News Update:
http://us4.campaign-archive1.com/?u=6d9133088733c220dff716f59&id=e4c910f077&e=276eecccbc
Topic: Swedish ASIC miner company kncminer.com - page 1103. (Read 3050068 times)
So getting to the bottom of the 4/8 VRM thing, it was just brought up within an interview for a new engineer, which was cool as I have an answer from Marcus and it was definitely not a cost saving issue. The reason; KnC bought the entire world's supply from General Electric of those VRMs and could not get anymore without a month's wait, meaning a lot of people would have been without their kit if every board required them. They were indeed a margin and as you can see not necessary to reach the promised spec. GE could not supply more for any money, and there was actually a scary period where it was believed 8 may be necessary, in which case then there would have been a problem fulfilling all orders.
Thanks! I think you did post them but I couldn't remember if it was in this thread or somewhere else..but thank you for indulging me
And that's quite different from the reading in Sweden @ 850-ish watts, so I think one can safely say that there was something different between those two, and by extension what eventually made it out for customers. And thanks for the background on the VRMs...very interesting and makes perfect sense in the timing of it and why it was rolled out without extensive testing. I withdraw my "reckless" statement and replace it with "unfortunately timed".And finally we have an answer. Not the endless ignorant mocking of orama about this issue. (planking anyone?)
KNC just had to own up and say, "guys, we are low on VRMs but good news we can still power the chip fine with 4. But, there is a power issue we are working to fix with firmware ASAP so please use a higher than 850W PSU to be safe or just wait until we get the firmware fixed before running your miner. We know you can't wait, so please keep the VRMs cool just in case. (they are running 30% over spec in regard to amps out) Heat on the chip will stay safely in spec, no worries there."
As I said all along, just own up to it. Thanks finally
Nice to get some background info about the VRMs thing.
Thanks!
These guys have been pulling allnighters. I hate you all. By the way, there is a reason as to why there was a slight delay. Rem I said Marcus is a perfectionist, and Sam is the one cracking whip about delivery being the greater priority. Well some middle ground was reached somewhere in between which I can't talk about. I'll let Sam take the honour there. 
I'm still curious what else took place in September..
More background stories would be really appreciated as it helps to understand better what KnC actually achieved and under which conditions..
edit:
If Sam would have some time left (not realistic I assume), he should write a book about it or even record an audio book together with all original characters involved - including O'rama (on open day & at last weekend of September chapter)
I'd even preorder one in hope BFL doesn't ship one earlier --with approx. same page-size, number of word per page and all those relevant specs-- as it might give me a better ROI (return on invested-time [fore reading])
"BFL: Preorders - The Golden Cash Cow!" - hey Josh if you read this and want to go for it, hereby I claim the copyright on this title
I mean it has potential to become part of history if Bitcoin will become relevant to the masses at some point in time.
And better they capture reality themselves than some smug making up a deformed story for profit only several years later..
Knc miners
Sold out
Sold out
Let's hope their next batch will be available in March and it will be a better version of their 28nm chip! Or at least wait until January.
agreed, you're right
I did not consider the bad situation for those with a <400GH/s miner taking a certain risk and now there are the ones coming late, have no risk and get working devices with >550GH/s.
I think, too, a discount would have been appropriate.
Still hoping for one on gen2...let's see.
I did not consider the bad situation for those with a <400GH/s miner taking a certain risk and now there are the ones coming late, have no risk and get working devices with >550GH/s.
I think, too, a discount would have been appropriate.
Still hoping for one on gen2...let's see.
(...)
Yep, I understood... In this way you are correct.
But if you make a revision for the new machines and boards and you can increase the hash speed with better quality components there will be a lot of bitching from the early customers (investors) if you sell it for the non investors. (For example who payed in the 7 days.)
So you have to make a good decision how to solve this without hurt your investors (such as me).
I think you know what I mean. I won't be bitching for compensate or something but give a better product from "our" money wouldn't be correct.
I know, I paid for 400, i got 560. That's nice but I (we) took risk, the new buyers not.
Still no offense, just talk. (As you know now why I'm took the question how many box per day when you were at the open day)
Yep, I understood... In this way you are correct.
But if you make a revision for the new machines and boards and you can increase the hash speed with better quality components there will be a lot of bitching from the early customers (investors) if you sell it for the non investors. (For example who payed in the 7 days.)
So you have to make a good decision how to solve this without hurt your investors (such as me).
I think you know what I mean. I won't be bitching for compensate or something but give a better product from "our" money wouldn't be correct.
I know, I paid for 400, i got 560. That's nice but I (we) took risk, the new buyers not.
Still no offense, just talk. (As you know now why I'm took the question how many box per day when you were at the open day)
I do not get your point.
Improvements take place in many steps, this is one of them.
You want KnC not to improve the hw any further ? Really ??
I can understand your point of view "I payed for initial funding, now the late comers have no risk and better products", but it misses the important aspect that some 1st batch customers already have reached break-even.
Difficulty increases, so improved hardware is an advantage for yourself in case you consider buying again.
And I highly doubt that a Nov Jupiter will return same amount of coins as an October one; some GH/s for batch#2 more will not change it.
Still no offense... I don't like trolls and offensive talking.
Improvements has to be done if any options for it. That's for sure.
As we paid for initial funding, we paid the risk, we invested into a no name company. After that we got prototypes. For me it was not a problem, I got a good one, I like to play it but there were a lot of guys with no technical skill to do the same.
I understand why they sent the boxes early without fine tuning. It was really better to hash away and play with it and not to wait a month until they tune them.
We paid for the the 2nd batch buyers with money and experience.
Personally I don't care about the improved gen1 cause I won't buy. I'll wait for gen2. But if someone from the early investors wish to buy from the improved gen1, a SMALL discount could be appriciated.
And yes, some of the batch 1 customers had broke even (me too) but don't forget why this happened....
So getting to the bottom of the 4/8 VRM thing, it was just brought up within an interview for a new engineer, which was cool as I have an answer from Marcus and it was definitely not a cost saving issue. The reason; KnC bought the entire world's supply from General Electric of those VRMs and could not get anymore without a month's wait, meaning a lot of people would have been without their kit if every board required them. They were indeed a margin and as you can see not necessary to reach the promised spec. GE could not supply more for any money, and there was actually a scary period where it was believed 8 may be necessary, in which case then there would have been a problem fulfilling all orders.
Thanks! I think you did post them but I couldn't remember if it was in this thread or somewhere else..but thank you for indulging me
And that's quite different from the reading in Sweden @ 850-ish watts, so I think one can safely say that there was something different between those two, and by extension what eventually made it out for customers. And thanks for the background on the VRMs...very interesting and makes perfect sense in the timing of it and why it was rolled out without extensive testing. I withdraw my "reckless" statement and replace it with "unfortunately timed".And finally we have an answer. Not the endless ignorant mocking of orama about this issue. (planking anyone?)
KNC just had to own up and say, "guys, we are low on VRMs but good news we can still power the chip fine with 4. But, there is a power issue we are working to fix with firmware ASAP so please use a higher than 850W PSU to be safe or just wait until we get the firmware fixed before running your miner. We know you can't wait, so please keep the VRMs cool just in case. (they are running 30% over spec in regard to amps out) Heat on the chip will stay safely in spec, no worries there."
As I said all along, just own up to it. Thanks finally
Does anyone know if it is OK to connect 2 of my modules from Jupiter 1 to Jupiter 2. I will move just the ribbon data cables over, so Jupiter 2 has 6 modules connected to it, but 2 of these are still powered by Jupiter 1 PSU.
Anyone?
I am sure you'll eventually try it, so let us know how it goes! I am suprised you have a board that actually supports 6 connections, my day 1 jup only has 4.
I just did. 6 boards using all 6 connection, using 1 V850 PSU.
Good to know. Thanks for testing. I asked KnC support if it was possible and they said the firmware did not support 6 modules
What's the hashrate for 6 modules ?
Strange... I clearly remember hno (OrSoC engineer) saying, on irc, that as long as the control board has 6 plugs the miners should work.
Does anyone know if it is OK to connect 2 of my modules from Jupiter 1 to Jupiter 2. I will move just the ribbon data cables over, so Jupiter 2 has 6 modules connected to it, but 2 of these are still powered by Jupiter 1 PSU.
Anyone?
I am sure you'll eventually try it, so let us know how it goes! I am suprised you have a board that actually supports 6 connections, my day 1 jup only has 4.
I just did. 6 boards using all 6 connection, using 1 V850 PSU.
Good to know. Thanks for testing. I asked KnC support if it was possible and they said the firmware did not support 6 modules
What's the hashrate for 6 modules ?
Well it is currently increasing.
The 4 boards together were hashing at 565GH/s stable. I added 2 more from my under-performing miner (420GH/s with 3 hashing at around 107GH/s and 1 at around 100GH/s)
So I expect the 6 boards to reach around 770GH/s which might not happen, because:
Jupiter 1 (the good one) was using 0.98.1 Beta provided by Bitcoinorama = 565GH/s
Jupiter 2 (the bad one) was using 0.98 = 420GH/s. It was even less with any other firmware.
Does anyone know if it is OK to connect 2 of my modules from Jupiter 1 to Jupiter 2. I will move just the ribbon data cables over, so Jupiter 2 has 6 modules connected to it, but 2 of these are still powered by Jupiter 1 PSU.
Anyone?
I am sure you'll eventually try it, so let us know how it goes! I am suprised you have a board that actually supports 6 connections, my day 1 jup only has 4.
I just did. 6 boards using all 6 connection, using 1 V850 PSU.
me too, it can work like that
Probably, but I am just testing my under-performing boards from my other Jupiter.
I also just shut it down and connected another PSU just to be on the safe side.
Does anyone know if it is OK to connect 2 of my modules from Jupiter 1 to Jupiter 2. I will move just the ribbon data cables over, so Jupiter 2 has 6 modules connected to it, but 2 of these are still powered by Jupiter 1 PSU.
Anyone?
I am sure you'll eventually try it, so let us know how it goes! I am suprised you have a board that actually supports 6 connections, my day 1 jup only has 4.
I just did. 6 boards using all 6 connection, using 1 V850 PSU.
that is really handy to know! considering I am a PSU short but have 4
Umm, that's cool, but I can assure you not that's clearly not he case on all units. I have pictures of the one I took to Atlanta using a Kill-A-Watt, and there were enough witnesses that were present, including members of this forum (Bargraphics, Phin Gage), and representatives from competing companies that saw this for their own eyes. Also there's plenty in this thread amongst the first recipients to verify otherwise.
And I'd be willing to bet a few satoshi's that the unit you took to Atlanta did not have the same firmware as those units that were subsequently shipped out to Day 1 customers onwards...and that the voltage was tweaked upwards in the interim thus increasing the power requirements (in the name of increasing performance/stability, I'm sure). Couple the fact that the actual physical design of the product was changed quite substantially (8 VRMs to 4 VRMs) and it's not exactly apples-to-apples
No, you are wrong. It was a standard unit I made when at the manufacturing facility. By that point I had not seen my hotel for a week, and certainly no one had had anytime or were confident enough to tweak a unit. Literally decided on going. Booked a flight at 11pm, the night before, for an 11am take-off. Then searched for a hotel, made a device, drove several hours back to Stockholm, washed, packed and made my way to the airport. Thing is no way was a clocked machine risked that early. We had only seen the chips 5 days earlier and just had to have something that worked. Why if something performed well, would it not be given to customers. Austin and Beccy Craig from Life on Bitcoin have had it in their possession since that Atlanta conference. Can we please stop this 8/4 VRM nonsense. The additional 4 vrms were surplus to reaching spec. They were in place in the beginning in case they were needed to achieve the spec. They weren't so they aren't.
Well that is surprising given the results that later cropped up with overcurrent PSU's and exploding capacitors..do you have a link to those kill-a-watt photos from Atlanta? I don't remember the values you saw there versus what was in the youtube video in Sweden. And just to clarify..when I said "tweaked" I meant subsequent production machines, not the one you took to Atlanta. In other words, my assumption was that the initial "beta" firmware or whatever you had then was running at say .8V and then when you guys decided on the production values for the boxes that were going to start shipping it was bumped to .9V for the added stability/speed benefits that brings. I wasn't suggesting that you custom-tuned the Atlanta box. But however it worked out in the order of events...there were definitely some growing pains in relation to power at the PSU and the power through the VRMs that had to be ironed out, as I know you are extremely familiar with
As for the 8/4 thing...well, you keep saying it didn't matter when the facts showed that it did. There was a material difference in how those things ran before/after the change which had to be accounted for in new firmware. Again, I don't think it was some malicious conspiracy or anything..but I do think it was a reckless change that was done without adequate testing which caused issues that had to be addressed further down the road.
Sure, as below, pretty sure I posted one of these before as I took a few for that reason. It was hard to balance the non-backlit LCD screen, with the bright Macbook Air Screen. This is with the Screen brightness turned down to it's lowest, so sorry there is two. Again there were a coupe of hundred people at the conference that witnessed the device's performance on the day.
So getting to the bottom of the 4/8 VRM thing, it was just brought up within an interview for a new engineer, which was cool as I have an answer from Marcus and it was definitely not a cost saving issue. The reason; KnC bought the entire world's supply from General Electric of those VRMs and could not get anymore without a month's wait, meaning a lot of people would have been without their kit if every board required them. They were indeed a margin and as you can see not necessary to reach the promised spec. GE could not supply more for any money, and there was actually a scary period where it was believed 8 may be necessary, in which case then there would have been a problem fulfilling all orders.
Thanks! I think you did post them but I couldn't remember if it was in this thread or somewhere else..but thank you for indulging me
And that's quite different from the reading in Sweden @ 850-ish watts, so I think one can safely say that there was something different between those two, and by extension what eventually made it out for customers. And thanks for the background on the VRMs...very interesting and makes perfect sense in the timing of it and why it was rolled out without extensive testing. I withdraw my "reckless" statement and replace it with "unfortunately timed". 
Knc miners
Sold out
Sold out
I do not doubt mods will occur, it's inevitable, and if all our boards start mysteriously dying of a similar symptom not present in our own they will be inspected for signs of stress, and if found to have been consequence of unofficial modding, then obviously RMA will be invalid.
Latter revisions of boards will undoubtedly be pushed harder as we learn the loads the chip is capable of, the chip itself is not the issue, but the boards themselves will require supporting components to undergo continuous revision.
Isn't 144GH/s per chip the maximum possible? I thought you can't push it faster because of the way it's built.
144ghs is the max based on each core running at 750mhz. if this was pushed higher, the hashrate could go way up.
...And Bananas could be Oranges if re-shaped, re-colored and re-flavored.
Yes, some units could run faster, some would run slower and others would not run at all.
You have to look at this the same way you would look at PC overclocking... You have to margin out the processor for best performance, error free performance and most efficient. then you could make a decision as to whether overclocking would be beneficial on your machine with a greater or lesser clock rate -- and perhaps greater or lesser voltage applied to the processor. The you would have to determine if the cooling were still adequate.
The reason some machine might perform better at a lower clock rate is if they were running with hardware errors that would be resolved by a slightly slower clock rate -- you have to look at total throughput.
The after all is said and done -- you look at the additional (or lower) power draw and see if you have better efficiency -- i.e. more earnings.
Finally -- running at the higher rates may actually damage the machine. If the damage would not be significant until after the machine has served it's useful mining life then a decision is easy.
Everything is a risk -- doing nothing -- or doing everything possible -- or a compromise.
All the decisions and non-decisions carry risk...
...Or you can flip coins and take chances.
PS: I think TexasLabRat is giving a pretty sensible approach to the power supply issue.
Does anyone know if it is OK to connect 2 of my modules from Jupiter 1 to Jupiter 2. I will move just the ribbon data cables over, so Jupiter 2 has 6 modules connected to it, but 2 of these are still powered by Jupiter 1 PSU.
Anyone?
I am sure you'll eventually try it, so let us know how it goes! I am suprised you have a board that actually supports 6 connections, my day 1 jup only has 4.
I just did. 6 boards using all 6 connection, using 1 V850 PSU.
Good to know. Thanks for testing. I asked KnC support if it was possible and they said the firmware did not support 6 modules
What's the hashrate for 6 modules ?
Does anyone know if it is OK to connect 2 of my modules from Jupiter 1 to Jupiter 2. I will move just the ribbon data cables over, so Jupiter 2 has 6 modules connected to it, but 2 of these are still powered by Jupiter 1 PSU.
Anyone?
I am sure you'll eventually try it, so let us know how it goes! I am suprised you have a board that actually supports 6 connections, my day 1 jup only has 4.
I just did. 6 boards using all 6 connection, using 1 V850 PSU.
me too, it can work like that
Does anyone know if it is OK to connect 2 of my modules from Jupiter 1 to Jupiter 2. I will move just the ribbon data cables over, so Jupiter 2 has 6 modules connected to it, but 2 of these are still powered by Jupiter 1 PSU.
Anyone?
I am sure you'll eventually try it, so let us know how it goes! I am suprised you have a board that actually supports 6 connections, my day 1 jup only has 4.
I just did. 6 boards using all 6 connection, using 1 V850 PSU.
Does anyone know if it is OK to connect 2 of my modules from Jupiter 1 to Jupiter 2. I will move just the ribbon data cables over, so Jupiter 2 has 6 modules connected to it, but 2 of these are still powered by Jupiter 1 PSU.
Anyone?
I am sure you'll eventually try it, so let us know how it goes! I am suprised you have a board that actually supports 6 connections, my day 1 jup only has 4.
From what I've read about GPU rigs it should be possible, as this one is running with 3PSUs but one mainboard only
http://www.techspot.com/gallery/member-galleries/p4037-here-ya-go-supersmashb.html
Does anyone know if it is OK to connect 2 of my modules from Jupiter 1 to Jupiter 2. I will move just the ribbon data cables over, so Jupiter 2 has 6 modules connected to it, but 2 of these are still powered by Jupiter 1 PSU.
Anyone?
I am sure you'll eventually try it, so let us know how it goes! I am suprised you have a board that actually supports 6 connections, my day 1 jup only has 4.
I do not doubt mods will occur, it's inevitable, and if all our boards start mysteriously dying of a similar symptom not present in our own they will be inspected for signs of stress, and if found to have been consequence of unofficial modding, then obviously RMA will be invalid.
Latter revisions of boards will undoubtedly be pushed harder as we learn the loads the chip is capable of, the chip itself is not the issue, but the boards themselves will require supporting components to undergo continuous revision.
Isn't 144GH/s per chip the maximum possible? I thought you can't push it faster because of the way it's built.
144ghs is the max based on each core running at 750mhz. if this was pushed higher, the hashrate could go way up.
(...)
Yep, I understood... In this way you are correct.
But if you make a revision for the new machines and boards and you can increase the hash speed with better quality components there will be a lot of bitching from the early customers (investors) if you sell it for the non investors. (For example who payed in the 7 days.)
So you have to make a good decision how to solve this without hurt your investors (such as me).
I think you know what I mean. I won't be bitching for compensate or something but give a better product from "our" money wouldn't be correct.
I know, I paid for 400, i got 560. That's nice but I (we) took risk, the new buyers not.
Still no offense, just talk. (As you know now why I'm took the question how many box per day when you were at the open day)
Yep, I understood... In this way you are correct.
But if you make a revision for the new machines and boards and you can increase the hash speed with better quality components there will be a lot of bitching from the early customers (investors) if you sell it for the non investors. (For example who payed in the 7 days.)
So you have to make a good decision how to solve this without hurt your investors (such as me).
I think you know what I mean. I won't be bitching for compensate or something but give a better product from "our" money wouldn't be correct.
I know, I paid for 400, i got 560. That's nice but I (we) took risk, the new buyers not.
Still no offense, just talk. (As you know now why I'm took the question how many box per day when you were at the open day)
I do not get your point.
Improvements take place in many steps, this is one of them.
You want KnC not to improve the hw any further ? Really ??
I can understand your point of view "I payed for initial funding, now the late comers have no risk and better products", but it misses the important aspect that some 1st batch customers already have reached break-even.
Difficulty increases, so improved hardware is an advantage for yourself in case you consider buying again.
And I highly doubt that a Nov Jupiter will return same amount of coins as an October one; some GH/s for batch#2 more will not change it.
Jump to: