ASIC Miner Blade Performance?

[stacksmasher]

Hello all!!   I started mining a few months ago with a bunch of those little USB Block Erupters and a pair of those little 5Gh BFL Jalapenos.

I sold all that hardware when the prices started getting crazy on ebay and decided to build a "Blade" setup with the backplane and powersupply feeding all the blades.

My question is about performance, I used to crunch wayyy more blocks with the Jalapenos but the blades have a much faster clock rate. 

I have confirmed the blades are running at full speed but it seems like they are lagging behind in the number of blocks per hour.


Thanks for any advice or help you can provide.

[aznatama]

difficulty setting?

[stacksmasher]

I am talking about the number of blocks I submit to the pool. I am using bfgminer as a proxy to submit work to bitminter. I would be willing to use another proxy as bitminter was one of the easiest for the usb devices.


What pool do most people use with the blade setups?  I assume "slush?"

[aznatama]

I'm assuming you don't mean you are submitting a full block right? since you'd need about 100+ blades to do that...

Do you mean shares submitted? I know difficulty setting affects how may shares you submit.  I'm new to blades as well, and I'm waiting for mine to arrive.  Storm in NE delaying shipment... =(

[sidehack]

I mine through a stratum proxy on Ozcoin, never had any issues.

Things that affect shares per minute with blades are vardiff, hashrate and efficiency. If you're reporting a high hashrate but low efficiency, your hardware is likely returning a lot of errors - probably need to cool them better, or get higher-quality 12VDC input. It could be that your pool has your vardiff throttled lower than you expected, which means the stratum proxy only returns shares above a certain difficulty metric - this setting keeps high-hashrate miners from overwhelming server connections.


What's the numbers on your Configuration page look like?

[stacksmasher]

Both are cool as a cucumber, I have a big box fan about 1ft away and the black heat blocks are cool to the touch. Blade 1 is the obvious turd here. Should I try moving it to another slot in the back-plane?

For power I have one of those 1000 watt HP server power supplies in the back-plane.

1  Blade is:

Total MHS:   08899
Received:   0000002139
Accepted:   0000001784
Per Minute:   119.93
Efficiency:   083.45%
Up Time:   0d,00h,14m,53s

and the other:

Total MHS:   10399
Received:   0000031360
Accepted:   0000029556
Per Minute:   141.52
Efficiency:   094.24%
Up Time:   0d,03h,28m,50s

[artbatista]

Both are cool as a cucumber, I have a big box fan about 1ft away and the black heat blocks are cool to the touch. Blade 1 is the obvious turd here. Should I try moving it to another slot in the back-plane?

For power I have one of those 1000 watt HP server power supplies in the back-plane.

1  Blade is:

Total MHS:   08899
Received:   0000002139
Accepted:   0000001784
Per Minute:   119.93
Efficiency:   083.45%
Up Time:   0d,00h,14m,53s

and the other:

Total MHS:   10399
Received:   0000031360
Accepted:   0000029556
Per Minute:   141.52
Efficiency:   094.24%
Up Time:   0d,03h,28m,50s

I had the same issue with one of mine. Remove the heat sink, polish the contact surface, reinstall with heat conductive compound.

It seems that many blades have very bad heatsink contact. The sink stays cool, the chips get hot.

If you run your blade with bfgminer proxy, you'll see a very high hardware error rate, probably in the 17-20% range.

My blade showed 82% efficiency and bfgminer reported 19% hardware error rate.

After working the heatsink, 97% efficient, full speed

Your mileage may vary, and for God's sake, be carefull with the heat compound. Most of them are electrical conductive.

Art

[sidehack]

Very true. All the blade heatsinks I've seen had very rough surfaces, lots of extrusion grooves. Oftentimes the boards were raised slightly above the surface too, which hurts thermal conduction even more.

[artbatista]

Very true. All the blade heatsinks I've seen had very rough surfaces, lots of extrusion grooves. Oftentimes the boards were raised slightly above the surface too, which hurts thermal conduction even more.

Yes. I followed your suggestion on one blade and it got a lot better. I shipped the other to you.



Art

[sidehack]

Ah, I didn't even realize that was you, didn't check username when reading. Good to see you got the one working well.

[stacksmasher]

What do you think of those little aluminum heat sinks from ebay? I was thinking of buying some just to see if they would help with cooling.

[aznatama]

The 20 packs?  they fit nearly perfectly on the USB Erupter chips, and worked well for my ESB's.  I get my blades in tomorrow, but I'd imagine the chips are the same, so it should be fine.  They're cheap, so worth a shit.  However, I think something larger would be better.  Perhaps a longer heatsing that would sit atop several chips?

[sidehack]

What I've done is get the 8x8x30 heatsinks, cut 4 in half and leave 8 whole. 1.5 heatsinks will span 4 chips on the blade.

[aznatama]

I'm having a similar problem now with my blades as well.  However, the heatsinks have enough air so that they're almost cool to the touch, so it doesn't appear to be a heatsink problem. (I'm funneling in air from outside, which around 35-40F.

I guess I'll pull one apart and inspect it.

[stacksmasher]

Both are cool as a cucumber, I have a big box fan about 1ft away and the black heat blocks are cool to the touch. Blade 1 is the obvious turd here. Should I try moving it to another slot in the back-plane?

For power I have one of those 1000 watt HP server power supplies in the back-plane.

1  Blade is:

Total MHS:   08899
Received:   0000002139
Accepted:   0000001784
Per Minute:   119.93
Efficiency:   083.45%
Up Time:   0d,00h,14m,53s

and the other:

Total MHS:   10399
Received:   0000031360
Accepted:   0000029556
Per Minute:   141.52
Efficiency:   094.24%
Up Time:   0d,03h,28m,50s

I had the same issue with one of mine. Remove the heat sink, polish the contact surface, reinstall with heat conductive compound.

It seems that many blades have very bad heatsink contact. The sink stays cool, the chips get hot.

If you run your blade with bfgminer proxy, you'll see a very high hardware error rate, probably in the 17-20% range.

My blade showed 82% efficiency and bfgminer reported 19% hardware error rate.

After working the heatsink, 97% efficient, full speed

Your mileage may vary, and for God's sake, be carefull with the heat compound. Most of them are electrical conductive.

Art

What compound do you recommend?

[Africa2003]

For me it looks like lack of cooling

My stats for 3 blades in chassis, with HP server PSU and cooling with big fan:

Total MHS:   10927
Received:   0000666112
Accepted:   0000662215
Per Minute:   149.31
Efficiency:   099.41%
Up Time:   3d,01h,55m,07s

[artbatista]

What compound do you recommend?

I use this:

http://www.newegg.com/Product/Product.aspx?Item=N82E16835242038

It's supposed to electrically non-conductive, but it's aluminum based, so I'd take that claim with a grain of salt.

Art

[sidehack]

I got a great deal on a bunch of Artic Silver ceramic about a year ago, that's what I've put on everything I've worked on.