Topic: [GUIDE] GridSeed 5-Chip USB, Blade & Black Miner Support/Tuning - page 48. (Read 308807 times)

Would something like this work? http://goo.gl/Ri1ooy

I have several units that were tightened enough from the factory to rip the thermal pad and that in turn is causing HW errors.

when i opened my miners i noticed that even some of the capacitors are a little bit higher than these chips. as these components are located on different sides milling need to be done in more places in my case, maybe in 2-3. i think i'll skip that and just use 5 thicker 2mm pads. the original pad is a real shitty one..


here's one sample how components may be touched with heatsink:

https://i.imgur.com/D9uyjjP.jpg

Thanks...and a milling machine would be ideal, but I don't have access to one. 

Nice solution nst6563, I've a milling machine so I can easily make a precise groove in the heatsink  I'm almost sure that the oscillator don't need cooling but I'm usure about the  U20 6pin chip since I didn't found the datasheet.

Probably simplest way is to briefly turn on the gridseed touch the chip and see if it run hot...if remain cool I'll mill all the heatsinks

Good job, buddy...!

I was too lazy to grab the Dremel in my garage...just bought some XBox shims...it is easier IMHO...

Edit...: in my experience it is NEEDED... 0 (zero) HW errors after thermal mod...stock to 1000MHz ... or Volt Mod... to 1150MHz...

Everybody is free to do it or not...@ his own risk...

ZiG

Here's my solution to the uneven contact of the heatsink for thermal bliss.  I simply used a dremel and a standard cutoff wheel and cut out a small section where those two offending components are.  The cut is about 3/4 the thickness of the cutoff wheel.  Now the base of the heatsink comes in complete contact with all 5 GC3355 chips, and the other two components make contact with the space I cut out.  I applied Arctic Silver Ceramique that I had on hand to the GC3355 chips, and a thermal pad for the other two components.

Realistically that groove can be MUCH smaller, however with a dremel cutoff wheel it had to come out in a half-circle shape from using the flat/top of the cutting wheel to make the groove (I didn't use the edge of the cutting wheel so the groove in the heatsink would be flat and not concave).  An easy way to mark the spot to mill out would be to put a dab of thermal paste on the top of those two components, set the heatsink down, then lift it back off.  I then marked the area with a pencil.

After cutting, I put a very small dab of thermal paste on the 5 gc3355 chips and test-fitted the heatsink.  It took 2 tests to get the groove deep enough, cleaning the thermal paste between each test.  Once I was satisfied it was making a good flat contact with all 5 I put a final application of thermal paste - very small dab- and screwed it all down.



I realize that it most likely didn't NEED this done....but why not?  It should have come from the manufacturer designed right in the first place.

The gap is created by the 2 small square components on the top of the picture...one is a quartz oscillator, which is taller...maybe like 1mm...it's marks are clearly visible on the original pad...


Today I posted my thermal solution to this thread...:

https://bitcointalksearch.org/topic/m.6205502


BTW, the blog, cited above is almost literally copying my original post about heat and thermals in the same thread...:

https://bitcointalksearch.org/topic/m.6027770


Hope this info is helpful...

ZiG   

Even with voltmod you don't need to do anything to the heatsink or to the thermal pads, if you are not dual-mining. Voltmodded it's still unlikely to exceed 10W, this heatsink is designed for 70W in sha256 mode. I have one running on my desk at 975 MHz without the heatsink and it doesn't seem to mind at all (chip surface below 40C at 22C ambient).

Infact the original pad is very thick (over 1mm), probably gridseed noticed the unleveled components and used that pad just to play safe.
I don't say that all modules have that issues, I've checked just on 2 and both where not leveled...

As you said jamieb81 a better/thinnest thermal pad could work...but thermal paste would have been much better

@suchmoon : Yes you right, mine too have some uneven deep marks on the pad...obviously I was interested to improve the dissipation due to the voltmod  

Don't bother. If you have decent airflow the heatsink doesn't matter (in scrypt-only mode). It will work at 850 MHz as is or even without heatsink at all. Overlocking beyond 850-900 is NOT possible with thermal modifications like this. You need extra voltage for that.

By the way when you take the heatsink off the chip imprints on the thermal pad show that you won't be able to make it up with thermal paste, at least mine had very clearly different depths.

I always suspected that that would not work, otherwise they probably would not have used pads in the first place.

Maybe you could use the stock pad, and put a little thermal paste in the 5 squares that formed in the stock thermal pad so that

it has better contact, maybe that would work better.

Or probably even better, just buy some new pads that are better.

Someone read the article about removing the thermal pad and put the paste in order to have better overclock?  It's this one: http://gridseed-blog.com/gridseed-overclocking-1000-mhz-without-voltmod/

Well, I suggest to avoid it because as I've thought the 5 chips and some others component aren't perfectly on the same plane, so the result is that one or more chips will not be even in contact with the thermal paste...WATCH OUT!

You can cleary see from this photo:


I've evenly tightened the screws more that one time but still there is a gap on most of the chips, in this case the U20 6pin chip is the culprit, it's a bit more high than other components so the heatsink surface touch only that chip.

Check /var/log/syslog for a pthread warning.  Cgminer is horrible with these gridseeds.

Check your USB hub, its power supply, USB cables. Swap parts out if you have spares, or start with a lower number of Gridseeds (5 or 10) and gradually add more until they start going off.

Has anyone had this problem???

My Gridseed IDs keep changing while mining on this one Rig. I have another rig of 20 with the exact same Scripta image (1-1). I followed the basic setup instructions for scripta+cgminer in start of this tread to the tee. 2 of my 3 gridseed rigs work fine, this one keeps doing this.

I don't get it.

Hi, I need some help please.
I'm running one GridSeed with Cgminer and I used Zadig to replace the driver.
I changed the conf and I let it run, but now I'm having problem because even if it is running at 361 kh/s it is not accepting shares!

I waited one hour and nothing new.

Does someone had the same problem?

Thank you

I've had it hashing without any issues for 2 days now. I only stopped it to change out a power supply.
And you will notice I only posted for the other user to use it for diagnostic purposes. Then to use the data to either qualify his/her theory that he/she had a defective unit or to use the data to use in CGminer.

BFGminer 3.99 is noted to be a real pain in the A at the moment, very unstable..

1a. Use BFGminer 3.99 with Gridseed support for Windows on 1 or all your miners.
1b. Scrypt only at 800.
2. Let it run on a pool that does not switch coins. Run for at least 1 hour then look at the Hardware errors (HW). In the Manage Devices, you can see which chip on which miner is giving you the hardware errors. Make note of each miner's ID and Chip ID that has the errors. Stop BFGminer and reset your miners for a few seconds (unplug and power off). Turn them all back on and reconnect.
3. Restart BFGminer 3.99 with 25 steps lower (in the first test of 800 minus 25 would be 775 for the second test so on and so forth until no HW errors exist).
4. Run for at least an hour (same pool as before) and make note of each device and any Hardware errors on which chips (if any).

Repeat steps 2, 3 and 4 until you have no errors on any chips on any device. You should now have a detailed list of your max settings for each unit you have. You can then run CGminer with each individual unit's max setting.
You may be able to do the diagnostic with CGminer also but I've only worked with BFGminer since day 1 of mining.

I have the hardware ID on each miner so I know which one is which rather than by com ports.
Also make sure the USB plugs are snug and flush on both the miners and hub/interface. If it wiggles around in the port, you can bet you won't get any good results from it.
And if you do have a defective unit, you now know which one it is and have test results to back up your claim.
FYI, I have some in the 400's and some in the 290's on 850. And not a single HW error after 6 hours of mining.
I plan on looking at the ones in the 290's to see what chips are under performing and disabling them or upping the frequencies on those to see if it jolts it back to where it should be.

yes: i physically disconnected and reconnected all of them (both power and usb) and now all are hashing as they are supposed to. strange.