Gridseed Experiment With No Fans

[BitcoinEXpress]

I wanted to see how long a Gridseed 5 Chip would last if I over heated it in order to answer the question if I could do away with the cheap noisy fans that use about 2W by themselves.



First Test
I took 20 5 chip Gridseeds from Zoomhash, removed the fans, placed them on their sides butted against each other and ran them at 850 mhz for 48 hours with out a single problem.

The setup was no fans with close proximity and temps hitting a max of 37C, which is well within normal was the result. Ambient temps inside the workshop was 30C or ~86 F. Temps were measured with wire temp leads to a digital thermometer placed every 5 Grids and was verified with a laser temp gauge. They were barely noticeable as warm when handled by hand. There were no HW errors and they worked perfectly. Conclusion, no fans needed in normal Scrypt mining.


Second Test

I wanted to see what it would take to kill one of these from over heating. I placed two units side by side and pointed an electric heater at them. Much to my surprise these bad boys laugh at 75C heatsink temps running with no problem at 850 mhz with only a few HW errors that was tolerable. After dropping the freq to 838 mhz it cleared all HW errors. Again fans not needed in Scrypt mode at 75C!

Wanting to kill this thing I broke out a 1500 W heat gun and heated it up to 102 C on the heat sink, HW errors increased but it was still going strong. I turned down the freq to 700 and all the HW errors disappeared while running at 102 C for over three hours. Determined to "win" I cranked it up to 145 C before HW errors were 100% at any freq.  At these temps the smell of melting resin was coming from it so I thought I had killed it but after cooling off, it fired right back up and was churning out 360 khs at 37C with no fans and still is.


Conclusion

The heat sinks are massively over engineered for Scrypt operation which is why Gridseed probably chose to go with a cheap POS low static pressure noisy fan. Quality fans could easily add $15.00 to the final price and aren't needed anyway in Scrypt operation. Since I have zero interest in SHA mining with GSD I didn't test in that mode. I have no doubt that SHA mining at 10X power consumption would kill these things really quick without fans.

My suggestion to Gridseed is to create a switch, software or hardware that will allow to turn off the fans and save about 2W per GSD in Scrypt mining. The bad news for me, I have over 600 of these to remove fans from.

My next test is going to be a blade in liquid oil submersion which I will document with video.

Has anyone else tested 5 Chip GSD with no fans?


~BCX~

[1715464835]

1715464835

[1715464835]

1715464835

[1715464835]

1715464835

[1715464835]

1715464835

[1715464835]

1715464835

[FalconCrypto]

Good info BCX thanks for sharing

[BitcoinEXpress]

Good info BCX thanks for sharing

Of course, proceed at your own risk and I have little doubt removing the fans kills your warranty LOL


~BCX~

[BitJohn]

Great stuff. I'm running mine fanless for weeks now zero issues. Not brave enough to try it on the blade.

[ilic]

lol nice thread, I love the dedication you show to try and kill one of these things!

[MisterGreen]

Great test run! 

I also removed the fan from my Gridseeds and they all running fine without any hardware errors.
If they can handle so much temperature i should try to run them with 900 Mhz instead of 850 Mhz. 

[BitJohn]

Great test run! 

I also removed the fan from my Gridseeds and they all running fine without any hardware errors.
If they can handle so much temperature i should try to run them with 900 Mhz instead of 850 Mhz. 

875 seems to be the magic number with no HW errors. I do have lots of issues with miners for windows though. Using mostly CPUminer for windows sandors versions work well.

[BitcoinEXpress]

Great test run!  

I also removed the fan from my Gridseeds and they all running fine without any hardware errors.
If they can handle so much temperature i should try to run them with 900 Mhz instead of 850 Mhz.  

875 seems to be the magic number with no HW errors. I do have lots of issues with miners for windows though. Using mostly CPUminer for windows sandors versions work well.

This brings up another thing I learned via trial and error. In cgminer since the WU are reported so differently it's near impossible to get an accurate efficiency rating. Hashrate alone is meaningless. It doesn't matter if your putting out 500khs if you're spewing so much garbage that the pool can only use 50% and you net a 250 khs WU at the pool which is exactly what happened when I modded one via the instructions at Cryptoblog.

I was able to get right ~495-500 khs local with an extreme over volt and pushing the freq to 1044. Unfortunately after stabilizing at Coinotron for 90 minutes I was only getting 255 khs to the pool with 495 khs reported locally. I had similar results at wemineltc.

What I discovered was in between 838 to 875 mhz you get about 95%+ efficiency, after 875 mhz, no matter resistor modded or not, you start dropping rapidly. At 900mhz you're looking at 85% maximum efficiency and anything near 1000 is 60% or less. The local hashrate looks great but it's what the pool sees is what you get paid on.


~BCX~

[ethought]

Thanks BCX, been meaning to try this myself, you saved me the hassle 

Just out of interest are you just cutting the fan power wires?

[MisterGreen]

Great test run!  

I also removed the fan from my Gridseeds and they all running fine without any hardware errors.
If they can handle so much temperature i should try to run them with 900 Mhz instead of 850 Mhz.  

875 seems to be the magic number with no HW errors. I do have lots of issues with miners for windows though. Using mostly CPUminer for windows sandors versions work well.

This brings up another thing I learned via trial and error. In cgminer since the WU are reported so differently it's near impossible to get an accurate efficiency rating. Hashrate alone is meaningless. It doesn't matter if your putting out 500khs if you're spewing so much garbage that the pool can only use 50% and you net a 250 khs WU at the pool which is exactly what happened when I modded one via the instructions at Cryptoblog.

I was able to get right ~495-500 khs local with an extreme over volt and pushing the freq to 1044. Unfortunately after stabilizing at Coinotron for 90 minutes I was only getting 255 khs to the pool with 495 khs reported locally. I had similar results at wemineltc.

What I discovered was in between 838 to 875 mhz you get about 95%+ efficiency, after 875 mhz, no matter resistor modded or not, you start dropping rapidly. At 900mhz you're looking at 85% maximum efficiency and anything near 1000 is 60% or less. The local hashrate looks great but it's what the pool sees is what you get paid on.


~BCX~

Ah! Thanks for sharing that info, too. This will save a lot of my nerves.