I'd hate to piss in the fire here, but at one time a company did make what was called a TMD fan which minimized the center deadspot by moving the fan motor(s) to 4 corners of the fan. I had one, it was VERY powerful for a 70mm fan, had a LOT of torque and produced very good pressure. Trouble is it didn't last very long. They don't make them anymore that I'm aware of. Back when I built my watercooling setup I had looked for some but couldn't find any. ah...found a link to a wiki: http://en.wikipedia.org/wiki/Tip-magnetic_drivingand a youtube : https://www.youtube.com/watch?v=kKofleGNE7g |
|
|
|
Can anyone post an example .BAT file that I could copy that will show how to setup Autotune function for each miner?
I have looked at the included AUTOtune .BAT example that came with cpuminer, but it only shows setup for 1 gridseed. I have to make it run 10 of them.
Thanks for all the help so far, and for all the great work done to help figure all this out. I am getting closer every day to a properly working setup!
minerd-sandor111.exe --gc3355=\\.\COM8,\\.\COM9 --gc3355-freq=\\.\COM8:1200,\\.\COM9:1225 --gc3355-autotune --url=stratum+tcp://ny.clevermining.com:3333 --userpass username:password That's what mine looks like. Hope it helps. |
|
|
|
to the super equiped pro's here:
anyone could measure USB power draw?
trying to figure out how many i could run stable without powered hub
cuz i would need a relai to keep using the controller auto-restart feature
That really depends...power draw with the fan running from 5v USB power or power draw with fan on stock 12v "jet engine mode"? If using the 5v USB power mod, I'd say no more than 2-3 on a non-powered hub since you're limited to 500mA spread across all ports in the usb hub. I think wolfey had measured the fan's power draw at one point. |
|
|
|
While on the subject of thermals if you really wanted to get down to the root of the cooling issue you'll have to either use thin copper shims or mill out part of the heatsink so the taller components don't prevent full contact with the gc3355 chips. If you chose to mill out part of the heatsink you will also notice many of the capacitors around the gc3355 chips are the same height if not just a tad higher so those will need to be insulated to prevent electrical contact with the heatsink. Shims are easier  |
|
|
|
|
If you're indecisive about 47k or 49.9k/50k you could use a pot, and adjust to your liking - I wouldn't recommend using a pot more than 50k though that way you can't turn it too high.
I would second wolfey's suggestion of using the latest build of cpuminer that Sandor posted in here recently as well. I never had much luck at all with bfgminer, and cgminer worked well enough for me but the per-chip auto-tuning feature that is in the latest cpuminer is hard to beat. You can also use a single instance of cpuminer now instead of running multiple instances for multiple gridseeds.
|
|
|
|
I tried the "simple" instructions that are listed in just about every readme/build.txt and I also tried these http://pastebin.com/p2cRgHxd ...no luck with either. Compiling this stuff in Windows is a real b*tch. I'm going to attempt it again after work on a clean system, the one I've been using is full of all kinds of crap and really needs to be wiped. I've got it to work, but I can't recall the exact steps I took... all I know is that it was a mess. BTW, and sorry for off-topic, I've just compiled an optimized build for Windows (disable FIFO & flow control) + pool hashrate display. https://www.dropbox.com/s/7z1cj75mhvojjah/minerd-sandor111.exeThis seems to push the pool hashrate beyond of what the chip is hashing at, or maybe it's just variance, but there is definitely some improvement. I'm testing this build out right now. Getting better results with it than the previous builds (including the one from cryto blog). The first couple builds would show 989KH/s client side. But while clevermining pool showed similar rates, manicminer pool showed half that. With this newest build, I'm getting consistent hashrate reporting both clientside and poolside for both pools I've tried. |
|
|
|
I tried the "simple" instructions that are listed in just about every readme/build.txt and I also tried these http://pastebin.com/p2cRgHxd ...no luck with either. Compiling this stuff in Windows is a real b*tch. I'm going to attempt it again after work on a clean system, the one I've been using is full of all kinds of crap and really needs to be wiped. |
|
|
|
Well, besides stating the 'obvious'..re: opinions, onion head...geesh....What do you know about what "Everyone"? Generalities are a lame way of trying to back up ones own opinion. I doubt very seriously that although there are obviously a few tinkerers on here - most pod owners are interested in getting on with making ROI from their miners. PERIOD! Also, I seriously doubt most folks will spend $100 to $300 on a hobbyists' tinker toy! Get real! This is an investment and its purpose is to make money with them. Not play with them. However, I'll let it be as you suggest ,,,that is until I need to bring it up again ;p Really? Are you and your ego that vulnerable or insecure that you must make everything a personal attack and resort to name calling? You bash me for "making generalities" (even though they are TRUE and in no way "backing up an opinion") yet turn right around and make your own generality? For you and many(most) others, it likely is an investment. However there are also those of us out there that it is both an investment AND a hobby. It is possible to have both. Also, there are a LOT of people that will spend $100-$300 on a "hobbyists' tinker toy". I've spent thousands on equipment to tinker with as part of a hobby, look at anyone that works on cars as a hobby, or computers (especially gaming), hell some people have aviation as a hobby and that's far from cheap. $100-$300 is pretty cheap when it comes to the scope of a hobby. So just like others have asked you to stop advertising in every post, I'm simply asking you to stop "attacking"/"bashing" (as well as the completely unnecessary name calling) other ideas that people come up with and are trying. It seems that no matter what someone else posts, whether it be their solder job, an idea, or other type of mod (successful or not), you seem to have a problem with it and return to preaching that the 49.9k on r139 mod is "the only mod anyone should worry about". |
|
|
|
Doesn't work = complaint, in my book re: SHA mining. The key word is IF.....not when. Personally I think SHA algo is on its way out due to security and some other issues I've read about. Any mod can be reversed. Who cares if it takes a few minutes more to revert R52 mod back to original? I think these other mods are just adding randomity and confusion to an otherwise straight forward and solved, simple mod that makes Scrypt only miners (the majority of seed users) higher profits per seed!  wolfey - your views and opinions are just that....yours. Everyone is well aware that in your opinion, everyone should use 49.9k at r139 and be done with it no matter if there is an easier method available or not. You don't need to preach to everyone in every post about "your" mod is the "best mod" (I mean really...who gives a sh*t who's mod it is anyway when it's a community effort). There are people out there that LIKE to mod. We LIKE to tinker with stuff, change things, explore other options. It's not about "being done with it and making money". We do the modding for FUN, for a hobby, because we like to. If you have issues with this, just leave the thread and let the modders mod since you're obviously "done with it and making money". |
|
|
|
*snip*
If I'm understanding this mod correctly. Reversing it will still require the same amount of work to reverse as reversing the R52 mod.
One thing that is better about this mod though is that all one has to do is remove the 17K resistor to put it back to original specs.
I for one do not look forward to soldering those two bridges again! I was happy we did away with that. It's a real pain to do, even for a master of soldering.
If all I have to do is solder on a 17K resistor to those two points, I agree that is it a better option than removing 'r139' (aka r52) and soldering in a replacement resistor only to later have to remove it then solder back in that tiny SMD resistor.
Still, why has no one yet figured out how to overclock SHA mode? Will any other value resistors allow this? A happy medium between SHA and Scrypt overclocking? I mean, it would be great to be able to overclock dual mode if the need ever arises.
The best thing to do is add a SPDT switch for MOD IN - MOD OUT.
I believe someone else envisioned this option a while back.
Yes, that was an idea I had and even hooked up a 50k pot to test things out. I just haven't had the chance to do anything more since the gridseed I've been testing things on hasn't been the same since it blew the 3022m mosfet while trying to dual mine at too high of a voltage/frequency (47k in place of r139 and running at 1100Mhz). The real idea I had was to do the single bridge mod, and replace or even add a another resistor in parallel to get the value of 49k so removal of r139 wouldn't be necessary. This way if one wanted to run in vmodded overclocked mode for scrypt only they could simply use the voltage=1 parameter in the command line options. Omission of the voltage=1 parameter would be running it in stock mode. However....I think it is possible to overclock the gridseeds and run in dual mode, just not anywhere near the level they are being voltmodded and overclocked for scrypt only. I think 1000Mhz would be possible. So...really...there are quite a few options to explore as far as modding the gridseed to allow for this. We could simply do the single bridge, and replace the stock resistors to the voltage=1 parameter could be used, or forget about the voltage=1 parameter and use a physical switch. Obviously both methods have their merrits.....but then again we could do BOTH, and change the value of r66 (36k - which is activated with voltage=1) to 50k, and change r139 (33k - stock or voltage=0) to 38k-40k. What this would (in theory) is allow the gridseed to run at a slightly higher voltage which should let it reach 1000Mhz for dual mode (and not let out the magic blue smoke), and enable the use of voltage=1 parameter to set the VID to use r139 (49.9k/50k) to raise the voltage enough to allow 1200Mhz+ for scrypt mode. Anyone is welcome to take this idea and run with it. I'm working on getting funds together for another gridseed since I can't really determine stability on the one the mosfet blew on. EDIT: Obviously the merit of using voltage=1 would be if you don't have physical access to the gridseeds and wanted to switch from dual mining to scrypt, it could be done via software in cgminer. Hey nst6563, when did you first have this idea and when did you write about it on here? About switching between voltages using the software parameter voltage=1? That's been a while since I had the idea, but I posted about it a few days ago (apr 19th). It's not really fully my idea, but just an expansion on the original bridge mod which enabled the use of the voltage=1. Also, it was *just* an idea - I only tried to dual mine with a modded gridseed once and blew a mosfet. I've repaired the damage and left the 50k resistor at r139 and only scrypt mined since. The other couple gridseeds I have are using 50k pots at r139, and the one coming today I may try this latest mod to cut r46 and place a resistor there while leaving the stock resistors in place. |
|
|
|
*snip*
If I'm understanding this mod correctly. Reversing it will still require the same amount of work to reverse as reversing the R52 mod.
One thing that is better about this mod though is that all one has to do is remove the 17K resistor to put it back to original specs.
I for one do not look forward to soldering those two bridges again! I was happy we did away with that. It's a real pain to do, even for a master of soldering.
If all I have to do is solder on a 17K resistor to those two points, I agree that is it a better option than removing 'r139' (aka r52) and soldering in a replacement resistor only to later have to remove it then solder back in that tiny SMD resistor.
Still, why has no one yet figured out how to overclock SHA mode? Will any other value resistors allow this? A happy medium between SHA and Scrypt overclocking? I mean, it would be great to be able to overclock dual mode if the need ever arises.
The best thing to do is add a SPDT switch for MOD IN - MOD OUT.
I believe someone else envisioned this option a while back.
Yes, that was an idea I had and even hooked up a 50k pot to test things out. I just haven't had the chance to do anything more since the gridseed I've been testing things on hasn't been the same since it blew the 3022m mosfet while trying to dual mine at too high of a voltage/frequency (47k in place of r139 and running at 1100Mhz). The real idea I had was to do the single bridge mod, and replace or even add a another resistor in parallel to get the value of 49k so removal of r139 wouldn't be necessary. This way if one wanted to run in vmodded overclocked mode for scrypt only they could simply use the voltage=1 parameter in the command line options. Omission of the voltage=1 parameter would be running it in stock mode. However....I think it is possible to overclock the gridseeds and run in dual mode, just not anywhere near the level they are being voltmodded and overclocked for scrypt only. I think 1000Mhz would be possible. So...really...there are quite a few options to explore as far as modding the gridseed to allow for this. We could simply do the single bridge, and replace the stock resistors to the voltage=1 parameter could be used, or forget about the voltage=1 parameter and use a physical switch. Obviously both methods have their merrits.....but then again we could do BOTH, and change the value of r66 (36k - which is activated with voltage=1) to 50k, and change r139 (33k - stock or voltage=0) to 38k-40k. What this would (in theory) is allow the gridseed to run at a slightly higher voltage which should let it reach 1000Mhz for dual mode (and not let out the magic blue smoke), and enable the use of voltage=1 parameter to set the VID to use r139 (49.9k/50k) to raise the voltage enough to allow 1200Mhz+ for scrypt mode. Anyone is welcome to take this idea and run with it. I'm working on getting funds together for another gridseed since I can't really determine stability on the one the mosfet blew on. EDIT: Obviously the merit of using voltage=1 would be if you don't have physical access to the gridseeds and wanted to switch from dual mining to scrypt, it could be done via software in cgminer. |
|
|
|
I just had a thought for the R46 pads since this is all alone, couldnt we solder a larger resister in the place?
PANASONIC - ERJP03F1692V - RESISTOR THICK FILM, 16.9KOHM, 1%, 0603 Price For: 25
I would imagine the 0603 is larger than the 0402?
But will this fit on the pads? PLease if someone can advise as I am in a posistion to purchase the larger ones if they work as this will be much easier right?
While the resistor itself may be larger you're still restricted to the size of the pad you're soldering to. So if it's longer than the 402's it could effectively make things a little more difficult because it will be blocking the pads to solder to. EDIT: I do not know the difference in size dimensions between 0402 and 0603 other than the 0603's are larger. |
|
|
|
|
I'm also trying this build with 2 gridseeds and client side I get ~989KH/s however poolside hasn't peaked above 730KH/s and it's been running for 4 hours now. With cgminer I would see around 990KH/s clientside and around 950KH/s-1.2MH/s poolside.
Is there something I haven't configured correctly? It doesn't seem that there is much to configure...but it never hurts to ask.
EDIT: Trying a different pool to see if I get similar results.
|
|
|
|
https://github.com/siklon/usb-miner/tree/master/software/cpuminerShort summary: * Added hashrate statistics * Added auto per-chip frequency tuning * Added JSON API at port 4028 -> send {"get":"stats"} to get stats for each miner/chip * Input any frequency you want -G, --gc3355=DEV0,DEV1,...,DEVn enable GC3355 chip mining mode (default: no)
-F, --freq=FREQUENCY set GC3355 core frequency in NONE dual mode (default: 600)
-f, --gc3355-freq=DEV0:F0,DEV1:F1,...,DEVn:Fn individual frequency setting
-A, --gc3355-autotune auto overclocking each GC3355 chip (default: no) It should work on Rpi, but I haven't tested it yet. Awesome! Are there Windows binaries available? I've attempted to build them in the past but always seem to run into some errors. |
|
|
|
Hmm i think i never buy usb devices ever besides stupid mouse and keyboard.
Been 4 days busy to get anything working at all
What a freaking pain in the behind again to get anything running as it should
I got a G-Blade 80 chips but this far i only got it running on my win 7 box with the bfg 3.99 GS version
Which btw runs only 3 Mh not anything near 5.2 Mh
And it constant wants to start the not able to mine ati gpu ....... could not find a command to disable the gpu from the command line in the non existing help files
I tried the cgminer but guess what i does not work ... Sigh
Freaking zadig installed (BTW I hate zadig) , with the latest blade compatible cgminer 3.7.2 to get YES you allready know "devices not found"
I remove the crap zadig by reloading the driver for the board
started the cfgminer and its running again at its half speed pace.
I looks like its simply not able to run normal on windows, i got a pi and that is also not working besides the very old GS 3.8.5 version which starts mining yes ..... BTC .... sigh
When i try the new 3.7.2 i get YES no devices found .....
Run cgminer 3.7.2, and when it says no devices found, unplug the usb to the miner, wait a few seconds, then plug it back in. It's no secret that cgminer has this problem. cpuminer doesn't seem to suffer from this issue and it doesn't require the stupid zadig process to replace the drivers either. |
|
|
|
*SNIP* Sorry I am not understanding pots? is that just resistor? Also Are we talking about this is based on Vmod3 right? I was thinking of Software switch but again, u said another good idea "physical switch" , 2 Wires on the 2 contacts in replace of the R52 resistor and those 2 wires lead you to maybe 4 different resistor with a 4 way switch hhaha.. Stock Mode - 47K Mode - 49.9K mode and then 52K mode BOOOM we now have SuperVmod- Giving us the choice to Dual mine, BTC mine, Scrypt Mine, + Scrypt mine OC in 3 OC steps. Since I am not a electronic guy even though I have soldered a few chips and leds on mobile phones back in the Nokia phone days.. I would need detailed instructions to even attempt this. Also seems like it would be an AWESOME mod but seems like alot of shit that needs to be done... Which is why I initially thought, By doing the Jumper/short mod + changing over 1 resistor to the perfect Kohm amount.. then enables us to flick the swtich with software. Both are awesome ideas if they work , The Physical switch is a definite one that will work but = Harder to do... The Software/hardware mod is easier but Not 100% it will work. I was looking at these miners and I thought of something to save room, I was thinking of pulling apart 2 x Gridseeds and removing the fan on one, somehow attaching the heatsink without the fan to the bottom of the other gridseed with the fan , 1 fan is pretty powerfull, this saves us doing the 5v fan mod as now you are only using the 1 fan for 2, so bascially cuts the fan power in half. And ALSO saves room as you have basically mashed 2x grids together and giving the cooling from the 1 fan. WIth the mod this will make 1 double whopper grid with 10 Chips= 1000Kh +  Something like this >>> https://www.dropbox.com/s/gaykii3dlqvqfjc/mashhh.jpgEdit: even better!!>>> I just thought of making it smaller by removing the sink directly under fan in the photo and having the back against one of the boards This makes it smaller by only having 3 peices of sinks and the fan, like this >>> https://www.dropbox.com/s/8q278fnjgzd18hj/mash2.png - and maybe cooler as the fan is now closer to both boards. Not bad looking for a future 1000Kh little beast  Yes, pot is short for Potentiometer which is a variable resistor. If I had 50k worth of "pot" I wouldn't be concerned about modding a gridseed The physical switch is definitely possible. It wouldn't really be all that hard either I don't think. The software parameter is the real interesting one I think. |
|
|
|
So where on this gridseed do I find the measurement point for the voltage?
Just for the hell of it and since I couldn't sleep, I did the bridge mod, then wired up a 50k pot to the 36K resistor that corresponds to VID1 (Voltage = 1 parameter) and a 50K pot to the 33K which corresponds to VID0. Since I know someone may ask...I removed the stock resistors. I set them both to stock resistances and noticed some odd behavior. At 800Mhz it seemed that without the Voltage=1 parameter the pod would hash away, get no accepted shares, and accumulate hw errors. At the same 800Mhz speed using the voltage=1 parameter the pod would hash away, get accepted shares, and still accumulate some hw errors. Not as many as without the voltage=1 parameter, but still errors all the same.
It could be that my pod is just going to throw hw errors due to some damage when the mosfet blew...so someone else may need to give the idea a shot and see how it turns out.
*edit* - yep...even with the 2 50k pots set to the stock gridseed values of 33k and 36k, and the clock speed dropped back to 800Mhz it generates errors at a frequency of about 1 per minute. Client side hashrate shows ~350kh/s but poolside shows around 327kh/s.
Hmmm yeh i know what you mean , its hard to tell if its the mods fault or its just your stuffed up mofset Grid, hmmm I am still waiting on all my accesories to start modding. someone with awesome soldering skills would most likely have fun trying this , hopefully someone does as this will give me an idea - I beleive having a mod that allows you to choose if you want higher voltage or not would be the most ideal mod of them all. This means you do not ever have to revert back to normal all you need to do is flick the switch and boom your back to stock mode Actually...if you REALLY wanted to just "flick a switch", you could wire up a 3-way switch with the resistance values you wanted, and mount the switch to the side of the gridseed (I stuck my 50k pots to the side of the fan and used the shortest wires I could) and label it. That way POS1 = stock, POS2=mod. The way mine is, I just unplug the gridseed and adjust the 50k pots. |
|
|
|
|
So where on this gridseed do I find the measurement point for the voltage?
Just for the hell of it and since I couldn't sleep, I did the bridge mod, then wired up a 50k pot to the 36K resistor that corresponds to VID1 (Voltage = 1 parameter) and a 50K pot to the 33K which corresponds to VID0. Since I know someone may ask...I removed the stock resistors. I set them both to stock resistances and noticed some odd behavior. At 800Mhz it seemed that without the Voltage=1 parameter the pod would hash away, get no accepted shares, and accumulate hw errors. At the same 800Mhz speed using the voltage=1 parameter the pod would hash away, get accepted shares, and still accumulate some hw errors. Not as many as without the voltage=1 parameter, but still errors all the same.
It could be that my pod is just going to throw hw errors due to some damage when the mosfet blew...so someone else may need to give the idea a shot and see how it turns out.
*edit* - yep...even with the 2 50k pots set to the stock gridseed values of 33k and 36k, and the clock speed dropped back to 800Mhz it generates errors at a frequency of about 1 per minute. Client side hashrate shows ~350kh/s but poolside shows around 327kh/s.
|
|
|
|
Right now I only have the one running at 1225Mhz. I have another one coming in a few days which I haven't fully decided if I want to just do the bridge and dual mine, or do the resistor swap and only scrypt mine. I'll definitely mill out that part of the heatsink so it makes full contact with all 5 of the gc3355 chips though.
If you try the bridge mod and dual mining please let me know. I have to make that decision too. I've been dual mining for a few weeks now and like the idea of it. With dual mining I've found the frequency is set and controlled by cgminer. Setting a different frequency in cpuminer has no effect at all when dual mining that I can tell. Max stable frequency dual mining for me is 750MHz. I was thinking maybe 800MHz or 850MHz could be possible with the bridge mod. 50MHz increments up to 900MHz in cgminer 3.8.5 for Gridseeds from what I see in the code. I practiced soldering on a scrap board today and it went well. It's easier when there's no pressure, nothing to mess up. I'm really not good at soldering but all tips mentioned here by the pros really helped. Practice on scrap:  As mentioned many times in this topic and in Jesse's Official Overclocking (voltmod) Guide For Gridseeds use flux. The solder flows in to replace the flux like magic. I know I can do and undue volt mod 1. I should be able to do volt mod 3 but I'm not sure if I could undue it in case I wanted to go back to dual mining. I'd really like to know how much we can get out of these while dual mining. Here's the idea I had....and right now it's just an idea... Do the single bridge mod so that you can activate the higher voltage by using the switch voltage=1 in the gridseed options. Then change the resistor that corresponds to that VID setting (which should be the 36k as referenced here https://bitcointalk.org/index.php?topic=519112.msg5767028#msg5767028 on post #33) to the higher value you're looking for (47.7k or 49.9k or even maybe higher). Then, should you wish to run stock voltage, you simply leave out the voltage=1 parameter. If you wish to mine scrypt only with increased hashrates, specify the voltage=1 parameter and the higher clockspeed. |
|
|
|
|
Show Posts
