I just wonder how fast these chips will go with ice cold air directed at the bottom side of the board. 
Topic: Petition: BFL SC Single should be 65Ghash and not 60Ghash (Read 10489 times)
I hope the thermal protection doesn't cut off if temps are too low either, as I'd like to run these in my (currently) GPU-heated garage I just wonder how fast these chips will go with ice cold air directed at the bottom side of the board.
I just wonder how fast these chips will go with ice cold air directed at the bottom side of the board. 
After seeing pics of the Single SC and Little Single SC-
Thermal Characteristics:
Correct me if I'm wrong, but I believe the limiting factor for overclocking these ASIC chips will be the plastic top layer on the chip itself.
The capability of that plastic to handle heat may be a limiting factor in overclocking.
Edit:
It also has thermal protection built in, so that might also be an issue if you are overclocking to a certain degree. hopefully there will be a way to disable it, though why anyone would want to is....a good question.
The Single SC and Little Single SC appear to be using the same board, same cooling. My best guess is that limiting or increasing speeds will be done by firmware, which can be upgraded or modified regardless of how many ASICS are present on a Single-SC type configuration.
After seeing pics of the Single SC and Little Single SC-
Thermal Characteristics:
Correct me if I'm wrong, but I believe the limiting factor for overclocking these ASIC chips will be the plastic top layer on the chip itself. The heat pipe cooling system seems to be overkill no matter if it's 1 or 8 ASICs, but this is to make up for not having a cooling system on the bottom side of the chip, or a chip that is more efficient with dissipating heat through the top layer. With regard to what it means for this thread, I believe a BFL Single SC should run and be just as overclockable as a Little Single. Neither the board, nor power supply should prove to be the limiting factors. The limiting factor should be the thermal characteristics within the plastic-clad ASIC chips themselves (as superior as these chips may be
).
Assuming everything above is true, 2 devices that have exactly half the power and use more watts from the outlet = not convenient. I understand the idea of reducing variance of earnings by spreading the load over more BFL units. That makes sense. But for people who want to reduce the amount of work, the amount of complexity of setup- the BFL Single SC is the best value. I guess it depends on where your priorities are.
To cool each of these units, I intend to use high air flow (large squirrel cage fan) and/ or waterblock cooling (if available), and whatever future ???s there may be. I want the maximum processing power in the smallest space, because much cooling will be focused on the air mass within that space. Also, hubs are just a weak point.
Since the hard drive analogy didn't drive it home, how about this one:
You own a metal recycling business. If, for about the same price, you could buy Two Ford Ranger Trucks that get 30mpg in order to haul 500lbs of scrap metal each or One Ford F-150 Truck that gets 21mpg and can haul 1000lbs of scrap metal, which one do you pick? Whatever the choice may be, do you ask Ford to increase the price of the Ranger, or limit its performance? It is already limited by inefficiency, because the ONLY job you intend to do (ASIC) is haul massive amounts of scrap metal from point A (scap metal site) to point B (your local metal recycling facility which happens to pay out in BTC). There is no room in this analogy for using the truck to run an errand picking up groceries at walmart.
MPG = miles per gallon = MH/W
After seeing pics of the Single SC and Little Single SC-
Thermal Characteristics:
Correct me if I'm wrong, but I believe the limiting factor for overclocking these ASIC chips will be the plastic top layer on the chip itself. The heat pipe cooling system seems to be overkill no matter if it's 1 or 8 ASICs, but this is to make up for not having a cooling system on the bottom side of the chip, or a chip that is more efficient with dissipating heat through the top layer. With regard to what it means for this thread, I believe a BFL Single SC should run and be just as overclockable as a Little Single. Neither the board, nor power supply should prove to be the limiting factors. The limiting factor should be the thermal characteristics within the plastic-clad ASIC chips themselves (as superior as these chips may be
). Assuming everything above is true, 2 devices that have exactly half the power and use more watts from the outlet = not convenient. I understand the idea of reducing variance of earnings by spreading the load over more BFL units. That makes sense. But for people who want to reduce the amount of work, the amount of complexity of setup- the BFL Single SC is the best value. I guess it depends on where your priorities are.
To cool each of these units, I intend to use high air flow (large squirrel cage fan) and/ or waterblock cooling (if available), and whatever future ???s there may be. I want the maximum processing power in the smallest space, because much cooling will be focused on the air mass within that space. Also, hubs are just a weak point.
Since the hard drive analogy didn't drive it home, how about this one:
You own a metal recycling business. If, for about the same price, you could buy Two Ford Ranger Trucks that get 30mpg in order to haul 500lbs of scrap metal each or One Ford F-150 Truck that gets 21mpg and can haul 1000lbs of scrap metal, which one do you pick? Whatever the choice may be, do you ask Ford to increase the price of the Ranger, or limit its performance? It is already limited by inefficiency, because the ONLY job you intend to do (ASIC) is haul massive amounts of scrap metal from point A (scap metal site) to point B (your local metal recycling facility which happens to pay out in BTC). There is no room in this analogy for using the truck to run an errand picking up groceries at walmart.
MPG = miles per gallon = MH/W
I vote with my dollars. Bump the Singles to 65 Gh and I'm a happy man!
Wouldn't it be hilarious if the 30GH units were identical in every way to the 60GH units, except that there are 4 processors instead of 8... which could be manually placed if you could get a hold of them.
Actually I am pretty sure that's how they will look like. They could even be completely identical from the outside, the only difference being the 30Hh/s unit has some faulty cores disabled.
That's much more likely - little single is the way to get rid of under-performing singles... disable some acis - downclock to get the right hash rate... or if a chip isn't stable at the 60gh range, just underclock it by 50% and sell it for half price.
The real petition title should read " BFL SC singleshould be shipping by now"
If it turns out they mine 3-5% faster, there will be a lot of pissed off people.
But even they perform a bit faster, two littles will also consume more energy while doing so. While the asics sure are the main energy dump, parts like voltage transformation, USB connectivity etc. also need energy, and when using two littles instead of one single you are doubling those, too.
The original post makes a point, but they should just increase the price of the little single. It will be more of a issue if 2x little singles perform better then 1x single, which is very possible. If it turns out they mine 3-5% faster, there will be a lot of pissed off people. I really don't see their logic in this, unless they plan to phase out the single. Letting people who ordered Jalapeno's upgrade and cut in line is also not cool. To those ignorant people complaining about this thread, this is not like walking into walmart and complaining about prices. This is like the guy behind you in line at walmart getting a better deal then you before you even got through the register. A business should always give the best offer possible up to the point of completing the transaction.
Thank you, good to see some people understand the point of the topic.
It's a good idea to increase the price, I didn't think of it, however orders have already been placed, so it's probably too late for that.
The original post makes a point, but they should just increase the price of the little single. It will be more of a issue if 2x little singles perform better then 1x single, which is very possible. If it turns out they mine 3-5% faster, there will be a lot of pissed off people. I really don't see their logic in this, unless they plan to phase out the single. Letting people who ordered Jalapeno's upgrade and cut in line is also not cool. To those ignorant people complaining about this thread, this is not like walking into walmart and complaining about prices. This is like the guy behind you in line at walmart getting a better deal then you before you even got through the register. A business should always give the best offer possible up to the point of completing the transaction.
speechless, I am simply speechless. 
It looks like at least 1 good thing came from this thread.
already asked, they wont
Wouldn't it be hilarious if the 30GH units were identical in every way to the 60GH units, except that there are 4 processors instead of 8... which could be manually placed if you could get a hold of them.
Actually I am pretty sure that's how they will look like. They could even be completely identical from the outside, the only difference being the 30Hh/s unit has some faulty cores disabled.
Maybe we can get them to sell us the chips for $10 a piece and we can build our own PCB's.
Wouldn't it be hilarious if the 30GH units were identical in every way to the 60GH units, except that there are 4 processors instead of 8... which could be manually placed if you could get a hold of them.
Actually I am pretty sure that's how they will look like. They could even be completely identical from the outside, the only difference being the 30Hh/s unit has some faulty cores disabled.
Wouldn't it be hilarious if the 30GH units were identical in every way to the 60GH units, except that there are 4 processors instead of 8... which could be manually placed if you could get a hold of them.
Actually, running a cooler system would reduce the failure rate. If that is the case then a 30Gh/s would have a lower fail rate on heat load alone...wouldn't it?
If that is the case, then mostly yes, but I my experience that seldom is the case
If the user does not extremely overclock her devices, in my experience the most common part of electronics to have a defect are capacitors and/or other parts of voltage regulation on a board. Considering that the 30Gh/s units will be mostly the same as the 60Gh/s ones (except the number of asics), the more units the user has the more capacitors etc she has, each having the same probability of failure. Double the capacitors in your system, and you double that probabilty.
[Speculation]
I have been thinking it over for some time.
What I think might have happened is that the old 40Gh/s rigs were probably not converted to 60Gh/s rigs. They were probably flashed....down to 30Gh/s. Hence the sudden introduction of the Little SC.
Meanwhile the new 60Gh/s Single SC probably comes from either a higher performing second batch of chips or from a second batch that is still in the process of being fabricated.
If you re-read the statement from BFL_Josh (Inaba) this hypothesis makes some sense.
They paid extra to expedite their ASIC run through the chip plant. Then he states that they had to adjust and carefully consider some of the components that make up the boards in the updated 60Gh/s systems.
If it were simply a straightforward flash of firmware...why mention the hardware components being sourced. Shouldn't that have been an issue in the prior stage of the development process?
----------------------------
It sounds to me like there are actually two batches of boards and they are (realistically speaking) almost identical except for a few minor hardware changes in components and the firmware revisions.
The reason that BFL probably doesn't want to double the amount of Little Singles (at this time) being ordered is because they don't have enough formerly 40GH/s systems that have been downgraded (if my hypothesis is correct). Clearly they would need to get rid of their overstock of previously spec'ed hardware at 40Gh/s. (Now re-spec'ed at 30Gh/s)
Which means turning lemons into lemonade by offering it to their Jalapeno customers as a free "upgrade". If that is the case, then it is a good way to offload a bunch of Single SCs that would never have made the 60Gh/s spec without changing parts to support the higher electrical load.
--------------------------
Here is the kicker though. What if BFL has tons of jalapenos laying around? I assume the bulk of their orders are for Jalapenos since they are significantly cheaper to the community.
If my speculation holds water, and they are giving the first downgraded Singles at $650. Either they are accepting a major loss, or the Singles are much cheaper to make than anyone realizes. Certainly cheaper than half the $1299 price.
If all this is anywhere near true, (and it is just speculation) then in the future they will have to give the jalapenos away at a cheaper price than $150. Unless they want to sit on stock that isn't going anywhere fast.
It also would mean that the Little Single owners are likely to get their orders first OR that BFL has them sitting in cartons sitting in some warehouse and is actually waiting for the 60Gh/s batch to finish being assembled. Which if so, means delays and missing the October and possibly November delivery dates.
I have been thinking it over for some time.
What I think might have happened is that the old 40Gh/s rigs were probably not converted to 60Gh/s rigs. They were probably flashed....down to 30Gh/s. Hence the sudden introduction of the Little SC.
Meanwhile the new 60Gh/s Single SC probably comes from either a higher performing second batch of chips or from a second batch that is still in the process of being fabricated.
If you re-read the statement from BFL_Josh (Inaba) this hypothesis makes some sense.
They paid extra to expedite their ASIC run through the chip plant. Then he states that they had to adjust and carefully consider some of the components that make up the boards in the updated 60Gh/s systems.
If it were simply a straightforward flash of firmware...why mention the hardware components being sourced. Shouldn't that have been an issue in the prior stage of the development process?
----------------------------
It sounds to me like there are actually two batches of boards and they are (realistically speaking) almost identical except for a few minor hardware changes in components and the firmware revisions.
The reason that BFL probably doesn't want to double the amount of Little Singles (at this time) being ordered is because they don't have enough formerly 40GH/s systems that have been downgraded (if my hypothesis is correct). Clearly they would need to get rid of their overstock of previously spec'ed hardware at 40Gh/s. (Now re-spec'ed at 30Gh/s)
Which means turning lemons into lemonade by offering it to their Jalapeno customers as a free "upgrade". If that is the case, then it is a good way to offload a bunch of Single SCs that would never have made the 60Gh/s spec without changing parts to support the higher electrical load.
--------------------------
Here is the kicker though. What if BFL has tons of jalapenos laying around? I assume the bulk of their orders are for Jalapenos since they are significantly cheaper to the community.
If my speculation holds water, and they are giving the first downgraded Singles at $650. Either they are accepting a major loss, or the Singles are much cheaper to make than anyone realizes. Certainly cheaper than half the $1299 price.
If all this is anywhere near true, (and it is just speculation) then in the future they will have to give the jalapenos away at a cheaper price than $150. Unless they want to sit on stock that isn't going anywhere fast.
It also would mean that the Little Single owners are likely to get their orders first OR that BFL has them sitting in cartons sitting in some warehouse and is actually waiting for the 60Gh/s batch to finish being assembled. Which if so, means delays and missing the October and possibly November delivery dates.
Second, it's not the $1 that matters, but value for money, which comes in the form of redundancy: if a Single fails and has to be sent back to BFL for repair, than it's better to have 2x30Ghash units than 1x60Ghash.
I'd say its the other way: Considering each complete unit has about the same probability of failure over a given timeframe, with 2*30GHash units you have double the probability of failure than with 1x60Ghash. I therefore demand the 1x60Ghash unit to be downclocked to 55Gh/s to reflect the time and money saved through lower failure rates
The only macro-level parts "moving" in a BFL system is probably the fans that cool the system. The electronics themselves are produced using the same method and process. If BFLs failure rate were 1 in 10 units. Then doubling the number of units to 2 would increase your odds of a failure to 2 in 20. But you would have a 18 in 20 chance that the hardware would be fine through its "useful life".
If a person makes an argument that doubling something increases the failure rate then they are admitting that the hardware itself has a very high rate of failure.
The argument used in [redundant] RAID configurations is built on the idea of increasing redundancy removes any one point of failure. It is also why most servers have more than 1 power supply for fail over.
-------------------------------------------
If you have 1 Single SC and you suffer a single failure you are down. Period.
If you have 2 Little Single then you suffer from one point of failure and still take in Bitcoins at a reduced rate. (But mining continues regardless).
I wasn't originally going to say anything but some of the ideas expressed previously in this and other threads are not well founded.
Second, it's not the $1 that matters, but value for money, which comes in the form of redundancy: if a Single fails and has to be sent back to BFL for repair, than it's better to have 2x30Ghash units than 1x60Ghash.
I'd say its the other way: Considering each complete unit has about the same probability of failure over a given timeframe, with 2*30GHash units you have double the probability of failure than with 1x60Ghash. I therefore demand the 1x60Ghash unit to be downclocked to 55Gh/s to reflect the time and money saved through lower failure rates
Long time lurker, infrequent poster Stopping in to check on the forums during this last year has been entertaining, but what I have to say is usually answered by reading other posts. However, this thread is especially stupid. It's good to finally be a part of it 
+2 Stopping in to check on the forums during this last year has been entertaining, but what I have to say is usually answered by reading other posts. However, this thread is especially stupid. It's good to finally be a part of it
Long time lurker, infrequent poster Stopping in to check on the forums during this last year has been entertaining, but what I have to say is usually answered by reading other posts. However, this thread is especially stupid. It's good to finally be a part of it
Stopping in to check on the forums during this last year has been entertaining, but what I have to say is usually answered by reading other posts. However, this thread is especially stupid. It's good to finally be a part of it Jump to: