Yes, that was the very first prototype board, only a very few
were made. Later H-CARDs have mounting holes.
In fact, only the very first two cards were ordered without mounting holes. Thanks for the clarification. |
|
|
|
Case work update:
- Power distribution and control PCB expected to be completed by early next week. Will support both Raspberry Pi or Beaglebone Black.
- Mechanical design is essentially complete, just need to complete part drawings and fire off to Steamboat
Still no price estimate, as this will depend on the quantity of parts ordered.
Changes since last design:
- 15 K16s per enclosure necessitated by a space added for an ATX PSU
- Ethernet connector added which will allow direct connection to the Raspberry Pi or Beaglebone Black for standalone operation
- Power distribution and control PCB will have both Molex and terminal strip connections for power
Case design could easily be adapted to support other PCBs (i.e. Bitburners)
Pics:
It's been a few days, and I was just curious if there were any more updates on the cases? With avalon starting to ship chips, I am very anxious to lock down how I am going to house and power my (16) boards. Timing and pricing are both important to me, but if I had to pick one, timing would be of the greatest concern right now. Thanks in advance! I would suggest that you make other arrangements as the case design project is parked. |
|
|
|
I'll just leave this here...  |
|
|
|
Case work update: - Power distribution and control PCB expected to be completed by early next week. Will support both Raspberry Pi or Beaglebone Black. - Mechanical design is essentially complete, just need to complete part drawings and fire off to Steamboat Still no price estimate, as this will depend on the quantity of parts ordered. Changes since last design: - 15 K16s per enclosure necessitated by a space added for an ATX PSU - Ethernet connector added which will allow direct connection to the Raspberry Pi or Beaglebone Black for standalone operation - Power distribution and control PCB will have both Molex and terminal strip connections for power Case design could easily be adapted to support other PCBs (i.e. Bitburners) Pics:   |
|
|
|
inb4 worldwide shortage of this part |
|
|
|
For those of you planning heatsinks, Intron provides a good reference. http://imgur.com/gUMVK0bPS: You are looking thru the board from the top side. Don't make the 'mirror error' too often made:) --Intron Hi Dave, I know this is an old post, but just needed to confirm something: the image at the link you included above shows 6 mounting holes on the PCB whereas the PCB in the picture in this link https://bitcointalk.org/index.php?topic=251966.msg2676985#msg2676985 seems to have none. Which is accurate? I assume that the photograph is of a prototype PCB which predated the mounting holes, but I'd like to know for sure. Thanks. |
|
|
|
So, has anyone noticed the dimensions? Looks like 20in x 16in x 8in is now stated on the Jupiter product page - that is a pretty drastic change from the originally stated 6in x 6in x 8in..
There is no way a 1000watt devices would have been 6in x 6in x 8in unless it was a fan heater, even then it's a bit small. Agreed. That said, it's still a _big_ bump in size, and I'm glad I noticed it as I'll need to reevaluate where I'm going to have room to put it. |
|
|
|
Lots bigger than I expected.
That's what she said. Sorry, I just couldn't let that one go  |
|
|
|
|
So, has anyone noticed the dimensions? Looks like 20in x 16in x 8in is now stated on the Jupiter product page - that is a pretty drastic change from the originally stated 6in x 6in x 8in..
|
|
|
|
Forgive me if this has been answered already. Each 3-k16 blade in this case, Is it a special board or just 3 seperate regular k16 boards mounted and hooked together. If I purchase the regular k16 boards at least 3 at a time will they work/mount in the case? Thanks
Just three K16 boards. You could put as few as one K16 or as many as 24 in the case. The max of 24 may end up being more like a max of 18 if we add a spot for a PSU. |
|
|
|
I know I am a bit late on this, but +1 on the case as well. Curious what the pricing would be, but given the PSU(s), cases, cabling, and time/effort involved, this seems like it would be a big win for those of us who bought more than a handful of boards. Is there any update on if this is going to be offered (or what price it might be offered at)? Didn't see anything on http://liquidsyndesigns.com or further mention in the thread... Case work is in progress. I received the first machined parts yesterday and am assembling prototypes. The PDU PCB design is underway, and I should have something to show next week. The Mineninja folks are on-board with development. Some pictures from today:   |
|
|
|
Can regulator could be replaced with 40A where will be able to run the 16 chip at 2,7Ghs?
Of course with heatsinks.
It might be difficult to find a pin compatible regulator. It appears, from a brief search, that this regulator series tops out at 30A. Regulator part number is TPS53355 Here's a post where Dave explains how to OC the board: https://bitcointalk.org/index.php?topic=251966.msg2745327#msg2745327 |
|
|
|
Is your K16 going to use the v0.3.0 revision board, and are you planning on using the 25 Amp buck reg (IR3847) for the banks or the original 16A one?
As i understand it there is no pcb design ready for the ir3847, it is not a drop in replacement, the p/s section of the pcb needs to be redesigned. I think I'm going to remove 4 chips from each board and overclock the remaining 12. Alternately it might be possible to build a separate pcb with the ir3847 on it. Exactly why I asked, it is a simple change to the PCB (and the program it was made in, is open source and 'gratis') so depending on if SB is a EE or not, it shouldn't be too difficult to swap out for him. I am wondering on making a small board with a IR3847 on it though, with traces going down to the correct positions for the K16 board, and remove the original buck reg. However, it's been 15 years since I really touched electronics, and everything is so tiny now  The IR3847 is a new part and stock seems to be virtually unavailable. There are many buck regulators available from other vendors though: http://www.digikey.ca/scripts/dksearch/dksus.dll?FV=fff40027%2Cfff8029a%2C15c0002%2C2040009%2C204000a%2C204000e%2C2040056%2C20400c3%2C2040230%2C2040437&k=buck+regulator&mnonly=0&newproducts=0&ColumnSort=-1000009&page=1&stock=0&pbfree=0&rohs=0&quantity=&ptm=0&fid=0&pageSize=25 |
|
|
|
As far as stacking goes, if you look at the rendered imaged of the stand-alone miner and that stacked versions you'll see that the fan is in a different orientation. That's what makes me think that the'll require at least some disassembly to stack.
Thanks,
-Jay
Don't read too much into the renders, they are illustrations more than anything. Which images are you referring to? |
|
|
|
For 0.65V you need exactly 1K resistor for R01F.
For 0.70V you need 1.5K.
With 1K resistor I get 1.26Amps @ 12.0 V, so 15Watts for an h-board.
[edit] Remember, with 16 chips, soldering that higher value resistor will quickly max out the amps that the regulator can handle.
Perfect. Thanks Dave! My heatsinks are ready |
|
|
|
When did the August delivery 25 Gh/s kit sell out?
Within 24hrs of the OP. |
|
|
|
Hi Dave, If you could provide the schematic section for the H-board power supply it would be very useful for those looking into overclocking. H-board regulator is this part Texas Instruments TPS53355. You can use it to OC the chips, but on the H-board 16 chips will max this thing out at .8v or so. Thanks Dave. Is the standard reference design used as per the datasheet? I guess what I'd like to determine is which resistor to adjust, and the math which determines the output voltage (based on the formulas in the datasheet). |
|
|
|
Hi Dave, If you could provide the schematic section for the H-board power supply it would be very useful for those looking into overclocking. |
|
|
|
Update:
Hosting:
The switched PDUs and PSUs are arriving early next week. We have chosen 550watt 80+gold PSUs to give both high efficiency and fine grain control for power cycling boards should the need arise. Each PSU will power between 13 and 15 boards, depending on the final power usage. The PSU will then be connected to our switched PDUs, which gives us the ability to power cycle the PSU remotely. The network infrastructure and configuration is nearing completion, and we will be ready to receive the units once they are assembled.
Hi SB, maybe you should consider revising your choice of the PSU or plan less boards per PSU. According to the best practices you shouldn't run a PSU at more than 70%. So a 550W PSU would be most efficient running at 385W. 12 boards running at 32W each would consume 384W and with slight overclocking even 12 boards would push past the 70%. +1 on the case. Is that 21 boards in there? I wonder what would it look like once you plug in all the cables. Cheers 24 x K16s. Cabling should be neater than you'd think as the design is based around a central controller and power distribution unit PCB (CCPDU). Two wires will run to each of the "blades" (composed of 3 X K16s) where the wiring will then branch out to individual boards. So only 16 power wires running throughout the chassis.Data comms are another thing, but the plan is to use an I2C chain terminating in a single USB connected K16, which will in turn be connected to a Beaglebone Black which will act as a controller and cgminer/bfgminer host. Based on Bkk's most recent changes to the K16 power distribution will be done via 6 pin molex ("PCIe") type connections. |
|
|
|
Update:
Hosting:
The switched PDUs and PSUs are arriving early next week. We have chosen 550watt 80+gold PSUs to give both high efficiency and fine grain control for power cycling boards should the need arise. Each PSU will power between 13 and 15 boards, depending on the final power usage. The PSU will then be connected to our switched PDUs, which gives us the ability to power cycle the PSU remotely. The network infrastructure and configuration is nearing completion, and we will be ready to receive the units once they are assembled.
Hi SB, maybe you should consider revising your choice of the PSU or plan less boards per PSU. According to the best practices you shouldn't run a PSU at more than 70%. So a 550W PSU would be most efficient running at 385W. 12 boards running at 32W each would consume 384W and with slight overclocking even 12 boards would push past the 70%. +1 on the case. Is that 21 boards in there? I wonder what would it look like once you plug in all the cables. Cheers 24 x K16s. Cabling should be neater than you'd think as the design is based around a central controller and power distribution unit PCB (CCPDU). Two wires will run to each of the "blades" (composed of 3 X K16s) where the wiring will then branch out to individual boards. So only 16 power wires running throughout the chassis. Data comms are another thing, but the plan is to use an I2C chain terminating in a single USB connected K16, which will in turn be connected to a Beaglebone Black which will act as a controller and cgminer/bfgminer host. |
|
|
|
|
Show Posts
