X6500 Custom FPGA Miner

[nbtcminer]

Just switch the 5v(red) and 12v(yellow) cables to the molex connecter you're powering the board with.

@Gar255:

Just tried that and it didn't work .. Upon reading back to page 30 of this thread, I remember that Fizzisst or TheSeven talk about how the Rev 2.0 has its 5v rail connected to the 12v barrel connector (I guess it ramps down the power?)

[TheSeven]

I think we're confusing the different kinds of Molex connectors here. I'll call them "HDD molex" and "fan molex" to tell them apart.

There are no voltage regulators in between the supply and fan pins, just some fuses that you don't need to worry about.
The fan molex connector is directly connected to the barrel and HDD molex connector (5V pin of the latter on rev. 2, 12V pin on rev. 3).
I.e. the fan will always be running at whichever voltage you supply the board with.

[nbtcminer]

I think we're confusing the different kinds of Molex connectors here. I'll call them "HDD molex" and "fan molex" to tell them apart.

There are no voltage regulators in between the supply and fan pins, just some fuses that you don't need to worry about.
The fan molex connector is directly connected to the barrel and HDD molex connector (5V pin of the latter on rev. 2, 12V pin on rev. 3).
I.e. the fan will always be running at whichever voltage you supply the board with.

@TheSeven:

Thanks for the speedy reply; I was confused about whether or not the fans were being supplied with 5v on the rev 2.0. Time to install my Silenx fans

[fizzisist]

Maybe the board schematic would be helpful for you? As TheSeven said, the "12V" line, which is really whatever voltage you are connecting to the power input, is simply connected to the two power connectors and the fan connectors.

As for you fan question, the board consumes about 18 W / 12 V, or 1.5 A. The fans consume 2 * 1.56 W / 12 V = 0.26 A. Yep, should be good with a 2 A regulator.

[tgmarks]

Do you have a board schematic for the rev 2 posted anywhere?

[TheSeven]

Do you have a board schematic for the rev 2 posted anywhere?

It's basically identical to the rev. 3 except for two differences:
- No molex fan connectors
- 5V and 12V pins of HDD molex connector swapped

[shackleford]

Here is my mineral oil setup I have been working on a long time. I  had a bunch of different problems and spent way more money on it then I care to admit.. My goals were for it were to be silent, clear (cosmetic), expandable, designed for convection. I like silent PCs and the last year of GPU mining was torture.
The basic design is a chimney column with thick material in the center with an outer shell with a thinner material and more surface area to try to dissipate the heat. I knew cooling could be an issue so the plan was to have additional cooling options for the “lid”.

The mining boards are secured to acrylic planks that fit into grooves and were supposed to act as “blades” with a power splitter and USB hub included on each blade and 3 6500’s and there are slots for four (so 12 6500’s in theory). The design had changed on the fly so it is not perfect, if I were to do it again I would have room for 4 on each blade and more room for the power connectors because the dimensions got pretty tight and I had to split off the USB hub (might get it back on though) and the power adapters cut into the other planned blade slot. It is connected to my mod Apple G4 Cube/AMD Brazos MiniItx Windows7 Media Center in my living room with a long USB cable. This computer was already on 24/7 and is low power.

When finished it took about 4 gallons to fill. When pouring in the Oil a loose cable from the 1 unused power adapter shorted out the bottom board. Fizzisist was very kind and repaired the board for me and switched my resistors for the voltage. Now with all three boards and a good ambient temperature I think my temperatures are decent. I am running the 200mhz bit stream with zero invalids. The overclocker bit stream gives me a ton of invalids at 200mhz for some reason so I don’t use it. The oil seems too thick to have a strong convection current and I thought that might be a problem so the internal lid has a 140mm fan for active convection and you can visibly see the current going from the top of the inner enclosure down along the sides of the outer enclosure and up the bottom of the inner.. it looks very cool. My temps right now with 70f ambient is (from bottom to top)~ 33c-36c-36c-37c-38c-38c. With two boards temps were in the low 30’s.

Oh and there is also a heat sync and a fan at the top that would be built in to the lid. So now that it is essentially done I will just be playing with dropping the temperatures and I have a bunch of ideas for that such as Peltiers and or a metal top with heat sinks. Expansion with additional miners will happen soon.

Thanks to Fizzisist and crew, you guys are great.

[Garr255]

Sweet setup! Yeah the whole FPGAmining crew is awesome

[fizzisist]

Here is my mineral oil setup I have been working on a long time.

That is downright beautiful! Fantastic work, and those temps are amazing. I especially like the 140mm fan for forced convection! I'm sure that will make quite the conversation piece in the living room.

How many invalids were you seeing with the overclocker firmware? It's normal to see some, and MPBM will automatically adjust to minimize those but maximize the hashrate. Some FPGAs can get up to 210 MHz, and I would imagine your setup would have higher than average performance. Just load up the firmware and set MPBM to start at 200 MHz, with 250 MHz for the max frequency. Let it run for a few hours and it should stabilize at a comfortable frequency with around 1% invalids. Please let us know what you get!

Also, would right angle barrel connectors make things easier? http://search.digikey.com/us/en/products/CA-2191/CP-2191-ND/568581

[shackleford]

Here is my mineral oil setup I have been working on a long time.

That is downright beautiful! Fantastic work, and those temps are amazing. I especially like the 140mm fan for forced convection! I'm sure that will make quite the conversation piece in the living room.

How many invalids were you seeing with the overclocker firmware? It's normal to see some, and MPBM will automatically adjust to minimize those but maximize the hashrate. Some FPGAs can get up to 210 MHz, and I would imagine your setup would have higher than average performance. Just load up the firmware and set MPBM to start at 200 MHz, with 250 MHz for the max frequency. Let it run for a few hours and it should stabilize at a comfortable frequency with around 1% invalids. Please let us know what you get!

Also, would right angle barrel connectors make things easier? http://search.digikey.com/us/en/products/CA-2191/CP-2191-ND/568581

I tried a right angle barrel adapters, that was another 30$ that was not useful. In theory that would work but the adapter still sticks out quite a bit and has the added negative of not being flexible. It is my fault for trying to work out all the dimensions ahead of time with some level of guessing.  When doing a setup like this you want it perfect…. absolutely perfect… before you pour the oil otherwise it becomes a sloppy mess fast.

Here are my results from running the OC firmware for 2.5 hr. the hottest board (at the top) is the best performer and even then it is pretty meh..Something just does not seem right how 1 firmware has invalids and the other has 0 at the same mhz over a long period of time. Are there any upcoming firmware enhancements? I guess I should run the OC firmware just for the protection. I like that if anything happens (like a cord getting caught in the convection fan) it will scale it down but I would like to see zero invalids like with the other firmware or at least marginaly better than 200mhz with some invalids. I will leave it running for a while longer but it is so slow to change. The 206mhz chip had zero invalids for almost 2.5 hr.

[TheSeven]

The overclocker firmware will not run without invalids by definition, because the overclocking algorithm would increase the clock frequency in that case. It will stop increasing the clock frequency when it hits 2% invalids or the warning temperature (45°C by default), and try to stabilize things from there on.

The ztexmerge_200 bitstream could probably run at even higher clock rates, but lacks the necessary overclocking logic. The overclocker bitstream is just a newer version with more features, which hasn't been optimized as heavily as the ztexmerge one yet.

[shackleford]

The overclocker firmware will not run without invalids by definition, because the overclocking algorithm would increase the clock frequency in that case. It will stop increasing the clock frequency when it hits 2% invalids or the warning temperature (45°C by default), and try to stabilize things from there on.

The ztexmerge_200 bitstream could probably run at even higher clock rates, but lacks the necessary overclocking logic. The overclocker bitstream is just a newer version with more features, which hasn't been optimized as heavily as the ztexmerge one yet.

Hmm, that is not the behavior I am seeing. My best chip is at .72% invalids @ 38c it had recently went from 208mhz back down to 206. four of my other chips have not exceded that criteria and are stuck below 200mhz

As I wrote that my best chip got an invailid and went down to 204mhz. It is still at .79% @ 38c. I am using the compiled windows version, I will try the python route if you think that will make a diffrence.

[TheSeven]

The overclocker firmware will not run without invalids by definition, because the overclocking algorithm would increase the clock frequency in that case. It will stop increasing the clock frequency when it hits 2% invalids or the warning temperature (45°C by default), and try to stabilize things from there on.

The ztexmerge_200 bitstream could probably run at even higher clock rates, but lacks the necessary overclocking logic. The overclocker bitstream is just a newer version with more features, which hasn't been optimized as heavily as the ztexmerge one yet.

Hmm, that is not the behavior I am seeing. My best chip is at .72% invalids @ 38c it had recently went from 208mhz back down to 206. four of my other chips have not exceded that criteria and are stuck below 200mhz

As I wrote that my best chip got an invailid and went down to 204mhz. It is still at .79% @ 38c. I am using the compiled windows version, I will try the python route if you think that will make a diffrence.

This might be a consequence of the UI showing an average value over the whole running time of MPBM, while the overclocking logic uses a short term average (since the last clock change). I should probably add the latter to the displayed stats as well.

[freshzive]

I am proud to report that according to BTC Guild one of my X6500 found a block

Cool :}
Now switch to P2Pool

Thought p2pool doesn't work well with MPBM (very high stales)? Or has that changed?

[c_k]

Works great for me

[TheSeven]

I am proud to report that according to BTC Guild one of my X6500 found a block

Cool :}
Now switch to P2Pool

Thought p2pool doesn't work well with MPBM (very high stales)? Or has that changed?

That was just some temporary issue with some pre-alpha releases of the v0.1.0 series, which had slightly higher stales on p2pool than the v0.0.4 series, which has been fixed long ago. And even that was <10% stales.

The only remaining thing that doesn't work well with p2pool is BFL's firmware...

[freshzive]

Any updates on when more units will be put up for sale?

[fizzisist]

Any updates on when more units will be put up for sale?

Actually, the heatsink-less units are in stock on Cablesaurus right now, but we're having a lot of difficulty with our heatsink supply. We finally found a supplier that can deliver large quantities of them, but those won't be here for a couple weeks. Still, Cablesaurus will be receiving a small shipment of units with heatsinks early next week.

[arklan]

pondering picking up some of these. what kind of time frame would i be looking at for an order of ten or so?

[fizzisist]

pondering picking up some of these. what kind of time frame would i be looking at for an order of ten or so?

For an order of 10 boards, we could ship the next business day. As I said above, heatsink availability may be a problem, though. Please email sales@fpgamining.com to discuss bulk prices!