896 mh/s firmware release - Butterfly Labs

[bulanula]

So I have tried the 896 firmware on every single I have upgraded so far and it has only worked on one. 864 seems to be the sweet spot for my singles. YMMV

Thanks for letting us know.

Is there any point releasing higher speed firmwares ?

What is the theoretical maximum for these chips that can be used if you had subzero cooling for example ?

[pieppiep]

Probably only BFL knows.
They made the design and compiled it to a bitfile, so they know the timing constraints.

[rjk]

My guess is that there may be a compilation error with some specific files, or that the speed just happens to tickle a bug in the hardware.

[SgtSpike]

So I have tried the 896 firmware on every single I have upgraded so far and it has only worked on one. 864 seems to be the sweet spot for my singles. YMMV

+1 for that.  It worked on 1/4, though I only ran it for a short while.

So I have tried the 896 firmware on every single I have upgraded so far and it has only worked on one. 864 seems to be the sweet spot for my singles. YMMV

Thanks for letting us know.

Is there any point releasing higher speed firmwares ?

What is the theoretical maximum for these chips that can be used if you had subzero cooling for example ?

BFL seemed to indicate that they ran chips up to 1.05 GH/s.

[e21]

Only have 1 single, but I was able to get the 880 MH/s firmware to run fine at a fairly high ambient temperature (~72-80*F in California), ran the diagnostics for about an hour, got 0 errors or throttles. (I was able to get the 872 MH/s firmware to run fine as well)

Tried the 892 MH/s firmware and I started getting errors but 0 throttles (using EasyMiner diagnostics, cgminer reports 0 HW). I noticed the errors didn't start until the single started to heat up, first 65 or so shares processed returned 0 errors, then when the reported temperature reached it's peak at about 48*C, (0 thottles still) I started getting about 1 error per 5 shares processed.. Seems to me like a component on the board besides the two chips is getting too hot and causing issues maybe? Going to try looking for hot spots with an infrared thermometer and applying stick-on RAM heatinks and see if that helps.

[kano]

So I have tried the 896 firmware on every single I have upgraded so far and it has only worked on one. 864 seems to be the sweet spot for my singles. YMMV

+1 for that.  It worked on 1/4, though I only ran it for a short while.

So I have tried the 896 firmware on every single I have upgraded so far and it has only worked on one. 864 seems to be the sweet spot for my singles. YMMV

Thanks for letting us know.

Is there any point releasing higher speed firmwares ?

What is the theoretical maximum for these chips that can be used if you had subzero cooling for example ?

BFL seemed to indicate that they ran chips up to 1.05 GH/s.

They ran a simulation that said they could get 1.05 GH/s

[SgtSpike]

So I have tried the 896 firmware on every single I have upgraded so far and it has only worked on one. 864 seems to be the sweet spot for my singles. YMMV

+1 for that.  It worked on 1/4, though I only ran it for a short while.

So I have tried the 896 firmware on every single I have upgraded so far and it has only worked on one. 864 seems to be the sweet spot for my singles. YMMV

Thanks for letting us know.

Is there any point releasing higher speed firmwares ?

What is the theoretical maximum for these chips that can be used if you had subzero cooling for example ?

BFL seemed to indicate that they ran chips up to 1.05 GH/s.

They ran a simulation that said they could get 1.05 GH/s

Ahhh, much different from actually running a chip that fast!  Especially given the results of real-world testing of the 896 firmware, I doubt that we'll see anything much faster - 896 seems to be beyond the limit for most chips already.

[Inspector 2211]

896 seems to be beyond the limit for most chips already.

TWO chips.
Quite likely, with TWO completely unrolled, pipelined miners on each of them.
Or a "sea of miners" on them, as brilliantly demonstrated by Bitfury on the Spartan6 platform.

[SgtSpike]

896 seems to be beyond the limit for most chips already.

TWO chips.
Quite likely, with TWO completely unrolled, pipelined miners on each of them.
Or a "sea of miners" on them, as brilliantly demonstrated by Bitfury on the Spartan6 platform.

Huh?  It was beyond the limit for 3 of mine already.

[Inspector 2211]

896 seems to be beyond the limit for most chips already.

TWO chips.
Quite likely, with TWO completely unrolled, pipelined miners on each of them.
Or a "sea of miners" on them, as brilliantly demonstrated by Bitfury on the Spartan6 platform.

Huh?  It was beyond the limit for 3 of mine already.

I was just trying to point out that the 896 MH/s are spread out over TWO FPGAs, not just one.
Maybe you meant to say "beyond the limit for most Singles".

[SgtSpike]

896 seems to be beyond the limit for most chips already.

TWO chips.
Quite likely, with TWO completely unrolled, pipelined miners on each of them.
Or a "sea of miners" on them, as brilliantly demonstrated by Bitfury on the Spartan6 platform.

Huh?  It was beyond the limit for 3 of mine already.

I was just trying to point out that the 896 MH/s are spread out over TWO FPGAs, not just one.
Maybe you meant to say "beyond the limit for most Singles".

Ahhh, yes indeed.  Beyond the limit for most Singles.

[Beaflag VonRathburg]

896 seems to be beyond the limit for most chips already.

TWO chips.
Quite likely, with TWO completely unrolled, pipelined miners on each of them.
Or a "sea of miners" on them, as brilliantly demonstrated by Bitfury on the Spartan6 platform.

Huh?  It was beyond the limit for 3 of mine already.

I was just trying to point out that the 896 MH/s are spread out over TWO FPGAs, not just one.
Maybe you meant to say "beyond the limit for most Singles".

Is it the limit of the chips, hardware to keep it cool, outside temperature or some combination of the previously mentioned?

[Inspector 2211]

896 seems to be beyond the limit for most chips already.

TWO chips.
Quite likely, with TWO completely unrolled, pipelined miners on each of them.
Or a "sea of miners" on them, as brilliantly demonstrated by Bitfury on the Spartan6 platform.

Huh?  It was beyond the limit for 3 of mine already.

I was just trying to point out that the 896 MH/s are spread out over TWO FPGAs, not just one.
Maybe you meant to say "beyond the limit for most Singles".

Is it the limit of the chips, hardware to keep it cool, outside temperature or some combination of the previously mentioned?

Hard to say, probably a combination of things.
Does anyone have a walk-in freezer where he could test whether freezing temperatures help?

In any case, there is a thermal resistance Tr1 from the die to the heat spreader and a thermal resistance Tr2 from the heat spreader to the heat sink and finally there is a thermal resistance Tr3 from the heat sink to the room air.
Tr1 cannot be modified, but we have already established that Tr2 is not at its optimal level, because the heat sink does not contact the whole heat spreader, only part of it. Also, I suspect that co-planarity of the FPGAs, or rather the lack of it, may contribute to a higher-than-desired Tr2 value.

[lomax]

Sailing along at 892MH/s here, with no increase in temps.

Waiting for the next firmware BFL, keep pushing the limits I say.

[BTCurious]

Hard to say, probably a combination of things.
Does anyone have a walk-in freezer where he could test whether freezing temperatures help?

Someone with throttling issues put his BFL in a cooling chamber and ran it at 0°C, and still had throttling issues.

[Scared]

So I took one of my singles and have it running at 50c. I updated the firmware to 896 and confirmed with Easyminer that it ran without any hardware errors. It ran with an Avg Speed of 897. I then ran it with cgminer for the same amount of time that Easyminer ran and got 876.9. I repeated this three separate times and got the same results. Why am I getting such a significant difference between cgminer and Easyminer? I understand from previous posts that cgminer doesn't report the hardware errors but Easyminer has shown 0 HW errors every time I've run it. 

Any suggestions?

[BFL-Engineer]

So I took one of my singles and have it running at 50c. I updated the firmware to 896 and confirmed with Easyminer that it ran without any hardware errors. It ran with an Avg Speed of 897. I then ran it with cgminer for the same amount of time that Easyminer ran and got 876.9. I repeated this three separate times and got the same results. Why am I getting such a significant difference between cgminer and Easyminer? I understand from previous posts that cgminer doesn't report the hardware errors but Easyminer has shown 0 HW errors every time I've run it. 

Any suggestions?

The speed reported by Easyminer indicates how fast the unit can process 4Billion range of nonces. The speed
reported by cgminer includes network latency, jobs that have no nonce result, LongPoll cancellation of all jobs
in queue, etc... All of these affect your productivity in terms of MH/s.


Regards,
BF Labs Inc.

[BTC-engineer]

Any updates regarding faster firmwares (>896)?

[e21]

So I took one of my singles and have it running at 50c. I updated the firmware to 896 and confirmed with Easyminer that it ran without any hardware errors. It ran with an Avg Speed of 897. I then ran it with cgminer for the same amount of time that Easyminer ran and got 876.9. I repeated this three separate times and got the same results. Why am I getting such a significant difference between cgminer and Easyminer? I understand from previous posts that cgminer doesn't report the hardware errors but Easyminer has shown 0 HW errors every time I've run it. 

Any suggestions?



The speed reported by Easyminer indicates how fast the unit can process 4Billion range of nonces. The speed
reported by cgminer includes network latency, jobs that have no nonce result, LongPoll cancellation of all jobs
in queue, etc... All of these affect your productivity in terms of MH/s.


Regards,
BF Labs Inc.

Yes, I had my single running the 872 MH/s firmware, which averaged out to about 855 MH/s in cgminer, then I updated to 892 MH/s, which then averaged out to about 876 MHs for me as well. The "U" also increased by about .3, from 10.8 to 11.1

[JWU42]

A "U" of 11.1 with the 892 FW is low...

This is a 864 FW - running for 3 hours

Code:

BFL 5:  42.8C         | 859.0/852.0Mh/s | A:2202 R:0 HW:0 U: 12.10/m

And this is the lowest U of 3 units at 816 FW (other two are 11.35 and 11.60)

Code:

 BFL 1:  44.7C         | 810.0/803.3Mh/s | A:1972 R:0 HW:0 U: 10.73/m