kano
Legendary
Offline
Activity: 4592
Merit: 1851
Linux since 1997 RedHat 4
|
|
May 13, 2013, 07:18:59 AM |
|
They don't update here anymore? Just advertise?
My impression is that they're sorting out the chips that ended up lower speed than predicted and trying to figure out how to ship everyone their units without having to pay out the ass for their number error. Then they'll start selling the new, slightly slower, much more expensive units. The new power draw figures aren't out yet (probably in case they have to change again) but it looks like the 5GH/s jalapeno is going to be 30w per the last thing I saw. For reference the new units support up to 16 chips on one board with higher power draw overall. It looked like the Ars Technial jalapeno was 50W... seems like there is still some inconsistency to the models being shipped currently. Again, I expect the guy can't measure power properly. He's the only one with 50W - so unless there is something wrong with his Jalapeno, it's simply a case of using a crappy Watt meter or not being able to read.
|
|
|
|
btharper
|
|
May 13, 2013, 05:19:31 PM |
|
They don't update here anymore? Just advertise?
My impression is that they're sorting out the chips that ended up lower speed than predicted and trying to figure out how to ship everyone their units without having to pay out the ass for their number error. Then they'll start selling the new, slightly slower, much more expensive units. The new power draw figures aren't out yet (probably in case they have to change again) but it looks like the 5GH/s jalapeno is going to be 30w per the last thing I saw. For reference the new units support up to 16 chips on one board with higher power draw overall. It looked like the Ars Technial jalapeno was 50W... seems like there is still some inconsistency to the models being shipped currently. Again, I expect the guy can't measure power properly. He's the only one with 50W - so unless there is something wrong with his Jalapeno, it's simply a case of using a crappy Watt meter or not being able to read. While very possible it could also be the difference between measured at the wall and expected/drawn by each chip (although I would doubt a 40% power waste). Some of what they had mentioned about changes to the metal layer on later batches of chips (read also: not the ones shipping now) is that the later chips had improved power profiles. So these might be the early chips that eat more power. He may just have a bad unit or bad measurement, or there might be higher variability and/or power draw in the first units. TL;DR - I would at least expect later units to draw less J/GH/s
|
|
|
|
kano
Legendary
Offline
Activity: 4592
Merit: 1851
Linux since 1997 RedHat 4
|
|
May 14, 2013, 01:15:14 AM |
|
They don't update here anymore? Just advertise?
My impression is that they're sorting out the chips that ended up lower speed than predicted and trying to figure out how to ship everyone their units without having to pay out the ass for their number error. Then they'll start selling the new, slightly slower, much more expensive units. The new power draw figures aren't out yet (probably in case they have to change again) but it looks like the 5GH/s jalapeno is going to be 30w per the last thing I saw. For reference the new units support up to 16 chips on one board with higher power draw overall. It looked like the Ars Technial jalapeno was 50W... seems like there is still some inconsistency to the models being shipped currently. Again, I expect the guy can't measure power properly. He's the only one with 50W - so unless there is something wrong with his Jalapeno, it's simply a case of using a crappy Watt meter or not being able to read. While very possible it could also be the difference between measured at the wall and expected/drawn by each chip (although I would doubt a 40% power waste). Some of what they had mentioned about changes to the metal layer on later batches of chips (read also: not the ones shipping now) is that the later chips had improved power profiles. So these might be the early chips that eat more power. He may just have a bad unit or bad measurement, or there might be higher variability and/or power draw in the first units. TL;DR - I would at least expect later units to draw less J/GH/s I got one of the Jalapenos in the 2nd small batch sent out after the first 2 Jalapenos - so well before him. Mine reads 32W at the wall (5.4Gh/s)
|
|
|
|
k9quaint
Legendary
Offline
Activity: 1190
Merit: 1000
|
|
May 14, 2013, 02:22:52 AM |
|
They don't update here anymore? Just advertise?
My impression is that they're sorting out the chips that ended up lower speed than predicted and trying to figure out how to ship everyone their units without having to pay out the ass for their number error. Then they'll start selling the new, slightly slower, much more expensive units. The new power draw figures aren't out yet (probably in case they have to change again) but it looks like the 5GH/s jalapeno is going to be 30w per the last thing I saw. For reference the new units support up to 16 chips on one board with higher power draw overall. It looked like the Ars Technial jalapeno was 50W... seems like there is still some inconsistency to the models being shipped currently. Again, I expect the guy can't measure power properly. He's the only one with 50W - so unless there is something wrong with his Jalapeno, it's simply a case of using a crappy Watt meter or not being able to read. While very possible it could also be the difference between measured at the wall and expected/drawn by each chip (although I would doubt a 40% power waste). Some of what they had mentioned about changes to the metal layer on later batches of chips (read also: not the ones shipping now) is that the later chips had improved power profiles. So these might be the early chips that eat more power. He may just have a bad unit or bad measurement, or there might be higher variability and/or power draw in the first units. TL;DR - I would at least expect later units to draw less J/GH/s I got one of the Jalapenos in the 2nd small batch sent out after the first 2 Jalapenos - so well before him. Mine reads 32W at the wall (5.4Gh/s) How hard is it to misread a kill-a-watt? Perhaps it is a multi-chip unit with the chips under-clocked.
|
Bitcoin is backed by the full faith and credit of YouTube comments.
|
|
|
kano
Legendary
Offline
Activity: 4592
Merit: 1851
Linux since 1997 RedHat 4
|
|
May 14, 2013, 03:36:11 AM |
|
... I got one of the Jalapenos in the 2nd small batch sent out after the first 2 Jalapenos - so well before him. Mine reads 32W at the wall (5.4Gh/s)
How hard is it to misread a kill-a-watt? Perhaps it is a multi-chip unit with the chips under-clocked. Clearly for the guy at Ars - misreading a kill-a-watt is pretty easy. Mine is a multi chip (2) unit (30 cores) that is under-clocked.
|
|
|
|
btharper
|
|
May 15, 2013, 04:57:08 AM |
|
... I got one of the Jalapenos in the 2nd small batch sent out after the first 2 Jalapenos - so well before him. Mine reads 32W at the wall (5.4Gh/s)
How hard is it to misread a kill-a-watt? Perhaps it is a multi-chip unit with the chips under-clocked. Clearly for the guy at Ars - misreading a kill-a-watt is pretty easy. Mine is a multi chip (2) unit (30 cores) that is under-clocked. Any options as far as overclocking go? And any news posted about the temperature it runs at?
|
|
|
|
Bicknellski
|
|
May 15, 2013, 05:15:40 AM |
|
They don't update here anymore? Just advertise?
My impression is that they're sorting out the chips that ended up lower speed than predicted and trying to figure out how to ship everyone their units without having to pay out the ass for their number error. Then they'll start selling the new, slightly slower, much more expensive units. The new power draw figures aren't out yet (probably in case they have to change again) but it looks like the 5GH/s jalapeno is going to be 30w per the last thing I saw. For reference the new units support up to 16 chips on one board with higher power draw overall. Klondike.... 16 avalon chips 4.512 GH/s @ 32 watts. Hmmm funny if they work and the start mining en masse by August. I'd be pissed if that happens and I was a BFL customer waiting at the end of the queue. It be nice to see a Garzik like review of any of these "prototypes" that BFL shipped. Funny none have been given a real review. I wonder why?
|
|
|
|
btharper
|
|
May 15, 2013, 06:53:05 AM |
|
As far as wanting to read the kill-a-watt, wired just posted the full image with their review, reading 43.5, meaning a sightly higher reading around 50 may well just be current inter-unit variability. Future variability is anyone's guess though.
|
|
|
|
kano
Legendary
Offline
Activity: 4592
Merit: 1851
Linux since 1997 RedHat 4
|
|
May 15, 2013, 09:28:29 AM |
|
... I got one of the Jalapenos in the 2nd small batch sent out after the first 2 Jalapenos - so well before him. Mine reads 32W at the wall (5.4Gh/s)
How hard is it to misread a kill-a-watt? Perhaps it is a multi-chip unit with the chips under-clocked. Clearly for the guy at Ars - misreading a kill-a-watt is pretty easy. Mine is a multi chip (2) unit (30 cores) that is under-clocked. Any options as far as overclocking go? And any news posted about the temperature it runs at? No clocking control. Mine runs at around 46C +/- 2ish ... 5.4+GH/s ... 32W But seriously - the guy considers himself worth giving a report and he comes up with 50W Sounds more like "Oh I just looked at it sideways and misread it - but does it really matter I'm within 80% of the actual value ..." He's certainly on the list of people I'd take no notice of anything he says. ... and 50W is 'slightly' higher than 43.5W? Seriously? Basically it all means his Kill-A-Watt is screwed (50W ... 43.5W ?!?) and he needs to have a reliable one before reporting ... but hell, why would anyone expect reliable info from someone reporting something ...
|
|
|
|
Fiyasko
Legendary
Offline
Activity: 1428
Merit: 1001
Okey Dokey Lokey
|
|
May 15, 2013, 05:46:36 PM |
|
... I got one of the Jalapenos in the 2nd small batch sent out after the first 2 Jalapenos - so well before him. Mine reads 32W at the wall (5.4Gh/s)
How hard is it to misread a kill-a-watt? Perhaps it is a multi-chip unit with the chips under-clocked. Clearly for the guy at Ars - misreading a kill-a-watt is pretty easy. Mine is a multi chip (2) unit (30 cores) that is under-clocked. Any options as far as overclocking go? And any news posted about the temperature it runs at? No clocking control. Mine runs at around 46C +/- 2ish ... 5.4+GH/s ... 32W But seriously - the guy considers himself worth giving a report and he comes up with 50W Sounds more like "Oh I just looked at it sideways and misread it - but does it really matter I'm within 80% of the actual value ..." He's certainly on the list of people I'd take no notice of anything he says. ... and 50W is 'slightly' higher than 43.5W? Seriously? Basically it all means his Kill-A-Watt is screwed (50W ... 43.5W ?!?) and he needs to have a reliable one before reporting ... but hell, why would anyone expect reliable info from someone reporting something ... Perhaps his BFL unit is internally damaged and its "messing around" with its own power usages due to a faulty component?
|
|
|
|
erk
|
|
May 30, 2013, 04:44:45 AM Last edit: May 30, 2013, 12:08:49 PM by erk |
|
29 May 2013 - ASIC Update
Almost there... I'm attaching a picture of our test board. We've almost got the firmware finalized, as you can see, we've brought power usage down quite a bit. This is a 50 GH/s unit with unbinned chips (meaning the chips were randomly selected, so some of the chips have non-functional engines. What this means is the unit can perform faster if you use binned chips with known good engines).
I think the picture speaks for itself, and I apologize for the quality... just took a quick snap while we were working. However, you can see that our power usage is under 5w/GH/s. This should apply across all product lines. I plan on testing out the short boards tomorrow. We should be getting in many of the reworked short boards tomorrow as well as some long boards. Next week, we expect a much larger shipment of both long and short boards to arrive (possibly even this Saturday, but that's still up in the air.)
In any event, we should have the firmware finalized by tomorrow night or Friday at the latest and ready to go. https://forums.butterflylabs.com/announcements/692-bfl-asic-status-3.htmlSome progress, though the power consumption seems a lot higher than I would have expected. They must be using 25% faster clock rate than the ones in the 5GH/s unit. 250Watts for 50GH/s is about 40% more than I calculated, factoring in the extra fans. Still I guess it's way better than the 620watt 66GH/s some reviews have reported the Avalon doing. The 5GH/s unit has been shown to have 2 x ASIC chips on the board, with space for 8 total, so I presume the 50 GH/s double board has 16 x ASIC chips, so they would need to overclock them 25% or they would only get 40GH/s out of 16 chips. /EDIT Actually the 250 watts is a pretty good figure, I worked it out wrong before, was thinking little single. You are getting effectively 10 Jalapenos worth of hash rate and they would be around 300watts worth if you bought 10 of those. The single seems more efficient.
|
|
|
|
smoothie
Legendary
Offline
Activity: 2492
Merit: 1473
LEALANA Bitcoin Grim Reaper
|
|
May 30, 2013, 05:25:02 AM |
|
29 May 2013 - ASIC Update
Almost there... I'm attaching a picture of our test board. We've almost got the firmware finalized, as you can see, we've brought power usage down quite a bit. This is a 50 GH/s unit with unbinned chips (meaning the chips were randomly selected, so some of the chips have non-functional engines. What this means is the unit can perform faster if you use binned chips with known good engines).
I think the picture speaks for itself, and I apologize for the quality... just took a quick snap while we were working. However, you can see that our power usage is under 5w/GH/s. This should apply across all product lines. I plan on testing out the short boards tomorrow. We should be getting in many of the reworked short boards tomorrow as well as some long boards. Next week, we expect a much larger shipment of both long and short boards to arrive (possibly even this Saturday, but that's still up in the air.)
In any event, we should have the firmware finalized by tomorrow night or Friday at the latest and ready to go. https://forums.butterflylabs.com/announcements/692-bfl-asic-status-3.htmlSome progress, though the power consumption seems a lot higher than I would have expected. They must be using 25% faster clock rate than the ones in the 5GH/s unit. 250Watts for 50GH/s is about 40% more than I calculated, factoring in the extra fans. Still I guess it's way better than the 620watt 66GH/s some reviews have reported the Avalon doing. 11.5 months and counting and still they don't have their shit together?
|
███████████████████████████████████████
,╓p@@███████@╗╖, ,p████████████████████N, d█████████████████████████b d██████████████████████████████æ ,████²█████████████████████████████, ,█████ ╙████████████████████╨ █████y ██████ `████████████████` ██████ ║██████ Ñ███████████` ███████ ███████ ╩██████Ñ ███████ ███████ ▐▄ ²██╩ a▌ ███████ ╢██████ ▐▓█▄ ▄█▓▌ ███████ ██████ ▐▓▓▓▓▌, ▄█▓▓▓▌ ██████─ ▐▓▓▓▓▓▓█,,▄▓▓▓▓▓▓▌ ▐▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▌ ▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓─ ²▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓╩ ▀▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▀ ²▀▀▓▓▓▓▓▓▓▓▓▓▓▓▀▀` ²²² ███████████████████████████████████████
| . ★☆ WWW.LEALANA.COM My PGP fingerprint is A764D833. History of Monero development Visualization ★☆ . LEALANA BITCOIN GRIM REAPER SILVER COINS. |
|
|
|
dan99
|
|
May 30, 2013, 06:00:21 AM |
|
29 May 2013 - ASIC Update
Almost there... I'm attaching a picture of our test board. We've almost got the firmware finalized, as you can see, we've brought power usage down quite a bit. This is a 50 GH/s unit with unbinned chips (meaning the chips were randomly selected, so some of the chips have non-functional engines. What this means is the unit can perform faster if you use binned chips with known good engines).
I think the picture speaks for itself, and I apologize for the quality... just took a quick snap while we were working. However, you can see that our power usage is under 5w/GH/s. This should apply across all product lines. I plan on testing out the short boards tomorrow. We should be getting in many of the reworked short boards tomorrow as well as some long boards. Next week, we expect a much larger shipment of both long and short boards to arrive (possibly even this Saturday, but that's still up in the air.)
In any event, we should have the firmware finalized by tomorrow night or Friday at the latest and ready to go. https://forums.butterflylabs.com/announcements/692-bfl-asic-status-3.htmlSome progress, though the power consumption seems a lot higher than I would have expected. They must be using 25% faster clock rate than the ones in the 5GH/s unit. 250Watts for 50GH/s is about 40% more than I calculated, factoring in the extra fans. Still I guess it's way better than the 620watt 66GH/s some reviews have reported the Avalon doing. 11.5 months and counting and still they don't have their shit together? I guess their shit need time to processed
|
|
|
|
Phinnaeus Gage
Legendary
Offline
Activity: 1918
Merit: 1570
Bitcoin: An Idea Worth Spending
|
|
May 30, 2013, 06:56:17 AM |
|
29 May 2013 - ASIC Update
Almost there... I'm attaching a picture of our test board. We've almost got the firmware finalized, as you can see, we've brought power usage down quite a bit. This is a 50 GH/s unit with unbinned chips (meaning the chips were randomly selected, so some of the chips have non-functional engines. What this means is the unit can perform faster if you use binned chips with known good engines).
I think the picture speaks for itself, and I apologize for the quality... just took a quick snap while we were working. However, you can see that our power usage is under 5w/GH/s. This should apply across all product lines. I plan on testing out the short boards tomorrow. We should be getting in many of the reworked short boards tomorrow as well as some long boards. Next week, we expect a much larger shipment of both long and short boards to arrive (possibly even this Saturday, but that's still up in the air.)
In any event, we should have the firmware finalized by tomorrow night or Friday at the latest and ready to go. https://forums.butterflylabs.com/announcements/692-bfl-asic-status-3.htmlSome progress, though the power consumption seems a lot higher than I would have expected. They must be using 25% faster clock rate than the ones in the 5GH/s unit. 250Watts for 50GH/s is about 40% more than I calculated, factoring in the extra fans. Still I guess it's way better than the 620watt 66GH/s some reviews have reported the Avalon doing. 11.5 months and counting and still they don't have their shit together?
|
|
|
|
johnyj
Legendary
Offline
Activity: 1988
Merit: 1012
Beyond Imagination
|
|
May 30, 2013, 11:16:32 AM |
|
250w at wall is impressive, but 76c degree on the chips! They are pushing it too hard, might need some powerful fans with 200+ cfm, it seems it's not a PWM fan like avalon used
|
|
|
|
Bitcoinorama
|
|
May 30, 2013, 12:48:13 PM |
|
Good to see positive progress for those invested, and for those doubting their financial situation, that one rig alone appears to have netted them $8.5k this month!!! That has to cover at least some overheads...
Do like their GUI.
|
Make my day! Say thanks if you found me helpful BTC Address ---> 1487ThaKjezGA6SiE8fvGcxbgJJu6XWtZp
|
|
|
creativex
|
|
May 30, 2013, 12:58:47 PM |
|
Awesome. Only 7+ months late and 200w over spec.
|
|
|
|
gyverlb
|
|
May 30, 2013, 01:28:11 PM |
|
Some progress, though the power consumption seems a lot higher than I would have expected. They must be using 25% faster clock rate than the ones in the 5GH/s unit. 250Watts for 50GH/s is about 40% more than I calculated, factoring in the extra fans. Still I guess it's way better than the 620watt 66GH/s some reviews have reported the Avalon doing.
Not really comparable. Avalon is a full system, not a hashing device. Reports on 230V are 560W (not sure what the difference is for BFL devices betweens 110V/230V). If you substract a device on which cgminer can run, you're probably left with hundreds of 66GH/s devices at 500W running at customers' location compared to 50GH/s for 250W that aren't yet delivered. With Avalon you even have the choice of cheap Klondike devices (see my sig, I'm betting some of my BTCs on it) delivered in 3 months if ordered now (10 weeks for Avalon chips, 2 weeks for assembly + shipping). BFL won't even provide guesstimates on delivery delays.
|
|
|
|
Photon939
|
|
May 30, 2013, 01:54:12 PM |
|
Hashing controllers are trivial now, the avalon runs on a cheapo $15 china wifi router and Mineforeman's minepeon distro for the raspberry pi gets you a mining controller for $35 that uses 2 watts.
|
|
|
|
mezzomix
Legendary
Offline
Activity: 2702
Merit: 1261
|
|
May 30, 2013, 02:03:09 PM |
|
With Avalon you even have the choice of cheap Klondike devices (see my sig, I'm betting some of my BTCs on it) delivered in 3 months if ordered now (10 weeks for Avalon chips, 2 weeks for assembly + shipping). BFL won't even provide guesstimates on delivery delays.
True. The people waiting for this Klondike device may have their miner long before the june/july 2012 BFL orders are shipped. It doesn't matter that a single consumes "only" 250W if you ordered in june 2012, paid nearly 200 BTC and receive the machine in 2014 (or never).
|
|
|
|
|