alh
Legendary
Offline
Activity: 1846
Merit: 1052
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
July 27, 2015, 10:00:18 PM |
|
Putting aside the various types of thermal paste and such, how do I determine the physical location of a specific ASIC based on it's number and Loop number? I have one ASIC (i.e. ASIC #7) that runs hotter than all the others. I am pretty sure that 0-3 are one board, and 4-7 are on the other. I am almost certain that ASIC #7 is the "last before the" hot air exits the SP20. I have though that when I ramp back up after the summer months here in the USA, I might consider doing a "paste job", but would like to do it on the correct one.
Ant specific guidance or diagrams on this?
|
|
|
|
sidehack
Legendary
Offline
Activity: 3360
Merit: 1859
Curmudgeonly hardware guy
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
July 28, 2015, 01:09:53 AM |
|
I'll be honest, I have nothing productive at all to add to this discussion. But I read the subject title every time I refresh the forum page and it's making me hungry. http://www.merriam-webster.com/dictionary/repast
|
|
|
|
Sweminer777 (OP)
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
July 28, 2015, 07:09:59 AM |
|
For starters, you should NOT have used anything with SILVER near ASIC and I think that is half of the problem...I also noticed that you are not running it in the enclosure it came in You should only have a very small amount of grease on the top of the chip and nothing anywhere else...I see in the data you posted that the volts are very low on 3 of the chips Use another form of thermal compound without metal in it like MX-4. Keep in mind that due to the half-assed grease job there may be leftover grease in places it was never meant to be...at least on my miners it was everywhere..the sides of the heatsinks, between the chips, everywhere...very messy and the dust loves to stick to that grease too... I used MG chemicals non conductive grease as it calls for quite a bit of it if you remove the grease between the copper shim and the aluminum during the process. I re-did a couple of machines which were throttling at .62v prior to pasting and they run a lot better now...I have pics somewhere of the process will upload when I can..the original job was horrible ![Shocked](https://bitcointalk.org/Smileys/default/shocked.gif) Are these the same settings you had prior to re-pasting and you were getting around 1400GH/s before this? Also since you are undervolting there is no reason to have such a spread from .67 start to .69 max unless you are in a hot location..if that is the case just drop the max voltage to .68 In the field "restart miner if rate below" ... what do you have in that field, and why do you have 272w for the max watts on all 4 loops when the realistic value you should have for that voltage is more like 230w per loop ![Huh](https://bitcointalk.org/Smileys/default/huh.gif) Why did you take the boards out of the original case BTW...there is a reason the airflow is over the boards in the way they are laid out..you may be cooling the top of the board ok but the back with all the power components needs cooling as well You have 22 hardware errors in half an hour...u definitely did something wrong in the process...did u get rid of the glue that was holding the heatsinks down? possibly damaging something in the process? ' I have 4 fans on it, one in front, 1 on back, one at top and one under, they are set parallel. i am on a closed envirement where sound is issue thats why i opted for taking it of the case i will post some pictures.. I have only put paste on the chips like you said, i was considering the AS5 own warning. I am running a 1000w cooler master m1000 that could be the issue as some one said here?. I bought my sp20 for 250us whit an rm1000 included, wich was more than a bargain. but now i am quite low to upgrade the PSU but im still considering getting a server psu maybe. So i guess that psu is main reason for such low voltages, it wont pull more than 700w from wall at any voltage i have tested, it's been very instable aswell on some volts.
|
|
|
|
Sweminer777 (OP)
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
July 28, 2015, 07:15:33 AM |
|
we must have gone to different schools
Im still waiting for you to re-paste the other 7 chips on your SP20 and post the before and after results Mr Scientist
Not before you leap down that cliff and let us know the gravitation acceleration that was measured. ![Wink](https://bitcointalk.org/Smileys/default/wink.gif) Some scientist you are...you should know that gravity is a constant ...want an apple to race with? ![Tongue](https://bitcointalk.org/Smileys/default/tongue.gif) Back to topic...what were your results from before and after with your 8x as conductive as MG grease...I am very curious to know your results as you actually went through all the trouble to remove 1 whole heatsink assembly...soooo what were the results from using your super duper AS5...was it 8x cooler? any numbers at all from your clearly extensive testing to back anything up or just another lame signature spam campaign hmmmm Im done ...thanks for the laugh ![Grin](https://bitcointalk.org/Smileys/default/grin.gif) Glad that you found humor in this thread. Sigs are cool, and a fact on this group ![Wink](https://bitcointalk.org/Smileys/default/wink.gif) I do not think I compared MG thermal stuff with AS5. All I did was remove one heatsink, took a pic, replaced the grease with AS5, screwed it back in. The voltage on the said chip did not rise immediately to the intended level of 0.65V because the thermal gap did not close. Then I applied epoxy adhesive and put pressure of the heatsink until the epoxy cured. It closed the thermal gap and the chip worked fine at 0.65V. Just like it did before the heatsink was removed. What I can say is that most likely AS5 has equal thermal conductivity compared to the OEM grease. 8X thermal conductivity does not imply that temps will be reduced by 8X. (Final temps also depends on the temperature differential between chip and heatsink). I think it means that AS5 will conduct 8X the amount of heat than MG paste. i did not remove the thermal greese from the blates between the heatsinks and the copper, i left that, as i read on some treadh that someone had a big issue whit that. The greese looks fine ,its alot and it keeps dripping when its hot , the boards were full of greese but didnt seem to have any bad effect, i still cleaned it, i just repasted the chip between the copper plate but nothing more than that, i put just a drip as it said that the AS5 is very think and it was really consistent , i had to press abit to get some good spreading on the chips ![Smiley](https://bitcointalk.org/Smileys/default/smiley.gif)
|
|
|
|
goxed
Legendary
Offline
Activity: 1946
Merit: 1006
Bitcoin / Crypto mining Hardware.
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
July 28, 2015, 11:18:08 AM |
|
we must have gone to different schools
Im still waiting for you to re-paste the other 7 chips on your SP20 and post the before and after results Mr Scientist
Not before you leap down that cliff and let us know the gravitation acceleration that was measured. ![Wink](https://bitcointalk.org/Smileys/default/wink.gif) Some scientist you are...you should know that gravity is a constant ...want an apple to race with? ![Tongue](https://bitcointalk.org/Smileys/default/tongue.gif) Back to topic...what were your results from before and after with your 8x as conductive as MG grease...I am very curious to know your results as you actually went through all the trouble to remove 1 whole heatsink assembly...soooo what were the results from using your super duper AS5...was it 8x cooler? any numbers at all from your clearly extensive testing to back anything up or just another lame signature spam campaign hmmmm Im done ...thanks for the laugh ![Grin](https://bitcointalk.org/Smileys/default/grin.gif) Glad that you found humor in this thread. Sigs are cool, and a fact on this group ![Wink](https://bitcointalk.org/Smileys/default/wink.gif) I do not think I compared MG thermal stuff with AS5. All I did was remove one heatsink, took a pic, replaced the grease with AS5, screwed it back in. The voltage on the said chip did not rise immediately to the intended level of 0.65V because the thermal gap did not close. Then I applied epoxy adhesive and put pressure of the heatsink until the epoxy cured. It closed the thermal gap and the chip worked fine at 0.65V. Just like it did before the heatsink was removed. What I can say is that most likely AS5 has equal thermal conductivity compared to the OEM grease. 8X thermal conductivity does not imply that temps will be reduced by 8X. (Final temps also depends on the temperature differential between chip and heatsink). I think it means that AS5 will conduct 8X the amount of heat than MG paste. i did not remove the thermal greese from the blates between the heatsinks and the copper, i left that, as i read on some treadh that someone had a big issue whit that. The greese looks fine ,its alot and it keeps dripping when its hot , the boards were full of greese but didnt seem to have any bad effect, i still cleaned it, i just repasted the chip between the copper plate but nothing more than that, i put just a drip as it said that the AS5 is very think and it was really consistent , i had to press abit to get some good spreading on the chips ![Smiley](https://bitcointalk.org/Smileys/default/smiley.gif) Did you put some Epoxy glue to the heatsink in order to make it stick back to the glued in portion and not be wobbly?
|
Revewing Bitcoin / Crypto mining Hardware.
|
|
|
Sweminer777 (OP)
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
July 28, 2015, 01:57:26 PM |
|
Did you put some Epoxy glue to the heatsink in order to make it stick back to the glued in portion and not be wobbly? [/quote] I did not remove the copper plate from the heatsink. I only put AS5 on the cores and it did well the job for them sticking to it.
|
|
|
|
goxed
Legendary
Offline
Activity: 1946
Merit: 1006
Bitcoin / Crypto mining Hardware.
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
July 28, 2015, 03:08:35 PM |
|
Did you put some Epoxy glue to the heatsink in order to make it stick back to the glued in portion and not be wobbly? I did not remove the copper plate from the heatsink. I only put AS5 on the cores and it did well the job for them sticking to it. [/quote] AS5 is a non adhesive thermal compound, it does not have good adhesive properties. There are 2 diagonally opposite screws to tighten the heatsink, ideally there should be 4, in order to apply adequate pressure and close the thermal gap. if you are not using glue on the white spots where the heatsink was factory glued, your heatsink might not have good thermal contact with the chip.
|
Revewing Bitcoin / Crypto mining Hardware.
|
|
|
Sweminer777 (OP)
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
July 28, 2015, 03:18:40 PM |
|
Did you put some Epoxy glue to the heatsink in order to make it stick back to the glued in portion and not be wobbly? I did not remove the copper plate from the heatsink. I only put AS5 on the cores and it did well the job for them sticking to it. AS5 is a non adhesive thermal compound, it does not have good adhesive properties. There are 2 diagonally opposite screws to tighten the heatsink, ideally there should be 4, in order to apply adequate pressure and close the thermal gap. if you are not using glue on the white spots where the heatsink was factory glued, your heatsink might not have good thermal contact with the chip. [/quote] My temps are fine on closed enviroment. is even better than before repaste. i was doing 80-85. Dropped to 68-69 after this, i have not glued anything, it just sits whit preasure.
|
|
|
|
goxed
Legendary
Offline
Activity: 1946
Merit: 1006
Bitcoin / Crypto mining Hardware.
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
July 28, 2015, 04:30:28 PM |
|
Did you put some Epoxy glue to the heatsink in order to make it stick back to the glued in portion and not be wobbly? I did not remove the copper plate from the heatsink. I only put AS5 on the cores and it did well the job for them sticking to it. AS5 is a non adhesive thermal compound, it does not have good adhesive properties. There are 2 diagonally opposite screws to tighten the heatsink, ideally there should be 4, in order to apply adequate pressure and close the thermal gap. if you are not using glue on the white spots where the heatsink was factory glued, your heatsink might not have good thermal contact with the chip. What was your hashing rate and chip voltages before repasting? My temps are fine on closed enviroment. is even better than before repaste. i was doing 80-85. Dropped to 68-69 after this, i have not glued anything, it just sits whit preasure. [/quote]
|
Revewing Bitcoin / Crypto mining Hardware.
|
|
|
Sweminer777 (OP)
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
July 29, 2015, 06:07:24 AM |
|
1250~ Doing about same . voltages i have been swapping, from 0.66 up to 0.74. but doing the same it wont go for more. I only got a 1000w psu so that might be the isssue. i bought the sp20 for 250us included a rm1000 psu, and i am running on an old m1000 from cooler master wich has been more stable than the RM1000. I am looking onto server psu for upgrade , not sure if i should spend my bitcoins on one or save them and work for some real money . As said previous the SP20 is not inside the original case as it was to loud, i will take photos later on the night to show my uggly setup haha ![Shocked](https://bitcointalk.org/Smileys/default/shocked.gif)
|
|
|
|
Marvell1
Legendary
Offline
Activity: 2520
Merit: 1129
Leading Crypto Sports Betting & Casino Platform
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
July 29, 2015, 03:03:45 PM |
|
so bottom line it for us here what did you use that lowered your temps AS5 then ? or MG ?
I have have a large repasting project for my Scrypt Asics , a 10-15 degree reduction in temps would be great for me as well
|
..Stake.com.. | | | ▄████████████████████████████████████▄ ██ ▄▄▄▄▄▄▄▄▄▄ ▄▄▄▄▄▄▄▄▄▄ ██ ▄████▄ ██ ▀▀▀▀▀▀▀▀▀▀ ██████████ ▀▀▀▀▀▀▀▀▀▀ ██ ██████ ██ ██████████ ██ ██ ██████████ ██ ▀██▀ ██ ██ ██ ██████ ██ ██ ██ ██ ██ ██ ██████ ██ █████ ███ ██████ ██ ████▄ ██ ██ █████ ███ ████ ████ █████ ███ ████████ ██ ████ ████ ██████████ ████ ████ ████▀ ██ ██████████ ▄▄▄▄▄▄▄▄▄▄ ██████████ ██ ██ ▀▀▀▀▀▀▀▀▀▀ ██ ▀█████████▀ ▄████████████▄ ▀█████████▀ ▄▄▄▄▄▄▄▄▄▄▄▄███ ██ ██ ███▄▄▄▄▄▄▄▄▄▄▄▄ ██████████████████████████████████████████ | | | | | | ▄▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▄ █ ▄▀▄ █▀▀█▀▄▄ █ █▀█ █ ▐ ▐▌ █ ▄██▄ █ ▌ █ █ ▄██████▄ █ ▌ ▐▌ █ ██████████ █ ▐ █ █ ▐██████████▌ █ ▐ ▐▌ █ ▀▀██████▀▀ █ ▌ █ █ ▄▄▄██▄▄▄ █ ▌▐▌ █ █▐ █ █ █▐▐▌ █ █▐█ ▀▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▄▀█ | | | | | | ▄▄█████████▄▄ ▄██▀▀▀▀█████▀▀▀▀██▄ ▄█▀ ▐█▌ ▀█▄ ██ ▐█▌ ██ ████▄ ▄█████▄ ▄████ ████████▄███████████▄████████ ███▀ █████████████ ▀███ ██ ███████████ ██ ▀█▄ █████████ ▄█▀ ▀█▄ ▄██▀▀▀▀▀▀▀██▄ ▄▄▄█▀ ▀███████ ███████▀ ▀█████▄ ▄█████▀ ▀▀▀███▄▄▄███▀▀▀ | | | ..PLAY NOW.. |
|
|
|
alh
Legendary
Offline
Activity: 1846
Merit: 1052
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
July 29, 2015, 04:16:51 PM |
|
My understanding is that Spondoolies didn't originally add a "glue bond" on the SP20, but did later in order to prevent the heatsink from rocking during shipment. I haven't heard that the glue bond was to actually improve thermal performance. It's too bad they didn't just originally use a 4-screw heatsink design.
|
|
|
|
Sweminer777 (OP)
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
July 29, 2015, 08:05:27 PM |
|
so bottom line it for us here what did you use that lowered your temps AS5 then ? or MG ?
I have have a large repasting project for my Scrypt Asics , a 10-15 degree reduction in temps would be great for me as well
I used AS5, i only put like a drop ans squished gently the heatsink back after screwing the 2 screws as it get loose, i watched on the side to see the paste to spread a bit. Re: Repasting sp20 spoondolies? Today at 04:16:51 PM Reply with quote #32 My understanding is that Spondoolies didn't originally add a "glue bond" on the SP20, but did later in order to prevent the heatsink from rocking during shipment. I haven't heard that the glue bond was to actually improve thermal performance. It's too bad they didn't just originally use a 4-screw heatsink design. Yeapp, what i understand the glue have no thermal benefits and is only for shipment requirments to prevent damage on the chips. a 4 screw thermal sinks would have been good, i was even checking for watercooling possiblilities but whit 2 holes the water blocks wont hold tight, will need to do some modifications that might not even be worth the time and money. So i will just let the sp20 run untill it dies.
|
|
|
|
goxed
Legendary
Offline
Activity: 1946
Merit: 1006
Bitcoin / Crypto mining Hardware.
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
July 29, 2015, 11:29:51 PM |
|
The chip that I removed the heatsink from, on the SP20, would not reach > 0.60V until I applied epoxy glue to the glue spots and pressed it until it stuck w/o wobble. The 2 screws do a crappy job of applying uniform pressure on the die. Air is a great thermal insulator. Ever seen a heatsink bolted on to a large die GPU with just 2 diagonal screws?![](https://ip.bitcointalk.org/?u=http%3A%2F%2Fimg.techpowerup.org%2F131117%2FIMG_8012008.jpg&t=665&c=v2t43jLEPXGR-g)
|
Revewing Bitcoin / Crypto mining Hardware.
|
|
|
philipma1957
Legendary
Offline
Activity: 4228
Merit: 8394
'The right to privacy matters'
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
July 30, 2015, 12:44:54 AM |
|
My understanding is that Spondoolies didn't originally add a "glue bond" on the SP20, but did later in order to prevent the heatsink from rocking during shipment. I haven't heard that the glue bond was to actually improve thermal performance. It's too bad they didn't just originally use a 4-screw heatsink design.
classic example of 'stupid cheap' or to be politically correct: 'pennywise pound foolish'
|
|
|
|
alh
Legendary
Offline
Activity: 1846
Merit: 1052
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
July 30, 2015, 03:55:47 AM |
|
It's kind of odd too considering all the good, and not cheap, choices that Spondoolies made on the SP20. I've seen the 2-screw diagonal heatsinks on small chips on a motherboard, but never the CPU. Maybe the chips weren't level enough on the board and it need a little "give" to avoid cracking the ASIC.
|
|
|
|
Sweminer777 (OP)
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
July 30, 2015, 11:42:47 AM |
|
It's kind of odd too considering all the good, and not cheap, choices that Spondoolies made on the SP20. I've seen the 2-screw diagonal heatsinks on small chips on a motherboard, but never the CPU. Maybe the chips weren't level enough on the board and it need a little "give" to avoid cracking the ASIC.
spoondolies have already stated that cracked chips won't reduce the perfomance, my chips were also cracked, perfomance have not been reduced. The thing as they "hang" vertically, the weight counts down onto one side, the heatsinks are really heavy for what they are. The sp20 is well build but it could have been better, small details were left out. as said previously 4 holes would have made the possibilitie for water cooling , blocks wont sit tight whit only 2 holes. It is possible to build some kind of holder but it will have to go around the entire board as you can't touch mostly any part of the board, everything is covered whit sensetive electronics. You can see that Jew enginering was aswell thinking on the economics a well build board whit crucial details left out, and those details were ment to cut down cost and maximise profit.
|
|
|
|
alh
Legendary
Offline
Activity: 1846
Merit: 1052
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
July 31, 2015, 05:22:37 AM |
|
Has anyone tried turning an SP20 on it's side? It would seem you could eliminate the "hanging heatsink" if you rotated the SP 20 so that the blades are horizontal and not vertical. It would look a bit odd, but I wonder how it would change the temps.
|
|
|
|
Sweminer777 (OP)
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
August 02, 2015, 02:43:56 PM |
|
Has anyone tried turning an SP20 on it's side? It would seem you could eliminate the "hanging heatsink" if you rotated the SP 20 so that the blades are horizontal and not vertical. It would look a bit odd, but I wonder how it would change the temps.
I did think about it, but i removed it from entire case, it's really hard to put it on something else if it's not original case, there is no kind of extra hole to put them firmly onto some plates. But from side might reduce the tension, i don't know about the heat, fan is to noisy and the case makes the noise louder, my main problem was the noisy thats why i opted for removing it from the case.
|
|
|
|
Prelude
Legendary
Offline
Activity: 1596
Merit: 1000
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
August 02, 2015, 03:34:17 PM |
|
The chip that I removed the heatsink from, on the SP20, would not reach > 0.60V until I applied epoxy glue to the glue spots and pressed it until it stuck w/o wobble. The 2 screws do a crappy job of applying uniform pressure on the die. Air is a great thermal insulator. Ever seen a heatsink bolted on to a large die GPU with just 2 diagonal screws?
Once the 2 screws are completely tightened you're saying there isn't enough pressure on the die? Would it be possible to add a washer to each screw between it and the top of the heatsink to increase pressure?
|
|
|
|
|