Topic: X6500 Custom FPGA Miner - page 28. (Read 220050 times)

Seems like the fans are doing a good job but I will take a closer look. How do Identify which is FPGA is which on the board? The heat sinks don't feel warm at all. Could it require cooling from underneath the board?

I will probably order my next x6500 without heatsinks since I will try to find my own cooling solution.

That seems really strange, especially that both boards are exhibiting the same behavior. I never saw an inclination for one FPGA to have more invalids than the other in my tests. It's not clear in your pictures, is the fan moving less air over FPGA 0? That is the one on the left when the USB connector is pointing at you. If so, you might want to try a 92 mm fan to get a little more coverage. By the way, those new mounting holes at the bottom left and right corners on rev. 3 are spaced perfectly for a 92 mm fan.

The new board has mounting holes which should be compatible with any Northbridge heatsinks. There are two 3 mm diameter holes placed on a 59 mm diagonal. These are usually used with "pushpins." At first glance at least, it looks like all of the heatsinks on this page are compatible: http://www.frozencpu.com/cat/l3/g40/c16/s500/list/p1/Air_Cooling-Chipset_HeatsinksCoolers-Northbridge-Page1.html

nbtcminer, Alpha looks like a good company, I'll look around there. Thanks.

@Fizzisist:

Now I know you have to be cost efficient with your heatsink options but have you ever considered using any of the Alpha heatsink models? I remember using Alpha heatinks way back in the day and they were pretty darn awesome (fully copper). They now sell a pretty decent line of chipset coolers which might be worth looking at:

https://www.micforg.co.jp/cgi-local/an/wse4.cgi?webpage=cat_cse.html

That looks really nice!

We would need to know if the mounting holes (if any) will fit any coolers in order to decide whether to order these sans heat sinks.

BTW, I am now sure my heat sinks are mounted correctly on my two x6500 but I am still getting >2% invalids on one of the FPGAs, curiously only on one of each board the one assigned 0, the other one is almost nil of invalids. Would this be a miner issue, or maybe power is not arriving evenly to both the FPGAs? The overall % of invalids averages to 1.5%

Also, a quick production update. The assembly shop is just waiting on the FPGAs to arrive, which apparently have been slightly delayed. They're expecting to receive them on Thursday or Friday, and assembly will take one week from that time. That puts the arrival on roughly 3/3. Then, it'll be about a week of testing before we start sending out any boards.

We're also running into a bit of a heatsink shortage. We're looking at various options, and hopefully will have something picked out before the boards get here. If you're planning to pick up an X6500 from this batch, but you'd like to BYOH (bring your own heatsink), please let us know here so that we have an idea of how many to pack up sans heatsinks before shipping to Cablesaurus. I think we'll be offering $15 off for that option.

And finally, here's a little preview of the board:

Sorry to hear that freshzive! You might want to try li_gangyi's suggestion:


Good luck!

Have you tried Isopropyl (rubbing) alcohol to loosen the bond of the adhesive to the chip?

@99%:
Your second unit (ending in WQWH) looks like it has some very high invalids. I woud recommend reapplying your heatsink or using thermal epoxy. I think Fizzisist mentioned it earlier but you should be under 2.0% invalids or else your unit might be heating up too much. Personally I recommend using a 120mm fan with a minimum of 30 CFM. I think the 120mm fans cool every thing on the board vs the 80mm's only cool the heatsinks.

If its already attached to the FPGA chip you are pretty much up the creek. Don't try removing it unless ur  ready / willing to sacrifice the chip.

1.) What kind of heating issues are we talking about here? Bubbles under the heatsink or too much epoxy applied?

2.) What kind of performance set backs are you experiencing? Please post some stats so we can better help you.


Cheers,

nbtcminer

Is there any product that can be used to remove thermal adhesive safely from the FPGA? One of mine got mis/overapplied and it's causing some overheating issues..

First of all, thanks for the answers, O_Shovah. Everything you said is correct. Our board should have some protection for overvoltage, but it's better to stay away from that. 9V would really be ideal. And feel free to try all of the bitstreams on fpgamining.com/bitstreams. They all use the same protocol so the miner software won't know the difference.

For the point he didn't answer, #3, the truth is that I'm not sure. It seems to be very difficult to avoid a fan completely. You'd have to experiment with this to find out, but I would recommend just adding a fan and cranking up that clock.

I've been really curious about immersion cooling these guys, so please share if you happen to try this!

A a fellow german x6500 owner. Welcome to the club

Just to answer 2-3 of your questions.

Regarding Nr 1

The voltage regulators are rated as a maximum regular input of 12 V  overswings to 12V+ will expectidly reduce the lifetime of the voltage regulators.
I expect the power supply you posted to have a certain amount of ripple in its output voltage. Therfore i recommend you  to use the 9V setting of the power supply.
This will keep you on the far save side.(power supplys that output precisely 12V are incredebly expensive)

Please also have a look at one of my earlier posts https://bitcointalksearch.org/topic/m.710994
There i took thermal images reviewing heat radiation and also measured the 12 V rail power consumption.

So you would end up at a current of ~1.65 A (excluding fan) @ 9V supply voltage per board running at 400Mh/s


Regrading Nr 2

To my knowledge all bitstreams currently avaidable are compatible with all revisions of the board and i have tested them all with the Modulaer Python Bitcoin Miner and they worked.

Regarding Nr 3

On one of my modding projects we submerged a dual GPU gaming PC featuring a 750 W power supply in Mineral oil.
The passive cooling proved sufficient for all componets.
You just must not insert any parts with high voltages (ATX power supply) into the oil as it will become conductive then.

I asume it will be sufficient cooling for the x6500. But please slowly test bitstream by bitstream and control temperatures to make shure.

Hey! I got my X6500 a couple of days ago and am looking into setting everything up now. First off: Awesome product! So cool to see how you guys have put this together and continue to improve on it!

I have a few newbie questions, it would be great if someone could help me out:

1) I plan on using this power adapter: http://www.amazon.de/Schaltnetzteil-5-000-stabilisiert-60950-profitec%C2%AE/dp/B000NZO28I/ref=sr_1_2?s=ce-de&ie=UTF8&qid=1329685887&sr=1-2 (sorry, the link is in German). Would it be safe to run that one in the 12 V setting? Because somewhere it was mentioned that it was important to make sure, that it would definitely not output a voltage higher than 12 V. I don't have the equipment to check that. Is this a big risk or basically just applies if I get the cheapest of cheap power adapters?

2) Can I use one of the older (lower mhz) bitstreams for testing? Or are they somehow incompatible with the current mining software?

3) What is the highest mhz bitstream that can be safely used _without_ a fan? So just the board with the standard heat sinks in place.

4) If you run one of the higher-mhz bitstreams (180 / 200), would it be possible to cool them by simply submerging the board in a container with mineral oil? And would such a setup still require a pump or maybe a fan that cools the oil? Or do you think that it could work completely passive?

Thanks in advance!

I suppose theoretically yes, but the fans are blowing so much air convection and all that is really a very minor effect. (Just thought of a "you know you are a bitcoin miner" when you NEVER type minor correctly the first time.)

fizz,

since heat rises what about if he just flipped his setup upside down.  would that lower his temps?

99Percent, that is an awesome setup! I can't tell you how much I love seeing these photos from everyone!

That said, it looks to me like you have pretty high invalids on FPGA0 on each board, especially the second board. I would like to see under 2 %, or less. I'm not sure the cause though. You certainly have enough airflow! You might want to try re-seating those heatsinks. I've seen this happen before, where a heatsink might have been bumped during shipping or something, and there's a little bit of an air gap under it. Also check that the heatsink is centered over the FPGA.

I would do this by taking the board and placing it upside down on your table (so that the heatsinks are resting on the surface). Then, use your thumbs to apply even pressure straight down on the board right at the spot on the board under the FPGAs. You don't want to put pressure at an angle at all. Press on it for maybe 15 seconds, using a good amount of force, but be careful that you don't slip sideways because that could pull the heatsink in a dangerous way, damaging the FPGA.

If that doesn't lower your invalids, you might want to try thermal epoxy. If all else fails, maybe a very slightly reduced clock rate would keep them happy. We are very close to releasing new firmware which will allow you to change the clock in 2 MHz increments without reprogramming. By decreasing the clock just a tiny bit, you might drastically reduce invalids which would make your effective hashrate even higher. More importantly, though, invalids are probably a sign of an FPGA that is too hot, and it might shorten its lifespan.

i can buy a small dorm refrigerator off Craigslist for $50.  why not stick 10 inside?

the only thing to worry about is condensation and i've heard different things about that. does anyone know the thermodynamics behind this and if there would be condensation?  the nice thing about a frig is you can adjust the temp to whatever you want.

If I could figure out a way to fit like 5-10 in a small cupboard with a cheap netbook without overheating I'd get a bunch of em

Air is blowing down.

I was surprised to see these x6500 don't warm up that much. You can try stacking them up with fans stacked up also on the side. I think the ideal solution would be side intake fan blowing downwards much like GPU cards, using a fan connector on the board (or maybe just already built in). I hope a future revision allows for this. Also a way to daisy chain the USB ports would be nice, and the power adapters to be stackable too.

I don't know about all the stats showing on the miner software, I'll appreciate if someone can point out issues.