Cairnsmore1 - Quad XC6SLX150 Board

[norulezapply]

Update

Ok first units are out. The first few days will be a very slow build up so a little patience is required. We are aiming for most units to ship with at least with at least in-built programming if not some level of bitstream so we will take 1 or 2 days to try and get that right. This also happens to tie up with our proper packaging that arrives hopefully on Friday.

Hi Yohan.

I got the email for the early shipment of the board, but I was wondering. If these boards are already shipping, without yet having a working bitstream or in-built programming option, does this mean I will have to purchase a JTAG programming cable to reprogram the boards once the bitstream is released?

Or will the shipping of them be postponed til Friday with the proper packaging and (hopefully) in-built programming and/or working bitstream?

Thanks and keep up the great work

[1713557376]

1713557376

[1713557376]

1713557376

[1713557376]

1713557376

[1713557376]

1713557376

[yohan]

If the build in programming is working you won't need a programming cable and that is our no1 aim to have that working this week. That makes it easier for everyone and we can then update bitstreams as we have better ones. Secondary aim is to have a working bitstream at the same time. We have held the main ship until Friday not only to try and achieve these things but also because our main packaging does not arrive until then. This small delay makes some sense for us to get these things right. However if any of our customers are proficient in FPGA programming and have the tools we can ship them tomorrow in adhoc packaging. Anyone who has their early ship notification and that feels that they are in that category just let us know.

The few people that have boards already for development have cables with them so we can keep them up to date in the short until the in-built programmer takes over.

We have some customisations to CGminer in hand for our own bitstream but if we get the Icarus bitstream hack working first then it should just look like 2 x Icarus. The Icarus way is a temporary solution and was always planned that way. We have thought from the beginning that we can offer something much better in performance and you can all speculate what that might be. Until it works I won't say much more. It could be a bad idea and big damp squib after all. The team are working hard on all of these things and just a little patience is needed whilst we get them right.

Production wise it is shaping up nicely. Our new Cairnsmore1 final assembly and test area gets fitted and kitted out tomorrow and after that it gets a bit easier for us to push large numbers of boards through these stages.

[dirtycat]

What are people using to mine?

... does it work with cgminer as an Icarus? (i.e. as 2 Icarus /dev/ttyUSBx or 2 COM ports?)
the latest (a few weeks old) cgminer git pull I have up may also help with that if it's a different speed to an Icarus Rev3
(the new --icarus-timing option)

Though, the fact that it is 4 chips (and not 2) may? need some minor changes to cgminer
(if it doesn't appear to the OS as 2 tty/COM ports)

IIRC nobody has this thing mining... YET!.. cant wait to see the results! im so excited!

[MXRider]

Call me crazy but I just refused for an early board I don't see the point having one until it's completely ready with bitstreams and all.

Mr Glasswalker. Post those number of yours

[Garr255]

Call me crazy but I just refused for an early board I don't see the point having one until it's completely ready with bitstreams and all.

Mr Glasswalker. Post those number of yours

I agree with your decision. A car isn't a car without an engine. (Or motor for those of you who were going to argue...)

[Glasswalker]

Mr Glasswalker. Post those number of yours

FYI: This board is one of the early "Developer" boards. So no bitstream on it yet (As yohan mentioned).

I still have to synthesize a bitstream, and flash it to the board over JTAG to get this hashing. Then I can post numbers. That's why I said it would take me a few days

But yes I'll get numbers posted as soon as possible.

Let's call it a race, see if I can beat Enterpoint to releasing numbers for their own product lol (which they will probably win lol)

[Inspector 2211]

FYI: This board is one of the early "Developer" boards. So no bitstream on it yet (As yohan mentioned).
I still have to synthesize a bitstream, and flash it to the board over JTAG to get this hashing. Then I can post numbers. That's why I said it would take me a few days

 

I fear that you will find out that synthesizing a bitstream for the Spartan6-150 is not as easy as some people think it is.

Yes, achieving 120 MH/s or 140 MH/s per FPGA may be fairly easy, but 210 MH/s (ZTEX) is not easy at all.

Achieving 260 MH/s (EldenTyrell) or 300 MH/s (Bitfury) is outright difficult.
"Outright difficult" ::= "the most difficult thing you've ever done in your life"

[nbtcminer]

"Outright difficult" ::= "the most difficult thing you've ever done in your life"

Speaking of which, isn't Eldentyrell's bitstream due sometime today / tomorrow?

[Inspector 2211]

"Outright difficult" ::= "the most difficult thing you've ever done in your life"

Speaking of which, isn't Eldentyrell's bitstream due sometime today / tomorrow?

On June 1st, I think.

But Bitfury's 300 MH/s bitstream and then, in turn, BFL's ASIC announcement have, as far as I see it, dramatically reduced the price he can reasonably charge for it. Furthermore, some FPGA boards (ZTEX boards, for instance) have DC/DC converters that are too weak to even run it.

Edit: It turns out, his website will go live in the wee morning hours of May 31st.

[norulezapply]

"Outright difficult" ::= "the most difficult thing you've ever done in your life"

Speaking of which, isn't Eldentyrell's bitstream due sometime today / tomorrow?

According to the following website - yes.

http://www.tricone-mining.com

[2112]

I fear that you will find out that synthesizing a bitstream for the Spartan6-150 is not as easy as some people think it is.

I somehow got an impression that the fault is just a swap of RxD/TxD lines. This should be doable with "fpgaeditor" on the ".ncd" file. But I may be wrong.

[Inspector 2211]

I fear that you will find out that synthesizing a bitstream for the Spartan6-150 is not as easy as some people think it is.

I somehow got an impression that the fault is just a swap of RxD/TxD lines. This should be doable with "fpgaeditor" on the ".ncd" file. But I may be wrong.

1. If he takes someone else's bitstream and runs it verbatim, that's easy as pie, as long as pinout and input clock are compatible.

2. But a bitstream cannot be edited at all, unless someone (like ngzhang, or Stefan) chose to provide the placed&routed .ncd file to a competitor. As you correctly point out, the .ncd file can be edited and a bitstream can be generated from it.

3. But if he literally tries to QUOTE synthesize a bitstream UNQUOTE - that's quite tough.

[Glasswalker]

Correct. As far as I know, ngzhang has not provided an ncd file. In addition a bitstream needs to be installed on the Control FPGA as well (Spartan 3E). To route traffic from the USB to the 4x matrix FPGAs.

So I'll be synthesizing (from HDL source) the bitstream for the Control chip (custom code), and for the 4 matrix chips, based on a modified version of the Icarus firmware (which I modified from the github source.)

Yes it's quite tough, and I have limited time. Which is why I said Enterpoint would likely beat me (won't stop me from trying though).

Ultimately I've been working on another bitstream specifically designed for this board (for some time now). But this is a quick fix to get it mining just so I can get some performance numbers for it. The other bitstream is quite some time away still. And I have no idea how my bitstream will perform. Since the Icarus bitstream performs in a known way on icarus hardware, the modified version should give an interesting comparison on the Cairnsmore1 board.

Anyway, no promises. I'm poking at it as I have time. If I succeed, I'll post numbers ASAP.

[2112]

As far as I know, ngzhang has not provided an ncd file.

Please double check. I currently have no access to my main computer, but I do distincly remember that he provided the design in an relatively easily editable form. The main hashing pipelines were created using Synopsys and the placed&routed netlist from that step was fed to the Xilinx ISE as a black box to be sorrounded by the glue/interface logic. The only reason that I couldn't play with it was that my main machine couldn't be upgraded with enough RAM to get sensible speed in the toolchain.

[Lethos]

Looking forward to seeing some epic results from these boards.
So do these 4 spartan 6's actually have the potential to do 250 - 300Mhash/s each?
I understand it's hard so won't get my hopes up.
However it's just a software thing that makes is possible?

[wildemagic]

So do these 4 spartan 6's actually have the potential to do 250 - 300Mhash/s each?

More likely 190-210Mh/sec each.

Others have speculated that customized bitstreams can run the fpa faster and more efficiently, but that is just speculation at the moment.

Expect to see around 800Mh/sec from these boards.

kind regards

[Lethos]

So do these 4 spartan 6's actually have the potential to do 250 - 300Mhash/s each?

More likely 190-210Mh/sec each.

Others have speculated that customized bitstreams can run the fpa faster and more efficiently, but that is just speculation at the moment.

Expect to see around 800Mh/sec from these boards.

kind regards

I'd be happy with that, I know it's all speculation until real results are posted and proven.
I'll be getting two of these in the next month I believe. Unfortunately I am a programmer with no experience with this area of coding, so I'm not opted to be involved with the developer boards but it will be nice to one day know the bitstream could be improved. Maybe next time once I learn how it's done.

[BTC-engineer]

FAN spec looks good but I looked up the nearest distributor and no listing of the 12cm model that I could so it sort of suggests that it's very new and maybe not quite available yet.

We are committed anyway in a large way to the Arctic F12. It's pretty quiet and more to the point available in large numbers to us in a short leadtime. When you start looking at the quantities we are buying many of them have to take the boat from china and that's usually 6 weeks min delivery.

If people want they have the option to change the fan themselves it will be simple enough to do that.

My recommendation is stay with the Arctic F12.

My experience regarding the Silenx fan vs. the Arctic F12 fan is the following.
If you compare the datasheets the fan from Silenx looks in most things better than the Arctic (or others), especially the sound level and air flow.
I've compared the Silenx fans with fans from other vendors including Arctic and I would say that the specified data from Silenx is not correct.
The Silenx fan is good, but the '15db'-Version is comparable to the Arctic F12. The real air-flow is not so good like from the ARCTIC F12 and the sound level is only slightly better. According the datasheet the air-flow and sound level should be much better. I've done Silenx orders from two different supplies to decrease the risk of just having a copy or broken fan.

[Glasswalker]

Looking forward to seeing some epic results from these boards.
So do these 4 spartan 6's actually have the potential to do 250 - 300Mhash/s each?
I understand it's hard so won't get my hopes up.
However it's just a software thing that makes is possible?

Currently EldenTyrell has a custom bitstream which can push these same chips to 250MHash/s approx.

And BitFury has a custom bitstream pushing (liquid cooled I believe) LX150s (same chips as Cairnsmore) to 300MHash/s or more.

But it should be pointed out that neither of those has been "independently verified" by a third party.

So yes the chips have the potential to reach much higher hash rates using a new bitstream (essentially firmware for an FPGA).

But right now, I think Enterpoints goal is to get "icarus-like hashing per chip" (so 190-200mhash) out of them to get them in customers hands, with an easy method of usb updating the boards so that once a new/improved bitstream is available, they can be updated easily.

[seriouscoin]

The pre-order price was said to be until the end of June. However the pre-order is now closed? 

Can you clarify if we can still preorder the board?