We will try and do some pictures of inital stack options maybe today and that will given some ideas to people of what is possible. We are still working on things like brackets for side blow fans and other stacking heights but these will probably take a few weeks to sort out and to have a full range available. Again this is an evolving area and a little patience is needed whilst we come up with the best solutions. It's not the highest priority on our list to do but we will get there and make something that works for customers.

Ok I will try and do a "state of the nation" outline:

USB- Drivers for the USB will hopefully be in place today. Basically we have are tidying up drivers that we have working so not really a risk but some time required. The FT4232 chip is very versatile and we have notionally one of the 4 channels that it supports allocated to JTAG, one to SPI driving, and two for coms. The FT4232 has internal micro-engines to support JTAG and SPI functions. At the moment the JTAG can be used to program the main array FPGAs and Flash memories that store bitstreams and we have this working for Linux and Windows. We may extend the SPI functions to directly program array Flash memories but that isn't working as yet. That's a nice to have and actually more of use to us in production maybe than anyone else.

Programmer- In-built programmer is working. See above USB description. Just needs board powered and the standard supplied USB lead wih some software.

Bitstreams - We have a twin,or even triple, approach here. The first of these is these is building the standard VHDL/Verilog dersigns that are arround for our pinout and all of the units shipped have a 50MHz version that is working and can earn Bitcoins. With this approach we can run Icarus style pairs directly wired or "patched" through the controller. We have a range of options on I/O that we can use. Building these bitstreams takes typically 1-2 days to attempt a build and they don't always complete so it is a slow process. We are trying a few things here to stabilise the standard design which is highly sensitive to build variation. This is our initial way of providing a bitstream.

Seperately we are looking at better, original design, bitstreams. This will deliver a much better solution both in stability and probably performance. It's going to take a few weeks and maybe a bit more to get these solutions to a useful level but these will be our long term offering.

When requested we are also supporting any 3rd party bitsream development.

Documentation- This with be very patchy for the first couple of weeks but it will start to appear soon. As we are evolving bitstreams and software it may also not be totally up to date for the first few weeks, or even complete, but we will do our best to make it useful and keep it on track.

Software- We have some initial CGminer extensions that hopefully will be available today. As we evolve controller and array FPGA designs there will be updates and variations so it will be a changing item over the next few weeks.

Stacking Kits- We are starting to get elements of these together but more will appear over the next few weeks. Basic stacking pillars are available now and up/down ribbon cables hopefully from today. Power linkages we are looking at to see what works best and that is tied a bit to development of a power distribution board that will allow a high effiency ATX power supply o/p to be used in a cost effective way to power a stack of boards.

Power Supplies- Because of the weight and hence shipping cost we decided not offer as standard a power supply. Given the wide range of ways that a Cairnsmore1 can be powered we think it's better for customers to source locally. We hope to run some tests on specific units an make a recomended list and some of this will be linked to what we do as power distribution boards. Some way down the line we might offer a turnkey rack that incorporates a power supply but that is more likely to be Cairnsmore2 than Cairnsmore1.

Production- We are moving into a stable manufacturing phase now. The hardware side of the product is locked down and the in-built programmer covers firmware updates in the field. The general target is to have the major part of the "development" over by the end of June. It's not an exact science predicting development cycles so that might vary a little. That's also our general target to achieve performance levels of about 200MH/s per array FPGA. More advanced things will come beyond that.

In terms of unit delivery times we will try and pull forward delivery dates if possible but we won't promise that. I do expect that to start improving for new orders in the next few weeks as we catch up on the initial demand and eventually units will become a stocked item for next day delivery into Europe / US and maybe 3-5 working days for those unlucky Australians on the other end of planet.

Ok so hopefully I have covered most questions in that lot and at the point of sounding boring I will say it's still a development product. Hopefully by now that doesn't surprise anyone. It's still very rough and ready as a product but that will improve over the next few weeks and longer.

Just to clarify we are currently working with an Icarus style design that we have rebuilt to accomodate the pinout mistake on the downstream port. We are currently playing with several versions to try and get performance improved. We are migrating to a version of this where the link actually goes through the controller allowing many different variations of what we might do.

We are looking at other possibilities as well,

Orders will flow out every day unless we have shortage like the expected one for a couple days this week on heatsinks. A very large batch of batch of heatsinks should arrive Friday, fingers are crossed, and then we are back up and running pumping out boards.

We are running real test mining now on a low performance level. There will be some related relevant stuff going on the webpage probably today. It's going to take anything from a few days to a few weeks to get the bitstream up a decent performance level. I am sure there will be some bug clearing to do as well. Remember this is still very much in the development phase and we are now only on 45th calendar day since we started.

Now that the hardware manufacturing side is moving into a more steady state level we will have more tme to look at bitstreams and related items and we will start to fill in the gaps.

Yohan

Ok not a total answer but we have sorted out pricing for a couple of the stacking items:

Item1 - Pack of 4 x 75mm stacking pillars - GBP £4 / USD $6.40 / 5.20€ + vat (if applies) - order code CAIRNSMORE1-STACK-75MM
Item2 - Up/Down ribbon cable - GBP £4 / USD $6.40 / 5.20€ + vat (if applies) - CAIRNSMORE1-UPDOWN

Item1 is already in stock. Item2 should be available in a couple of days.

Yohan

At the moment we are roughly on schedule. Maybe a few days behind on the ideal. We have set our shipping schedule at something that we think is realistic. There will always be production issues that cause glitches and we have set the schedule to hopefully account for those. There is some possibility that we might do better that we have promised I won't ever guarantee that. It's our margin for when something unexpected happens as it usually does.

The Icarus thing was just to give us a fast path to working tools and bitstreams. Unfortunately we made a mistake in the downstream pin numbers so the bitstream part didn't quite work. That's one of the downsides of doing the design in a very fast manner as we have. We could have done a respin of the PCB but that would have cost 2-3 weeks on the schedule and related costs so our engineering solution is to fix it at the bitstream level.

It's important to stress that the Icarus thing was only intended to be a temporary thing and the board can working in number of ways that is not Icarus. There are direct I/O from each FPGA to the controller and a lot of different logical structures can be implemented. That's where Cairnsmore1 gets interesting particularly when paired up with the up/down interface and we can build arrays that operate as one device. Some of this is ideas testing for what we might do in Cairnsmore2 and Cairnsmore3 and we will develop these ideas as we go along.

Equally we could have looked at the Ztex way. The pinout of Ztex was not checked but there is a reasonable chance we can use that with a rebuild and other work as well. That's probably not worth doing as it duplicates up other things we are working on. The main thing about Cairnsmore1 is that I think we have got a design that is capable of competing with any other current offering and maybe also doing a lot better.

I think everything that didn't meet courier cutoff time today go keyed into Monday's courier pickup. Apologies to those who's unit didn't escape our cluches today.

The inital bitstream is just just an Icarus build just rebuilt to our pinout.

NET "osc_clk" LOC = "J1";
NET "RxD" LOC = "D1";
NET "TxD" LOC = "B1";
NET "extminer_txd<0>" LOC = "B21";
NET "extminer_rxd<0>" LOC = "B22";
NET "led[0]" LOC = "A17";
NET "led[1]" LOC = "B18";
NET "led[2]" LOC = "A18";
NET "led[3]" LOC = "A16";
NET "dip[0]" LOC = "A4";
NET "dip[1]" LOC = "D6";
NET "dip[2]" LOC = "C6";
NET "dip[3]" LOC = "C8";

The current controller build can supply 50MHz, 100MHz, 150MHz or 200Mhz depending on dip switch setting. Eventually will have something a little more advanced but that is it today. The in-built programmer features can update the controller as well as the main array. That gives us an on-going path to make improvements and for customers to apply them to their units.

The problem we keep having with Icarus builds is the intermittant working of the coms interface and that's assuming a build completes which often it doesn't. This is principally down to the way the coms is designed and can be improved and we might do that at some point depending on how our other approaches work out.

Yoahn after this initial shipment of a single board to 50 people when is the next shipment due? Also do you have performance numbers? thanks.
[/quote]

Todays main significance was it was the first main ship that we have done and we expect to run a reasonable percentage of that every working day , except for August, from now until the end of September and maybe beyond. As such there is no batch shipment as such we just turned on the tap and we now leave it running for as long as is needed. There may be some minor hiccups due to component shortages and eventually it will tail off for whatever reason. Todays shipment was a little unusual in the sheer number of people getting boards and we didn't get all them out. Simple logistics got there. It takes about 10 minutes to enter an address and related data for a courier and that is a limiter on the number of shipments per day. However after the next couple of days of shipping we move into the shipping phase where are shipping several units together and the number of shipments falls.

On the units that shipped today they all have the 50MHz bitstream and the in-built programmer is working. We will be working on improving that over the weekend and through next week but as the build times are lengthy and often fail it may be a few days before we get the next jump in performance.

All customer boards will be Issue 1.1 or later.

I will try and deliver some more hardware porn later today after we get todays shipment out. It's going to be a bit busy today.

Sorry about that. If I found the correct entry you should a response now. It was entered on the pre-order but the response email somehow didn't get out.

Our final early ship offer has gone out this morning. Anyone that had a June indication should have had one of these. If you think you should have had one check your spam box and if it's not there do contact us. From the end of next week we will be ship order balances in the order we received them. If you placed more than 1 order those orders will be available roughly on the schedule that was indicated to you. I think we are now on day 42 since we announced the concept and started work(do correct if I can't count, it does seem a lot longer) and I am pleased with the delivery of the design from our team.

We have taken the product a long way this week and we think that we are a few days away from showing at least a major percentage of it's potential. Building bitstreams is so much fun (not) but we will have more design team staff, and machines, available from next week with the Cairnsmore1 shipping to make that happen. We might also have a few surprises yet for our competitors but I would want to do an entire BFL so I will leave you to salivate over what that might be.

Oh heck I'm going to do a BFL and not tell you about it but it's entirely the answer to the meaning of life and boundless prosperity and we are probably never going to deliver it.

On a more serious note we have been doing some load tests on the board and the heatsink and fan combination, in the standard setup, is working extremely well. Some of the design features that we incorporated to deliver heat to the rear of the board are also working well. This will actually be of significant benefit in stacking arrangements either with fans in conventional fashion or push/pull side blow that will take air over the rear side. We also think a small rear heatsink could be of benefit there and we will actually try that out in the next few weeks. Rear heat extraction maybe also be a good way with water coolers but more on that after we do some more experiments. Don't expect this data for at least a few weeks.

Production wise we are moving out of the fairly hap hazard thing that is initial production week. We are still improving the processes and even the test equipment and that will be an ongoing process for a few weeks yet. Aligned with that should an increase in numbers of units delivered every week.

The way we test these is a test harness that is derived from CGminer but runs in isolated mode. So it sends in commands in the same and we check with the golden nonce that it all works.

Update

Ok today saw the first bitstream in operation working correctly in all FPGA positions. We found a magic combination of coms stability and tool settings to allow a Icarus style bitstream to build. Now don't to excited as to achieve the coms stability we made the serial interface and hash core all run on a common clock of 50MHz so that's way off on the performance. However we are migrating this now to 100Mhz operation as the next stage and crossed fingers that will work. After that we will improve the coms modules and fix the problem properly so we can go faster again.

We needed a small tweek for stability on the in-built programmer but that tweek has worked and we are now checking that is ok on a large sample of boards.If that sample test proves positive we will move to full shipping release tomorrow.

Depends where in the queue you are what your ETA is. At the moment we are not far off our indicated shipping dates.

On bitstreams we probably won't release our advanced versions bitstream source code. There will be far too much work simply to hand this away to our competitors.

We might get a bitstream into the board this week based on a re-build of standard stuff but for higher performance one of our own is probably about 4 weeks away. That's also dependent on what else we have to do in the next few weeks on other projects. There may be third party solutions available before our optimised solution is available. So don't expect spectacular bitstream results from us this week. You are going to have to be a bit patient on this. Our target this week is ensuring that the in-built programmer is totally rock solid, getting the production line more optimised, actually shipping boards and trying to get a standard bitstream working to our satisfaction. That's all on top of a large number of UK people being on holiday this week.

Although we have not had a massive amount of time to work with standard bitstreams but we have had some progress and can get a test nonce to run now using our CGminer based test harness. We still have some coms issues to resolve and that is the main area of work this week at the bitstream level.

We have been looking at high power ATX PSUs and have a few coming in this week to sample and try out. We will use them on our burn in rig. There are several that offer 8 PCIe connections plut Disk Drive connections and we are looking at how best to extract the power and deliver to the boards. One way might be a combined switch on and distribution board and we are looking at whether that is worth doing for our own test purposes or even more generally.

Yohan

Enterpoint. like all the hardware producers, has a ready made charging model in the shape of the boards and that is initial intention that's a straight forward sale. Obviously if the bitstream gets ripped off and used elsewhere we might have to look at that again. It's different for firmware and software producers that don't have the hardware dongle.

We might. or not, end up with the fastest bitstream done by ourselves but the key to anything being scccessful is a good hardware platform and I think we have that in Cairnsmore1. It's up to customers which bitstream provider they want to use and how much they might pay for that. We don't have any problems with this model.

Regards
John