Topic: Cointerra Hardware Support **Unofficial - page 16. (Read 56948 times)

I have faced many problems with my machines... half board not working, 1 core not runing, water pump showing 0 flow rate.

I did some experiments over the past week, and your +- 200C problem happened to me once and only once:---   Did you switch the 3-pin cable connections of your 2 water blocks on the board?  (i.e. the bottom water block takes power from the the top, and the top block taking power from the bottom) 

The water blocks HAVE to take power from the exact 3-pin supply that it was originally shipped, that's how the board communicate with the block and check the temperature.  If you mess with 3-pin cables, the board will get confused.

Well I can say when mine overheated it read low temps 40°C. I will say this, when they overheated they shut off before I got the GUI opened up. As for throttling, mine throttles ever since firmware upgrade 0.7.6. It still throttles with temps in the 50-60's.
 

Well on the 2nd application I left the LP on the water block.  I was even able to get it all nice and smooth on there in precisely the right spots.  So on my 2nd attempt, I did in fact have LP on the blocks.  Plus as I observed, a bit of LP was already getting between the gaps on application #1.  So more LP will just make things worse.

Also, the thing is behaving the same way with LP and Noctua NT-H1.

Something has become a bit more permanently effed up now.  ;-(

Yes, I tried PL 1.  The temps are still way out of whack (200C).  Power stepping doesn't seem to affect that one chip's temp at all.  I don't think the chip is actually that hot.  I think the temp sensor on one of the chips is hopelessly effed up.  Either way, the machine is totally gimped whereas it had been my "good box" since day one.

Also, when it throttles and tries to reset, it registers NEGATIVE TEMPS for the other chips for a second or two  Methinks some math is getting messed up in their firmware.

My other chips, however, all appear in the 30s.  I'm guessing this is because they're not hashing (ie., throttled).  But can one chip throttle the other 7?  Even the 4 on the other board?

I wonder if its not flowing right since you only applied it to the chip? I would try putting it on both. Also have you tried to power step it down to 1 to see if it will run a while before it overheats?

Well, I did the LP twice and one round of Noctua today.  No luck.

I applied only to the chips, just as I did on my other machine's board.  That machine is now running better than new.

I used about half the LP just as I did on my other machine (remember I'm only doing one board per machine, not an entire machine).  If I were to use any more it would definitely flow into the gaps.  It already did this today so I used a bit less on my 2nd round of LP application.

The results on the machine I did yesterday have been great.  It's running nice and cool in the 60s.  So I thought I should deal with my other machine with the slowly rising board temp (gets into the 80s and 90s depending on various).  I still had all the parts and tools out on my bench and liked the idea of all my machines running with temps in the 60s.

Unfortunately, my local Lowes didn't have any M3 screws longer than 1/4" so I decided to use my chopstick jig again and wrap up this little project.  Managing these Cointerra rigs is getting to be a pain.

The application of Liquid Pro went smoothly, even better than yesterday.  I turned on the machine and everything seemed great.  Nice low temps in the 60s just as one might expect.  So I began cleaning up my workbench very much in a good mood.  Shortly after I noticed that only ONE of the machine's green "miner" LEDs was on.  Perhaps I forgot to plug in that (or some other)  cable, I thought.

I shut the machine down and took a look.  Nope, all the cables were in order.

Then I started the machine back up.

Both lights came on and by looking at the machine alone, everything seemed fine.

Then I went to the web interface.  This is where the bad news began.  I was showing one chip with temps in the 180s even spiking past 200C!  WTF?!?!?!?!

I immediately shut down the machine.  I had read on the CT forums about this before they took them down.  Folks were blaming it on wonky temp sensors.  I don't recall if there was ever a solution.

I cannot understand how this problem just somehow started.  Perhaps no good deed really does go unpunished as they say.

I took the water blocks off again to see if anything was amiss with the Liquid Pro.  To be honest, some did appear to spill over the edges of one or two of goldstrike chips.  So I cleaned them all up and put the blocks back on.

Started the machine and had the same problem.  Now, it was even worse.  Aside from the oddball temperature rating, CTA0 is no longer hashing ... or only occasionally.  It seems to be throttling back due to the extremely high (and likely bogus) temp reading.  For a few seconds after it throttles and tries to reset it will get a few hashes in and register some gH.  Then it throttles completely and does nothing.

I took the water blocks off again and CLEANED OFF ALL THE LIQUID PRO.  It's not so hard to get off when it's still liquid.  

I wanted to test how things might be if I went back to regular thermal past (Noctua).  So I reapplied the Noctua and put everything back together.

Started the machine again and still had the problem.  CTA0 won't hash because it thinks (maybe it's not crazy either) that one of the chips is overheating beyond the point of reason ... hence keeps throttling itself to 0.

I have no idea what's gone wrong.  One guess is that some little drop of Liquid Pro must have landed in a very inconvenient spot on the board.  I swabbed the whole thing down with no luck.  Plus I was very careful so I doubt this one.

Another idea is that these boards are just plain fragile as hell and that something got messed up by simply moving the water blocks on and off.  This doesn't explain how things appeared to be running great until the LED stopped and I restarted the machine.

Regardless, my formerly "good" machine is now my 1/2 dead machine and it's a major bummer.  I wish I had left everything alone and not even bothered with Liquid Pro now.

It seems as though the temp sensors are built into the chips and there appears to be nothing I can do to correct the issue.  Did I mention that it's a huge bummer?

Maybe I'll put my dead board up for sale on eBay.  I saw a guy parting his out not long ago and I wish I had grabbed it.

The screws work great. Especially if you're doing it by yourself. I got mine at a local nut and bolt shop. Cost me a dollar for 8 of them. Just make sure you get M3 screws.

I like the screw based jig.  Maybe I'll head out to find some appropriate screws and shroud them with some tubing this afternoon.  I've only done 1 of my boards, so I have three more eventually to do.  Only one has temps above 80C at the moment but it would still be nice to have everyone in the 60s.  Better for the chips as well as hash/watt ratio.

When did mine I bought four M3 screws 1.5 inch long cut the heads off and wrapped masking tape around each one (about 4.25mm) leaving 3/16 or so of threads showing. Screwed them in each hole and installed the water block. Removed one at a time and installed the hold down screws corner to corner finger tight. Then tightened them the rest of the way corner to corner 1/4 turn at a time.

Having applied TIM to my goldstrike chips twice now, I've noticed a few things that make the application a bit more difficult than your normal CPU or GPU rig.  I'm hoping the following observations might save some folks time and/or help their application of Liquid Pro or any other TIM.

1.  The mount points, ie. the screws, for the water block aren't the easiest to deal with.  By that I mean specifically keeping the water block stationary as you put the four mounting screws back in can be prone to more movement than you might want.  The very first time (using Noctua paste) I noticed that the water block was moving around quite a bit as I screwed it back on.  The second time, I tried to be more careful about that but found it quite a challenge.  You might even consider it somewhat of a royal pain in the neck.

2.  The washers for each screw are aluminum.  So they don't get picked up by a magnetic screw driver.  Even worse, they have a tendency to fall into hard-to-reach places as you deal with the screws and are generally just an annoyance.  The net result is that they add a bit more unwanted chaos to an already delicate process.

3.  Even though both my machines seem to have some from the factory, I do not believe there should be any TIM on the outer square "ring" as such.  This is more of an observation than anything else.  I'm referring to the metal square that immediately encloses your 4 goldstrike chips under the cooling block.

My solution to #1 was to build a jig that helped tremendously.  Here's a pic.

https://docs.google.com/file/d/0B5ki_Xeim-C2RVl3bUtNWWQ5bkE

NOTES ON THE JIG:  In the picture those are four chopsticks sharpened on each end.  Not expensive ones but not the super cheap have-to-break-them-apart ones either.  You basically just need some long round pieces of wood or metal.

To make your own, first a) pick your material to hold the chopsticks (foam, cardboard, styrofoam, wood, clay, a sponge, etc...) then b) measure the distance between the screw holes (I forget as I'm typing this) then c) draw a square on some graph paper with holes matching those distances then d) poke your sticks through and into your chopstick-holding material of choice.  Make sure you can slide each of the sticks UP and DOWN with some good friction so that they'll stay in place.  They should not slide on their own.

For #2 I just put the tiniest dab of crazy glue between the screw and the washer (NOT BETWEEN THE WASHER AND THE WATER BLOCK PLATE).  Do this with the SCREWS OUTSIDE THE MACHINE.  Cut a Qtip in half and use the stick side as an applicator if that helps.  I went out and did some shopping as I waited for them to dry.  I'm not 100% clear on the minimum time you want here.  Just make sure they're glued on and all the glue is dry.  It's not so much to bond the washer and screw.  It's more to just prevent the stupid washer from falling onto the board as you're trying to (very precisely) screw it back in place.  So you really only need the tiniest drop of glue.

For #3 I just left the outer square bare, no TIM or anything on it.  Like I said, just an observation here.

NOTE:  to remove the old TIM I found using a good solvent and purifier essential.  I use the 2 step Arcticlean and have been quite happy with it.

1.  First apply your Liquid Pro.  What an amazing material.  Behaves like solder but at room temperature and takes a LONG time to solidify.  The qtip they provide works great.  I had a small, flat, stiff, paintbrush handy just in case I had any spills over the edge of a chip.  BTW:  the brush turned out to be NOT AS GOOD as the qtip for actual application onto the chips.

2.  Then insert the long ends of the (chop)sticks into the water block's metal brackets where the screws normally go.  Your foam or whatever material you used should be on top with the water block underneath it and the motherboard below.  The sticks should be just touching the board (assuming you can't get four holes in one - see step 4 below).

3.  Have someone hold the water block up and safely away from the four goldstrike chips while you try step 4 next.  You don't want it falling down onto your nice new mirror-like application of Liquid Pro and messing it all up.  You also don't want anyone putting their fingerprints on the super clean and purified copper bottom of your block.  Care needs to be taken here.

4.  Insert each pointed stick end into the corresponding screw hole.

5.  Now you have four vertical "rails" upon which to slide down the water block and keep it in place.

6.  Slide the water block down gently into place, applying light pressure to the top as you insert the screws.

7.  Insert screws one at a time as follows.

8.  Slide ONE of your chopsticks up a few inches while keeping the other three firmly in place along with your cooling block.

9.  Insert ONE screw (carefully) into the now vacant hole.  Just finger tighten it.

10.  Slide ONE other chopstick upward, the one diagonally across from the one you just slid up and replaced with a screw.

11.  Repeat with the remaining chopsticks (and screws).

12.  Then go back around diagonally tightening the screws a little at a time until they're all snug.

I found this method kept the water block much more solidly in place as I screwed it back down.  I had my wife help me as some of the steps above are better with more than two hands.

I suspect the jig would help with general application of any TIM but I've been wanting to give Liquid Pro a try so I did.  Everything I've read suggests it is in fact better than even the best pastes.

I'm going to let it run at PL 7 for a day, then 8 then 9.  So far, at PL 7 my results are great.  

Prior to application, I wasn't even able to run at PL 7 any more.  I had to step down to 5, eeking out 1.1 th or so with my "hot" CTA spiking up to 90C at which point I'd throttle back to PL 4 for a while.  Ugh.

Now, at PL 7 my "hot" CTA1 is at 60.4 - 43.75 (used to cook up to 90 or so at PL 5).

And my former "cool" CTA0 is at 61.68 - 52.41.

I'm very happy with that improvement.  I hope it continues to hold as I creep back to PL 9, which this machine hasn't seen in about a month.  

I noticed when applying regular pastes that they only worked for a short while.

We'll see.

PS.  I did use the scotch pad ever so slightly on the copper bottoms of the cooling blocks.  I figured some tiny micro scratches might help the bonding/curing process.

Ive noticed all my machines are hashing lower then spec for the past week or two. They are 1.5Gh/s 1.5Gh/s and 750Gh/s (one miner died). Not sure why the constant drop for so long. Temps, fans, pumps all appear normal

UPDATE: Lost power for a couple hours and when it returned the machines appear to be hashing and normal speeds, I guess a reboot helped them

I did not use the scotch pad. I went full throttle after applying with no problems. I did use the provided q-tip for spreading. Put it all back together and fired up! Just be careful with the LP and take your time, DO NOT get that stuff anywhere other than on the chip surface. Good luck and report back!

When I applied LP to mine I started it at power step 8. I let it run about a day then went to power step 9. After a reset to power step 9 my temps dropped about 5°C. As for the Scotch Brite (rough pad) I didn't use it.

I went full throttle directly. Thats the only way you can see if you applied it correct or not. Running now a week without any issues since that.

Sad that CT took down their forums.  I had been reasonably active there like many others.

Anyhow, I've been nursing my two TMIV's along since April.  One continues to crank out a reasonably stable 1.6 th with decent core temps.  The other is hobbled down to 1.2 th as one core gets far too hot.  I've applied Noctua NT-H1 twice.  Worked well for a while, but then temps went back up.  Decided to give Liquid Pro a try.

Now I've got my LPro ready to go.  I'm hoping I'll see gains similar to what others have claimed here as well as a solution that keeps working (hopefully).

My question is, do I need to let the LPro cure before firing my machine back up at all?  Or should I just run it at "low" (PL 1-3) power while the stuff cures for 40hrs or so.  Or even better, can I just go right up to full throttle (PL 9)?

I know the material cures, I'm just not clear on what care needs to be taken during that process.  

The documentation provided by Cool Laboratory is both a bit short and sometimes torqued by the German translation.  They make no mention of curing, or oddly even needing to use the included "rough" pad which I'm guessing some may use to rough up the surface?  

I think I may skip that step on my Goldstrike chips though.  Maybe more something for CPU/GPUs where there is a larger smooth surface to deal with.

I was able to fix it after having it unplugged for 24 hours.  It appears that it was "heat related"  Now that the box had time to cool off....It is running "hot" but normal.   I will be replacing the paste.

OK, so yesterday I finally got some Artic Silver 5 and did some repair (did not apply Liquid Pro, yet, because noticed that once applied hard to remove.  Will do Liquid Pro if everything is fine).

1 pump stopping issue:
I have faced this issue for a few of my miners.  Just, 1 pump got 0 pump speed reading.
- I replaced the thermal gel with Artic Silver 5.  No use;
- I remove the white plastic inside copper.  No use;
- I switch the power between front water block and rear water block !!! it runs !!! for 30 seconds ... then it gets EXTREMELY hot and shut.

Conclusion:
- since switching the power supplies between the two water blocks will trick the board in believing that the faulty core is receiving a functional water block (but indeed not), it will start runing, but due to heat, will shut down very soon.

Further Trial (please recommend water block replacement):
- this has greatly narrow down the problem to : If you have 1 out of 4 core not hashing, most probably, the water pump on that core is burnt / broken !!!  I think replacing the water block may fix the problem easily.  Can someone recommend a suitable common water block which could be bought from local computer store easily???  I will try replacing the water block, and provide you guys with the result once tested.

Nothing plugged into the board and it still doesn't work. No leds and the orange led lights up on the power supply.

 Unplug one pump and see if it works, if not plug it back in and unplug the other pump and test again. I have seen 2 users that by merely pulling the pump power, half a board works.