Biodom
Legendary
Offline
Activity: 3920
Merit: 4413
|
|
June 19, 2014, 09:22:40 PM |
|
There is some very compelling evidence that supports this hypothesis. But the most glaringly obvious is the single PCI power connector. They're not so stupid to design it to handle 420w. But they were compelled to squeeze as much hash out of each chip that they threw away common sense.
With regards to the single PCIe connector and 425W... Understandably, this is WAY over the 150W limit and they WILL be liable if there will be fires because KnC deliberately made this a single connector. Perhaps, they are scrambling now to get a second one in....hence the hash while you wait. From a practical point of view (try it on your own risk-this is NOT a rec)-many people had run antminer S1 dual blade on Corsair CX500M. This PSU has a single PCIe cable with a split (each half of the split feeds 180-200W blade). When I used it on one of my overclocked S1 (rated at ~400W), the cables were quite worm, but it still run OK, however the machine would not tolerate overclocking further (to 422-435W), so this was probably the culprit. The conclusion: there might be difficulties in running a single PCIe connector at 425W, but up to 400W worked (at least for a short time-I since switched to a different PSU)
|
|
|
|
tarmi
Legendary
Offline
Activity: 1232
Merit: 1011
|
|
June 19, 2014, 09:31:23 PM |
|
are those PSU's for the Neptunes?
No they're the five ASIC boxes. Those things plus a controller and a pile of PSUs are the "neptune". I really don't get this company, how could they change the design of the final products at the last moment, without even telling the customers, they showed a single box when they did the pre-sale and now you have 5 boxes, why didn't they just stick with the 28nm and make a bunch of single miners, they just could have hired bitmain to do the final product, looks like the upcoming Antminer S1 will be similar looking. They could have used the same 28nm chip and undervolt them, and it would have been out 5 months ago, when difficulty was 1.5 Billionyes they could have, but they needed a data center crowdfunded.
|
|
|
|
Tigggger
Legendary
Offline
Activity: 1098
Merit: 1000
|
|
June 19, 2014, 09:33:38 PM |
|
There is some very compelling evidence that supports this hypothesis. But the most glaringly obvious is the single PCI power connector. They're not so stupid to design it to handle 420w. But they were compelled to squeeze as much hash out of each chip that they threw away common sense.
With regards to the single PCIe connector and 425W... Understandably, this is WAY over the 150W limit and they WILL be liable if there will be fires because KnC deliberately made this a single connector. Perhaps, they are scrambling now to get a second one in....hence the hash while you wait. From a practical point of view (try it on your own risk-this is NOT a rec)-many people had run antminer S1 dual blade on Corsair CX500M. This PSU has a single PCIe cable with a split (each half of the split feeds 180-200W blade). When I used it on one of my overclocked S1 (rated at ~400W), the cables were quite worm, but it still run OK, however the machine would not tolerate overclocking further (to 422-435W), so this was probably the culprit. The conclusion: there might be difficulties in running a single PCIe connector at 425W, but up to 400W worked (at least for a short time-I since switched to a different PSU) I spoke with one of their engineers on IRC about this. The information he provided was that the 150w listed in the specs for PCI-E isn't a limit it's the minimum requirement for it to be PCI-E compatible, and that that good quality PSU's can handle it fine. I'm going the Dual EVGA 1300W route as I have them handy, early order so will report back how it goes.
|
|
|
|
midyatspor
Member
Offline
Activity: 107
Merit: 10
|
|
June 19, 2014, 09:38:47 PM |
|
There is some very compelling evidence that supports this hypothesis. But the most glaringly obvious is the single PCI power connector. They're not so stupid to design it to handle 420w. But they were compelled to squeeze as much hash out of each chip that they threw away common sense.
With regards to the single PCIe connector and 425W... Understandably, this is WAY over the 150W limit and they WILL be liable if there will be fires because KnC deliberately made this a single connector. Perhaps, they are scrambling now to get a second one in....hence the hash while you wait. From a practical point of view (try it on your own risk-this is NOT a rec)-many people had run antminer S1 dual blade on Corsair CX500M. This PSU has a single PCIe cable with a split (each half of the split feeds 180-200W blade). When I used it on one of my overclocked S1 (rated at ~400W), the cables were quite worm, but it still run OK, however the machine would not tolerate overclocking further (to 422-435W), so this was probably the culprit. The conclusion: there might be difficulties in running a single PCIe connector at 425W, but up to 400W worked (at least for a short time-I since switched to a different PSU) I spoke with one of their engineers on IRC about this. The information he provided was that the 150w listed in the specs for PCI-E isn't a limit it's the minimum requirement for it to be PCI-E compatible, and that that good quality PSU's can handle it fine. I'm going the Dual EVGA 1300W route as I have them handy, early order so will report back how it goes. the problem is not going to be the PSU but the cable/connector from the PSU to the Unit.
|
|
|
|
samsonn25
|
|
June 19, 2014, 09:41:25 PM |
|
There is some truth to this.
A amd r9 295x2 pulls 690 watts and uses a 6+8 pin dual power connector. Another version of the 290 dual card pulls 740 watts.
|
|
|
|
faetos
|
|
June 19, 2014, 09:52:18 PM |
|
There is some very compelling evidence that supports this hypothesis. But the most glaringly obvious is the single PCI power connector. They're not so stupid to design it to handle 420w. But they were compelled to squeeze as much hash out of each chip that they threw away common sense.
With regards to the single PCIe connector and 425W... Understandably, this is WAY over the 150W limit and they WILL be liable if there will be fires because KnC deliberately made this a single connector. Perhaps, they are scrambling now to get a second one in....hence the hash while you wait. From a practical point of view (try it on your own risk-this is NOT a rec)-many people had run antminer S1 dual blade on Corsair CX500M. This PSU has a single PCIe cable with a split (each half of the split feeds 180-200W blade). When I used it on one of my overclocked S1 (rated at ~400W), the cables were quite worm, but it still run OK, however the machine would not tolerate overclocking further (to 422-435W), so this was probably the culprit. The conclusion: there might be difficulties in running a single PCIe connector at 425W, but up to 400W worked (at least for a short time-I since switched to a different PSU) I spoke with one of their engineers on IRC about this. The information he provided was that the 150w listed in the specs for PCI-E isn't a limit it's the minimum requirement for it to be PCI-E compatible, and that that good quality PSU's can handle it fine. I'm going the Dual EVGA 1300W route as I have them handy, early order so will report back how it goes. Let me know if you want some Texas brisket to cook on those bad boys! I hope it works out for out and there aren't any problems.
|
|
|
|
sikke
|
|
June 19, 2014, 10:18:11 PM Last edit: June 19, 2014, 10:33:52 PM by sikke |
|
so you wish to pull 400watts from random psu 3+3 pci-e 12V 11A from each pair of pins. Ya rly.
|
|
|
|
rograz
|
|
June 19, 2014, 10:30:21 PM |
|
There is some very compelling evidence that supports this hypothesis. But the most glaringly obvious is the single PCI power connector. They're not so stupid to design it to handle 420w. But they were compelled to squeeze as much hash out of each chip that they threw away common sense.
With regards to the single PCIe connector and 425W... Understandably, this is WAY over the 150W limit and they WILL be liable if there will be fires because KnC deliberately made this a single connector. Perhaps, they are scrambling now to get a second one in....hence the hash while you wait. From a practical point of view (try it on your own risk-this is NOT a rec)-many people had run antminer S1 dual blade on Corsair CX500M. This PSU has a single PCIe cable with a split (each half of the split feeds 180-200W blade). When I used it on one of my overclocked S1 (rated at ~400W), the cables were quite worm, but it still run OK, however the machine would not tolerate overclocking further (to 422-435W), so this was probably the culprit. The conclusion: there might be difficulties in running a single PCIe connector at 425W, but up to 400W worked (at least for a short time-I since switched to a different PSU) I spoke with one of their engineers on IRC about this. The information he provided was that the 150w listed in the specs for PCI-E isn't a limit it's the minimum requirement for it to be PCI-E compatible, and that that good quality PSU's can handle it fine. I'm going the Dual EVGA 1300W route as I have them handy, early order so will report back how it goes. Molex (company making the pci-e power connectors) dictates the max spec per pin to 8A according to an earlier posted PDF, so 400W is outside even the most optimistic estimates! Not saying it wont work, but why are they even gambling with this when it would cost them a few bucks tops to use 2 of them.
|
|
|
|
samsonn25
|
|
June 19, 2014, 10:37:48 PM Last edit: June 19, 2014, 10:52:38 PM by samsonn25 |
|
YES. Especially after they pocket 12k per miner.
Even average high end gpu pulling 250-300 watts has 2 power connectors.
|
|
|
|
faetos
|
|
June 19, 2014, 10:48:12 PM |
|
my surprised face on ppl just now figuring out how cheap and crappy knc are:
:|
|
|
|
|
Biodom
Legendary
Offline
Activity: 3920
Merit: 4413
|
|
June 19, 2014, 10:50:35 PM |
|
Molex (company making the pci-e power connectors) dictates the max spec per pin to 8A according to an earlier posted PDF, so 400W is outside even the most optimistic estimates! Not saying it wont work, but why are they even gambling with this when it would cost them a few bucks tops to use 2 of them.
OK, so 6 pinsx8A=48AX12=576W per each PCIe cable max? To me it sounds a bit too much, or maybe not each pin (out of 6) counts? I am very far from being an electrical engineer, sorry.
|
|
|
|
Cablez
Legendary
Offline
Activity: 1400
Merit: 1000
I owe my soul to the Bitcoin code...
|
|
June 19, 2014, 10:57:17 PM |
|
Molex (company making the pci-e power connectors) dictates the max spec per pin to 8A according to an earlier posted PDF, so 400W is outside even the most optimistic estimates! Not saying it wont work, but why are they even gambling with this when it would cost them a few bucks tops to use 2 of them.
OK, so 6 pinsx8A=48AX12=576W per each PCIe cable max? To me it sounds a bit too much, or maybe not each pin (out of 6) counts? I am very far from being an electrical engineer, sorry. Each PCIe connector has three 12v wire 'pairs' i.e. one hot line (yellow) with associated ground wire when connected. This then yields 3 pins at 8A which is 24A@12V= 288W per connector at rated max. Where does the rest of the power come from or is it even needed is the question?
|
Tired of substandard power distribution in your ASIC setup??? Chris' Custom Cablez will get you sorted out right! No job too hard so PM me for a quote Check my products or ask a question here: https://bitcointalk.org/index.php?topic=74397.0
|
|
|
samsonn25
|
|
June 19, 2014, 11:00:50 PM |
|
Even 8 pin connector is only 384 watts.
|
|
|
|
Syke
Legendary
Offline
Activity: 3878
Merit: 1193
|
|
June 19, 2014, 11:19:34 PM |
|
Even 8 pin connector is only 384 watts.
But to get to that rating you have to custom wire the connector. Standard PSUs only have 3 hot wires, even on a 8-pin connector. 288w is the limit.
|
Buy & Hold
|
|
|
rograz
|
|
June 19, 2014, 11:21:01 PM |
|
Even 8 pin connector is only 384 watts.
Actually the 8 pin ones still only have 3 yellow so you are still limited to 3x8, why the pci-e standard calls for extra ground cables on the 8 pin I'm not sure (less load on motherboard ground perhaps?)
|
|
|
|
Biodom
Legendary
Offline
Activity: 3920
Merit: 4413
|
|
June 19, 2014, 11:29:31 PM |
|
Molex (company making the pci-e power connectors) dictates the max spec per pin to 8A according to an earlier posted PDF, so 400W is outside even the most optimistic estimates! Not saying it wont work, but why are they even gambling with this when it would cost them a few bucks tops to use 2 of them.
OK, so 6 pinsx8A=48AX12=576W per each PCIe cable max? To me it sounds a bit too much, or maybe not each pin (out of 6) counts? I am very far from being an electrical engineer, sorry. Each PCIe connector has three 12v wire 'pairs' i.e. one hot line (yellow) with associated ground wire when connected. This then yields 3 pins at 8A which is 24A@12V= 288W per connector at rated max. Where does the rest of the power come from or is it even needed is the question? ^^^This sums it up perfectly.
|
|
|
|
rograz
|
|
June 19, 2014, 11:39:45 PM |
|
Molex (company making the pci-e power connectors) dictates the max spec per pin to 8A according to an earlier posted PDF, so 400W is outside even the most optimistic estimates! Not saying it wont work, but why are they even gambling with this when it would cost them a few bucks tops to use 2 of them.
OK, so 6 pinsx8A=48AX12=576W per each PCIe cable max? To me it sounds a bit too much, or maybe not each pin (out of 6) counts? I am very far from being an electrical engineer, sorry. Each PCIe connector has three 12v wire 'pairs' i.e. one hot line (yellow) with associated ground wire when connected. This then yields 3 pins at 8A which is 24A@12V= 288W per connector at rated max. Where does the rest of the power come from or is it even needed is the question? ^^^This sums it up perfectly. Nah you forgot this,
|
|
|
|
Bogart
Legendary
Offline
Activity: 966
Merit: 1000
|
|
June 20, 2014, 12:07:52 AM |
|
This was on one of my CoinCraft Desk modules, which draws around 200W. Replacing the connectors was a pain in the butt. It took me about 4 hours from start to finish.
|
"All safe deposit boxes in banks or financial institutions have been sealed... and may only be opened in the presence of an agent of the I.R.S." - President F.D. Roosevelt, 1933
|
|
|
|
samsonn25
|
|
June 20, 2014, 12:16:26 AM |
|
I was thinking that too. Also is it a real mole or copy brand. That goes for the metal connector and also the quality of the plastic for the housing.
|
|
|
|
|