GekkoScience BM1384 Project Development Discussion

[TheRealSteve]

If that change was made to help us, it was done without my knowledge. I don't think we've talked to cgminer devs yet, but you'd have to ask Novak for specifics.

No, it was specifically to deal with some U3 issues.  For a long time, -ck didn't have a U3 to test against (not sure if he does again now), and the previous version of cgminer was giving people trouble (using a version older worked okay though).

Note that he's got some choice words on the U3 that make it clear he's no fan of the comm chip's behavior and non-ease of identification of the U3:
https://bitcointalk.org/index.php?topic=827356.msg11598332#msg11598332
https://bitcointalk.org/index.php?topic=827356.msg11036830#msg11036830
https://bitcointalk.org/index.php?topic=28402.msg10538698#msg10538698
https://bitcointalk.org/index.php?topic=28402.msg10039320#msg10039320
Only this last one probably semi-applies to the Compac as well, but you've already touched on that in an earlier post (identifying as a U3).
I have no idea why the comm chip would misbehave on the U3 - bog standard CP2102 used on several miner designs and even larger numbers of other consumer devices.

Once novak has things fleshed out on his end, it certainly can't hurt to send -ck one for an extra set of eyeballs and testing (if he's interested); I'd be happy to frontcover the costs.

[1714214441]

1714214441

[1714214441]

1714214441

[sidehack]

Yep, the Icarus driver is pretty much a disaster at this point, with all the different little tricks required to detect from a couple dozen different machines using the same interface. We're going to try to write our own driver, and probably do something with the CP2102's manufacturer and device codes for ready identification.

[philipma1957]

Yep, the Icarus driver is pretty much a disaster at this point, with all the different little tricks required to detect from a couple dozen different machines using the same interface. We're going to try to write our own driver, and probably do something with the CP2102's manufacturer and device codes for ready identification.

that would be a nice touch. making these more fun to work with.  I agree with you on temps. my 1 stick can do freq 218 with a fan in 91f temps so if your hub is a bit weaker or you want to go fanless.  freq 150 - 175 should be okay.

[AJRGale]

Yep, the Icarus driver is pretty much a disaster at this point, with all the different little tricks required to detect from a couple dozen different machines using the same interface. We're going to try to write our own driver, and probably do something with the CP2102's manufacturer and device codes for ready identification.

changing the VID/PID on them things shouldn't be hard to do, to be USB compliant, you may want to register it?
as i have mentioned before, if you are using them cp2*** chips, please use ESD protection on it (if you haven't done it yet, also people asking about it here maybe good info in there?)

if not you could move onto the more bells and whistles, yet more expensive, "FT232R"? and call it a day?

Alright, looks like skinny it is. The PCB dimensions of the Compac are 1" by 2.5"; the Amita will probably be more like 1.3" by 4.1". The extra height is there to accommodate the doubled heatsinks, since there's one 25x40mm heatsink per chip. It'll look pretty much just like the Compac extended upward with a second sink above the first.

I'm probably gonna cut out of here in about two hours (making this an 8.5 hour day already) so let's see if I can't get the Amita PCB done in that amount of time.

Also, I'm going to give our tardy debtors until about Friday to get things ironed out with us (I expected it to be taken care of last Tuesday) before I think about short-term-borrowing some coin to cover the chip order I expected to have paid last week. So don't worry about it quite yet.

i gotta find dimensions of them 18 chip board (i believe thats the one getting made to fit the U1 yes?), im wondering if i could retrofit new-r-box sinks to them..

[TheRealSteve]

changing the VID/PID on them things shouldn't be hard to do, to be USB compliant, you may want to register it?

A proper VID registration is not chump change.  For something with such a small run, mainly used by techy people, and presuming they don't necessarily want to put a USB logo on the product.. it's just not worth it.  Thus my earlier suggestion of instead getting a PID from a PID vendor, or using the VID of a defunct company.  Plenty of articles on the internet about the pros/cons of doing so in terms of potential technical issues, whether or not entities have a right to sell the PIDs they do, and legal considerations.

if you are using them cp2*** chips, please use ESD protection on it  [...] if not you could move onto the more bells and whistles, yet more expensive, "FT232R"? and call it a day?

While ESD diodes on the CP2102 inputs is nice (AntMiner U1/U2, bi•fury use 'm), I've not had any on other products that didn't have them fail.  But that's anecdotal and there's plenty others who do swear there's issues.  If an alternative were to be chosen, though, FTDI wouldn't really be my first choice for the same reason that Prolific wouldn't be my first choice.  There's always MCP and wCH.  If anything, I'd use a CP2103 and make use of the 4xGPIO.. but that doesn't negate ESD concerns

[AJRGale]

changing the VID/PID on them things shouldn't be hard to do, to be USB compliant, you may want to register it?

A proper VID registration is not chump change.  For something with such a small run, mainly used by techy people, and presuming they don't necessarily want to put a USB logo on the product.. it's just not worth it.  Thus my earlier suggestion of instead getting a PID from a PID vendor, or using the VID of a defunct company.  Plenty of articles on the internet about the pros/cons of doing so in terms of potential technical issues, whether or not entities have a right to sell the PIDs they do, and legal considerations.

Eh, i didn't see the costs, yeah, ignore my statement..

if you are using them cp2*** chips, please use ESD protection on it  [...] if not you could move onto the more bells and whistles, yet more expensive, "FT232R"? and call it a day?

While ESD diodes on the CP2102 inputs is nice (AntMiner U1/U2, bi•fury use 'm), I've not had any on other products that didn't have them fail.  But that's anecdotal and there's plenty others who do swear there's issues.  If an alternative were to be chosen, though, FTDI wouldn't really be my first choice for the same reason that Prolific wouldn't be my first choice.  There's always MCP and wCH.  If anything, I'd use a CP2103 and make use of the 4xGPIO.. but that doesn't negate ESD concerns

well you should know my religious stance on esd protection now, ive said it enough times..

[philipma1957]

changing the VID/PID on them things shouldn't be hard to do, to be USB compliant, you may want to register it?

A proper VID registration is not chump change.  For something with such a small run, mainly used by techy people, and presuming they don't necessarily want to put a USB logo on the product.. it's just not worth it.  Thus my earlier suggestion of instead getting a PID from a PID vendor, or using the VID of a defunct company.  Plenty of articles on the internet about the pros/cons of doing so in terms of potential technical issues, whether or not entities have a right to sell the PIDs they do, and legal considerations.

Eh, i didn't see the costs, yeah, ignore my statement..

if you are using them cp2*** chips, please use ESD protection on it  [...] if not you could move onto the more bells and whistles, yet more expensive, "FT232R"? and call it a day?

While ESD diodes on the CP2102 inputs is nice (AntMiner U1/U2, bi•fury use 'm), I've not had any on other products that didn't have them fail.  But that's anecdotal and there's plenty others who do swear there's issues.  If an alternative were to be chosen, though, FTDI wouldn't really be my first choice for the same reason that Prolific wouldn't be my first choice.  There's always MCP and wCH.  If anything, I'd use a CP2103 and make use of the 4xGPIO.. but that doesn't negate ESD concerns

well you should know my religious stance on esd protection now, ive said it enough times..

I get a lot of ESD issues in my home in the winter (forced air heat)    I have shocked Antminer U1/U2 and ice furies more then one time.

I had over 300 U1/U2 sticks in house and was running up to 130 of them at one time.  I killed two or three with direct ESD to a stick.  But most of the time (more then 25 times_ they survived the ESD.  So I would rate the protection on them as decent.

[sidehack]

Also, point of interest. The temporarily-applied heatsink (rubber bands) on my test stick halfway came off sometime in the last 21 hours of running at 150MHz in my 80+F shop. It was tilted over so the shim was basically only contacting one edge of the chip - not a face, just the corner edge. The current was up to 600mA (from 570) and the heatsink and PCB were pretty darn toasty but cgminer reported zero HW in that time. I'm okay with that. So this guy has been running at ~610mV 150MHz drawing 4.94V/570-600mA with 168K accepted shares and zero HW for a shade over 24 hours.

Good enough for me. I think I'll build a few more today.

Also, on my U2 I can see a couple unpopulated pads on the USB signal lines that I assume are AJRGale's ESD protection. I do not have it on the Compac, but if it becomes a problem and we run a second batch I'll look into adding that.

[sidehack]

I built me 8 sticks today, and most of 'em appear to work. One of them won't start at 150MHz until it's cranked to something like 650mV (instead of 615) so I don't know about that guy. One of 'em is just straight-up misbehaving (chip issues maybe), but the rest seem to work. I have four right now running off a Pi and I'll see tomorrow how they look.

One of the eight sample heatsinks is misdrilled. Not too impressed with those odds.

[valkir]

Nice! Cant wait to see them and to get mine! 

[philipma1957]

I built me 8 sticks today, and most of 'em appear to work. One of them won't start at 150MHz until it's cranked to something like 650mV (instead of 615) so I don't know about that guy. One of 'em is just straight-up misbehaving (chip issues maybe), but the rest seem to work. I have four right now running off a Pi and I'll see tomorrow how they look.

One of the eight sample heatsinks is misdrilled. Not too impressed with those odds.

are they at the burger address?


http://eligius.st/~wizkid057/newstats/userstats.php/1BURGERAXHH6Yi6LRybRJK7ybEm5m5HwTr


 I found it do I  guess test address is different this time.

[sidehack]

Naw, I was in a bit of a hurry to finish up and still get to the post office on time (unfortunately I arrived seconds after closing time) and I think the config defaulted to the BEUSB address on BTCDig that TheRealSteve set up for the erupter contest. I haven't bothered to make an account there but it's handy for testing since the vardiff starts at 2 instead of 128.

If the shop VPN hadn't gotten hosed in a storm-related router crash last week (and I haven't fixed that part of the config yet, whoops) I could log in and change it. But I didn't. Oh well. Maybe tomorrow I'll figure out the issues with the other sticks, hook the hub up to a beefier 5V supply and run all 8 together over the weekend at the 1BURGER address. That'd be 66GH. It's fun to think I could push these to 11GH off about 3.6W, when an AM Blade was top of the line gear just two years ago and took 24 times that power to do slightly less work.

Think I should push them at all? If I can keep 'em cool at 225MHz (what, about 4.5W each?) that's right at 100GH which was a loaded Blade backplane, and even counting the power taken by the fan and the Pi this would still come in at half the draw of a single Blade. Hooray technology!

[AJRGale]

Naw, I was in a bit of a hurry to finish up and still get to the post office on time (unfortunately I arrived seconds after closing time) and I think the config defaulted to the BEUSB address on BTCDig that TheRealSteve set up for the erupter contest. I haven't bothered to make an account there but it's handy for testing since the vardiff starts at 2 instead of 128.

If the shop VPN hadn't gotten hosed in a storm-related router crash last week (and I haven't fixed that part of the config yet, whoops) I could log in and change it. But I didn't. Oh well. Maybe tomorrow I'll figure out the issues with the other sticks, hook the hub up to a beefier 5V supply and run all 8 together over the weekend at the 1BURGER address. That'd be 66GH. It's fun to think I could push these to 11GH off about 3.6W, when an AM Blade was top of the line gear just two years ago and took 24 times that power to do slightly less work.

Think I should push them at all? If I can keep 'em cool at 225MHz (what, about 4.5W each?) that's right at 100GH which was a loaded Blade backplane, and even counting the power taken by the fan and the Pi this would still come in at half the draw of a single Blade. Hooray technology!

then good old 500W foot heaters.. i was half tempted to buy an old set of them, someone here in aus is selling them for $100ish for 10 blades, and they a shat on by 8 little hand made sticks!

i better keep working on my quantum entangle light based ALU

[TheRealSteve]

I think the config defaulted to the BEUSB address on BTCDig that TheRealSteve set up for the erupter contest. I haven't bothered to make an account there but it's handy for testing since the vardiff starts at 2 instead of 128.

Too bad it's one of the few pools that'll even go that low, at least reliably

That pool has a semi-public output of hash rates, but no fancy public graphs, but this graph off the account page should all be GekkoScience testing anyway:

[philipma1957]

I think the config defaulted to the BEUSB address on BTCDig that TheRealSteve set up for the erupter contest. I haven't bothered to make an account there but it's handy for testing since the vardiff starts at 2 instead of 128.

Too bad it's one of the few pools that'll even go that low, at least reliably

That pool has a semi-public output of hash rates, but no fancy public graphs, but this graph off the account page should all be GekkoScience testing anyway:

thanks for info and charts.

[philipma1957]

Naw, I was in a bit of a hurry to finish up and still get to the post office on time (unfortunately I arrived seconds after closing time) and I think the config defaulted to the BEUSB address on BTCDig that TheRealSteve set up for the erupter contest. I haven't bothered to make an account there but it's handy for testing since the vardiff starts at 2 instead of 128.

If the shop VPN hadn't gotten hosed in a storm-related router crash last week (and I haven't fixed that part of the config yet, whoops) I could log in and change it. But I didn't. Oh well. Maybe tomorrow I'll figure out the issues with the other sticks, hook the hub up to a beefier 5V supply and run all 8 together over the weekend at the 1BURGER address. That'd be 66GH. It's fun to think I could push these to 11GH off about 3.6W, when an AM Blade was top of the line gear just two years ago and took 24 times that power to do slightly less work.

Think I should push them at all? If I can keep 'em cool at 225MHz (what, about 4.5W each?) that's right at 100GH which was a loaded Blade backplane, and even counting the power taken by the fan and the Pi this would still come in at half the draw of a single Blade. Hooray technology!

 my 1 stick at 225Mhz with fans is a bit warm.  but I have been running it steady at 218.75 with the fan in an 80- 90 f garage.

I would think running them at freq 218.75  will give you about 94-96gh and if you use fans they will not over heat.

[sidehack]

That chart's not too encouraging. Looks like the setup dropped out shortly after I fired it up. I wonder if it was the hub, the Pi or the power.

[Ikron]

The pi was infamous for its lousy usb reliability. I wouldn't be surprised if that is still the case.

[cavaliersrus]

my b+ 512 is still running strong from last time i was there sidehack

digging the updates on the miners your working on

[philipma1957]

may be his hub.  he had 8 sticks pulling 4 watts each.  I know he has a rebuilt custom hub it may have given up the ghost.

Anyone that  ran a ton of sticks knows hubs can stress out when you get to the top end of the hub.