Ok, thanks... i guess heat will not spread too high only from overclocking without highervolting so i think i would try overclocking too. Normally i overclock and downvolt my gpus to lengthen their lifetime while rising speed... maybe thats possible for the miner too..

Regarding the scope - it is for sure important tool to have, but you might want to consider buying a logic analyzer, which is way more useful when troubleshooting digital communication. 125 ns in the Avalon protocol translates to 8 MHz, there are 100 MHz (*) logic analyzers available for about $300 (or less expensive clones on eBay), with virtually unlimited sample memory, because they use the PC (over USB).

(*) Sample 2 channels at 100 MHz, 4 channels at 50 MHz etc.


And question about the NOR gate used for the clock signal: does it have Schmitt trigger inputs?

I have a cheap Logic Analyser already. Good for 8x 200MHz or 16x 100MHz.
The scope has a much deeper memory. But also it will be useful for other than logic capture which was an area I was missing.

No schmitt trigger on NOR gate. But there is on the clock buffers, not that it needs it.

---
Regarding under volting - I believe the Avalon info posted was that 6.6W/GH was already slightly under volted at 1.15V so it will be interesting to see how this goes during testing. I thought briefly about adding an I2C digital resistor to the buck reg. output control but decided playing with that idea would have to wait til later. I don't know if the feedback circuit is too sensitive for that type of thing. Right now to adjust voltage you would need to change a resistor. For testing purposes I could try a small trimmer and see if there is an optimal value. One thing at a time though.

Yes. I try to delay the extracted clock output long enough for the setup time on the data input. It may need some trial and error there - and the scope could be useful to see what it's doing.

Until you sell so many chips at once that the network hashrate increases by over 200TH/sec from your chips alone

Which is what they are doing

Because that's what you do.

During a goldrush, you sell shovels.

It's also orders of magnitude easier, less work, less risk, and potentially profitable, to sell chips to people who then make the miners or whatever.

It's like saying "Why does intel make CPUs when they could make computers"

I have a few Chips that can be used as sacrificial lambs for the BKKCoins torture chamber... Have at it when I send you some BKKcoins... be really nice to push the limits and see what you can get from passive and forced air cooling. Scotty give us more power! Any advantage we can get now will extend the useful life of these boards.

That should be right. Since any hash test is as good as another the fastest work change over should be "the time needed to push new data" and that would only occur if you truncated current work as close to the end as possible (since that gives you the least changes over long term).

I'll definitely have this in first version but I'm probably going to change how I do it. It won't be AW as that doesn't follow the protocol really. It would be just A with new work so it acts like W with queue clearing. To stop work without new work you would just use E Enable to disable work.

Another change: I'm going to use a single byte binary address instead of 3 digits. I thought maybe it would ease debugging but letter codes with binary after will be just as easy, and why increase data xfer for that. The <address><workid> pair then make a 16 bit unique identifier for each work sent out. Which is easier both in the PIC and in the driver to track.

------ other news ------
I'm looking through the bflsc.c code now and have started preparing a cmd line test utility based on usbutils to allow me to poke the PIC and test outside of cgminer.

I also submitted a request to Microchip for one of their free USB PIDs (product ids) as the alternative is randomly choosing some VID/PID to hijack.

With the limit of 32 watt power supplies over clocking may max out around 300MHS.
But if you only stuffed 14 ASIC chips you should be able over clock higher without overloading the 1.2 volt supply.
It will be fun to experiment once we get some hardware and software to play with.

It might not be that hard to switch out the regulator for a higher current one as well.  Not sure what the package is and if there is another one that fits though.

Perhaps if you have space, add the component pads for a third regulator, but don't populate it by default.  Can easily add one another one then, which would also split and reduce the load on each individual regulator.