Topic: X6500 Custom FPGA Miner - page 51. (Read 220050 times)

The xilinx timing and power analysis stuff is a bit crazy at times. If you're making the mistake of believing it that might be part of your lower hash rates.

The ideal setup for mining is not zero errors... if you can get 10% more speed but only lose 1% of the shares due to corruption then you're still 9% ahead. And false positives are trivially swept up by the supervising cpu.  Ideally you'd want to do some self-testing in order to auto calibrate the clock rate on each unit to get the highest yield possible from each chip.

Otherwise, you might just try making a bunch of inconsequential changes in order to perturb the routing and get better timing out of it.  These mining engines seriously tax the interconnect occupancy, so small differences may make big timing differences

.

So wtf... have you forensically analysed the code yet?

PS: our board looks cooler

Thanks, clearly you are getting better efficiency. The 1.2V regulators are not hot on this board, but they are slightly warm. According to the datasheet, they should have roughly 90% efficiency in our configuration. I doubt your regulator gets much better than that, if at all. Your advantage probably comes from the FPGA code, then. Interesting! I'm curious to see what you're doing differently that gives this better efficiency.

The design of the initial release required 5.9W at 128 MHz. (release number 110826 on http://www.ztex.de/btcminer).

If your 1.2V voltage regulators are becoming hot (IMHO you are using two), they are at least a major part of the problem.

You're right, the power supply design is different and could definitely vary in efficiency. A more even comparison might be to see what the power usage of your board is when mining at 125 MH/s. Could you test that out? I wouldn't be surprised if your design was still more efficient, but I doubt it would be 50% more so.

I partially agree.

According to your post X6500 requires 17W for 250 MH/s, that is 50% more than 8.5W per 190 MH/s. This difference could be caused by less efficient voltage regulators, but also by the FPGA design, which is of course not board dependent.

At about 200MH/s the FPGA reaches internal limits.  Power supply is critical and makes out up to 10% of performance.

I don't think it's fair to say that your board achieves greater MH/s or efficiency. The board design should not impact this, and both of our boards use the same FPGAs. At $/FPGA, our X6500 has you beat ($290 vs. $460). In addition, nothing is stopping an adventurous user from running your code on our board.

On the other hand, if you really have achieved 190 MH/s on this FPGA, kudos to you!

There is an other FPGA Board at https://bitcointalksearch.org/topic/ztex-usb-fpga-modules-115x-and-115y-215-and-860-mhs-fpga-boards-49180.

The Board currently achieves 190 MH/s on a single Spartan 6 XC6SLX150-3. For amounts of 5 or larger it is cheaper (i.e. more MH/s per $, see https://bitcointalk.org for volume discounts) and it is about 50% more efficient.

The Software is Open Source, see http://www.ztex.de/btcminer an reached a mature level. A cluster mode with hot-plugging allows to run many FPGA board from a single software instance.

Boards are available immediately.

Definitely a bug. Also dont delte your browser cookies because it deletes your glbse account for some reason.

I had this happen to me before when buying some other shares. It will just sit in your order list forever. I think the best thing to do is cancel that order and place it again. I should bring that up to nefario as a possible bug.

I've ordered some shares and they show in my orders as btc_short.... what the heck does that mean?  How do I pay for the shares? I've got the BTC in my account..... just need to pay.

Thanks

So far, the X6500 prototype is performing very well. As you've seen above, I've been mining with it at 250 MH/s (125 MHz clock) for almost 24 hours now and haven't seen any problems. At that speed, it consumes 17 W. The software could use some polish and there are still more tests to do, but the important thing is that the board works. One other thing is to test higher clock rates. I'm trying to synthesize 133 MHz right now.

There is some bad news, though. The first is that we will not be able to meet the original deadline estimate. It looks like it will take about 3 weeks from the time we place the orders to the time Cablesaurus will be able to ship them to you guys. Hopefully the next piece of news makes up for this delay: we will drop the prices to $580 for the X6500 and $390 for the X6000!

The other piece of bad news is that Paypal sucks. I'm sure that's no surprise to most of you, this being the Bitcoin forum. It turns out that pre-selling stuff is not allowed, and Paypal sent Cablesaurus a nasty email telling him to cut it out. This throws a wrench in the plan that we would essentially pool this order by collecting payments before the boards weremade. We're now planning to put up a big chunk of our own money to get these orders out, but we won't be able to cover all of the costs on our own. We're looking for help from those customers willing to pre-pay in BTC now. We can offer an additional $20/board discount to those customers. We will not be able to accept any payments through Paypal until the boards are ready to ship.

At this point, we're ready to begin accepting BTC pre-payments, for customers that have already put down a deposit and for new orders. If you're interested, please PM myself and/or Cablesaurus. We expect to ship in about 1 month, depending on how quickly these pre-order payments come in.

Screen shot of the mining software (I know it ain't pretty, but it works!):



Edit: After 33 minutes, it has submitted 118 shares, which is the predicted number of shares for a hash rate of 254 MH/s. 3 shares were rejected.

Cool! Yeah, GLBSE takes some getting used to for sure. I should really create a separate thread for that...

And it's done: https://bitcointalksearch.org/topic/fpga-mining-contract-49166

for the love of pete, now it shows up 

I want to buy 20 shares, I've not used GLBSE before, I've set up an account and transferred 5 BTC to buy shares with.... problem is my bitcoin client shows 6 confs and still my BTC is not showing up in my GLBSE account.

I've ordered the shares but obviously can't pay for them till the flippin BTC shows up.... and yes I've not made any mistakes with the address for depositing the BTC :-)

Big news! The X6500 prototype is fully mining now at 250 MH/s! We ran into some hiccups getting the software working, and had to wait a week for the board to fly from Singapore to California so that I could test out code changes much more quickly. It arrived last night and I got it fully working tonight!



I'll post some better photos not taken with a cell phone and some screen shots tomorrow.

It's mining now (earning money for FPGA.contract shareholders ). It will run overnight and I'll spend the day tomorrow running a lot of different tests and cleaning up the software. First thing is to measure the actual power usage.

Edit: Looks like it's using exactly 17 W. Not too shabby at 14.7 MH/J!

I'm interested to buy 10 shares, but I can't find it on GLBSE. Maybe there will be an option to pay with hashing power on Eligius? Let's say you will create a worker for user that want to buy shares and every 0.25BTC that worker earn will be converted to 1 share...

EDIT:
Ok, I've found it and ordered 20 shares...

That is absolutely not an issue. The thing is, I'm doing that not for myself but as a attempt to get a foot in the BTC economy, well sort of. Since I haven't the cash to by a board that is...

Just wondering how the deadline is, if the cases are supposed to be available when the boards are shipping... My progress has been kind of stalling lately but I'll attempt to get something together in time if you tell me when...  

Do give us more information when you are ready, we are definitely interested.