Search content
Sort by
Showing 10 of 10 results by xaxik
Re: 1GH/s, 20w, $700 (was $500) — Butterflylabs, is it for real? (Part 2)
One question. Do you think it is possible for anyone with the experience necessary to design a custom FGPA board to get the power draw so wrong like BFL claims they did.
They might just rougly estimate the power draw.
It seems as if they took information from this document:
http://www.ti.com/general/docs/gencontent.tsp?contentId=70774
-x
Re: 1GH/s, 20w, $700 (was $500) — Butterflylabs, is it for real? (Part 2)
Part of my EP3SL150 design:
http://i.imgur.com/FXfnN.png
Part of the BFL sandpapered "bitcoin processor" design:
http://i.imgur.com/ejws0.png
The BFL "bitcoin processor" is Altera Stratix III of unknown size.
The EP3SL150 FPGA configuration bitstream is around 5MB long.
On their board is 8MB flash.
EP3SL150 uncompressed bitstream size is 47 Mbit.
EP3SL200 and 260 is 93 Mbit, which is 11.6 MB.
It doesn't fit into 8MB flash uncompressed, but it may
fit into it with compression.
So my guess is, that they use EP3SL200 or 260.
It is definitely possible to put 3 fully unrolled pipelines into the 260
(at 150MH/s each) and, with some tuning, gain the advertised speed.
To me, the case is closed.
BFL just monetizes their access to chips at prices unavailable to anyone else.
Now, why they just don't resell these $4k+ chips on the market??
-x
Btw, thanks for this image:
http://i.imgur.com/FXfnN.png
Part of the BFL sandpapered "bitcoin processor" design:
http://i.imgur.com/ejws0.png
The BFL "bitcoin processor" is Altera Stratix III of unknown size.
The EP3SL150 FPGA configuration bitstream is around 5MB long.
On their board is 8MB flash.
EP3SL150 uncompressed bitstream size is 47 Mbit.
EP3SL200 and 260 is 93 Mbit, which is 11.6 MB.
It doesn't fit into 8MB flash uncompressed, but it may
fit into it with compression.
So my guess is, that they use EP3SL200 or 260.
It is definitely possible to put 3 fully unrolled pipelines into the 260
(at 150MH/s each) and, with some tuning, gain the advertised speed.
To me, the case is closed.
BFL just monetizes their access to chips at prices unavailable to anyone else.
Now, why they just don't resell these $4k+ chips on the market??
-x
Btw, thanks for this image:
Quote
Someone asked that I measure the actual chips - they are 30mm^2.
Thanks Inaba.Are they exactly 30 or rather 29mm^2?
Some chips from altera are 29mm and some are 31mm...
I'm just curious, what may be unter the
Re: FPGA development board "Icarus" for mining purpose - pre-order started.
Thanks a lot.
Those packages look like 1mm grid thermally enhanced flipchip BGA ... guesstimating by the 0.1" headers ... 29x29mm FF780
so if these are SIII, pretty much "could be anything from a L50 to a E260"
One thing kinda throwing me off... if those are HCs, 1.1V core doesn't make sense. HCII is 1.2V and HCIII and IV are 0.9V.
As for "what SIII would you need for that"... depends on how much time you spend optimizing, first guess would be 2 unrolled engines with 1 pipeline stage per round, each running @ 250MHz... gut feeling says a L150.
So, random guesses:
a) these are HCIIIs, they prototyped with SIII and forgot to update the label for vcore.
or
b) only SIII based prototypes exist, once they get enough pre-orders HCIIIs get ordered (about 8wk if you rush assembly... possible.)
or
c) they somehow got a whole bunch of large SIIIs really cheap.
or
d) something entirely different.
so if these are SIII, pretty much "could be anything from a L50 to a E260"
One thing kinda throwing me off... if those are HCs, 1.1V core doesn't make sense. HCII is 1.2V and HCIII and IV are 0.9V.
As for "what SIII would you need for that"... depends on how much time you spend optimizing, first guess would be 2 unrolled engines with 1 pipeline stage per round, each running @ 250MHz... gut feeling says a L150.
So, random guesses:
a) these are HCIIIs, they prototyped with SIII and forgot to update the label for vcore.
or
b) only SIII based prototypes exist, once they get enough pre-orders HCIIIs get ordered (about 8wk if you rush assembly... possible.)
or
c) they somehow got a whole bunch of large SIIIs really cheap.
or
d) something entirely different.
I guess that these are engineering/preproduction samples of EP5AGXA3.
- they are 1V1
- I was able to succesfully sythesize 250MH/s with almost no effort (just combined two cores from here) into Stratix III/150 (which is 65nm). Never run it live though.
- Arria could be twice as fast for its 28nm -> 500 MH/s
- it is supplied in 27mm F672 package
@inaba: What size are the chips anyway? It can be measured really easily,
without pulling the heatsink off...
Just my 2c.
-x
All cards are tightly packed together with no extensions inside an Antec 1200. A scythe 135cfm fan is blowing directly onto the video cards. I'd also like to mention that it's also quiet, you can place it in another room without hearing it in the next.
I know I can overclock these to get 1GH by shoving fanspeeds up higher, but how much higher can they run before wearing down the fans?
I know I can overclock these to get 1GH by shoving fanspeeds up higher, but how much higher can they run before wearing down the fans?
Something I was going to build until I figured out, that the cards would barely
cover electricity costs in the place, where I live...
UPDATE: My card now steadily cranks out 324.6Mh/s @ 1010mhz core with 360mhz mem, all at stock voltage. Updating the phatk kernel is key to maxing out hash rate, going to the current version bumped me up a solid 10Mh/s, not bad for simply updating phatk.
Yeah, I was able to squeeze 300 Mhash/s @ 970 MHz core with 600 MHz mem from my 5830,
however I'm paying $0.30 per kWh, so basically the card worked just to cover electricity cost...
Oh, these power hungry 5830s.
The geeky "will work for bandwidth" should be "will work for kilowatthours" nowadays...