ZTEX USB-FPGA Modules 1.15x and 1.15y: 215 and 860 MH/s FPGA Boards

[Inspector 2211]

A testing release has been published: http://www.ztex.de/btcminer/ZtexBTCMiner-120130.jar . See the software thread (https://bitcointalk.org/index.php?topic=40047.msg722099#msg722099) for details.

I'm happy to report that the "15d3a" version works for me now - it starts at 200 MHz, goes to 204, then 208, then 212, but the error rate is too high at 212
and it settles at 208.

Good job, Stefan!
1.0e3 thanks!

[Turbor]

Going to give it a try this evening. 120126 and d2 work very well for me. Glad i was able to bring nr6 to life.

[ztex]

A BTCMiner release has been published, see the software thread (https://bitcointalk.org/index.php?topic=40047.msg738240#msg738240) for details.

[Unacceptable]

What kind of hash rates does this device get?

USB-FPGA Module 1.11c (XC6SLX25, SG 3, 64 MB RAM)

http://shop.ztex.de/product_info.php?products_id=52

At $175 I might be willing to try a few of them,any kind of warranty??

[DeathAndTaxes]

What kind of hash rates does this device get?

USB-FPGA Module 1.11c (XC6SLX25, SG 3, 64 MB RAM)

http://shop.ztex.de/product_info.php?products_id=52

At $175 I might be willing to try a few of them,any kind of warranty??

The 1.15x has 150K LUTs that FPGA has 25K LUTS so theoretical max would be maybe 30 MH/s BUT the 1.15X uses a 150K LUT FPGA so it can unroll the SHA-256 double loop.  This chip doesn't have the "space" so you get all the overhead of the rolled algorithm so ballpark maybe 20 MH/s.  Of course you would need to find/build a bitstream for it. 

What is weird is you said you would be willing to try a FEW?  You could get a 1.15x for the price of 3 of those and gain all the community support, plus higher performance chip. 

[Inspector 2211]

He may be interested in writing his own Verilog code, and with the LX150 device he'd have to BUY  the toolchain, as the LX150 is not supported by the free toolchain. But I recommend buying at least the LX75 device. Module 1.15b. It can be used for mining AND for trying out new code variants. Maybe I'll buy a 1.15b myself. The discussion with wondermine in a different thread gave me some ideas on how to speed up the "add" operations.

[DeathAndTaxes]

He may be interested in writing his own Verilog code, and with the LX150 device he'd have to BUY  the toolchain, as the LX150 is not supported by the free toolchain. But I recommend buying at least the LX75 device. Module 1.15b. It can be used for mining AND for trying out new code variants. Maybe I'll buy a 1.15b myself. The discussion with wondermine in a different thread gave me some ideas on how to speed up the "add" operations.

Agreed.  If you want to do your own research "on the cheap" a LX75 (75K LUTs) is the best chip.  It is the largest chip which can still use the free ISE.   A 25K LUT chip is simply too small to be useful.

[Unacceptable]

What kind of hash rates does this device get?

USB-FPGA Module 1.11c (XC6SLX25, SG 3, 64 MB RAM)

http://shop.ztex.de/product_info.php?products_id=52

At $175 I might be willing to try a few of them,any kind of warranty??

The 1.15x has 150K LUTs that FPGA has 25K LUTS so theoretical max would be maybe 30 MH/s BUT the 1.15X uses a 150K LUT FPGA so it can unroll the SHA-256 double loop.  This chip doesn't have the "space" so you get all the overhead of the rolled algorithm so ballpark maybe 20 MH/s.  Of course you would need to find/build a bitstream for it.  

What is weird is you said you would be willing to try a FEW?  You could get a 1.15x for the price of 3 of those and gain all the community support, plus higher performance chip.  

No I'm not a software guy,I just think for the kind of money being asked per unit ($414-$569 & $580) I need reassurance that if something dies,due to no fault of my own (possible defect),that I will have some kind of recourse,for at least 3-6 months.Heck my 05 impala has 125,000 miles & I got an extended warranty for large repairs(engine,trans,A/C) for $200 a year.

$175 is about my limit for risk,one at a time(buy one,let it mine & buy another with those earnings),if the hashrates were worth it.At 30 mh/s it isn't enough for a ROI in any decent timeframe,for me anyhow.I was hoping around 150 mh/s,but its all good.

I haven't noticed anyone raging about DOA's or major malfunctions with Ztex or Icarus,so I might be a bit anal about a warranty for no reason.I'm still thinking it over

What kind of mh/s does the LX75 get.

BTW,I'm not tryin to knock anyone's products,so please don't take it as I were.

Thanks !!!

[antirack]

I wouldn't be too worried about that, ZTEX has been around for years and they are a German company with business registration, phone numbers, liabilities, laws, etc. Not some outlet operating from a backyard in Cambodia.

In fact, if I am not wrong, German law minimum warranty is 6 months and they have to provide replacement parts and defects liability for 2 years (means they have to fix it, for a reasonable cost, for 2 years even if it is your fault etc). Correct me if I am wrong.

Yes, of course this all won't be of any use if the company goes bankrupt. But in Germany it's not easy to just go bankrupt and disappear, trust me on this. Germany is not for no reason one of the most restrictive economies in the world (in comparison to the USA or Hong Kong for instance, or other countries on the top of the list of freest economies). For instance, the director of a German company is criminally liable if the company goes bankrupt and this is indeed enforced even for small scale operations. I don't know many countries where this is the case.

Warranty information:
http://www.ztex.de/agb.e.html

[Unacceptable]

  Thanks for finding that info

So as I understand it,there's a 2 year warranty??

7.2  The customer shall examine the goods upon receipt for completeness and any defects. In the event of identifiable defects, the customer shall inform Company ZTEX GmbH thereof in writing within 14 days of receipt. Non-identifiable defects shall be reported in writing immediately upon their identification. The warranty period shall commence upon receipt of the goods by the buyer and shall remain in effect for 24 months pursuant to statutory provisions. Commercial transactions shall be subject primarily to Section 377 HGB [Handelsgesetzbuch, German Commercial Code].

If so,I guess I might give one a shot

[CA Coins]

Per Ztex in his emails, there is a 2-year EU mandated warranty on the boards that covers defects in the manufacturing.  the boards are tested before shipping, and you will be covered if it is DOA.  He has mentioned that it doesn't cover improper usage of the board.  I figure if the boards fries from you sticking in the wrong voltage power (see one of the threads here), you won't be covered.  
The build of the boards look pretty solid.  I have had no issues with it even after repeatedly connecting/disconnecting the power/usb connectors.  The fpga runs pretty cool and the software downclocks based on errors so I am not too worry about frying the fpga.  Feels much safer (maybe unfounded) than running the 6950 at 75C.  However, I don't think anybody knows (or does anybody know) how long these boards are expected to run, but they feel solidly built to me.    

[rupy]

ztex said he expected these to run 5 - 10 years... I guess depending on heat?

[CA Coins]

ztex said he expected these to run 5 - 10 years... I guess depending on heat?

I think 5-10 years would be pretty sweet.  Let's see if BTC survives.  I would think these things are made to run for awhile but I am not sure how heat and overclocking would affect longevity.

I did some thermal measurements on the boards, using the stock Xilence heatsink/fan with a open case and also with a big desk fan blowing directly on them. 

Code:

Room temp:  19C
Boards open to air

Between board and heatsink: 30C
Bottom of heatsink:  29C
Top of heatsink: 24C

With fan blowing directly on board
Between board and heatsink: 27C
Bottom of heatsink:  27C
Top of heatsink: 22C

With cooler weather and better airflow, this is the best I have gotten so far:
2012-02-13T22:45:36: bus-0-0: ztex_ufm1_15d3-2012-L1-A1: f=204.00MHz,  errorRate=0.29%,  maxErrorRate=0.74%,  hashRate=203.4MH/s,  submitted 17 new nonces,  luckFactor=1.03
2012-02-13T22:45:36: bus-0-0: ztex_ufm1_15d3-2012-L1-A2: f=204.00MHz,  errorRate=0.48%,  maxErrorRate=1.10%,  hashRate=203.0MH/s,  submitted 14 new nonces,  luckFactor=0.87
2012-02-13T22:45:36: bus-0-0: ztex_ufm1_15d3-2012-L1-A3: f=204.00MHz,  errorRate=0.12%,  maxErrorRate=0.41%,  hashRate=203.7MH/s,  submitted 12 new nonces,  luckFactor=0.93
2012-02-13T22:45:36: bus-0-0: ztex_ufm1_15d3-2012-L1-A4: f=200.00MHz,  errorRate=0.00%,  maxErrorRate=0.00%,  hashRate=200.0MH/s,  submitted 13 new nonces,  luckFactor=0.82
2012-02-13T22:45:36: bus-0-0: ztex_ufm1_15d3-2012-L1-B1: f=208.00MHz,  errorRate=0.00%,  maxErrorRate=0.00%,  hashRate=208.0MH/s,  submitted 20 new nonces,  luckFactor=1.22
2012-02-13T22:45:36: bus-0-0: ztex_ufm1_15d3-2012-L1-B2: f=208.00MHz,  errorRate=0.56%,  maxErrorRate=1.39%,  hashRate=206.8MH/s,  submitted 12 new nonces,  luckFactor=0.70
2012-02-13T22:45:36: bus-0-0: ztex_ufm1_15d3-2012-L1-B3: f=204.00MHz,  errorRate=1.13%,  maxErrorRate=2.03%,  hashRate=201.7MH/s,  submitted 15 new nonces,  luckFactor=1.04
2012-02-13T22:45:36: bus-0-0: ztex_ufm1_15d3-2012-L1-B4: f=204.00MHz,  errorRate=0.13%,  maxErrorRate=0.26%,  hashRate=203.7MH/s,  submitted 12 new nonces,  luckFactor=1.03
2012-02-13T22:45:36: bus-0-0: ztex_ufm1_15d3-2012-L1-B5: f=204.00MHz,  errorRate=0.02%,  maxErrorRate=0.41%,  hashRate=204.0MH/s,  submitted 14 new nonces,  luckFactor=1.06
2012-02-13T22:45:36: bus-0-0: ztex_ufm1_15d3-2012-L1-B6: f=200.00MHz,  errorRate=0.00%,  maxErrorRate=0.26%,  hashRate=200.0MH/s,  submitted 13 new nonces,  luckFactor=0.75
2012-02-13T22:45:36: bus-0-1: ztex_ufm1_15d3-2012-L1-A0: f=208.00MHz,  errorRate=0.58%,  maxErrorRate=1.02%,  hashRate=206.8MH/s,  submitted 12 new nonces,  luckFactor=0.95
2012-02-13T22:45:36: bus-0-1: ztex_ufm1_15d3-2012-L1-A5: f=204.00MHz,  errorRate=1.03%,  maxErrorRate=1.66%,  hashRate=201.9MH/s,  submitted 13 new nonces,  luckFactor=0.87
2012-02-13T22:45:36: bus-0-1: ztex_ufm1_15d3-2012-L1-A6: f=204.00MHz,  errorRate=0.00%,  maxErrorRate=0.50%,  hashRate=204.0MH/s,  submitted 17 new nonces,  luckFactor=1.12
2012-02-13T22:45:36: bus-0-1: ztex_ufm1_15d3-2012-L1-A7: f=204.00MHz,  errorRate=0.18%,  maxErrorRate=0.52%,  hashRate=203.6MH/s,  submitted 14 new nonces,  luckFactor=0.99
2012-02-13T22:45:36: bus-0-1: ztex_ufm1_15d3-2012-L1-A8: f=208.00MHz,  errorRate=0.05%,  maxErrorRate=0.26%,  hashRate=207.9MH/s,  submitted 15 new nonces,  luckFactor=0.85
2012-02-13T22:45:36: bus-0-1: ztex_ufm1_15d3-2012-L1-A9: f=208.00MHz,  errorRate=0.73%,  maxErrorRate=0.89%,  hashRate=206.5MH/s,  submitted 15 new nonces,  luckFactor=1.14
2012-02-13T22:45:36: bus-0-1: ztex_ufm1_15d3-2012-L1-B0: f=204.00MHz,  errorRate=1.09%,  maxErrorRate=1.46%,  hashRate=201.8MH/s,  submitted 10 new nonces,  luckFactor=0.97
2012-02-13T22:45:36: bus-0-1: ztex_ufm1_15d3-2012-L1-B7: f=208.00MHz,  errorRate=1.17%,  maxErrorRate=1.78%,  hashRate=205.6MH/s,  submitted 12 new nonces,  luckFactor=0.92
2012-02-13T22:45:36: bus-0-1: ztex_ufm1_15d3-2012-L1-B8: f=204.00MHz,  errorRate=0.04%,  maxErrorRate=0.35%,  hashRate=203.9MH/s,  submitted 17 new nonces,  luckFactor=1.18
2012-02-13T22:45:36: bus-0-1: ztex_ufm1_15d3-2012-L1-B9: f=200.00MHz,  errorRate=0.29%,  maxErrorRate=1.11%,  hashRate=199.4MH/s,  submitted 12 new nonces,  luckFactor=1.05

[ztex]

2 year EU warranty is the minimum guarantee which is ensured by EU law. This covers defects which existed at the time of delivery.  Since the FPGA boards are functionality tested there is not much room for this kind of errors.

In practice you can expect the FPGA board to be replaced if improper use is at least unlikely, e.g. if only a small percentage of a cluster fails within 2 years. (If more fails there are probably cooling problems).

The total run time of all LX150 FPGA boards produced by my company is about 50 years. AFAIK until now no board failed during proper use. So there is a good chance that the board run as long their power costs are less than their mining revenues.

[Turbor]

Well I've finally bitten the bullet, and ordered a single 1.15d with 'experimental board' (i.e. power supply rig) for my first intended 5-board frame design. And hopefully, as the GPU / logic board / PSU / RAM / drive systems are sold off, I'll be able to afford to buy more ZTEX FPGAs.

I haven't gone for the 1.15x models for two reasons. Firstly is the availability, but secondly, the 1.15x board is very specific to particular applications, and with the power regulation hardware on-board, requires better cooling.

The 1.15d needs to be plugged into a separate board underneath - I'm hoping that my experience with 12-GPU frame rigs will allow me to build a 5-board 'tube' with a couple of PC fans at each end, using the standard heatsinks. Forced air, enclosed by perspex so the cool circuit boards can be shown off :-)

Firstly a BIG shout out to Stefan for supplying and delivering my FPGA, experimental board and cabling within a day... he sent it on the Thursday evening from Germany and I picked it up in England by lunch-time on Friday. That's impressive stuff.

The kit looks nicely built - it doesn't look like I can make any use of the microSDHC card slot built onto the 1.15d, sadly, because the FPGA is too close to the card slot and the heatsink required will block any use of the card slot. OK, it'll have to be programmed via USB - not the end of the world...

Stefan - if you're reading this - yes I wanted immediate delivery and you're probably wondering why I haven't got the unit up and running yet... well, the software is only supplied for Windows (hurl) and Linux. My main machines all run Mac OS X - so I've downloaded the source to your SDK and dependencies and trying to port to OS X. It shouldn't be a problem at all, unless the only way of loading the bitstream to the FPGA is to use the binary Java distributions you have for download (Windoze and Linux only). I don't want to have to buy the expensive FPGA SDK, for obvious reasons. However, all your free tools *ought* to compile under Mac OS X - at least on Intel powered Macs. Will let you know if I hit any roadblocks...

Other than that, superb customer service, and blunt but accurate advice about what kit I needed to buy. I'm not an electronics engineer, and my questions could have got Stefan treating me like the guy who fried his voltage regulators. But he was polite and responded to my noob questions quickly.

Looks financially like I'll get the first FPGA board going, then build a funky perspex case, sadly only be able to afford another single unit at the end of the month (i.e. full price ) - but when I can release the capital to buy in bulk then I'll be trying to take advantage of the significant volume discounts. With English electricity costs, and only so much winter left to take advantage of (no gas central heating switched on this year!), I really need to switch from thirsty GPU rigs (9 Ghash worth... expensive power bills) to FPGAs.

Hopefully mine will be one of the 'good chips' and run over 200 Mhash using vertical air cooling... without spending more of my power budget on fans than actual hashing power! (just one of my frame rigs uses 60W worth of mains-powered fans... imagine all of those watts pumped into FPGAs....)



And cheeers Stefan, fantastic customer service and the kit looks top quality. Will ensure that I use a reasonably decent power source for the boards though....

The 1.15d should get you 208 to 212 MHz with good cooling. I cool mine with a 120mm fan from above and a 92mm from the side. At 21 degree C the boards run 208 MHz at almost zero % error rate. Heatsink placement is a bitch. I have one board that i can only use with a clamp, holding down the heatsink (and that with arctic alumina adhesive...  ). If you use the pads, i would use a clamp for some time to be sure the heatsink sticks to the chip. 1.15x boards run a bit faster due to their bigger heatsink. And yes, the boards are heavy duty. I ripped off the heatsink countless times just to find out that it won't work without clamp. Very good quality.

[CA Coins]

Hey Catfish, welcome to the ztex clan.  Good luck with the 1.15d!

I only have 1.15x, but cooling is not really a big issue for me.  I even tried blowing the board directly with a big fan and it didn't affect it much.  The good thing is that I had it run in a semi-enclosed space with ambient temperature up to 27C and it hashed just fine.  The bad part is that when I cooled the ambient down to 22C with good airflow, it didn't really matter too much (only 2 of 20 boards clocked 4Mhz higher, but hopefully the airflow will keep them running longer).  Either way, good luck.  I think Turbor has had some really good experience in getting these bad boys to clock pretty high.  And, Turbor, thanks for the advice on the heatsink.  I re-applied thermal paste on two of the units, quite painful, but it didn't seem to matter too much.  From what Stefan said, under 30C the speed is not really affected by thermals.

[Turbor]

From what Stefan said, under 30C the speed is not really affected by thermals.

Yes, it's 4 MHz more or less for some boards. At 16 degree they do 220 and 216 for example and at 21 degree they clock down to 216 and 212. But i prefer a lower clock rate with less errors !

[ztex]

The kit looks nicely built - it doesn't look like I can make any use of the microSDHC card slot built onto the 1.15d, sadly, because the FPGA is too close to the card slot and the heatsink required will block any use of the card slot. OK, it'll have to be programmed via USB - not the end of the world...

microSD card can be used if you mount the heat sink within the frame in the silk screen: http://www.ztex.de/usb-fpga-1/usb-fpga-1.15.e.html#hs

I don't want to have to buy the expensive FPGA SDK, for obvious reasons. However, all your free tools *ought* to compile under Mac OS X - at least on Intel powered Macs. Will let you know if I hit any roadblocks...

The ZTEX SDK is free: http://www.ztex.de/firmware-kit/. The Xilinx ISE is not required.

For porting BTCMiner to a other OS's the JNI library has to be compiled for that target OS, see http://wiki.ztex.de/doku.php?id=en:software:porting. Under Mac

Code:

make -f Makefile.macosx

in the libusbJava-src directory of the SDK  should do the job.  Copy the library to the working directory of your BTCMiner and it should work.

[ztex]

Heatsink placement is a bitch.

Heat sink installation on USB FPGA Modules 1.15d should not be difficult if you follow the instructions at http://www.ztex.de/usb-fpga-1/usb-fpga-1.15.e.html#hs.

Only two additional things have to be considered (in comparison to the 1.15x board):

• Heat sink removal is critical. Always unmount it by twisting (see http://www.ztex.de/usb-fpga-1/usb-fpga-1.15.e.html#hs ) tilting may damage the PCB (IMHO that happened to Turbors board)
• Some airflow has to be ensured, either by case fans or by using the active cooling upgrade in the shop

[Turbor]

Heatsink placement is a bitch.

Heat sink installation on USB FPGA Modules 1.15d should not be difficult if you follow the instructions at http://www.ztex.de/usb-fpga-1/usb-fpga-1.15.e.html#hs.

Only two additional things have to be considered (in comparison to the 1.15x board):

• Heat sink removal is critical. Always unmount it by twisting (see http://www.ztex.de/usb-fpga-1/usb-fpga-1.15.e.html#hs ) tilting may damage the PCB (IMHO that happened to Turbors board)
• Some airflow has to be ensured, either by case fans or by using the active cooling upgrade in the shop

The board is working fine. It just needs some pressure from above. I can live with that It's the faster board of the two btw...