-Redacted-
|
|
July 03, 2013, 03:40:40 PM |
|
Actually I think people were warned to specifically not insulate the top of the chip. I can't remember where, but I distinctly remember reading that a heatsink on top of the chip can actually cause damage. Maybe that was because the pressure of the heatsink can cause chips to come loose from the board or something, but I do remember being warned not to try to exhaust heat out of the top.
The packaging of these chips is tuned to have specific thermal gradient ranges they will operate correctly under - if you actively cool the tops of the chips it can actually end up damaging the chip/ripping loose lead bonds/fracturing the packaging because the temperature differential between the top and bottom of the chip is too large..
|
|
|
|
knedle
Member
Offline
Activity: 99
Merit: 10
|
|
July 03, 2013, 06:20:35 PM |
|
So in short for all those who think about applying some more radiators on top - if it says don't do something, it means don't do something. You can do that if you want, but it's your problem if something goes bad.
|
|
|
|
Sophokles
|
|
July 03, 2013, 06:59:43 PM |
|
So in short for all those who think about applying some more radiators on top - if it says don't do something, it means don't do something. You can do that if you want, but it's your problem if something goes bad.
Nobody suggested doing this. Question was, is it possible to stack them top towards top, with heatsinks on each bottom side.
|
|
|
|
bigbeninlondon
|
|
July 03, 2013, 07:13:06 PM |
|
So in short for all those who think about applying some more radiators on top - if it says don't do something, it means don't do something. You can do that if you want, but it's your problem if something goes bad.
Nobody suggested doing this. Question was, is it possible to stack them top towards top, with heatsinks on each bottom side. Why not stack them according to the design? The front page shows a schematic.
|
|
|
|
ujka
|
|
July 03, 2013, 07:24:08 PM |
|
Why not stack them according to the design? The front page shows a schematic.
I still want to 'stack' them in a cube: https://bitcointalk.org/index.php?topic=223680.msg2362815#msg2362815In that cube fans are positioned to ensure good airflow both inside the heatsink tunnel, and over the top of the boards. It's a Klondike design, but I think it can work here, too.
|
|
|
|
grabber
Member
Offline
Activity: 67
Merit: 10
|
|
July 03, 2013, 08:03:09 PM |
|
Youmare the man! Cant wait to pick up my boards Changed that dammed chip and: SUCCESS!
|
|
|
|
SebastianJu
Legendary
Offline
Activity: 2674
Merit: 1083
Legendary Escrow Service - Tip Jar in Profile
|
|
July 04, 2013, 12:35:18 AM |
|
I would like to see the overclocking abilities before i decide if i use burnin or klondike designs.
|
Please ALWAYS contact me through bitcointalk pm before sending someone coins.
|
|
|
burnin (OP)
Sr. Member
Offline
Activity: 243
Merit: 250
ALTCOM Ab9upXvD7ChnJxDRZgMmwNNEf1ftCGWrsE
|
|
July 04, 2013, 01:14:20 AM |
|
Todays status update: Overall Project status: Working hard to get the boards into Production & final touches to the order websiteDevelopment status: Yesterdays 24hr Burn in was completed with 0,3% HW errors. Implemented Temperature monitoring today and tested Overclocking capabilities a bit. 20x 256Mhz = 5.120Ghash/s - 36 Watts 142Mhash/W 20x 320Mhz = 6.400Ghash/s - 50 Watts 128Mhash/W Talked to the guys from Anfi-Tec today to get the watercooling solution going. >>>>And for all the CAD gurus out there: 3D MODELS<<<<https://www.dropbox.com/s/wl0t2anb000tz8q/3dmodel.zip
|
|
|
|
AMD FTW
Sr. Member
Offline
Activity: 317
Merit: 250
GET IN - Smart Ticket Protocol - Live in market!
|
|
July 04, 2013, 02:04:21 AM |
|
Todays status update: Overall Project status: Working hard to get the boards into Production & final touches to the order websiteDevelopment status: Yesterdays 24hr Burn in was completed with 0,3% HW errors. Implemented Temperature monitoring today and tested Overclocking capabilities a bit. 20x 256Mhz = 5.120Ghash/s - 36 Watts 142Mhash/W 20x 320Mhz = 6.400Ghash/s - 50 Watts 128Mhash/W Talked to the guys from Anfi-Tec today to get the watercooling solution going. >>>>And for all the CAD gurus out there: 3D MODELS<<<<https://www.dropbox.com/s/wl0t2anb000tz8q/3dmodel.zipOr one could build a custom aquarium for their rigs and stack the PCB's close and use some 3M Novec 7000 coolant. I need to find a cheaper source for this as last year it was around $275 a gallon and now its like $375 a gallon. https://www.youtube.com/watch?v=Z_X_hgtlJpA
|
|
|
|
Exoskeleton
|
|
July 04, 2013, 02:59:27 AM |
|
The 3M Novec 7000 coolant with a custom enclosure has been at the top of my list of cooling solutions for a while. If we can get a board tested in it and the results are good you can count me in for 2 gallons in a group buy. That is only if a group buy would lower the price by a reasonable amount. Id be curious what the cheapest rate we could get is and perhaps if someone could get a discount for us as well.
|
|
|
|
Bicknellski
|
|
July 04, 2013, 03:25:01 AM |
|
The 3M Novec 7000 coolant with a custom enclosure has been at the top of my list of cooling solutions for a while. If we can get a board tested in it and the results are good you can count me in for 2 gallons in a group buy. That is only if a group buy would lower the price by a reasonable amount. Id be curious what the cheapest rate we could get is and perhaps if someone could get a discount for us as well.
https://bitcointalk.org/index.php?topic=248184.0Add your ideas here or just lurk... we are trying to work out a solution for cooling for overclocking using submerged oil cooling.
|
|
|
|
bezzeb
Member
Offline
Activity: 103
Merit: 10
|
|
July 04, 2013, 06:39:00 AM |
|
Actually I think people were warned to specifically not insulate the top of the chip. I can't remember where, but I distinctly remember reading that a heatsink on top of the chip can actually cause damage. Maybe that was because the pressure of the heatsink can cause chips to come loose from the board or something, but I do remember being warned not to try to exhaust heat out of the top.
The packaging of these chips is tuned to have specific thermal gradient ranges they will operate correctly under - if you actively cool the tops of the chips it can actually end up damaging the chip/ripping loose lead bonds/fracturing the packaging because the temperature differential between the top and bottom of the chip is too large.. You're on the right track, but the risk is mechanical, not thermal. We're not dealing with a huge amount of heat here like you would get with a high end CPU or GPU, and the temperature differential won't do anything exciting. (Unless you were totally bonkers and were using liquid nitrogen or something... badideadontdoit) These chips are just designed to cool from the bottom, not the top. You can't turn screws with hammers, use the right tools for the job.
|
|
|
|
tigerbit
Member
Offline
Activity: 80
Merit: 10
|
|
July 04, 2013, 07:15:15 AM |
|
Implemented Temperature monitoring today and tested Overclocking capabilities a bit. 20x 256Mhz = 5.120Ghash/s - 36 Watts 142Mhash/W 20x 320Mhz = 6.400Ghash/s - 50 Watts 128Mhash/W
Serious question to the overclockers: how do you determine your profitability? i.e. why bother if the efficiency drops off a cliff? More efficient is better, right? Thinking about it, I guess you could look at the ROI on the extra output against the power overhead. By that I mean the first 5.1 Gh/s cost you 36W, the next 1.3 cost you 14W so whilst that is profitable at electricity and difficulty levels then its worth doing. When its not, don't. But how much is sensible to invest in cooling to achieve higher rates? Is there a rule of thumb? You'd need to work out an ROI and set a profit expectation. Has anyone looked at this?
|
If you found this useful modest tips welcome BTC: 15noAopoPUcA4D4dTJihgDVE8axZh8VWia
|
|
|
rethaw
|
|
July 04, 2013, 07:35:36 AM |
|
Implemented Temperature monitoring today and tested Overclocking capabilities a bit. 20x 256Mhz = 5.120Ghash/s - 36 Watts 142Mhash/W 20x 320Mhz = 6.400Ghash/s - 50 Watts 128Mhash/W
Serious question to the overclockers: how do you determine your profitability? i.e. why bother if the efficiency drops off a cliff? More efficient is better, right? Thinking about it, I guess you could look at the ROI on the extra output against the power overhead. By that I mean the first 5.1 Gh/s cost you 36W, the next 1.3 cost you 14W so whilst that is profitable at electricity and difficulty levels then its worth doing. When its not, don't. But how much is sensible to invest in cooling to achieve higher rates? Is there a rule of thumb? You'd need to work out an ROI and set a profit expectation. Has anyone looked at this? In general, increases in difficulty affect your ROI long before electricity costs. EDIT: i.e. hashpower is worth more today than it is tomorrow. A flat 10% decrease in efficiency pales in comparison to the 10% difficulty changes 3 times a month.
|
|
|
|
sprint347
|
|
July 04, 2013, 08:59:24 AM |
|
Burnin, please confirm you will still be building 10 chip miners
|
|
|
|
tarmi
Legendary
Offline
Activity: 1218
Merit: 1010
|
|
July 04, 2013, 09:00:35 AM |
|
yes, no words about 10 chip modules, and I was planning to buy few.
|
|
|
|
sprint347
|
|
July 04, 2013, 09:07:12 AM |
|
I ordered chips for a Bitburner X 4 weeks ago.
If it turns out there will not be a 10 chip model anymore I'm screwed.
|
|
|
|
Schrankwand
|
|
July 04, 2013, 09:09:20 AM |
|
Todays status update: Overall Project status: Working hard to get the boards into Production & final touches to the order websiteDevelopment status: Yesterdays 24hr Burn in was completed with 0,3% HW errors. Implemented Temperature monitoring today and tested Overclocking capabilities a bit. 20x 256Mhz = 5.120Ghash/s - 36 Watts 142Mhash/W 20x 320Mhz = 6.400Ghash/s - 50 Watts 128Mhash/W Talked to the guys from Anfi-Tec today to get the watercooling solution going. >>>>And for all the CAD gurus out there: 3D MODELS<<<<https://www.dropbox.com/s/wl0t2anb000tz8q/3dmodel.zipAre you going to provide the water cooling solution with your shop, too? I mean, if they can get it done as fast. That would be beyond awesome, actually. I am flipflopping on your normal and the air cooling solution variation. I guess the boards will increase their price with heatsink to around 90-100€ per piece (calculated from 80€ for bare solution and 100 for full solution with fan). Add a pump, a few cables and a radiator and we are done...
|
|
|
|
Micky25
Legendary
Offline
Activity: 974
Merit: 1000
|
|
July 04, 2013, 09:12:22 AM |
|
I ordered chips for a Bitburner X 4 weeks ago.
If it turns out there will not be a 10 chip model anymore I'm screwed.
why not buy some additional chips? They are cheap at the moment and a 20 chip model will be much more efficient and cost- effective.
|
|
|
|
sprint347
|
|
July 04, 2013, 09:20:10 AM |
|
I ordered chips for a Bitburner X 4 weeks ago.
If it turns out there will not be a 10 chip model anymore I'm screwed.
why not buy some additional chips? They are cheap at the moment and a 20 chip model will be much more efficient and cost- effective. I've been trying to buy 10 chips from SebastianJu's 5th batch to complete a XX, with no success. If I order any new chips they will come from the 6th batch which have not been ordered from Avalon yet. It doesn't make sense having 10 chips 'wait around' for another 10 to come from the following batch. I'm better off selling the 10 I've ordered and start again. All I need is assurance that the 10x model is going ahead as per the first post. I've asked three times now, just need a yes/no!
|
|
|
|
|