X6500 Custom FPGA Miner

[fizzisist]

I want to buy a few of these, but does it make sense to buy these if i get free electricity anyway and could buy used cards instead? power is not such a factor.

Maybe heat is.

I found out the hard way that it is incredibly difficult to dispose of the heat generated by, say, eight mining rigs in a small room (window wide open to the point of being an invitation for thieves) or just three mining rigs in a small room where the window doesn't open.

Also, space might be a factor. I've never seen anyone quoting hardware based on MH/volume, but the X6500 is roughly 8x9x2.5 cm^2, or 2.22 MH/cc. Bitcowok's planned setup is 30 of them in a what looks like a standard ATX tower. That would give about 12 GH/s, and consume about 500 W. I'd say that opens up the possibility of mining in a lot more apartments, dorm rooms, etc.

[fizzisist]

Are you still offering volume discount on these units.

Yes, the bulk prices in the first post are still valid! Email sales@fpgamining.com to arrange a bulk order.

what's your shipping costs to germany?
thx

This would be a question for Cablesaurus. Email sales@cablesaurus.com, or I'll let him know to reply here in the thread.

[bitcowok]

fizzisist: do you have any of the blue heatsinks left over from the last batch? could you put them up on cablesaurus when the heatsink-less boards become available? Alternatively, where did you source them?

My arctic silver adhesive arrived a while back, just got around to doing the heatsinks on mine.

couple pics of the gluing procedure just for fun:

http://jabawok.net/gallery/d/29245-2/20120311_002.jpg
http://jabawok.net/gallery/d/29242-2/20120311_001.jpg

Results: previously 166 max board now does 180 (with 4% invalid), and previously 180 max board now does 200 (with 0% invalid).

I had to crank the fans up to 12V (were on 7V) to keep the invalids down on the under-performing board, which is consuming about 3 more watts.

[fizzisist]

Yay, glad to be in the FPGA subforum!

fizzisist: do you have any of the blue heatsinks left over from the last batch? could you put them up on cablesaurus when the heatsink-less boards become available? Alternatively, where did you source them?

My arctic silver adhesive arrived a while back, just got around to doing the heatsinks on mine.

couple pics of the gluing procedure just for fun:

...

Results: previously 166 max board now does 180 (with 4% invalid), and previously 180 max board now does 200 (with 0% invalid).

I had to crank the fans up to 12V (were on 7V) to keep the invalids down on the under-performing board, which is consuming about 3 more watts.

Yes, I have a bunch of those heatsinks. I'll set up a rev3 board with them for a test (it's nice to get actual temperature measurements instead of just guessing which heatsink works better). I'll see if Cablesaurus will sell them. There aren't a lot though, maybe just enough for the rest of your build.

Glad you got some improvements with the epoxy! I'm surprised you're still stuck down at 180 MHz with that board, though... Are both FPGAs acting up, or is one of them worse than the other? You may want to try another experiment with "tunneling" the air. If you took a piece of cardboard and angled it down to sort of funnel the air from the fan down to a smaller cross-section at the opposite end of the board, it should increase the air speed over the heatsinks. Lemme see if I can sketch that...

Ok, maybe something like this (airflow moving left to right):

[allinvain]

Maybe something like this would be ideal: http://www.silverstonetek.com/product.php?pid=257&area=en

[Vanderbleek]

I'm not sure if that would help -- I think it might just impede the flow of air (narrowing the aperture would increase back-pressure).

[PulsedMedia]

Maybe something like this would be ideal: http://www.silverstonetek.com/product.php?pid=257&area=en

designed to increase pressure. Anytime air pressure increases heat happens.
Tho not sure if it's enough to matter in this level of pressure, most likely not, and the temp increase is in in the 0.0x range if even that.

[allinvain]

Maybe something like this would be ideal: http://www.silverstonetek.com/product.php?pid=257&area=en

designed to increase pressure. Anytime air pressure increases heat happens.
Tho not sure if it's enough to matter in this level of pressure, most likely not, and the temp increase is in in the 0.0x range if even that.

I don't think it's quite that strong to actually case heat increase due to pressure/air friction. I think it's better for channeling air across the heatsink instead of blowing it about. However having said that I think you still need some airflow across the rest of the PCB - especially the underside of it. Ideally I'd like to have a giant heatsink on the underside of the PCB as well  and with a bit of airflow across both the top and the underside of the PCB (as well as the fpga heatsink of course) I think this board can run at 200 MHz with 0% errors and even in a mildly hot climate.

[Cablesaurus]

what's your shipping costs to germany?
thx

Shipping is actually calculated automatically at the time of checkout, or prior to checkout while in your cart.

All FPGAs are shipped Express via USPS from Virginia, USA (Zip Code 22827.) Depending on the package size if ordering other supplies as well, this could vary greatly so it's hard to be more specific. You can calculate automatically from the site though to double check!

I'll be posting a new link on the site, to avoid confusion with any old links, once we're ready to roll.

[fizzisist]

I finally have some hard numbers on power consumption, and it confirms great O_Shovah's results shown here.

I did these tests on an X6500 rev3, equipped with Zalman ZM-NBF47 heatsinks. Airflow was provided with an 80 mm fan at the side, just like with the standard rev2 setup, except with a humongous-ized version of those little blue ones.



I measured the current and voltage right at the board, so this excludes inefficiencies of your PSU. I also checked many of these measurements at varying times during the day, so big changes in ambient temperature and never saw a measurable difference. I suppose there would be a difference if your heatsinks weren't doing their job, though. I plan to repeat the entire set of measurements on a different board with both the same heatsinks and a different set of heatsinks to see how much variation there is between FPGAs and between heatsinks.

It's interesting to see how much of an effect the different bitstreams have, despite clock rate. Clearly, the ztexmerge bitstreams are not just faster, but more efficient.



At 400 MH/s (which most people are using apparently), the power consumption is just 16.4 W.



Interestingly, all of the ztexmerge bitstreams have a linear relation between power and clock speed, so they all have basically the same efficiency. The 200 MHz build gets 24.4 MH/J.

[PulsedMedia]

Those coolers somehow looks epic on that board! *thumbs up*

Nice results

[nbtcminer]

Nice V3 is out!! I'll take some pics on monday when I get back to work! Also I'm testing out Fizzisist's fork of the MPBM and man oh man does it look nice so far! Kudos to the X6500 team on a good job thus far!

[allinvain]

Any word on when we can expect to see them on cablesaurus, and what cooler they'll offer as stock (the zalman one?)?

[tgmarks]

Any word on when we can expect to see them on cablesaurus, and what cooler they'll offer as stock (the zalman one?)?

+1 especially for the zalman

[99Percent]

IIt's interesting to see how much of an effect the different bitstreams have, despite clock rate. Clearly, the ztexmerge bitstreams are not just faster, but more efficient.

Are you getting invalids with these heatsinks?

[tritium]

Does anyone have any experience on ordering these to the UK? price, customs, delivery etc

[shackleford]

I am planning an enclosure for 6500’s (up to 4). I want this to be very self-contained with only ports on the back for 1 USB and 1 Power and then have it distributed to cards. Is it recommended to have a powered USB hub?.  The plan is to have USB distributed with a 4 port hub internally but if it needs to be powered it will complicate things for me but should be doable.

For the power I wanted to get an adapter that splits 1 into 4 such as http://www.aartech.ca/cb-cctv-p4-power-splitter-1-input-4-output-barrel-connectors.html?utm_source=googlebase&utm_medium=googlebase&utm_campaign=export_feed

For the power I was looking at something like this ( 102w (12v/8.5A) AC-DC Power Adapter) http://www.mini-box.com/110w-12v-8-5A-AC-DC-Power-Adapter   with an adapter to change the barrel size.  I am not seeing very much that will provide the power I am looking for. I have used these adapters and power supplies from Mini-Box for other projects and I was thinking of just going with an internal power supply but it seems unnecessary and I am not sure how the power would be divvied up from it.

Still looking but I was hoping somebody may have input on the power/usb.

[fizzisist]

I am planning an enclosure for 6500’s (up to 4). I want this to be very self-contained with only ports on the back for 1 USB and 1 Power and then have it distributed to cards. Is it recommended to have a powered USB hub?.  The plan is to have USB distributed with a 4 port hub internally but if it needs to be powered it will complicate things for me but should be doable.

For the power I wanted to get an adapter that splits 1 into 4 such as http://www.aartech.ca/cb-cctv-p4-power-splitter-1-input-4-output-barrel-connectors.html?utm_source=googlebase&utm_medium=googlebase&utm_campaign=export_feed

For the power I was looking at something like this ( 102w (12v/8.5A) AC-DC Power Adapter) http://www.mini-box.com/110w-12v-8-5A-AC-DC-Power-Adapter   with an adapter to change the barrel size.  I am not seeing very much that will provide the power I am looking for. I have used these adapters and power supplies from Mini-Box for other projects and I was thinking of just going with an internal power supply but it seems unnecessary and I am not sure how the power would be divvied up from it.

Still looking but I was hoping somebody may have input on the power/usb.

Everything about your plan sounds great to me! I don't think you will need the powered USB hub. In my testing, I tend to make sure I have my hubs powered, but I have used them unpowered occasionally and everything seems fine. Here's what's drawing power from the USB cable:

FT232R (15 mA typ.)
3 x LEDs (3 x 8.5 mA)
2 x temp sensors (2 x 0.28 mA)

So, less than 50 mA. The 4 boards should be fine off of a bus-powered hub.

You're power adapter and idea to split to 4 of them seems perfect to me, too.

Good luck with your build and please share your progress with us!

[fizzisist]

IIt's interesting to see how much of an effect the different bitstreams have, despite clock rate. Clearly, the ztexmerge bitstreams are not just faster, but more efficient.

Are you getting invalids with these heatsinks?

In the last run I did with that board before I took off the heatsinks, I ran 200 MHz for 26 hours without a single invalid on either FPGA. Unfortunately, we won't be shipping units with those heatsinks. The main problem is that we can find them in any reasonable quantity. Second is that they really are a little bit too big. The two mounting holes at the top edge of the board (the power side) are just barely blocked by those huge fins, so to stack with those heatsinks would probably require filing off a little bit of the fins to make room for a standoff. Not a difficult task, but it just means it isn't really the right fit for us to be including those as the stock option. It's looking like we will most likely be going with the Logisys Deepcool style heatsinks.

That said, if you want to use those Zalman ones or some other type, we'll also offering X6500s without heatsinks, at a discount. The first ones shipping out (early next week) will be without heatsinks and, depending on demand, plan to continue offering some without heatsinks.

[julz]

Any word on when we can expect to see them on cablesaurus, and what cooler they'll offer as stock (the zalman one?)?

Well I can *see* the heatsinkless board on cablesaurus again now. It even says "In Stock"... but when I go to buy it, it says:
"Products marked with *** are not available in the desired quantity or not in stock!"
(desired quantity 1 or 10 - doesn't seem to make a difference)

Stop teasing, Cablesaurus!