FPGA development board "Icarus" - DisContinued/ important announcement

[DrHaribo]

Currently beta testing Icarus support for the BitMinter client:
https://bitcointalk.org/index.php?topic=31163.msg1008906#msg1008906

[1714293907]

1714293907

[1714293907]

1714293907

[Luke-Jr]

Nonces starting with three nullbytes (0x000000??) are never reported. Nonces starting with three FF (0xFFFFFF??) always get 01 for the fourth byte, which is usually an incorrect result. This had me confused for a while. Error in the bitstream?

Adding Icarus support to my miner. But I can't get it to ever return a nonce starting with three null bytes. I feed it data for which there is such a solution, but it's not working.

Is that normal? Does it start scanning from nonce 0x00000100 ?

yes， it didn't scan the nonce range form 0. approx: 132

Can you give an exact answer?
Since that will slightly affect the timing for when the nonce range is aborted and also how many nonces were actually processed when a nonce is found.
i.e. also details of how it affects both FPGAs

About 253 according to https://github.com/progranism/Open-Source-FPGA-Bitcoin-Miner/blob/master/projects/X6000_ztex_comm4/hdl/fpgaminer_top.v#L155

Note it's probably not the exact same bitstream.

[wildemagic]

However, 11.2s is calculated based on the Icarus mining at 2.6316ns per hash
(which of course means 5.2632ns per pair of hashes)

Once you have a new correct nanosecond value for your faster bitstream (lets say it was 2.2222ns), you simply start cgminer with the option:
--icarus-timing 2.2222

Ok so doing a temp update to the 200mhz bitstream I recalc'd and get these values :

2.500 10.744

so would I be right to set --icarus-timing 2.5000=10.8

or do i leave the second value at the default and just have the 2.5000 without the =10.8 ?

Preliminary results seem to have bumped the U/rate up from 20-21 to about 21-22, so its the expected 5%
Power consumption seems unchanged.

[update]

Setting --icarus-timing 2.5000=10.8 yielded results of about 1200 shares per hour, similar to the default 190mhz bitstream which usually nets about 1260 shares per hour
Setting --icarus-timing 2.5000 is yielding about 1328 shares per hour which is the expected 5% bump going from the 190mhz up to the 200mhz bitstream

kind regards

[kano]

As per the FPGA-README, the number you supply would be 108 (not 10.8)

However, 108 is too high for a 2.5ns bitstream - as I mentioned before 2.5ns would be less than that (10.6=106)

If 2.5ns is correct (or lower than correct) then it will calculate the 2nd number and that will work fine.

Certainly you could use 80 (8.0s) and that would ensure it's never idle.

Seriously, if you want it optimised accurate you need to at least run --icarus-timing short (for about an hour) and see the answer it comes up with

[wildemagic]

As per the FPGA-README, the number you supply would be 108 (not 10.

Apologies, I was using 108 not 10.8

Seriously, if you want it optimised accurate you need to at least run --icarus-timing short (for about an hour) and see the answer it comes up with

I did that, the results turned out to be 2.5000 10.7xx

Thanks for your confirmation.

kind regards

[rudrigorc2]

As per the FPGA-README, the number you supply would be 108 (not 10.

Apologies, I was using 108 not 10.8

Seriously, if you want it optimised accurate you need to at least run --icarus-timing short (for about an hour) and see the answer it comes up with

I did that, the results turned out to be 2.5000 10.7xx

Thanks for your confirmation.

kind regards

are your ICA from the last batch?

[wildemagic]

No, I have 1xv2 and 3xv3

kind regards

[wildemagic]

Hrm, varying results from the 200mhz beta bitstream.  Seems faster, but then its erratic like the cores are having errors sometimes.  CGMiner does not report any HW errors tho.
I think the 200mhz stream might be an older bitstream @ high speed.  Next up ill give the v4 bitstream a go and see what the results are.

4x120mm fans on the way and ill get some better temps on the cores and try again at high speed.

I will also try and get the Tricone mining bitstream going over jtag when i get better cooling and report back here with that too.

kind regards

[wildemagic]

The 190mhz v2 and v3 default shipping bitstreams result in results of approximately 1490-1530MH/s for the 4 boards.

The 200mhz bitstream showed results of approximately 1520-1650MH/s for the 4 boards.
BUT large swings in speed at times significantly reduced hashing speeds due to overheating im guessing (only have the v2 40mm fan and 3xv3 50mm fans installed for cooling)

Now testing the 190mhz v4 bitstream with some interesting results.  Seems fast so far sitting pretty solidly on 1620MH/s. U=22 or about U=5.5 per board.
Im pretty sure there are some efficiency improvements in the v4 bitstream that the 200mhz beta doesnt have.  
From size comparisons it seems the 200mhz bitstream is based on the v3 bitstream (perhaps ngzhang can confirm)

If I knew how I would like to synthesize a v4 bitstream at higher speed, but i suspect this is not an easy process.

kind regards

[ngzhang]

 

V4 ..

will have a final bitsteam update next week(i hope). and completely change the architecture.
please wait for that before do any optimization effort.

[wildemagic]

will have a final bitsteam update next week

please make some mcs files and not just .bit files so that i can perminently flash the fpgas.

if possible could you also make msc files of the .bit files in the wiki so that at least v3 v4 and beta200 are perm flashable?

kind regards

[xiangfu]

+1

BTW:
if there some picture about your '1.7kW solar array' and your mining farm ?
  4 x Icarus 1 x 5850  all powered by a 1.7kW solar array and 3G wireless in outback Australia.

will have a final bitsteam update next week

please make some mcs files and not just .bit files so that i can perminently flash the fpgas.

if possible could you also make msc files of the .bit files in the wiki so that at least v3 v4 and beta200 are perm flashable?

kind regards

[wildemagic]

BTW:
if there some picture about your '1.7kW solar array' and your mining farm ?

The farm is a bit boring, just a pc and 4 icarus cards.  I will be building an enclosure for the icarus before summer to improve the 4x120mm fan airflow.

But I do have an old pic of my panels http://img713.imageshack.us/img713/4863/2011121318opt.jpg and batteries/inverters http://img828.imageshack.us/img828/1208/2011121318opt1.jpg
I have 2 additional 125w panels and 2 additional 60w panels now.

kind regards

[ngzhang]

will have a final bitsteam update next week

please make some mcs files and not just .bit files so that i can perminently flash the fpgas.

if possible could you also make msc files of the .bit files in the wiki so that at least v3 v4 and beta200 are perm flashable?

kind regards

certainly, will release both .bit and .mcs files.

i will release .mcs files for all bitsteams, unless it's a UNrecommended test bitsteam (like 200m beta).

[wildemagic]

i will release .mcs files for all bitsteams, unless it's a UNrecommended test bitsteam (like 200m beta).

The 200mhz one is the one I really want, Im running really good cooling.

Are your other bitstreams going to approach the performance of the 200mhz beta ?

kind regards

[ngzhang]

i will release .mcs files for all bitsteams, unless it's a UNrecommended test bitsteam (like 200m beta).

The 200mhz one is the one I really want, Im running really good cooling.

Are your other bitstreams going to approach the performance of the 200mhz beta ?

kind regards

 
wait for a few days.

[luffy]

it would also be great if Icarus can be programmed from PC via USB!
is it possible in the near future?
or else a wiki about Icarus programming is welcomed
thanks

[wildemagic]

it would also be great if Icarus can be programmed from PC via USB!
is it possible in the near future?

I also would like this, as it means that I can try the tricone stream and wouldnt have to fiddle with that damn xilinx cable.

kind regards

[wildemagic]

Well after installing the icarus modules under 120mm fans, I have noticed considerable improvement in stability for the 200Mhz beta bitstream :

Code:

ICA 0:                | 399.8/395.5Mh/s | A:5618 R:22 HW:0 U: 5.55/m
ICA 1:                | 399.8/395.4Mh/s | A:5557 R:25 HW:0 U: 5.49/m
ICA 2:                | 399.6/395.3Mh/s | A:5530 R:29 HW:0 U: 5.46/m
ICA 3:                | 399.7/395.3Mh/s | A:5574 R:22 HW:0 U: 5.50/m

I use the  --icarus-timing 2.5000=100 command line option and things seem to be working well enough.
All and all its about 1600mh/s for the 4 units and improvement of exactly 5% as per the increase from 190mhz-200mhz.

If I knew how I would like to do a 210 or faster and see what I could net.  Unfortunately I have NFI <grins>

Looking forward to nzhangs announcement in regards to the new bitstream.

kind regards

[xiangfu]

Hi

You rock. this is totally cool. I real want that.

BTW:
if there some picture about your '1.7kW solar array' and your mining farm ?

The farm is a bit boring, just a pc and 4 icarus cards.  I will be building an enclosure for the icarus before summer to improve the 4x120mm fan airflow.

But I do have an old pic of my panels http://anonymouse.org/cgi-bin/anon-www.cgi/http://img713.imageshack.us/img713/4863/2011121318opt.jpg and batteries/inverters http://anonymouse.org/cgi-bin/anon-www.cgi/http://img828.imageshack.us/img828/1208/2011121318opt1.jpg
I have 2 additional 125w panels and 2 additional 60w panels now.

kind regards