BW LK-1402 Compac Prototype x 3 (minimum bid 0.08BTC)
I'll start the bidding for a second working stick at 0.15BTC
Topic: GekkoScience miscellaneous auction for 16nm miner dev funding - page 3. (Read 3995 times)
Quote
BW LK-1402 Compac Prototype x 3 (minimum bid 0.08BTC)
I assume the price is for 1 of the 3 available right?I'll start the bidding for 1 of them @ 0.1 BTC
So I have a crew working on development for 16nm miners intended specifically for consumers. This includes a single-chip stickminer, a 10-chip pod miner and ~400W S1-chassis-compatible boards. But dev costs money and to help raise money I'm auctioning off some of the test setup stuff from previous miner design work.
This includes three prototypes of a BW LK1402 Compac for which drivers are currently being written. The chip manufacturer has opted not to sell ASICs so no more will be made, but the software was already well under way when this was found out. The original plan was to pay the coder a percentage of sales, but if there's no production there can't be sales. The proceeds from the auction of the three LK1402 Compac prototypes will help make up for that; he plans to finish out the driver which means these sticks should be functional miners.
Here's a list of what I've got up for grabs, in what quantities.
A3218 Compac - Unpopulated PCB x 12
These are some test PCBs for Compac stickminers designed around Avalon's A3218 ASIC. Unfortunately there wasn't enough available information to really get things to communicate properly. One complete prototype was assembled and sent to the programmer but it never went anywhere.
BM1385 Compac, fully assembled x 1
This Compac is built around a Bitmain BM1385 ASIC harvested from an S7 miner. I could never get the chip to talk, and since it's a 0.35mm pitch QFN and the chips suck to harvest anyway I wasn't too upset about it. I only built the one. It did successfully test a new power layout, so that's handy.
BM1385 Compac - Unpopulated PCB x 4
The above-described miner, just bare PCBs.
Populated V1.1 Amita x 4
The "Amita" is the pod-miner product line I want to actually exist someday. I did go through multiple BM1384 prototypes with varying levels of success, but since the BM1384 is painfully outdated I gave up on it. The platform mostly became a testbench for power, stability and manufacturing precision. These boards are built around 8x BM1384 ASICs in a bucked string topology and are, for the most part, populated with all node-level components and a functional power circuit. They probably don't work for one reason or another. Some were rewired to scab onto an S5 controller.
Unpopulated V1.1 Amita x 3
The Amita is intended to have a USB connection and built-in microcontroller for fan speed and voltage adjustment. However, we didn't get around to writing the code for that part. All the footprints are there on this initial version, except I screwed up one of the jacks.
Unpopulated V1.3 Amita x 8
The original (V1.1) test pod was designed for USB but that part was never implemented. On this second version I was more focused on just having a closed solution that worked, so I redid the power layout and removed all microcontroller stuff in exchange for an 18-pin jack, simulating an 8-chip S5 board the S5 controller would work with transparently. Ideally it would have been mass-producible but I still had some problems with stable voltage levels at the ASICs. Couple that with the BM1384 being a year old at that point and it wasn't really feasible.
Assembled V1.1 Amita x 1 (starting bid 0.1BTC)
This is an initial-test-version Amita which has been fully assembled with a proper cooler. I completed two of these but am keeping the first for my own archive. This one was wired up a little more stably to function off an S5 controller. The original concept was to make it compatible with Intel LGA1155 CPU coolers and with Freezer7 coolers that a lot of people might have dozens of laying around after the Technobit HEX4M debacle. This board is assembled with one of those.
Assembled V1.3 Amita x 1 (starting bid 0.1BTC)
I'll be giving up on CPU cooler compatibility on future Amita projects because it wastes a lot of board space trying to fit arbitrary standards, but the BM1384 Amita in both prototyped versions still used that concept. The ASICs are compressed into a 3x3cm square to fit under a Freezer7 cooler base or inside the center billot of a standard Intel cooler. This V1.3 Amita is fully assembled (including 18-pin jack and S5-compatible temperature sensor) but I don't recall offhand how functional it actually was.
Hacked S5 8-chip test board x 1
This was my original bucked string topology test. It was a concept Novak and I came up with right around the same time ASICMiner (remember those guys?) announced the BE300 test specs, which was really what got us interested in miner design. We spent a full two days discussing the features and attributes we'd like to see in a miner. We wanted to combine voltage adjustability with the reduced parts count of string topology, so the obvious result is a high-voltage buck converter on a short string (like the S7 and S9 now use in a fixed-voltage form). My initial test of this used a hacked TPS53355 circuit rewired for variable 2.4-3.2V (instead of fixed 0.8V) powering a rewired section from an S5 board, which worked pretty well once some power kinks were ironed out. This became the basis for the Amita prototype boards.
Unpopulated 18-chip string test board x 2
The eventual goal of my miner design projects is to bring to market an S1-chassis-compatible board kit using modern chips. Thanks to Bitmain's early (and subsequently abandoned) leveraging of interchangeable parts, a board that'll fit on an S1 will also work on an S3 and an S5. These chassis have been tested to dissipate upwards of 500W without a lot of the reliability concerns which have plagued their later high-density designs, and they can be found for pretty cheap on the secondhand market. Our initial idea was to make each S1 a four-board kit (two boards per heatsink) where each board could operate as a ~100W standalone, not unlike what RockMiners did with several miner chassis designs based around the same PCB. This concept survives in a few half-S1 prototype PCBs designed for external power, basic UART signaling (intended for an external USB adapter) and 18 chips in a 3p6s string configuration.
Lemon BM1384 Compac x 12 (minimum bid 0.16BTC)
I'd like to sell these as a single unit if possible. I've gone through something like three thousand of these stickminers from start to finish - all our manufacturing is done in-house - and I've had a few that I just got tired of fiddling with. Some might have PCB issues, some might just be boneheaded oversights on my part, but most are pretty much fully populated (though none have heatsinks, since those aren't installed until they pass initial testing). I don't know how many someone industrious could get working again.
Amita V1.1 stencil
This is a 30x40 stainles steel stencil for the Amita V1.1 stencil. Darn thing had enough pads (several hundred) that it was well worth buying a proper stencil instead of trying to grease by hand. Most of the parts were in the same place on the V1.3 so it was reusable.
BM1385 Compac stencil
This is also a 30x40 stainless steel stencil, for the BM1385 Compac prototype. I ended up reusing the power layout on a couple future versions so this one was reusable as well. The Compac has few enough parts that I didn't really need it that much though.
BW LK-1402 Compac Prototype x 3 (minimum bid 0.08BTC)
The software is still in the works on this - an initial cgminer version already exists, but we're having trouble getting the initial prototype to hash on all cores. Even with all the overhead, right now it's still more than twice the hashrate of a BlockErupter at less than half the power. When it's working properly it should top out around 50GH at maximum overclock. Hopefully we can make that happen.
BW LK-1402 Compac unpopulated PCB
I only had 4 sample ASICs and will keep the first prototype for my archives, which means at most three more functional sticks will be made unless BW decides to open up sales. That leaves me with a single unpopulated PCB someone can have.
Of course I always have regular power supply gear and custom cabling and other things available for sale over here (https://bitcointalksearch.org/topic/fs-gs-server-psus-boards-kits-stickminers-gpu-riser-power-made-in-usa-940317) and I really like selling that stuff because it's my actual income. But if this auction is a success it'll greatly help to fund dev efforts on future miner projects. Especially the BW sticks, which will pay the programmer for the work he did on 'em.
I've put minimum bids on things I'd like to get at least that much for, probably because of the materials and stuff in 'em. Shipping will have to be arranged between me and the buyer, shipped from MO, US 65401. Sales must be paid for within 24 hours of the auction ending, paid to address 1GekkosciLeaey8Na9siC8oD5HcMtLnWwd
The action will end Monday 10/24 at 5PM CST. Any bids submitted after that time won't count. Standard rules apply - quantity @ price in a post on this thread counts as a submission. If the highest bidder doesn't take 'em all the next-highest bidder can match price. Any items not paid for within 24 hours of the auction will be offered to the next highest bidder. Alright guys, be civil.
This includes three prototypes of a BW LK1402 Compac for which drivers are currently being written. The chip manufacturer has opted not to sell ASICs so no more will be made, but the software was already well under way when this was found out. The original plan was to pay the coder a percentage of sales, but if there's no production there can't be sales. The proceeds from the auction of the three LK1402 Compac prototypes will help make up for that; he plans to finish out the driver which means these sticks should be functional miners.
Here's a list of what I've got up for grabs, in what quantities.
A3218 Compac - Unpopulated PCB x 12
These are some test PCBs for Compac stickminers designed around Avalon's A3218 ASIC. Unfortunately there wasn't enough available information to really get things to communicate properly. One complete prototype was assembled and sent to the programmer but it never went anywhere.
BM1385 Compac, fully assembled x 1
This Compac is built around a Bitmain BM1385 ASIC harvested from an S7 miner. I could never get the chip to talk, and since it's a 0.35mm pitch QFN and the chips suck to harvest anyway I wasn't too upset about it. I only built the one. It did successfully test a new power layout, so that's handy.
BM1385 Compac - Unpopulated PCB x 4
The above-described miner, just bare PCBs.
Populated V1.1 Amita x 4
The "Amita" is the pod-miner product line I want to actually exist someday. I did go through multiple BM1384 prototypes with varying levels of success, but since the BM1384 is painfully outdated I gave up on it. The platform mostly became a testbench for power, stability and manufacturing precision. These boards are built around 8x BM1384 ASICs in a bucked string topology and are, for the most part, populated with all node-level components and a functional power circuit. They probably don't work for one reason or another. Some were rewired to scab onto an S5 controller.
Unpopulated V1.1 Amita x 3
The Amita is intended to have a USB connection and built-in microcontroller for fan speed and voltage adjustment. However, we didn't get around to writing the code for that part. All the footprints are there on this initial version, except I screwed up one of the jacks.
Unpopulated V1.3 Amita x 8
The original (V1.1) test pod was designed for USB but that part was never implemented. On this second version I was more focused on just having a closed solution that worked, so I redid the power layout and removed all microcontroller stuff in exchange for an 18-pin jack, simulating an 8-chip S5 board the S5 controller would work with transparently. Ideally it would have been mass-producible but I still had some problems with stable voltage levels at the ASICs. Couple that with the BM1384 being a year old at that point and it wasn't really feasible.
Assembled V1.1 Amita x 1 (starting bid 0.1BTC)
This is an initial-test-version Amita which has been fully assembled with a proper cooler. I completed two of these but am keeping the first for my own archive. This one was wired up a little more stably to function off an S5 controller. The original concept was to make it compatible with Intel LGA1155 CPU coolers and with Freezer7 coolers that a lot of people might have dozens of laying around after the Technobit HEX4M debacle. This board is assembled with one of those.
Assembled V1.3 Amita x 1 (starting bid 0.1BTC)
I'll be giving up on CPU cooler compatibility on future Amita projects because it wastes a lot of board space trying to fit arbitrary standards, but the BM1384 Amita in both prototyped versions still used that concept. The ASICs are compressed into a 3x3cm square to fit under a Freezer7 cooler base or inside the center billot of a standard Intel cooler. This V1.3 Amita is fully assembled (including 18-pin jack and S5-compatible temperature sensor) but I don't recall offhand how functional it actually was.
Hacked S5 8-chip test board x 1
This was my original bucked string topology test. It was a concept Novak and I came up with right around the same time ASICMiner (remember those guys?) announced the BE300 test specs, which was really what got us interested in miner design. We spent a full two days discussing the features and attributes we'd like to see in a miner. We wanted to combine voltage adjustability with the reduced parts count of string topology, so the obvious result is a high-voltage buck converter on a short string (like the S7 and S9 now use in a fixed-voltage form). My initial test of this used a hacked TPS53355 circuit rewired for variable 2.4-3.2V (instead of fixed 0.8V) powering a rewired section from an S5 board, which worked pretty well once some power kinks were ironed out. This became the basis for the Amita prototype boards.
Unpopulated 18-chip string test board x 2
The eventual goal of my miner design projects is to bring to market an S1-chassis-compatible board kit using modern chips. Thanks to Bitmain's early (and subsequently abandoned) leveraging of interchangeable parts, a board that'll fit on an S1 will also work on an S3 and an S5. These chassis have been tested to dissipate upwards of 500W without a lot of the reliability concerns which have plagued their later high-density designs, and they can be found for pretty cheap on the secondhand market. Our initial idea was to make each S1 a four-board kit (two boards per heatsink) where each board could operate as a ~100W standalone, not unlike what RockMiners did with several miner chassis designs based around the same PCB. This concept survives in a few half-S1 prototype PCBs designed for external power, basic UART signaling (intended for an external USB adapter) and 18 chips in a 3p6s string configuration.
Lemon BM1384 Compac x 12 (minimum bid 0.16BTC)
I'd like to sell these as a single unit if possible. I've gone through something like three thousand of these stickminers from start to finish - all our manufacturing is done in-house - and I've had a few that I just got tired of fiddling with. Some might have PCB issues, some might just be boneheaded oversights on my part, but most are pretty much fully populated (though none have heatsinks, since those aren't installed until they pass initial testing). I don't know how many someone industrious could get working again.
Amita V1.1 stencil
This is a 30x40 stainles steel stencil for the Amita V1.1 stencil. Darn thing had enough pads (several hundred) that it was well worth buying a proper stencil instead of trying to grease by hand. Most of the parts were in the same place on the V1.3 so it was reusable.
BM1385 Compac stencil
This is also a 30x40 stainless steel stencil, for the BM1385 Compac prototype. I ended up reusing the power layout on a couple future versions so this one was reusable as well. The Compac has few enough parts that I didn't really need it that much though.
BW LK-1402 Compac Prototype x 3 (minimum bid 0.08BTC)
The software is still in the works on this - an initial cgminer version already exists, but we're having trouble getting the initial prototype to hash on all cores. Even with all the overhead, right now it's still more than twice the hashrate of a BlockErupter at less than half the power. When it's working properly it should top out around 50GH at maximum overclock. Hopefully we can make that happen.
BW LK-1402 Compac unpopulated PCB
I only had 4 sample ASICs and will keep the first prototype for my archives, which means at most three more functional sticks will be made unless BW decides to open up sales. That leaves me with a single unpopulated PCB someone can have.
Of course I always have regular power supply gear and custom cabling and other things available for sale over here (https://bitcointalksearch.org/topic/fs-gs-server-psus-boards-kits-stickminers-gpu-riser-power-made-in-usa-940317) and I really like selling that stuff because it's my actual income. But if this auction is a success it'll greatly help to fund dev efforts on future miner projects. Especially the BW sticks, which will pay the programmer for the work he did on 'em.
I've put minimum bids on things I'd like to get at least that much for, probably because of the materials and stuff in 'em. Shipping will have to be arranged between me and the buyer, shipped from MO, US 65401. Sales must be paid for within 24 hours of the auction ending, paid to address 1GekkosciLeaey8Na9siC8oD5HcMtLnWwd
The action will end Monday 10/24 at 5PM CST. Any bids submitted after that time won't count. Standard rules apply - quantity @ price in a post on this thread counts as a submission. If the highest bidder doesn't take 'em all the next-highest bidder can match price. Any items not paid for within 24 hours of the auction will be offered to the next highest bidder. Alright guys, be civil.
Jump to: