The combined sidehack-novak usb stick review thread. AKA GekkoScience BM1384

[HerbPean]

That's pretty cool !

Great job guys

[sidehack]

I certainly wouldn't mind getting ahold of that cgminer build. Novak's gonna be busy all week catching up on his own projects and I need to be manufacturing PSU gear so I doubt either of us are gonna have much time for cgminer coding.

Also, by now there should be at least six of the new final-prototype sticks delivered to US testers (three more went to people overseas). Hopefully this week enough of them will post info that we can get a consensus whether or not they're sellable. Might open an order queue as early as Friday.

[TheRealSteve]

PM sent - forum's having some glitches, postponing cgminer build thread until tomorrow/day after depending on hours.

[sidehack]

Running my stock 4.9.0 at 250MHz I'm seeing 13.7GH; running your custom build at 250MHz I got up to 11.5GH with a lot of resets. The power draw was what I expected for 250MHz but the hashrate and WU came in low. Not sure what's up.

[TheRealSteve]

Unfortunately I can't even provide a proper comparison build; all the cgminer code outside of the frequency options is the same, so any differences would come from not having the exact same libraries used in the official builds, kludge workarounds for compile errors, etc.  Might find out later if there's some egregious difference - if I get around to posting that windows build thread, get some feedback, etc.

Might want to compare against 4.9.2, since that's the codebase used - but I'd be surprised if there's a performance difference between official 4.9.0 and official 4.9.2

[-droid-]

Hopefully no magic smoke was released in an untimely coincidence.  I'll see about documenting the cgminer build (mis-)steps somewhere (not this thread, way too o/t).  If anybody wants to play with my binary, shoot me a PM - or just wait for sidehack/novak's build

Not trying the build yet but photos of my gear at freq 250.  I am impressed with these little sticks.

say 2.2 amps x 4.75 volts =  9.9 watts giving 27.46gh  this is 0.36 watts per gh at a high overclock.

you should look into being a hand model

[sidehack]

Unfortunately I can't even provide a proper comparison build; all the cgminer code outside of the frequency options is the same, so any differences would come from not having the exact same libraries used in the official builds, kludge workarounds for compile errors, etc.  Might find out later if there's some egregious difference - if I get around to posting that windows build thread, get some feedback, etc.

Might want to compare against 4.9.2, since that's the codebase used - but I'd be surprised if there's a performance difference between official 4.9.0 and official 4.9.2

The only comparison I have to 4.9.2 is the stock compile on the Pi that didn't enumerate the sticks at all. I'll see about digging up a stock Windows 4.9.2 after I'm done messing with 4.9.0 code on Linux.

[philipma1957]

Hopefully no magic smoke was released in an untimely coincidence.  I'll see about documenting the cgminer build (mis-)steps somewhere (not this thread, way too o/t).  If anybody wants to play with my binary, shoot me a PM - or just wait for sidehack/novak's build

Not trying the build yet but photos of my gear at freq 250.  I am impressed with these little sticks.

say 2.2 amps x 4.75 volts =  9.9 watts giving 27.46gh  this is 0.36 watts per gh at a high overclock.

you should look into being a hand model

Actually I have wonderful fingernails.  Many women have told me that would pay to have them.

[alh]

Actually I have wonderful fingernails.  Many women have told me that would pay to have them.

You might not want to repeat that too often to your Wife..... 

[Jake36]

Unfortunately I can't even provide a proper comparison build; all the cgminer code outside of the frequency options is the same, so any differences would come from not having the exact same libraries used in the official builds, kludge workarounds for compile errors, etc.  Might find out later if there's some egregious difference - if I get around to posting that windows build thread, get some feedback, etc.

Might want to compare against 4.9.2, since that's the codebase used - but I'd be surprised if there's a performance difference between official 4.9.0 and official 4.9.2

The only comparison I have to 4.9.2 is the stock compile on the Pi that didn't enumerate the sticks at all. I'll see about digging up a stock Windows 4.9.2 after I'm done messing with 4.9.0 code on Linux.

4.9.2 upgraded the USB reset for the U3's, (-U3 will USB reset on no shares for 2 seconds instead of 1.), which might be why it's not enumerating the sticks ??.

[philipma1957]

Unfortunately I can't even provide a proper comparison build; all the cgminer code outside of the frequency options is the same, so any differences would come from not having the exact same libraries used in the official builds, kludge workarounds for compile errors, etc.  Might find out later if there's some egregious difference - if I get around to posting that windows build thread, get some feedback, etc.

Might want to compare against 4.9.2, since that's the codebase used - but I'd be surprised if there's a performance difference between official 4.9.0 and official 4.9.2

The only comparison I have to 4.9.2 is the stock compile on the Pi that didn't enumerate the sticks at all. I'll see about digging up a stock Windows 4.9.2 after I'm done messing with 4.9.0 code on Linux.

4.9.2 upgraded the USB reset for the U3's, (-U3 will USB reset on no shares for 2 seconds instead of 1.), which might be why it's not enumerating the sticks ??.

I get that on my 4.9.2 build.  I really struggled getting zadig to load drivers after that I am good on the two sticks up to freq 250

Also you need a really good hub to drive sticks at freq 250 … my sticks do 1.07 amps and 1.13 amps and pull .36 watts 

So 250 is not bad as a safe top end.

[sidehack]

So, we got some cgminer 4.9.0 edited to take up to 300MHz in 6.25MHz increments. I haven't started tuning the stick for min stable voltage, but I got it to light up at 300MHz off 730mV core, pulling 5.03V 1.47A off the line which gives us a stick-level power of 7.39W for 16.5GH at 0.448W/GH, putting its device-level efficiency prett much on par with an S5. With the fan I had on it, the heatsink was leave-your-fingers-on-it-indefinitely hot.

I'm okay with that.

[EDIT] For a pair of sticks tuned a bit better, I'm seeing 1.38A and 1.42A for an average draw of 1.40A, so 7W or 0.425W/GH

[TheRealSteve]

Very nice   Is that already a cross-compile for Windows (for phil et al)?  I wonder what the issue might be with 4.9.2 - no problems here (stock or custom), but that's not on a U3 / Compac.

[philipma1957]

Very nice  Is that already a cross-compile for Windows (for phil et al)?  I wonder what the issue might be with 4.9.2 - no problems here (stock or custom), but that's not on a U3 / Compac.

I have not yet tried to do your linked software. I  got  too busy. with some other stuff.

Now that sidehack is doing 300mhz and getting   .44 watts   use a fan and just a bit to hot to touch  5.03volts at 1.47 amps = 7.39 watts


 and I did 250mhz getting .36 watts  use a fan and you can hold them just warm  1.1 amps x 4.73 volts = 5.203 watts


I am thinking 275 may be a good compromise.  along with  6 watts or 6.3 watts per stick


To all please remember these 250, 275, 300 settings will need a very good hub and a fan



In the long run I may do a 5-10 stick fun miner and run at  freq 218.75  they will not need a bridge for power and should cool well with my fans.

[sidehack]

If by "just a bit too hot to touch" you mean "leave your fingers on it indefinitely", then sure. Because they were definitely not too hot to touch.

[philipma1957]

If by "just a bit too hot to touch" you mean "leave your fingers on it indefinitely", then sure. Because they were definitely not too hot to touch.

that is exactly what I mean.

I can take the sticks set on freq 250 and hold them with 2 fingers  for 5 or more minutes (allowing the fan to still hit and cool the stick)

I was guessing the 300 freq would be too hot to do this for 1 or 2 minutes

with one stick at 5.2 watts

the other at 7.39  that extra 2 watts is too much for long term holding of the sticks.

[TheRealSteve]

Building cgminer on Windows with MinGW, step-by-step (2015-07-06) - includes the freq change steps.  Keep in mind sidehack reported lower hash rate / more errors from a build following the steps laid out herein

[philipma1957]

Building cgminer on Windows with MinGW, step-by-step (2015-07-06) - includes the freq change steps.  Keep in mind sidehack reported lower hash rate / more errors from a build following the steps laid out herein

is this what you dropped linked to me?
a lot of work you did man!

[sidehack]

Novak and I noticed some discrepancies with the hex values I was using and the hex values from S5 code, which probably caused most of the issues. The sticks, when given an invalid hex, jumped up to freakish current draw without being detected.

I'll link an archive with the modified files for cgminer 4.9.0 that we have working here in a bit. I haven't tested every frequency step but I know 287.5MHz and 300MHz work.



[EDIT] http://gekkoscience.com/misc/compac/compac-cgminer.zip  contains modified cgminer.c and driver-icarus.c which should compile into a cgminer 4.9.0 which gives 6.25MHz granularity from 100MHz to 300MHz for the U3, and consequently for our Compacs

[xZork]

philipma, what are you doing with 1-2-3 blocks?