier and factory, and place some order for the 3rd batch.
it's sure there will be some BOM changes to increase power efficiency, enhanced cooling, and so on.
so later 2-3 days i will very busy. PM and email reply will much slower. please forgive me. confirm an accurate manufacture plan is the most important thing to do now.
regards everyone.
I thought the Chinese New Year holiday is 15 days long ?!?!
Wikipedia seems to say so: http://en.wikipedia.org/wiki/Chinese_new_year
So, if day 1 was Jan. 23rd, day 9 is Jan 31st, day 10 is Feb 1st and day 15 is Feb. 6th
January 31, 2012, 11:33:47 PM
hi, guys.
the holiday is finally over.
i'm going to have some meeting with my supplier and factory, and place some order for the 3rd batch.
it's sure there will be some BOM changes to increase power efficiency, enhanced cooling, and so on.
so later 2-3 days i will very busy. PM and email reply will much slower. please forgive me. confirm an accurate manufacture plan is the most important thing to do now.
regards everyone.
the holiday is finally over.
i'm going to have some meeting with my supplier and factory, and place some order for the 3rd batch.
it's sure there will be some BOM changes to increase power efficiency, enhanced cooling, and so on.
so later 2-3 days i will very busy. PM and email reply will much slower. please forgive me. confirm an accurate manufacture plan is the most important thing to do now.
regards everyone.
Jump to: