Topic: Block Erupter USB - Overclocking/ hacking ? - page 29. (Read 168765 times)
Could small peltier units be used to cool the back plate?
It looks like having hardware errors on these erupters is pretty normal. Has anyone managed to get theirs to run 24/7 without any hardware errors? I am not sure its a heat related issue because I've attached some small heatsinks to the ASIC chip and have server fans blowing onto them, but still the occasional error.
Even with the clip-on heatsinks touching both sides, and having a fan on them, I still get just under 1% HW errors.
I noticed the 5V at the USB connector is actually 4V. The other voltages on the pads at the end of the stick are OK though. Not sure why the 5V is low, as it measures 5V on the next molex connector along (I have a 550W PSU powering my RPi rig, nowhere near the 17A the 5V line is meant to produce). Maybe something in the USB hubs is lowering the voltage. I haven't beefed up the power lines inside my hubs yet.
It looks like having hardware errors on these erupters is pretty normal. Has anyone managed to get theirs to run 24/7 without any hardware errors? I am not sure its a heat related issue because I've attached some small heatsinks to the ASIC chip and have server fans blowing onto them, but still the occasional error.
Anyone tried sticking some small heat sinks on the top?
It helped a lot on my Raspberry Pi...
It helped a lot on my Raspberry Pi...
Yes, I have three "RAM heatsinks" on top of the stock cooler plate, and one "MOS heatsink" on the BE100 chip. Used the same heatsinks for the chips in my rPI too.
Anyone tried sticking some small heat sinks on the top?
It helped a lot on my Raspberry Pi...
It helped a lot on my Raspberry Pi...
Yes, I have three "RAM heatsinks" on top of the stock cooler plate, and one "MOS heatsink" on the BE100 chip. Used the same heatsinks for the chips in my rPI too.
I think a better option would be just to dunk these in transformer oil I think that's the oil way to get enough cooling to overclock. The way these are designed I just don't think there is enough thermal dissipation though the board to overclock they struggle as is. Heat is dissipated through the bottom pad at the die. I thinks most of it is collecting around the base of the chip. The only way to cool that is oil.
And how do you cool your mineral oil? I've wondered about that for a while.
http://www.youtube.com/watch?v=qW0MbZPfmiQ
http://www.brazetek.com/24x24-water-to-air-heat-exchanger
Its funny you posted that I just brewed a batch of their brown ale yesterday. Any way any type of heat exchanger would work I may try it and see how well it transferred the heat.
hmm. so ya need a heat exchanger and some kind of pump. sounds like this is getting industrial.
I think a better option would be just to dunk these in transformer oil I think that's the oil way to get enough cooling to overclock. The way these are designed I just don't think there is enough thermal dissipation though the board to overclock they struggle as is. Heat is dissipated through the bottom pad at the die. I thinks most of it is collecting around the base of the chip. The only way to cool that is oil.
And how do you cool your mineral oil? I've wondered about that for a while.
http://www.youtube.com/watch?v=qW0MbZPfmiQ
http://www.brazetek.com/24x24-water-to-air-heat-exchanger
I think a better option would be just to dunk these in transformer oil I think that's the oil way to get enough cooling to overclock. The way these are designed I just don't think there is enough thermal dissipation though the board to overclock they struggle as is. Heat is dissipated through the bottom pad at the die. I thinks most of it is collecting around the base of the chip. The only way to cool that is oil.
And how do you cool your mineral oil? I've wondered about that for a while.
I think a better option would be just to dunk these in transformer oil I think that's the oil way to get enough cooling to overclock. The way these are designed I just don't think there is enough thermal dissipation though the board to overclock they struggle as is. Heat is dissipated through the bottom pad at the die. I thinks most of it is collecting around the base of the chip. The only way to cool that is oil.
Its 13x13x6 mm
I think there is enough space in there for one heat sink on the BE 110.
Cutting it is an option too.
I think there is enough space in there for one heat sink on the BE 110.
Cutting it is an option too.
What size is that? We are talking 5-6mm chips on this. The RPi RAM chip that sits on top of the CPU is about 12mm square.
Anyone tried sticking some small heat sinks on the top?
http://archive.benchmarkreviews.com/images/reviews/cooling/Zalman_VF900Cu/Zalman_VF900Cu_RAM_Heat_Sink.jpg
It helped a lot on my Raspberry Pi...
http://archive.benchmarkreviews.com/images/reviews/cooling/Zalman_VF900Cu/Zalman_VF900Cu_RAM_Heat_Sink.jpg
It helped a lot on my Raspberry Pi...
I've had these things for just under 24hrs. I've had one running since It came out of the mailbox. The heat sinking though these things is terrible, there has to be a 10*c difference from the top of the chips to the bottom heatsink i'd say this is going to be the limiting factor on overclocking. I bolted one straight to a heatsink without the stock one and the heatsink gets warm but there is just a small difference in temp for stock to the larger heatsink. I'd like to know the thermal conductivity of the sheet between the board and the heatsink i'm not sure if its this or the board itself.
They do run hot. I think everyone has fans pointed at them. The photos of the original batch with the spiky heatsink would have been better. What temps are you getting?
I assumed the sheet was silicone to take up any gaps between the heat spreader and the back plate due to non-flatness. Silicone is meant to be a good conductor of heat.
The clip on heatsinks I posted earlier make a big difference as they touch the chip on the top, as well as contacting the heat spreader.
I've had these things for just under 24hrs. I've had one running since It came out of the mailbox. The heat sinking though these things is terrible, there has to be a 10*c difference from the top of the chips to the bottom heatsink i'd say this is going to be the limiting factor on overclocking. I bolted one straight to a heatsink without the stock one and the heatsink gets warm but there is just a small difference in temp for stock to the larger heatsink. I'd like to know the thermal conductivity of the sheet between the board and the heatsink i'm not sure if its this or the board itself.
If your just buying in single quantities there's no way it would be worth your cost. I order from digikey a lot and I typically get my order within 2 to 3 days. I'll probably throw some 14 and 16mhz crystals in my next order I keep alot of extra stock just because its cheaper if you buy in quantity and convenient to have on hand.
Yup, min £20 for debit card orders on farnell. £12 delivery on digikey and mouser.
Maplins are useless for anything like this - nothing remotely like it in stock, and if there was, they would be about £6.99.
Maplins are useless for anything like this - nothing remotely like it in stock, and if there was, they would be about £6.99.
Check digikey or mouser or element 14 or one of the other hundreds of electronics distributers
Those crystals can be had for under a dollar in small quantities so getting the is not a problem. The crystals used are more than just a crystal it has power ground enable and clock out, a normal crystal need two caps and this has them built in. The real question is how fast can we go. I should have two of these arriving today for experimentation. I'll report back what I find.
The ones I have seen on ebay are over £3.50, there isn't a 14.0MHz one, and I can't be arsed asking each seller what the dimensions are, as they seem to lack the ability of putting it in the listing. The dimensions they list are "SMT", which is no use at all. And I don't even know if the pinouts are standard.
Adding 15% to 0.008BC a day isn't going to pay for itself. Especially with the current price drop... I might do one just for kicks.
Edit:
This one looks the part, 5x3.2mm http://uk.farnell.com/txc/7b-14-31818maaj-t/xtal-14-31818mhz-18pf-smd-5x3-2/dp/1841973 but I'm pretty sure the minimum order quantity and delivery charges will make it not worthwhile.
Those crystals can be had for under a dollar in small quantities so getting the is not a problem. The crystals used are more than just a crystal it has power ground enable and clock out, a normal crystal need two caps and this has them built in. The real question is how fast can we go. I should have two of these arriving today for experimentation. I'll report back what I find.
Jump to: