Pages:
Author

Topic: GekkoScience BM1384 Project Development Discussion - page 48. (Read 146665 times)

legendary
Activity: 966
Merit: 1003
If your willing to sell bare metal boards that retro-fit the R-box I'm in!....Especially if they don't need significantly more power than my current R-Boxes already use.
legendary
Activity: 3374
Merit: 1859
Curmudgeonly hardware guy
I have no reason to consider fan cost, if I assume people are going to buy them to put them in existing New R-box that already have fans. If you're building your own thing, yeah you're probably right about one fan being better.
legendary
Activity: 966
Merit: 1003
Yes, I like the size of the R-box Boards. You have enough R-box cases to set two together and see what I described. Thing is, If, I can get just the finished boards, I can get my own case fabricated easily enough. The two board w/140mm fan is more or less something to consider, might be more cost effective than a bunch of 72mm fans.
legendary
Activity: 3374
Merit: 1859
Curmudgeonly hardware guy
So what you're saying is, that board AJRGale asked about and then in a different enclosure which doesn't currently exist and I'm not going to make. You like the idea of AJR's board?
legendary
Activity: 966
Merit: 1003
I have a Dozen R-boxes running, got them all from CrazyGuy. If he hadn't sold out, I would have more of them. My opinion is two boards, the size of the R-Box with a single 140mm fan,  instead of four 72mm fans. Kind of like Half an R4. Keep the 6 pin PCI-E connector per board as well as the USB per board. I haven't had anywhere near the difficulties with my R-Boxes that others have and maybe my three 850Q SuperMicro PSU's have something to do with that but I love the size and USB connection.
legendary
Activity: 3374
Merit: 1859
Curmudgeonly hardware guy
AJRGale:

How many is a one-off? You looking for the same power dissipation as the R-Box? The software would be pretty much identical, but the hardware design for this would actually be easier than our 18-board if I can VRM a bank of chips instead of stringing them. Probably increases the parts cost a bit, hits about 5% on efficiency, but would be a lot easier to design and test.

If we stock clock 'em at 300MHz and run 15 chips that's a power dissipation of about 120W and hashrate of about 250G. 3 chips per VRM, five VRMs. I could grab the standby TPS53355 or I could clone the Compac VRM (which would be a bit cheaper, more versatile and give a better top-end). Stock speed would get us about 26A off the VRMs. With the Compac VRM I could build a stronger buck (that thing's seriously overbuilt; the inductor would melt before any silicon failed) you could push farther. 400MHz on the chips would get you more like 38A, which could push pretty hard but with a stout-enough inductor (or two in parallel?) you could do it. 400MHz would get you 330GH and pull upward of 180W, which is still doable on a single 6-pin but good luck keeping it cool with the stock fans.

Fun part would be volt-setting. I'd probably stock set 'em at around 740mV but make 'em adjustable on the full range, 600mV to 800mV. And get this - with the TPS51219 from the Compac, it takes in an external voltage reference to match on the output. Very few buck drivers do that. Most of 'em use a fixed internal reference and a divided feedback, but this guy was specifically designed for driving low-voltage high-current bucks so it takes direct feedback and an external divided reference. That means I can actually string all five VRMs off a common reference and have a single point of adjustment - either an internal potentiometer or a digital pot set from command-line.

If I can do hardware pot, heck I could have that board designed in a day. I've got three-wide nodes already drawn up, and the buck circuit drawn up (several times). If I needed to add thermal cutoff would be a bit of a trick, but I could probably wire a thermistor attached to the top heatsink into a Schmitt trigger on the EN lines of every buck, which kicks them off at about 90C and turns 'em back on below 80. Should also drop the reset lines so the chips re-initialize.

Prototype costs would not be zero. I could have most of the thing drawn up in a day, but getting proto PCBs to test would be a significant expense. Getting board dimensions and stuff is going to be easy, since I also have three of these dead. I also have seven working and four untested, courtesy of CrazyGuy.


So, here's the part where I ask how much it's worth to you to have that designed. Because I could do it. I'll still have to lean on Novak for the driver code, but honestly it'll be less work than our TypeZero driver if we leave it with hardware volt-setting (which will be a lot easier to do in hardware also, no micro and digital pot and the associated firmware).
legendary
Activity: 4256
Merit: 8551
'The right to privacy matters'
I'm also hoping more than just Philipma1957 post information about the sticks in the review thread, since there should be at least six people besides myself who have final-version prototypes in hand. The sooner I get feedback on the design, the sooner I can decide whether to tweak further or open up a sales queue.

And I got today a little package too! Smiley
Feedback coming soon!

I Look forward to your thoughts  on  them.  You have had stick building practice so I would like to read what you think of them.
hero member
Activity: 767
Merit: 500
that would be 59W for 115GH or 0.51W/GH all in. Not too bad.

Not too bad at all, especially if you compare to the U3, which for the same wattage on 4 of the older chips was only netting about 60GH/s, so you've basically doubled up on it (and without any of the hardware issues that come with a U3).

And this is where I just realized I'll never use my U3s again...  Shocked

Hmm.. i wonder if i could fit 4-6 of these inside a u3 chassis, with a little laptop heat-piping Hmmm...
definitely getting 10/20 of these inside an old "new-r-box" chassis

actually.. sidie baby, how much for a one off run of these chips on a "new r-box" clone board? Cheesy
I have 2 dead ones: sinks, fans, board (with shitty quality attempts at repairing the cp2012 chips) and chassis, you know there will be little market for these, maybe me, dave, phil, and steve? ( i think they are the ones that have these rockxie boxes)
vs3
hero member
Activity: 622
Merit: 500
I'm also hoping more than just Philipma1957 post information about the sticks in the review thread, since there should be at least six people besides myself who have final-version prototypes in hand. The sooner I get feedback on the design, the sooner I can decide whether to tweak further or open up a sales queue.

And I got today a little package too! Smiley
Feedback coming soon!
legendary
Activity: 4256
Merit: 8551
'The right to privacy matters'
that would be 59W for 115GH or 0.51W/GH all in. Not too bad.

Not too bad at all, especially if you compare to the U3, which for the same wattage on 4 of the older chips was only netting about 60GH/s, so you've basically doubled up on it (and without any of the hardware issues that come with a U3).

And this is where I just realized I'll never use my U3s again...  Shocked

ebay sales bro. list them now.
legendary
Activity: 1274
Merit: 1000
that would be 59W for 115GH or 0.51W/GH all in. Not too bad.

Not too bad at all, especially if you compare to the U3, which for the same wattage on 4 of the older chips was only netting about 60GH/s, so you've basically doubled up on it (and without any of the hardware issues that come with a U3).

And this is where I just realized I'll never use my U3s again...  Shocked
legendary
Activity: 3374
Merit: 1859
Curmudgeonly hardware guy
So, updates.

I have right now a couple sticks running on a modified cgminer which allows frequencies up to 300MHz (16.5GH per stick) pointed at the 1BURGER. The two sticks should be running 33GH total; Eligius as of right now reports a 3-hour average of 32.4GH so I'm okay with that.

My cooling setup is a single 120mm fan pulling 0.7A off 12V. With the hub I have, it should be enough to comfortably cool at least six or seven sticks. Assuming 2W for a Pi, 8W for the fan and 7W per stick for seven sticks, that would be 59W for 115GH or 0.51W/GH all in. Not too bad. The rest of my test parts arrived today which means I can build a few more sticks when I get some free time, if I want to push them more.

I'm busy this week with some consulting and manufacturing, but I'm hoping to do initial tests on the 18-board prototype on Thursday. I'm also hoping more than just Philipma1957 post information about the sticks in the review thread, since there should be at least six people besides myself who have final-version prototypes in hand. The sooner I get feedback on the design, the sooner I can decide whether to tweak further or open up a sales queue.

Amita PCB design has been started, probably 80% complete actually.

As of a few hours ago, my chips from Bitmain were in Kentucky awaiting some customs rigamarole.
legendary
Activity: 3374
Merit: 1859
Curmudgeonly hardware guy
I should be here by 9.
hero member
Activity: 735
Merit: 500
★YoBit.Net★ 350+ Coins Exchange & Dice
hey hack are you going to be avail tomarrow sometime im going to be in area and would like to stop by with that s2 board that is had those issues sir ?ill be home around 7-8 am and ill head there shortly after ?
hero member
Activity: 868
Merit: 1000
Hi
Will the 18 chip board be under licence in Eu/Uk or will that come direct from your stable?
legendary
Activity: 3374
Merit: 1859
Curmudgeonly hardware guy
If any of y'all testers want the tracking number for your package, PM me and I'll dig it up.
donator
Activity: 2352
Merit: 1060
between a rock and a block!
That seems accurate. I approve.

Additionally, I got confirmation of stock from Bitmain and the chips are paid for. TXID f530b8bde269d1c2433e4603a6b4b0db753938fe32f4cb52518bb5f8abddb5b6

I'll be sending out I think nine sticks this afternoon, eight to the testers and one to German guy.
I told my postman to call me when your package gets to my post office LOL

edit: yay!! legendary tag has arrived...
hero member
Activity: 767
Merit: 500
You were my next alternate if anyone else had backed out.

I've had a lot of people asking to do reviews (seriously a lot, including a bunch of n00bs and folks I've never heard of), and a lot of people asking to buy sample sticks. Gonna say this now so I don't have to repeat myself a bunch of other times (again) - if you're not one of the 8 listed above, you're not getting a sample stick. Go ahead, offer me a hundred bucks for one. You're not getting it. More than half of the ones I have materials for are being sent out to reviewers, and we're keeping the rest for in-house stuff like software dev and posterity and stuff. So please, stop asking.

oh shi- i forgot to throw my PM in too! what was i thinking, im such a n00b! ah well, looks like i miss out on free stuff like everyone else
Tongue

i could see that happening, i can see you just giving them a simple "No." for a message back, if you didn't just delete the message.

you know, i was going to buy dogies last 2 items he has for sale, screw him! im saving for these! Tongue
legendary
Activity: 4256
Merit: 8551
'The right to privacy matters'

Oh Shut out to bitmaintech for selling you the chips.

 And they are pretty good at delivering gear.


Give them a shout out instead, they'll appreciate it more.  Cheesy

done spell check strikes again.
legendary
Activity: 872
Merit: 1010
Coins, Games & Miners
Some behaviour suggestions in this thread peeps, just to keep things chugging along nicely:

  • Don't try to disclose the amount paid by the team for the chips, they had to sign an NDA for that (afaik) and we don't need that contract being tossed to the garbage bin, just keep calm about that, keep your calculations to you.
  • Don't effing ask when will it be ready. It will be ready when it's done.
  • Want to support? 1BURGER address is for that. Want to support further? buy your PSUs and adapter boards from these guys, they make excellent electronics and high quality cabling.

There... had to say it....
Pages:
Jump to: