Block Erupter USB - Overclocking/ hacking ?

[ktbken]

mjgraham

Glad it worked
What resistor are you using with the 17Mhz oc ?
I feel more soldering in my future 

[mjgraham]

Well I am not sure, was running it from a lab supply while I was experimenting, I remember it was right around 1.4v although it was over 3A so it wont work the the current regulator, however now that it works I might not need so much voltage before when the software was not working I spent a lot of time "trying everything" finally just left it. Monday I plan to do a better job working on voltage/current/speed parameters. Been running 4 hrs or so.
 
A:567 R:0+0(none) HW:14/.61%

[Bluestreak66]

Friedcat said at higher speed there is a chance that the micro is not capturing the nonce which would explain why speeds may not be as high as they should. I'm hoping he may post some more info on here.

Well I am not sure, was running it from a lab supply while I was experimenting, I remember it was right around 1.4v although it was over 3A so it wont work the the current regulator, however now that it works I might not need so much voltage before when the software was not working I spent a lot of time "trying everything" finally just left it. Monday I plan to do a better job working on voltage/current/speed parameters. Been running 4 hrs or so.
 
A:567 R:0+0(none) HW:14/.61%

I never tried 17mhz but I have a 18mhz installed and it will not run @ below 1.4v.

[mjgraham]

Friedcat said at higher speed there is a chance that the micro is not capturing the nonce which would explain why speeds may not be as high as they should. I'm hoping he may post some more info on here.


I never tried 17mhz but I have a 18mhz installed and it will not run @ below 1.4v.

Well a 17 does not exist most likely I am using a signal generator, however I am going to have a nice chart to post, so far I am at 20 Mhz however were getting up there on voltage and heat density is making me worry. I can tell you so far were looking at 1.50V @ 3.992A for 20 MHz. I started at 12 MHz and made a chart with the common available oscillators and worked with each one until no HW errors in 5~10 min, long term hard to say. Should be done testing by tomorrow, maybe the fan wont quit in the middle of the night and a fire break out  Although so far at this test there are 4% HW errors. Is the maximum that everyone believes 1.8v? I think it is going to take a peltier to go to much farther then that adds another level of something to mess up.

As for the 1.4V at 18 MHz that is right on with what I have at that point.

[Bluestreak66]

Friedcat said at higher speed there is a chance that the micro is not capturing the nonce which would explain why speeds may not be as high as they should. I'm hoping he may post some more info on here.


I never tried 17mhz but I have a 18mhz installed and it will not run @ below 1.4v.

Well a 17 does not exist most likely I am using a signal generator, however I am going to have a nice chart to post, so far I am at 20 Mhz however were getting up there on voltage and heat density is making me worry. I can tell you so far were looking at 1.50V @ 3.992A for 20 MHz. I started at 12 MHz and made a chart with the common available oscillators and worked with each one until no HW errors in 5~10 min, long term hard to say. Should be done testing by tomorrow, maybe the fan wont quit in the middle of the night and a fire break out  Although so far at this test there are 4% HW errors. Is the maximum that everyone believes 1.8v? I think it is going to take a peltier to go to much farther then that adds another level of something to mess up.

As for the 1.4V at 18 MHz that is right on with what I have at that point.

Awesome, Good to see results are matching up. Are you getting the correct hash rates as you increase speed? As I said in one of my first post, I think heat dissipation will be the limiting factor. Also we need to come up with a better solution than a lab power supply to power these with. There are some cheap dc/dc converters on ebay I'm curious if they would provide a clean enough output to power the chip or several in parallel. I had a 70w peltier on mine running 16 and then 18mhz, It worked great until condensate formed the it locked up and the thing froze over.

[mjgraham]

Yes power is going to be a problem (well one problem) , I found a replacement regulator for the one on the board , going to need a new inductor also. The new regulator has a ground pad so that works out have to take off the silk screen but it *should* be a drop in replacement, the inductor not sure yet. I have been running it at 20 MHz for 5 hours now and the stats are
A:641 R:1+0(.16%) HW:39/1.6%
The bad thing that is at 1.525 V and 4.004A so it would be max on the new regulator, most all the other regulator chips either have the pad as output or different pin out. If the chart holds up were going to be at 1.75V and 5.5A for a 24 MHz run, if it will. I am going to put a heat sink on top of the chip tomorrow, running about 137 F now on top.

[vgo]

Yes power is going to be a problem (well one problem) , I found a replacement regulator for the one on the board , going to need a new inductor also. The new regulator has a ground pad so that works out have to take off the silk screen but it *should* be a drop in replacement, the inductor not sure yet. I have been running it at 20 MHz for 5 hours now and the stats are
A:641 R:1+0(.16%) HW:39/1.6%
The bad thing that is at 1.525 V and 4.004A so it would be max on the new regulator, most all the other regulator chips either have the pad as output or different pin out. If the chart holds up were going to be at 1.75V and 5.5A for a 24 MHz run, if it will. I am going to put a heat sink on top of the chip tomorrow, running about 137 F now on top.

Great job, thanks mjgraham.

[Lucko]

So you need to run thous with cgminer 3.1.1 and use --icarus-options 115200:1:1 --icarus-timing short. Can you run it on newer?

[mjgraham]

So you need to run thous with cgminer 3.1.1 and use --icarus-options 115200:1:1 --icarus-timing short. Can you run it on newer?

I have not tried it, I am using bfgminer 3.1.4 using the "-S all" option seems to be automatic after that point.

[Josue]

Yes power is going to be a problem (well one problem) , I found a replacement regulator for the one on the board , going to need a new inductor also. The new regulator has a ground pad so that works out have to take off the silk screen but it *should* be a drop in replacement, the inductor not sure yet. I have been running it at 20 MHz for 5 hours now and the stats are
A:641 R:1+0(.16%) HW:39/1.6%
The bad thing that is at 1.525 V and 4.004A so it would be max on the new regulator, most all the other regulator chips either have the pad as output or different pin out. If the chart holds up were going to be at 1.75V and 5.5A for a 24 MHz run, if it will. I am going to put a heat sink on top of the chip tomorrow, running about 137 F now on top.

Thankss¡¡¡¡¡¡¡¡ Waiting for your numbers.

[pr9me]

Following this!

Quite interesting...as I have been contemplating doing this myself. Your results will be initiative I need to attempt or not attempt this.

Good luck!

[mjgraham]

Well didn't get as far as I wanted to today, I am running at 22.12 Mhz 1.65V and 4.8A going to let it run tonight, I want to get to 2x speed, I am worried thought the next step in the graph shows the voltage being 1.8V, Heat is the main problem, or I guess the area the heat comes out at, I mean the P4 heatsink I am using is rated for 125W and I am putting 8W in so it only raises 2 deg. F with the fan on but the area under the chip is quite hot. So far I really have tried to keep all as stock on the cooling, I have not removed the solder mask and using the stock thermal pad. I did try a peltier today but I feel that has to be watched to closely, something goes wrong you got either a fire or ice ball. Although I don't know if you could cool the heatsink the BE was mounted to if that would help. At any rate 1 Hr at 22.12 Mhz I have.  

A:130 R:0+0(none) HW:11/2.2%

first time I have seen this

[2013-08-27 15:31:58] Staged work underrun; increasing queue minimum to 2

Maybe I will be done tomorrow.
Thanks

Update. Latest test run


Summary of runtime statistics:
                   
 [2013-08-27 20:06:00] Started at [2013-08-27 14:42:40]                                 
 [2013-08-27 20:06:00] Runtime: 5 hrs : 23 mins : 20 secs                   
 [2013-08-27 20:06:00] Average hashrate: 631.1 Megahash/s                   
 [2013-08-27 20:06:00] Solved blocks: 0                   
 [2013-08-27 20:06:00] Best share difficulty: 10.1k                   
 [2013-08-27 20:06:00] Share submissions: 702                   
 [2013-08-27 20:06:00] Accepted shares: 701                   
 [2013-08-27 20:06:00] Rejected shares: 1 + 0 stale (0.00%)                   
 [2013-08-27 20:06:00] Accepted difficulty shares: 2792                   
 [2013-08-27 20:06:00] Rejected difficulty shares: 4                   
 [2013-08-27 20:06:00] Hardware errors: 63                   
 [2013-08-27 20:06:00] Network transfer: 462.6 / 110.7 kB  ( 23.9 /   5.7  B/s)                   
 [2013-08-27 20:06:00] Efficiency (accepted shares * difficulty / 2 KB): 9.97                   
 [2013-08-27 20:06:00] Utility (accepted shares / min): 2.17/min
                   
 [2013-08-27 20:06:00] Unable to get work from server occasions: 0                   
 [2013-08-27 20:06:00] Work items generated locally: 6782                   
 [2013-08-27 20:06:00] Submitting work remotely delay occasions: 0                   
 [2013-08-27 20:06:00] New blocks detected on network: 35
                   
 [2013-08-27 20:06:00] Summary of per device statistics:
                   
 [2013-08-27 20:06:00] ICA0       | 5s:374.2 avg:630.9 u:618.1 Mh/s | A:701 R:1+0(.14%) HW:63/2.2%                   
 [2013-08-27 20:06:00]                   

[devalo]

             
 [2013-08-27 20:06:00] Average hashrate: 631.1 Megahash/s                   

Well, THAT good Sir, is impressive! Congrats!

[Trillium]

[2013-08-27 15:31:58] Staged work underrun; increasing queue minimum to 2

That's to do with the 'queue' parameter in the miner client. Either set it to a very high number (~1000) which will use a lot of CPU (relatively) or set it to zero (0).

It is not really a 'error', it just seems to be with the rate the client goes through work units.

[mjgraham]

Well here are the stats at 24 Mhz. I cant imagine I am going to get much more w/o damage to it. 1.825V 5.748A. If we knew the point the gates burned up would be nice. Also I know lower is better what is an acceptable cutoff on hard ware errors?


Summary of runtime statistics:

 [2013-08-28 11:55:57] Started at [2013-08-28 08:52:58]
 [2013-08-28 11:55:57] Runtime: 3 hrs : 2 mins : 58 secs
 [2013-08-28 11:55:57] Average hashrate: 678.0 Megahash/s
 [2013-08-28 11:55:57] Solved blocks: 0
 [2013-08-28 11:55:57] Best share difficulty: 8.31k
 [2013-08-28 11:55:57] Share submissions: 1597
 [2013-08-28 11:55:57] Accepted shares: 1596
 [2013-08-28 11:55:57] Rejected shares: 1 + 0 stale (0.00%)
 [2013-08-28 11:55:57] Accepted difficulty shares: 1596
 [2013-08-28 11:55:57] Rejected difficulty shares: 1
 [2013-08-28 11:55:57] Hardware errors: 82
 [2013-08-28 11:55:57] Network transfer: 312.0 / 201.6 kB  ( 28.4 /  18.4  B/s)                
 [2013-08-28 11:55:57] Efficiency (accepted shares * difficulty / 2 KB): 6.36                  
 [2013-08-28 11:55:57] Utility (accepted shares / min): 8.73/min

 [2013-08-28 11:55:57] Unable to get work from server occasions: 0
 [2013-08-28 11:55:57] Work items generated locally: 3818
 [2013-08-28 11:55:57] Submitting work remotely delay occasions: 0
 [2013-08-28 11:55:57] New blocks detected on network: 19

 [2013-08-28 11:55:57] Summary of per device statistics:

 [2013-08-28 11:55:57] ICA0       | 5s:400.5 avg:677.8 u:624.4 Mh/s | A:1596 R:1+0(.06%) HW:82/4.9%
 [2013-08-28 11:55:57]


Well decided to try for 25 MHz needless to say I got to 1.875V and things went a little off. But going back it went back to working, I am sure I am right on the edge of smoke, unless someone knows how much voltage they can stand, I may have pushed my luck far enough in this run.

[ibminer]

Well here are the stats at 24 Mhz. I cant imagine I am going to get much more w/o damage to it. 1.825V 5.748A. If we knew the point the gates burned up would be nice. Also I know lower is better what is an acceptable cutoff on hard ware errors?

Summary of runtime statistics:

 [2013-08-28 11:55:57] Started at [2013-08-28 08:52:58]
 [2013-08-28 11:55:57] Runtime: 3 hrs : 2 mins : 58 secs
 [2013-08-28 11:55:57] Average hashrate: 678.0 Megahash/s
 [2013-08-28 11:55:57] Solved blocks: 0
 [2013-08-28 11:55:57] Best share difficulty: 8.31k
 [2013-08-28 11:55:57] Share submissions: 1597
 [2013-08-28 11:55:57] Accepted shares: 1596
 [2013-08-28 11:55:57] Rejected shares: 1 + 0 stale (0.00%)
 [2013-08-28 11:55:57] Accepted difficulty shares: 1596
 [2013-08-28 11:55:57] Rejected difficulty shares: 1
 [2013-08-28 11:55:57] Hardware errors: 82
 [2013-08-28 11:55:57] Network transfer: 312.0 / 201.6 kB  ( 28.4 /  18.4  B/s)                
 [2013-08-28 11:55:57] Efficiency (accepted shares * difficulty / 2 KB): 6.36                  
 [2013-08-28 11:55:57] Utility (accepted shares / min): 8.73/min

 [2013-08-28 11:55:57] Unable to get work from server occasions: 0
 [2013-08-28 11:55:57] Work items generated locally: 3818
 [2013-08-28 11:55:57] Submitting work remotely delay occasions: 0
 [2013-08-28 11:55:57] New blocks detected on network: 19

 [2013-08-28 11:55:57] Summary of per device statistics:

 [2013-08-28 11:55:57] ICA0       | 5s:400.5 avg:677.8 u:624.4 Mh/s | A:1596 R:1+0(.06%) HW:82/4.9%
 [2013-08-28 11:55:57]

Nice Job... can't wait to try these mods out.  Waiting on my crystals & resistors, but not really planning on attempting higher then 20Mhz
You have any pictures of your setup?... would be fun to see  

[mjgraham]

OK here is the chart I have been working on, green areas are possible with stock regulator , yellow I think I have found a replacement but going to have to change inductor also and red is external only. Now this is just for reference this is what worked for me, I did not go to far on the cooling aspect, no soldermask removal or anything other than the stock thermal pad. I am amazed that this chip can even handle this much power.



Overview of the setup, not much fancy, 12 AWG wires from the supply to the board.


Close up, see the power and clock input, temp sensor in the heatsink.


Signal generator.


This had been pretty interesting , would love to make a schematic of the device and make a new board, a lot of "if onlys"  
Hope this data can help someone.

[easynote]

And a DIY for Nooooooobs?

[viriat0]

And a DIY for Nooooooobs?

x2, Need more Hash!!

[sgrunger]

I wouldn't mind going over 500
Someone earlier stated, 16.384mhz, 1.2kohm R1 gets you 440, is there an easy 2 part replacement to get to 500+ ?

Edit - Ah just saw the chart. So, still only two parts to change, but you have to give it an external power source and ground ?