New R-Box Upgrade Kit

[philipma1957]

If were now talking 4-5 chip "pod" (or "ufo" like I used to call them), vs. 12-18 chip box, I vote pod.

While we're on that subject, what kinda price would we be looking at? 35-ish?

The Amita (two-chip stick) is coming in about $35. This'd have an additional two chips ($7) and a microcontroller for voltage adjustment and temp monitoring, so add at least another ten bucks.

 25 usd ----------------- 1 chip stick   gives about 13.5 gh max at freq 250---- as I have not gotten software to load higher freq values (poor software skills on me)

35 usd  ----------------  2 chip stick   should give about   32 gh max ---- once built


45 usd -50 usd ? -- 4 chip pod    should  give  about 80-90 gh max----  65gh with 35 watts 90 gh with 50 watts



110 usd --------------- 18 chip board gives about    300gh     at 120 watts


Just to compare pricing please make corrections.--- I have corrected as per sidehack's later post

[sidehack]

4-chip would be probably $45-50, depends on PCB cost and some other factors. Also I like how you say the 1-chip stick max at 13.5GH when I have two sticks running 16.5GH for the last couple days. I think you're just not trying hard enough...

The 4-chip would run 90GH if you could keep it cool. 65GH would run under 35W, 90GH more like 50. The 18-chip board I'd probably set stock frequency at 300MHz for a NRB, which would get you 300GH off about 120W at the machine. For 18 chips probably would run higher than $100, maybe $110 neighborhood. We'd save some in regulator cost but add more in chips.

[philipma1957]

thanks for corrections.


I listed the 1 chip at 13.75 gh as that is my max.

Since I can't get a higher freq yet. I am pretty much certain I can get it to do 300 since at 250 with a fan it is only 90f or so.

My software skills are brutal.

 So cut + paste not much else.

If you want to send an sdcard with your minera image that has all the higher freq's

 I could test a rasp pi.

Which rasp pi did you  use?

 the B the B+ or the newest model 2

I have a B and

 the newest model 2 4 usb hub

[sidehack]

I already posted a link to modified source code that goes from 100MHz up to 300MHz by 6.25MHz increments. All you'd have to do is replace the files in cgminer 4.9.0 source and compile. Don't forget Icarus support.

[alh]

If you are seriously thinking of a "4-chip" pod, do you want to adjust the poll at the top, or start a 2nd one?

If you go down the "4-chip pod" path, then it's right smack in the middle of the U3 space. While I don't own a U3, what I read makes it sound horribly unreliable, and difficult to get stable. Is it BM1382, and that part of the problem, or what?

My Rbox-little was very stable with a good power supply (i.e. not a 12V brick). It's only drawback was that it was close to 1W/GH at the wall. Once I got 3 working well with my Raspberry Model B (not B+ nor Pi2), it literally ran for 6 months without intervention. it was plenty quiet as well.

[TheRealSteve]

If you go down the "4-chip pod" path, then it's right smack in the middle of the U3 space. While I don't own a U3, what I read makes it sound horribly unreliable, and difficult to get stable. Is it BM1382, and that part of the problem, or what?

It's just not as well designed as it should have been, both hardware and driver interface (in that it doesn't really have one).  That's pretty unrelated to the hashing chip of choice (BM1382 is also used in the S3 - not much of a problem with those).  Short of a bitwise quirk there's no reason a 4-chip miner would have issues in terms of comms where 2 or 8 don't.

I think the nice thing about a pod miner is that they're really fairly simple in construction.  Fan, heatsink, board - all matching in size and if done well only needs screws through existing mounting holes to hold it together - done.

[philipma1957]

If you go down the "4-chip pod" path, then it's right smack in the middle of the U3 space. While I don't own a U3, what I read makes it sound horribly unreliable, and difficult to get stable. Is it BM1382, and that part of the problem, or what?

It's just not as well designed as it should have been, both hardware and driver interface (in that it doesn't really have one).  That's pretty unrelated to the hashing chip of choice (BM1382 is also used in the S3 - not much of a problem with those).  Short of a bitwise quirk there's no reason a 4-chip miner would have issues in terms of comms where 2 or 8 don't.

I think the nice thing about a pod miner is that they're really fairly simple in construction.  Fan, heatsink, board - all matching in size and if done well only needs screws through existing mounting holes to hold it together - done.

the plus side is a simple build  and nice to have one or two around for fun.

[TheRealSteve]

Should hop on over to the SFARDS thread as well if you haven't already

[philipma1957]

Should hop on over to the SFARDS thread as well if you haven't already

SFards  - J4bberwock is not this far along as sidehack
avalon  mini - is not  this far along


looks like

Sidehack -------- 1 chip usb miner orders real soon  as working tested sticks  have been passed out.

Sfards -----------J4bberwock  a 2 chip dual miner not yet no working model as of now

Avalon ---------- 4(?) chip mini  in August.

My goal is to review them all and own them all.

Then see about owning more of sidehacks bigger gear as it develops.

Maybe a few 2 chips sticks.
 The 4 chip  little r-box
the 18 chip big r-box

Should be fun.

[sidehack]

Just for you Phil, I'm gonna start packaging these things in Pokeballs.

[philipma1957]

Just for you Phil, I'm gonna start packaging these things in Pokeballs.

pokeballs look good

[xwtfbobx]

I for one have the big r-box, as well as the R-K miner (think it was essentially 4 110gh/s big r-boxes thrown on a tubular heatsink)  essentially that would allow me to upgrade 5 cards, and pretty much triple my entire hash power for less than a new s5 shipped from bitmain...  will have to look for sure but i do believe the heatsink and hash boards in the rk-miner and the big rbox are the same.

* after comparing the 2 it looks like the the r-k miners added a controller (possibly a teensy) on each hashboard,  but all in all same heatsink just 4x as many on the r-k miner and quite possibly the rockminer r-4

big rbox


r-k miner


r-4


so 3 miners utilizing the same form factor and heatsink :-)

[sidehack]

That, coupled with Jabberwock's apparent intention to make an SFARTS pod, kinda leans me toward the NRB board. Though a pod miner would still be cheaper and easier to prototype and manufacture, making it also quite attractive. Nuts.

[xwtfbobx]

That, coupled with Jabberwock's apparent intention to make an SFARTS pod, kinda leans me toward the NRB board. Though a pod miner would still be cheaper and easier to prototype and manufacture, making it also quite attractive. Nuts.

lol whatever the decision, i have 5 gridseed pods collecting dust as well... i  am game either way, just thought i would point out that rockminer used the same boards and heatsinks for quite a few different miners.

[QuintLeo]

What I'd like to see would be a replacement for the Gridseed Gblades - the original "5.2 Mh/s Scrypt" per pair ones not the blacks.

 I'm SURE, given the junk power regulation stuff on those, that there are a bunch around that have died or died on one "blade" but not the other.

 HS on those should easily handle up to 20 chips, depending on the voltage they're run at.

 Optimally voltage adjustable, but a 16-to-20 chip string should run pretty efficiently.
 I figure (based actual power usage of my S5s) that the board (with a gold PS) would use about 40% more power at the wall than the chip specs, and I get the following:

 16 string would be .75 volts   - 286 GH/s at 155 watts (might be marginal these heatsinks but should be OK with a good 90mm fan)
 18 string would be .666 volts - 200 GH/s at   80 watts (interpolating here, but this would be viable with the "specified" GBlade 12V/10A power supply per side)
 20 string would be .6 volts    - 165 GH/s at   58 watts (pretty close to a direct match on power consumption with the original GBlade, viable with a barrel power connector)


 The 16 string would be kind of like a "quarter S5", a hair more efficient but not a bit diff.
 The 18 string IMO would be the best tradeoff of performance vs. power usage and should run plenty cool on these HS with the original fan.
 The 20 string would be a VERY nice high-efficiency replacement.

 Board should be 4" x 8" to fit the original HS "pad", but could be longer on the 8" side for low-power components that don't really need the HS (like the UART and possibly a microcontroller). The pad would make it trivial to put the BM1384s on one side of it and make good heat transfer to the HS itself, put the other "taller" stuff on the other side (you have about 3/4 inch of space to work with if you design the boards to "interleave" or put tall components on a part of the board that sticks out past the HS).


 If you could get any of these configurations built and sold for under $120 or so they'd be a winner!

[philipma1957]

I had 17 of these at one time  the heatsinks can not handle 20 antminer s-3 chips.     more  of a size issue  the chips won't fit.

I liked this gear.  I used about 85 watts per unit with atx power.  they did not overheat using 85 watts.

 So they could do more like 10-12 chips   since these s-3 chips can pull 7 or 8 watts on the top end.  My sticks are doing freq 250 and pull around 5.5 watts

What I'd like to see would be a replacement for the Gridseed Gblades - the original "5.2 Mh/s Scrypt" per pair ones not the blacks.

 I'm SURE, given the junk power regulation stuff on those, that there are a bunch around that have died or died on one "blade" but not the other.

 HS on those should easily handle up to 20 chips, depending on the voltage they're run at.

 Optimally voltage adjustable, but a 16-to-20 chip string should run pretty efficiently.
 I figure (based actual power usage of my S5s) that the board (with a gold PS) would use about 40% more power at the wall than the chip specs, and I get the following:

 16 string would be .75 volts   - 286 GH/s at 155 watts (might be marginal these heatsinks but should be OK with a good 90mm fan)
 18 string would be .666 volts - 200 GH/s at   80 watts (interpolating here, but this would be viable with the "specified" GBlade 12V/10A power supply per side)
 20 string would be .6 volts    - 165 GH/s at   58 watts (pretty close to a direct match on power consumption with the original GBlade, viable with a barrel power connector)


 The 16 string would be kind of like a "quarter S5", a hair more efficient but not a bit diff.
 The 18 string IMO would be the best tradeoff of performance vs. power usage and should run plenty cool on these HS with the original fan.
 The 20 string would be a VERY nice high-efficiency replacement.

 Board should be 4" x 8" to fit the original HS "pad", but could be longer on the 8" side for low-power components that don't really need the HS (like the UART and possibly a microcontroller). The pad would make it trivial to put the BM1384s on one side of it and make good heat transfer to the HS itself, put the other "taller" stuff on the other side (you have about 3/4 inch of space to work with if you design the boards to "interleave" or put tall components on a part of the board that sticks out past the HS).


 If you could get any of these configurations built and sold for under $120 or so they'd be a winner!

[QuintLeo]

I'm talking Antminer S5 chips though (1384) not S3 chips (1382) - the 1384 is the SAME SIZE (8mm square) as the GC3355 chips on the GBlade, and fit into LESS board space since they don't have leads sticking out from the chips like the GC3355 does. I have never bothered looking at the specs on the 1382 chip as it's outdated technology now and no longer available anyway.

 Power isn't an issue on an 20 string, see the BM1384 specs at .6 volts (string). 18 string is closer to the .65 volt spec than the .7 but would fall somewhere inbetween.

 10-12 chips would NOT be doable as a string, though they could be made to work with a VRM setup - optimally a software-configurable one like the SP20E has for better flexability and longevity.

[philipma1957]

I guess I was thinking of another piece of gear when it came to chip size.

J4bberwock's thread:

https://bitcointalk.org/index.php?topic=576784.0

The 2 disassembled PCB

[sidehack]

I think the 4-chip pod is a more logical target. It should be possible to make a single board which fits both the Gridseed and U3 heatsinks; I'll have to get ahold of an RBox pod to see what it'd require but I know I can hit the other two for sure.

[notlist3d]

I think the 4-chip pod is a more logical target. It should be possible to make a single board which fits both the Gridseed and U3 heatsinks; I'll have to get ahold of an RBox pod to see what it'd require but I know I can hit the other two for sure.

I have to applaud your designs.  If you could make one to fit in greedseed and U3 heatsinks that would be a very very nice product.  The good thing is there are A LOT more gridseed pods then new rboxs.  So much this cheaper price for equipment to put it in.  Although I do see why some want R-box it has merit as the amount of chips.

I personally would enjoy playing with a 4 chip modded gridseed pod.   If you go though with this design I think it's a great idea.   Either way you choose though best of luck.  I have been playing with the 1 chip usb model still and has been fun, and still working good.