Topic: GekkoScience is finally building a 16nm Pod Miner - page 4. (Read 1314 times)
Right now it's impossible for anyone anywhere to buy this. The first batch I would like to keep US-only as much as possible, because we'll still be working out troubleshooting and tech support procedures and if something goes wrong I'd rather it went wrong with someone who spoke English and lived within ready mailing distance.
@Valnurat
@minefarmbuy
Bitshopper.de in Germany has done reselling for Gekkoscience hardware for a quite long time.
@minefarmbuy
Bitshopper.de in Germany has done reselling for Gekkoscience hardware for a quite long time.
Impossible? I though there were vendors in Germany. Otherwise, I'm sure there are plenty who can export to you.
Why is it possible to by when you live in Europe?
A comment on USB.
As implied by that these are built on NewPac tech, this board features Bitmain BM1387 (S9) ASICs. Bitmain chips don't have much of a work buffer, and when initialized each chip in the data chain is addressed in such a way that the nonce range is divided up somewhat evenly amongst them. If I understand correctly, the address basically becomes the upper-byte starting point for each chip to parse the 32-bit nonce range. Once that range has been exhausted, in order to prevent wasted work, a fresh pile of data needs to be sent in pretty quickly.
USB2.0 can transfer enormous amounts of data rather quickly using bulk transfer packets. However, the number of packets of any size transferred per second is limited. Because Bitmain chips can't pre-buffer work, and because each chip gets a portion of the load (as opposed to say Innosilicon chips, which can buffer two unique work units per ASIC), fresh work has to be pushed in in fairly small increments rather quickly. This uses up a lot of really small packets, and maxes out the limits of the bus at around 800GH. I think this is the maximum hashrate one could achieve even with multiple devices on the same hub.
When ASICBoost is enabled, the amount of work data delivered to the chips increases slightly but because each work unit is fully parsed with four unique midstates, the practical effect is work units only need to be updated one fourth as often. Because of timing constraints and increased packet sizes, VH's implementation updates work about 1/3.2 as often, so the maximum hashrate from a single USB connection is roughly 2.6GH which means two or three pods should be able to function at high speed off one hub like this.
Limitations like that are probably a primary reason why Bitmain shifted to a fancier controller board with an FPGA buffering work and doling it out on multiplexed serial lines. When a 32-bit nonce range is divided amongst 60-70 chips it gets burned through pretty quickly.
For future higher-hashrate projects we'll either have to look into board-level work buffering to ease the USB load (and require some pretty fancy firmware) or work with different chips with a more bus-efficient work buffering mechanism (like almost all of them besides Bitmain).
As implied by that these are built on NewPac tech, this board features Bitmain BM1387 (S9) ASICs. Bitmain chips don't have much of a work buffer, and when initialized each chip in the data chain is addressed in such a way that the nonce range is divided up somewhat evenly amongst them. If I understand correctly, the address basically becomes the upper-byte starting point for each chip to parse the 32-bit nonce range. Once that range has been exhausted, in order to prevent wasted work, a fresh pile of data needs to be sent in pretty quickly.
USB2.0 can transfer enormous amounts of data rather quickly using bulk transfer packets. However, the number of packets of any size transferred per second is limited. Because Bitmain chips can't pre-buffer work, and because each chip gets a portion of the load (as opposed to say Innosilicon chips, which can buffer two unique work units per ASIC), fresh work has to be pushed in in fairly small increments rather quickly. This uses up a lot of really small packets, and maxes out the limits of the bus at around 800GH. I think this is the maximum hashrate one could achieve even with multiple devices on the same hub.
When ASICBoost is enabled, the amount of work data delivered to the chips increases slightly but because each work unit is fully parsed with four unique midstates, the practical effect is work units only need to be updated one fourth as often. Because of timing constraints and increased packet sizes, VH's implementation updates work about 1/3.2 as often, so the maximum hashrate from a single USB connection is roughly 2.6GH which means two or three pods should be able to function at high speed off one hub like this.
Limitations like that are probably a primary reason why Bitmain shifted to a fancier controller board with an FPGA buffering work and doling it out on multiplexed serial lines. When a 32-bit nonce range is divided amongst 60-70 chips it gets burned through pretty quickly.
For future higher-hashrate projects we'll either have to look into board-level work buffering to ease the USB load (and require some pretty fancy firmware) or work with different chips with a more bus-efficient work buffering mechanism (like almost all of them besides Bitmain).
Those Tripplite industrial hubs are very nice. We use them on our systems as they are built like a tank, near bullet-proof, have tighter than normal plug retention plus can be fed off of our 24vdc system buss.
Sorry my session timed out before I could edit. So any USB A hub can connect a few of these then?
I'm an edit fiend today.
Thoughts on https://www.tripplite.com/usb-hub-10-port-hi-speed-2.0~U223010
https://www.tripplite.com/rugged-industrial-usb-2.0-high-speed-hub-15kv-esd-immunity-metal-case-mountable-4-port~U223004IND
I'm an edit fiend today.
Thoughts on https://www.tripplite.com/usb-hub-10-port-hi-speed-2.0~U223010
https://www.tripplite.com/rugged-industrial-usb-2.0-high-speed-hub-15kv-esd-immunity-metal-case-mountable-4-port~U223004IND
You'll have to be more specific, because I have no idea in what case a USB splitter would be useful for this device.
Ok, question edited for more clarity. Using my hub would be kind of a waste, because this guy draws almost exactly zero power from USB, but yes given the limitations of USB I believe you should be able to run at least 3 up to absolute maximum speed from a single source. (assuming ASICBoost is enabled, which is required to reach top speeds anyway)
Ok, question edited for more clarity. Using my hub would be kind of a waste, because this guy draws almost exactly zero power from USB, but yes given the limitations of USB I believe you should be able to run at least 3 up to absolute maximum speed from a single source. (assuming ASICBoost is enabled, which is required to reach top speeds anyway)
It's beautiful. Do you recommend your USB hub for multiples?
The power light is orange. My primary PCB supplier doesn't have an orange soldermask option and arranging it would have been super expensive while also going mostly unseen.
The 606 may not be made with Bitfury chips, but I'm still planning on orange.
You are exactly right about the reason. We are GekkoScience, after all.
You are exactly right about the reason. We are GekkoScience, after all.
I don't see any orange!
I want one.
the all-new Terminus R606, building on the work of both the R808 and NewPac USB
Some more details please on the new miner?
the all-new Terminus R606, building on the work of both the R808 and NewPac USB
Price is TBD. I'll probably be doing more coordination with resellers regarding an MSRP and the like but those negotiations largely haven't happened yet.
As far as I know, no KYC (or fried chicken) is required for buying from anyone I distribute to.
I may open limited direct sales on these, depending on who I know that needs a job, but won't make any promises. For now the safe bet is to talk to your preferred reseller.
As far as I know, no KYC (or fried chicken) is required for buying from anyone I distribute to.
I may open limited direct sales on these, depending on who I know that needs a job, but won't make any promises. For now the safe bet is to talk to your preferred reseller.
Who's chips are being utilized in this shiny new miner?
Price is TBD. I'll probably be doing more coordination with resellers regarding an MSRP and the like but those negotiations largely haven't happened yet.
As far as I know, no KYC (or fried chicken) is required for buying from anyone I distribute to.
I may open limited direct sales on these, depending on who I know that needs a job, but won't make any promises. For now the safe bet is to talk to your preferred reseller.
As far as I know, no KYC (or fried chicken) is required for buying from anyone I distribute to.
I may open limited direct sales on these, depending on who I know that needs a job, but won't make any promises. For now the safe bet is to talk to your preferred reseller.
Such low power consumption and high hash rate how about the future price 
Is it also required KFC verified when ordering and shipping this machine?
Happy April fools day!
Is it also required KFC verified when ordering and shipping this machine?
Happy April fools day!
This will look good next to my ltc Apollo’s.
Coins are way up today this should be nice to add for summer mining in the garage.
Coins are way up today this should be nice to add for summer mining in the garage.
Job well done ! I do appreciate people taking effort to bring back mining at home.
It's looking good! Who's chips are in it? Bitfury's?
Jump to: