Search content
Sort by
Showing 5 of 5 results by FractionalReserve
HX 850 and HX 1050 is of same design, only difference is output as far as I know. So maybe it would be better to say be careful with the Corsair HX series.
Bitcoinorama, if I understand you correctly the issue only occurs after the ATX has shut down autoamtically.
When you then restart them they will do something odd, (i.e Voltage spike or something, Current surge seems odd as this would be an err on the Jupiter side).
this odd thing they are doing are frying the caps (Voltage spike => current surge) ore somethin.
This is exactly how our board was fried.
ATX shut down due to overload (
?)
When restarted... boom, smokes and fireworks.
did I get you right?
Bitcoinorama, if I understand you correctly the issue only occurs after the ATX has shut down autoamtically.
When you then restart them they will do something odd, (i.e Voltage spike or something, Current surge seems odd as this would be an err on the Jupiter side).
this odd thing they are doing are frying the caps (Voltage spike => current surge) ore somethin.
This is exactly how our board was fried.
ATX shut down due to overload (
?) When restarted... boom, smokes and fireworks.
did I get you right?
Then why not leave the 8 VRMs on for that point??
why are you dodging that? what was the main gain for everyone on taking off the 4 VRMs?? You could keep shipping with 8 until you tested taking off 4.. .how can you try to keep avoiding this logic?
why are you dodging that? what was the main gain for everyone on taking off the 4 VRMs?? You could keep shipping with 8 until you tested taking off 4.. .how can you try to keep avoiding this logic?
Simple. KNC chose to use an off the shelf power module. This means rapid prototyping but it also means higher cost. Those modules run $25 each. 8 per board means $200 per board in cost just for the power supply. That is $800 per Jupiter. If they leave 4 of them off that is an extra $400 in profit per unit. 1500 units @ $400 ea = $600,000.
There is absolutely no other reason to change the extensively designed, tested and fully operational boards right in the middle of peak production.
Of course thats the reason.
As I understood it they ran the 8 VRM boards with only 4 VRMs activated and without any problems.
When they skipped mounting 4 of the VRMs in production, suddenly the Voltage levels differed.
That has now been fixed with Firmware 0.95 where Voltage offset function in the VRMs has been used to give correct output.
They had an issue, fixed it. No need to discuss this further?
If you have 4 VRM boards DO UPPGRADE TO 0.95 !!!
Haha, I said soon.
Is this soon enough for you guys =)
The new asic bords with only 4 VRMs [which were designed to use 8] are using to much power. KNC will release a bug fix for this in next (0.95?) Firmware hopefully arriving in short (Today?) A Jupiter in this Config will use app. 890W. So it can be difficult to start with a 850W ATX. (You can still run them with only 3 boards conected).
What happen to margins on margins. The boards were designed to have 320A of DC regulators (which drop the 12V supply down to the 0.75V used by the chip). Now if BFL cheaped out I could see that but KNC has talked for months and months and month about margins on margins so when they actually get to the finish line and start shipping products at the very last minute they decide to yank out all the margins and run the hardware at close to the theoretical limit and hope everything works.
So the solution to an overheating and overloaded power system is to push the power system even harder. To take 160A (which was designed to be 320A) of capacity and run it at 25% over the redline.
So what is is this negative margins on negative margins now?
Quote
The VRMs are also running over spec with app. 50+ A current each! They probably won't burn, but if you can cool them extra until the new firmware is realeased, do it!!
The Asics are tunning to hot (+70 deg C) as they get app, 0.9V instead of 0.7V as they are supposed to. Cool them as much as you can!!! The Problem is due to the VRMs not working according to spec but KNC will be able to fix it with new firmware (we did our own patch for our miners yesterday).
The Asics are tunning to hot (+70 deg C) as they get app, 0.9V instead of 0.7V as they are supposed to. Cool them as much as you can!!! The Problem is due to the VRMs not working according to spec but KNC will be able to fix it with new firmware (we did our own patch for our miners yesterday).
How about just start shipping the proper 8 VRM design and offer replacements to affected customers (who you shipped underperforming units).
Quote
Easiest fix to cool properly is to provide cool air, belo 20 deg C is a good idea. Open case + big fan will also work.
Yeah that seems like the solution. Open case, giant fan, hardware running overspec, and power consumption through the roof. Throw in the cost of high AC load to maintain 20 deg C cooling. Too bad there is no simpler solution like I don't know ...shipping the product as designed, promised, and sold.
Quote
PS, this does not apply to 8 VRM asic boards. No worries here.
So wouldn't the best solution to ship the boards with 8 VRMS as you initially promised they would?
Hi, I see you quoted me from the KNC forum. Just to get thing straight, I am not from KNC!
We just run a huge bunch of Jupiters in our private mining op. We discovered the problems and let KNC now of them asap.
In the meantime we made our own mod where we adjusted the VRMs voltage offset (by software), the 4 VRM boards are running just as good as the 8 VRM boards and poerconsumtion at wall is below 550W.
All that is needed is the next FW uppdate that will fix this. I talked to KNC development (not support!!!) and they are really rushing this!
No need to ship with 8 vrms as there are no need for them, only need for correct FW! Don´t see any point for them to go back to 4 VRMs when it can be easily adjusted in software.
0.94 has a bug that shuts down parts of the ASICS but never restarts them, thats wy 0.94 performes badly, that will also be in the next FW update.
insuuficent cooling will cause HW errors and shut down cores => bad perfomance
The low performance on some ASICs even if cooling is sufficient comes from another problem they have regarding the internal ASIC communication. As I understand they are on a goood way to fix this. next FW will be better but there will still be more perfomance gains in future FW versions. now they are trying to rush out a 0.95 that will fix the above issues and hopefully give some performance gains and not only solve the above issues! We will se and we all hope for the better.
Hello, haven´t had time to go thru the Newbie process before but as it just has changed I thought I might do some good here.
We are running a large amount of KNC Jupiter Miners and has resolved most problems during these first weeks and we have a large understanding for why issues are present on some miners.
Lets see once this newbi status is gone I will post some in the KNCMiner thread that might help some people out (or at least give them some clues whats going on).
Cheers
We are running a large amount of KNC Jupiter Miners and has resolved most problems during these first weeks and we have a large understanding for why issues are present on some miners.
Lets see once this newbi status is gone I will post some in the KNCMiner thread that might help some people out (or at least give them some clues whats going on).
Cheers