Topic: Some project 28nm? (Read 2085 times)

No word on that.

I agree that Artix-200 will be the FPGA of choice for mining, and the $64,000 question is whether anyone (hint, hint) can fit all 125 stages of a miner into the Artix-200's 156 columns, from left to right (or from right to left), without the dreaded U-turn.
Or, to be more precise, whether anyone (hint, hint) can get all 2 x 125 stages of TWO miners into it.
The two "parks" (Central Park and Golden Gate Park) for the ADC and the PCIe controller in the left half of Artix-200 certainly don't help with such an endeavor...


Yes, there is! There was a talk just on designing power supplies for -7 chips (Artix, Kintex, Virtex) and this point was stressed. However, most DC-DC controllers have an enable pin, thus enforcing a well-defined power-up sequence is actually quite easy.

Not so easy for power-down, but while the power-down sequence is NOMINALLY the reverse power-up sequence, it isn't critical in practice (because the "wrong" condition will only last for a short period of time, not long enough to cause damage).

There is a datasheet note that states: "On power-down, voltage w may only be higher than voltage x by y volts for z milliseconds, otherwise the long-term reliability of the chip will be affected", but the lecturer said, this is almost implicitly guaranteed in most designs and should not be of concern to most people.


I sure hope not!

But I agree with you: While Intel is now shipping 22 nm CPUs in volume, TSMC still hasn't solved the problems with the 28 nm process. Sigh.

Any word on whether or not the 200 chip will be among the first ones available?  I'm hoping the availability of that one will prevent people from using the smaller Artices, sort of like how nobody uses Spartan6-LX75's.  The Artix-100 is really funny-shaped.  Somebody beat that chip with the ugly stick.


My experience has been that sequencing doesn't matter if you program via JTAG -- just don't start the JTAG sequence until all supplies are stable.  Is there reason to believe Artix-7 will be any different?


They're probably having to adjust the prices upwards due to the 28nm yield trainwreck at TSMC.

I just snuck out of the X-Fest (Avnet/Xilinx seminar) to do some work in my nearby office, which doubles as my mining office.

The highlights of X-Fest, as far as they are relevant for FPGA mining, are:
- Artix will arrive by the end of this year  -  let's hope this is not a BFL-like time estimation 
- like all Xilinx-7 devices, Artix requires power supply sequencing and needs at least one more voltage compared to Spartan-6
- not even ballpark pricing for Artix was available

- this year, there are no discounts on eval boards or the ISE software license 

- but there is a low-cost Kintex-7 eval board: The so-called Mini-Module Plus with a Kintex XC7K325T-1FFG676 for $895
- it includes a full version of ISE, which is, however, locked to Kintex XC7K325T devices
- but it doesn't come with a power supply, which is $300 extra (several different vendors to choose from: TI, Maxim, GE, ST, AD)
- all of these power supplies only provide 6A for VCCINT, which may or may not be enough for mining
- and it doesn't include the base board, which is $500 (but not strictly required, if you wire the Mini-Module to the P/S yourself)

Will I buy the Kintex Mini-Module? I'm not sure. An Artix version of the mini module will arrive "at the end of the year" and that one will be much more interesting for mining / Bitstream development.

We have a Kintex design and silicon already and this silicon is unlikely to meet Bitcoin expectations pricing wise. We also have Arrria-V silicon and some parts there might be viable for Bitcoin but it needs a proper analysis to be sure. One Cyclone-V (150LE part) is supposed now to be available in small numbers and we might do something with that.

Tomorrow, I will attend X-Fest, which is a complimentary all-day Xilinx seminar sponsored by Avnet, and if they have the KC705  Kintex-7 eval kit on sale at a significant discount, I'll buy it.
But this board won't make a cost-efficient miner and I neither have the time nor the gumption to try to duplicate EldenTyrell's work.
So, yeah, I'll compile a miner for that board all right, but it'll be far from optimal and will not earn the board's purchase price back.

First let's see a few of the 28nm arch chips delivered, at a reasonable price, and then we can discuss making something with them.

I can't not in my possibility 
Exclusively only now a few project at 28nm chip

do u want to start one?

Project with kintex 7 or Stratix 5?