We’re pleased to bring you a new update.
Thank you for your patience! We know it has taken a while for new details to come out, and offer our most sincere apologies for any inconvenience. We’re as eager as you to bring news when it is ready for release, and above all deliver your device.
The power consumption of the Viper device was estimated to be 10W for a core voltage of 1.2V and core frequency of 200MHz. This estimate was based on simulation results that kept a power down feature during idle cycles for the internal SRAM in consideration. However since mining is quite the continuous activity, the idle cycles could not contribute enough for lower power needs. As such, power consumption went to approximately 18W at full speed and frequency.
As per the initial calculations, the current requirement for the core voltage was less than 9A per Viper device ( 10W / 1.2V = 8.33A ). The initial prototype was designed with a 50A power module from a leading semiconductor manufacturer for 4 Viper devices (4 x 9A = < 50A), keeping room for some over-clocking possibilities. Since the power consumption changed to 18W, the current required also changed to 15A per device which would require at least 60A for 4 Viper devices. We also noticed some voltage dips on the core voltage rail of around 200mV during the start of computation of one Viper device, this could affect the working of another Viper device connected to the same voltage rail. To cater to such requirements, single modules were pretty expensive. In order to save cost and do some more optimizations, another new design was created. It uses 15A power modules from another leading semiconductor manufacturer for each Viper device. However these devices have been behaving a little erratic and we have found them unstable during our QC tests.
The design is currently getting reviewed and failure analysis is being done by the experts at the power supply company. We are waiting for their response so we can get the correction done. A parallel design is also being done with another power supply having 20A capacity. We have tested the software, and the functionality is working properly which really pleases us. After this last issue gets sorted out, the bulk production can start and shipment will follow.
A few sample Viper devices were tested to identify issues with PLL, IOs, memory etc. with the test vectors generated. Devices were tested and categorized into those that pass and fail. When the same devices were tested with the ZIF socket board, the results observed with these devices were different. Devices which were marked as pass did not pass the functional test completely when tested with the ZIF socket. A few devices passed the functional test with the ZIF socket board and a few failed. Test vectors were updated to include BIST (built-in self-test) to identify device which pass and fail functionally.
On further analysis of the results observed with the tester, the ZIF socket and the test vector simulation results, it was found that the tester has a higher voltage drop and configuration of Viper PLL via JTAG with the tester not working properly. Test vectors were updated to configure PLL over UART and shipped back to test house along with the devices for a re-test. By using the new test vector and increasing the core voltage with the tester, test results of the ZIF socket board and the tester started matching properly. In short, this issue is now under control and it is being worked on as fast as we can.
We are sorry shipment has not yet occurred. We very much appreciate the situation and are working our hardest to deliver the best product possible with the resources we have available. We would like to reiterate our previous statement that all we can do is reward your patience with a product that is ready and optimized instead of rushing and then deliver potentiality faulty and under-performing hardware. We know you are eagerly waiting to get your hands on these devices, we share this enthusiasm and are quite impatient ourselves to start shipping, and hope you can understand we’re not here to take the easy way out: but deliver a high-quality well-thought of and well tested device. We are still providing the various enhancements to the final specifications of the devices including a 10 times increase of the hash rate. We plan to e-mail all customers as soon as we have the shipping date, but we cannot estimate one at this time. All we can say is that once we have these last issues resolved and assembly has started, shipment will follow as soon as possible afterwards. We already have our chip quantity in stock, so there will be no need for us to wait for chips. That stage has now been completed.
On that note, we’d like to express our gratitude for your immense patience and confidence, and extend our sincerest apologies for the annoying delays whilst this machine is being engineered to perfection. We know it seems that every month there is delay after delay, but please understand that development is outsourced and we are genuinely passing on information we are being told by the technical teams working on the project as it comes in.
We do our best to speed the process up and finalize the devices.
Stay tuned! We hope, and are fairly confident, that our next message will be that mass-production has started and be able to inform you that shipping will be underway.