896 mh/s firmware release - Butterfly Labs

[P_Shep]

120W on 1 single? That don't seem right.

[rjk]

120W on 1 single? That don't seem right.

My kill-a-watt doesn't lie.
However, I am using the stock power pack, and you are using an efficient PSU, so that could be where the difference went.

[P_Shep]

120W on 1 single? That don't seem right.

My kill-a-watt doesn't lie.
However, I am using the stock power pack, and you are using an efficient PSU, so that could be where the difference went.

Yeah, but 40W difference? Does that include the controller PC / 40W lamp too?

[rjk]

120W on 1 single? That don't seem right.

My kill-a-watt doesn't lie.
However, I am using the stock power pack, and you are using an efficient PSU, so that could be where the difference went.

Yeah, but 40W difference? Does that include the controller PC / 40W lamp too?

Ah heck, thanks for making me look again. It seems that the power must have gone out, and it reverted back to reading the voltage. 
However, on the old bitstream it was pulling about 80 watts, now on the faster one it is pulling about 90. Stock power brick is the only thing plugged into the kill-a-watt.

[kano]

Ah good - I was rather reluctant to try the higher clock bitstreams when I saw that figure ...
Glad to know it was a accidental mistake.

But ignoring some vague post by someone else about it ...
yes I'd expect the gain in MH/s (M%) vs power increase (P%) to certainly be P% > M%

... now where's the USB boot I want ...

[rjk]

... now where's the USB boot I want ...

Sorry, I googled for a while and came to the conclusion that it is only possible to create bootable Windows USB drives from within Windows, unless you do a lot of hacking within your *nix environment.
Additionally that Windows 7 is orders of magnitude easier to use on a USB than XP is.
Tips: Format as NTFS, otherwise not much is going to work. Chainloading with syslinux may be possible. dd if/of from an ISO to a device probably won't work. Sorry, I can't do much else for you unless you have a windows box somewhere. Using Virtualbox might work, but you can't use vbox OSE because it doesn't have USB support. Download the compiled version with non-free components built in if you want to use USB passthrough.

[SgtSpike]

I get 850 Mhps with the 862 firmware, and 856 Mhps with the 872 firmware.

Interesting.  Only 6/10 increase in hashing power.

Temps are interesting... I have a 6" fan blowing on them all.  The closest one to the fan is 52C, 2nd closes is 47C, third is 47C and fourth is 55C.  Definitely a variety of temps!  So far, no throttling though.

[jddebug]

I flashed 4 brand new singles with the 872 firmware. Plugged them into my Kill-O-Watt using stock power supply on each.

With the 832 firmware they took total of 322 watts.

With the 872 firmware they take a total of 336 watts.

Went from 3250 MH/s to 3400 MH/s.  EDIT: I had mistyped as 3340.

(cgminer) U went from 44 to U of 48

I believe I'm seeing some throttling occurring on 2 of them although the light never seems to blink. I'll flash them to the next level down and see that I get.

[kano]

Well those figures are as I expected: P% > M%
(4.35% > 2.77%) i.e. the power increase % is greater than the MH/s increase %

[jddebug]

Well those figures are as I expected: P% > M%
(4.35% > 2.77%) i.e. the power increase % is greater than the MH/s increase %

See my edit. I had mistyped the MH/s

[wogaut]

I flashed 4 brand new singles with the 872 firmware. Plugged them into my Kill-O-Watt using stock power supply on each.

With the 832 firmware they took total of 322 watts.

With the 872 firmware they take a total of 336 watts.

Went from 3250 MH/s to 3400 MH/s.  EDIT: I had mistyped as 3340.

(cgminer) U went from 44 to U of 48

I believe I'm seeing some throttling occurring on 2 of them although the light never seems to blink. I'll flash them to the next level down and see that I get.

So with those numbers you have about the same MH/W just dividing the numbers you posted (the throttling might result in some variability over time).
That is good to hear, as this would mean you just got 4.6% more Singles (if stable).

[e21]

1 Revision 3 single, using stock power supply, measured using Kill-a-Watt load meter:

81 Watts with the 832MH/s firmware @832MH/s
84.5 Watts with 872MH/s firmware @~870MH/s, 0 throttles, ~49*C average temp

Thanks BFL!

EDIT: throttles once every 10-15 minutes or so on 872 MH/s firmware; ambient temp is ~24*C right now. Going to try placing it in an area with better air-flow and see if throttling stops altogether.

EDIT: moved and temps stay under 48*C now, 0 throttles

[wogaut]

Boy, that firmware update almost gave me a heart attack. 
My single was happily mining at 832MH/s and 48deg C with zero errors, so I gave the update a try.
I ended up getting about 440MH/s average on the 864 update and 370MH/s average on the 872 update, with mining rates fluctuating widely between 75 and 850MH/s and error rate through the roof, while FPGA temps were about 34deg C.

So I flashed it back to the default 832MH/s firmware and it's happily mining again as before.

[BTC-engineer]

One of my singles worked with the 872 firmware without further improvements.
On the others I had to improve the cooling/airflow - now they are also working @872 without problems.

@BFL: Any plans to release further firmwares with a higher clockrate? I wonder why you stop with the 872 firmware. Could it be that something else (e.g. voltage regulators) could overheat because they are not or not sufficient covered by the existing one-point temperature monitoring? 

[jddebug]

One of my singles worked with the 872 firmware without further improvements.
On the others I had to improve the cooling/airflow - now they are also working @872 without problems.

@BFL: Any plans to release further firmwares with a higher clockrate? I wonder why you stop with the 872 firmware. Could it be that something else (e.g. voltage regulators) could overheat because they are not or not sufficient covered by the existing one-point temperature monitoring? 

I'm interested in what you did to improve cooling/airflow.

Thanks

[SgtSpike]

One of my singles worked with the 872 firmware without further improvements.
On the others I had to improve the cooling/airflow - now they are also working @872 without problems.

@BFL: Any plans to release further firmwares with a higher clockrate? I wonder why you stop with the 872 firmware. Could it be that something else (e.g. voltage regulators) could overheat because they are not or not sufficient covered by the existing one-point temperature monitoring? 

Probably because a large percentage of miners wouldn't even be able to run that clockrate without throttling, and people would complain even more than they are.

[BFL]

One of my singles worked with the 872 firmware without further improvements.
On the others I had to improve the cooling/airflow - now they are also working @872 without problems.

@BFL: Any plans to release further firmwares with a higher clockrate? I wonder why you stop with the 872 firmware. Could it be that something else (e.g. voltage regulators) could overheat because they are not or not sufficient covered by the existing one-point temperature monitoring? 

Probably because a large percentage of miners wouldn't even be able to run that clockrate without throttling, and people would complain even more than they are.

Two new faster firmwares have just been released.  As previously noted, these faster firmwares may or may not run without throttling depending on the individual tolerance of any particular unit and the ambient temperature.  Some units will work just fine with these firmwares.  Some will thottle.

880 mh/s
896 mh/s

You can find them on the driver download page:  http://www.butterflylabs.com/drivers/

[Cablez]

Give them time to upload it.  Working now.

[bulanula]

Thanks for the overclocking support but no other company does this when it could influence warranty issues.

Very strange but AWESOME !

[Fefox]

so far so good with 880, 2 units are running fine with no throttle... in the mid 50's for temps.


flashing 2 more now with 896