Awesome..I look forward to your update especially since my miner is also "troubled"  Thanks for sharing this info! early results look promising 30 mins elapsed running .94 but not overclocked (VRMs not reverted to high voltage) ran enabled cores to get all but 11 on one die and 48 on another quit cgminer as soon as it started to keep it from turning off more cores (only 1 more off from the asic_test total) (would need to keep it from autostarting) showing high avg range on BTCGuild ~470 seen range 450-490 even after flushworks, etc (start up was very high, was over 500 for a while) Block: ...a50ff57d #265409 Diff:268M ( 1.92Ph/s) Started: [22:43:44] ST:6 F:0 NB:4 AS:0 BW:[ 49/ 31 B/s] E:2750.94 I:36.63mBTC/hr BS:494k 4/708 41.0C | 500.2/487.0/458.1Gh/s | A:437 R:17+0(3.7%) HW:38008/ 14% -------------------------------------------------------------------------------- KNC 0: 37.0C | 137.7/132.2/126.0Gh/s | A:105 R: 1+0(.94%) HW: 9805/ 14% KNC 1: 37.0C | 143.5/140.0/137.5Gh/s | A:141 R:10+0(6.6%) HW: 4724/6.3% KNC 2: 35.0C | 125.5/117.5/103.1Gh/s | A: 98 R: 6+0(5.8%) HW:15977/ 23% KNC 3: 41.0C | 100.8/98.05/92.84Gh/s | A: 94 R: 0+0(none) HW: 7577/ 15% -------------------------------------------------------------------------------- I would like to revert the VRMs to high voltage and see how the Gh and HW errors increase Would like to also run this on .96 now with more cores turned on knowing that bfgminer doesn't turn them off and see if it gives lower HW% and even better avg. but I need to make an install package of bfgminer that will work on reboots.. I am recompiling and everything from scratch each time which is a waste (many directories get rewritten on reboots) I don't understand how you are getting .7 out of the VRMs on firmware 0.94. I'm running 0.94 and have .9+v and don't see .7 unless I run 0.95. |
|
|
|
more likely is BTC going to 2000usd! BTC value ≠ difficulty, although that seems to be the trend. Anyhow, I digress. We need to wait till the market stabilizes, it's too risky to buy/pre-order any ASIC right now. I am almost certain any ASIC bought today, will see very little to no ROI by the time the device gets to your door and/or after you mine back the initial investment cost. Forget about BFL's Monarch, given their track record.. it won't be released till Q3/Q4 of 2014. By then, it's like having a Jalapeno of today... or worst. I look at the size of the thing and the projected hashrate and think it's going to get too hot. BFL customer once, never again. Never say never, why burn bridges?  I am not the one burning bridges Yep, if I ever bought from them again I expect the outcome would much the same as Lucy holding the football to be kicked by, who is it, Linus? who ends up flat on his back. |
|
|
|
Starting bfgminer as described (of course I "q"uit the original cgminer first) results in a Segmentation fault and the TTY settings broken and a cleared screen. opkg install gdb
gdb --args ./bfgminer -S knc:auto -c /config/cgminer.conf -T -D
run ... wait for the crash ... I had been in direct contact with Luke-Jr (thanks for your time!) and it looks like he could find the problem and fix it. I think re-installing bfgminer from the GIT repository will fix the problem (if anyone else did experience it). I made some tests (using a binary compiled with debug symbols - I must re-compile it) and it did work for 1 hour. As some questions did pop up using bfgminer, I switched back to cgminer. Luke-Jr: Can you give us some more ideas about how your software works with a KnC miner? The old cgminer (or is it some other task running in background?) is disabling (and enabling) cores with too many HW errors. This can be visualized using that BertMod utility. bfgmiger is now doing this task by its own. You can see the number of active cores (2*4*48=384 for a Saturn) in the status line. It looks like you can manually "enable" disabled cores. Will they also be enabled over time automatically again? Is it possible that the BertMod statistics don't match the status of bfgminer? My DIE 0 is the one that always had been warmer and the one that starts to disable cores (without very good cooling). Using bfgminer you can see very good that this DIE has 7% HW errors compared to 2% of the other DIE 3. The effect of a fushwork is still a big impact on the hashrate. I can not really confirm that it's better than how cgminer is doing this. Luke-Jr: Is this a general problem that can not be fixed due to the nature of those multi hundred cores? From your programmers point of view, what advantage do you think bfgminer will have on those flushworks? trepex / Ralf I note your mention of good cooling. I mentioned in a post how having fans on the sides of the module when removed found the voltage and current going up on the cores and the few iffy cores came back and apparently stayed back. I forgot to mention one other thing. With the fans away I could reach in and test the temp with my finger. Thinking I had forgotten to try and remove a cover of a VRM when I was putting tantalum caps on the 12v rail, I considered that if they are snap on tops, heatsinks, then taking them off and putting heatsink compound might be possible but then maybe one just wasn't down far enough and since I could reach the VRMs I pressed down on each. The die with the iffy cores still runs more current than the others but is 100%. Maybe it wasn't the fan removal but the pressing down on the VRMs' tops. Edit: I made sure not to press down too hard as flexing the board could break solder joints at the Ball Grid Array points of the ASIC. |
|
|
|
I hear the figure 30PH/s tossed around for this coming January. Then to stay current we need to buy 10x the hashing power we have now. Ouch.
to stay current, you need to double the amount of your miners with every difficulty rise  Alas, we're competing against technical investors with big bucks who likely held off, and perhaps a LOT of them. |
|
|
|
more likely is BTC going to 2000usd! BTC value ≠ difficulty, although that seems to be the trend. Anyhow, I digress. We need to wait till the market stabilizes, it's too risky to buy/pre-order any ASIC right now. I am almost certain any ASIC bought today, will see very little to no ROI by the time the device gets to your door and/or after you mine back the initial investment cost. Forget about BFL's Monarch, given their track record.. it won't be released till Q3/Q4 of 2014. By then, it's like having a Jalapeno of today... or worst. I look at the size of the thing and the projected hashrate and think it's going to get too hot. BFL customer once, never again. Never say never, why burn bridges? |
|
|
|
more likely is BTC going to 2000usd! BTC value ≠ difficulty, although that seems to be the trend. Anyhow, I digress. We need to wait till the market stabilizes, it's too risky to buy/pre-order any ASIC right now. I am almost certain any ASIC bought today, will see very little to no ROI by the time the device gets to your door and/or after you mine back the initial investment cost. Forget about BFL's Monarch, given their track record.. it won't be released till Q3/Q4 of 2014. By then, it's like having a Jalapeno of today... or worst. I look at the size of the thing and the projected hashrate and think it's going to get too hot. |
|
|
|
|
If the powers that be go to adversely effect BTC, aside from having Washington DC make it outright illegal, they'll target BTC stability, pump and dump punishing investors again.
|
|
|
|
|
I hear the figure 30PH/s tossed around for this coming January. Then to stay current we need to buy 10x the hashing power we have now. Ouch.
|
|
|
|
Yep, probably won't make back the BTC equiv. of the miner's purchase. Good thing the value of BTC is going up. Of course it seems clear that buying BTC is the more profitable choice. As long as the value of BTC depends on Wall Street not developing a sense of ethics, BTC will rise.
with the stakes raised, there are more 'cost basis' tricks being done now... if you aren't changing your game you get left behind in other words, many people you may be comparing yourself to have a lower bar for ROI than just BTC in BTC out I absolutely agree. I look at the hit on my IRA this year which includes the spring crash, some bills, spending on mining gear, and then look at my total BTC*$value just to not feel so bad. |
|
|
|
Yep, probably won't make back the BTC equiv. of the miner's purchase. Good thing the value of BTC is going up. Of course it seems clear that buying BTC is the more profitable choice. As long as the value of BTC depends on Wall Street not developing a sense of ethics, BTC will rise.
Probably an unwise observation. Wall Street might say, humm, if BTC price is going to be used as an indication of our being ethical or not, then there are some tricks we can use to adversely effect the stability/value of BTC. |
|
|
|
|
Yep, probably won't make back the BTC equiv. of the miner's purchase. Good thing the value of BTC is going up. Of course it seems clear that buying BTC is the more profitable choice. As long as the value of BTC depends on Wall Street not developing a sense of ethics, BTC will rise.
|
|
|
|
Okay, received my jupiter today. Wast getting about 450gh/s with temperatures around 60 degrees C.
I updated to .96 and my temps went down to 40 degrees C. Hashrate stayed about the same.
I loaded up enablecore.bin and it gave me an error while rebooting. Now everytime I try to go to the miners setup page, it gives me a 404 page not found.
Any ideas??
**** Ignore ^. A hard reset did the trick.
Averaging 480 gh/s with temps around 50.
I sometimes wonder if when one first fires up a new miner and there's bad cores and high temperature, if it's better to let it run just like that for a couple of days. The rationale would be that if in shipment some ball grid away points cracked and separated but still making marginal electrical contact, there will be a higher resistance point there that will be hot, hence the high temperatures. It might get hot enough at that point or points to liquify the solder while nearby healthy solder points remain okay. Then after turning it off and cooling, when you restart that point has a solid solder contact. Mine has been running at 100% on all 4 cores overnight and maxed out at 137.5GH/s avg so I would expect max avg of a Saturn to be 275GH/s and Jupiter 550GH/s, 5 second averages not withstanding. |
|
|
|
|
Mining Bitcoins is like growing pines on a residential property. Ain't nobody's business 'till ya sell 'em.
|
|
|
|
|
On 0.94 still as it gives me the best cores. Kept dropping 1 to 3 cores on die 3. Tried removing some extra cooling one side at a time to see if that would make more cores drop then knowing which side I could find out which of the two VRMs might need extra cooling, if that was the fault.
I have two 12v fans one on each side facing the VRMs. All currents were in the mid-40 amps. Removed the fan from the outside and expected to see more disabled cores. The reverse, the three cores came back. I removed both 12 v fans, unplugging them dropped my power at the wall 3 watts. Now with no side fans cooling the VRMs the VRM output voltages all increased roughly 0.03 volts and the currents all went from mid-40amps to low-to-mid-50amps! And all cores enabled! So, these are designed to run warm.
|
|
|
|
UPS tried hard to stop me getting my package, refusing to let me settle my customs debts with legal tender, insisting on a certified cheque even though they themselves had already deliberately told me not to to come by until after five pm so they knew damn well it was too late to go get such a cheque, sending me to non existent credit unions to try to get such a cheque and so on, but finally by bumping it upstairs along the 800 number I got them to accept a debit card.
Both fans were off (it is a Saturn), but I had gotten the special screwdrives so I was able to clip them back in place.
I momentarily even saw the web interface, but when I told it my stratum URL for my p2pool it somehow told the power supply to shut down.
To get the power supply (a Cooler Master Ex2treme 725) to start again I have to remvoe the paperclip, so it seems to tell the supply to ignore the paperclip and shut down until the paperclip is removed and re-inserted.
The Saturn comes on for a minute or two, but before I can even ping it a bright white light shines and it apparently tells the power supply to shut down again.
The first bunch of times I thought maybe my paperclip was losing connection upon warming up or something but no it really does seem to have to be removed before it will work again.
Nothing is obviously disconnected or broken, and watching the white light more carefully along with the fans it looks like the fans do not get shut down the instant the white light shines, some number of seconds later it shines again at lower intensity just for an instant, or something else right near it flashes for a moment, and then the fans shut down, including the power supply's fan.
Any idea what-all that behavior might be trying to convey?
-MarkM-
It sound like something with your ASIC modules or the wiring to it or them. You can try unplugging both the ASIC modules at the Asic board, the PCIE connector. Turn on the supply, let it boot, with another machine you can ping the web address you expect it to be at. Maybe ten pings after the pings start getting returns, or maybe 15, open putty and ssh to the address and login. If you can then the BBB is working okay. Same with the web page. Shut it down. Connect one of the two ASIC modules. Boot and watch the pinging on another machine. If boot fails, unplug the module and plug in the other and try again. If you isolate the fault to one ASIC or the other, work from there. |
|
|
|
There's perhaps another aspect of the 8 VRMs vs 4VRMs. When I removed my module although there are standoffs for 6 screws, 4 board to standoff and 2 heatsink clamp to standoff, only 2 of the 4 board to standoff screws were in place. My fan like most had been knocked free in transit.
I wonder if the 8VRM boards, earlier construction, might have all 4 of the board to standoff screws and could that have meant less flexibility of the module pcb and fewer damaged BGA points.
I bought a zTex 1.15x early this year for $400. It was erratic and I sent it back. When it was returned it was very stable. I'm told they reflowed the FPGA. So, damage can happen to BGA points and can be fixed. |
|
|
|
|
There's perhaps another aspect of the 8 VRMs vs 4VRMs. When I removed my module although there are standoffs for 6 screws, 4 board to standoff and 2 heatsink clamp to standoff, only 2 of the 4 board to standoff screws were in place. My fan like most had been knocked free in transit.
I wonder if the 8VRM boards, earlier construction, might have all 4 of the board to standoff screws and could that have meant less flexibility of the module pcb and fewer damaged BGA points.
|
|
|
|
Okay, up and running with a couple of good sized tantalum caps on the module 12v rail. Will catch up on posts while it settles out on 0.94 then see if 0.95 is any better. Opps, forgot to enablecores as a few were down on 1 die.
Now running 0.94, that's .91-.94volts on the core, all cores 100% but I suspect I'll see cores dropping on die 3 as it's drawing 50.6amps while the others are drawing under 50amps. But the real test isn't if die 3 drops 1-3 cores or not, it's if the cores will stay almost as good on 0.95 with .7volts on the cores. Big difference, 270watts at the wall versus what was it 148? Okay, I'm already up to 129GH/s. 0.95 could never get pass 100GH/s so time to give it a test. ---------------- Okay, no good. Lots of cores being disabled. No way. Ah, too bad. My last two tantalum caps and it took nerves of steel to put stands of silver solder to the board. Went back to 0.94, enabledcores, all 100%, back to 270watts at the wall but I'll get full or almost full hashrate. |
|
|
|
Okay, up and running with a couple of good sized tantalum caps on the module 12v rail. Will catch up on posts while it settles out on 0.94 then see if 0.95 is any better. Opps, forgot to enablecores as a few were down on 1 die.
Now running 0.94, that's .91-.94volts on the core, all cores 100% but I suspect I'll see cores dropping on die 3 as it's drawing 50.6amps while the others are drawing under 50amps. But the real test isn't if die 3 drops 1-3 cores or not, it's if the cores will stay almost as good on 0.95 with .7volts on the cores. Big difference, 270watts at the wall versus what was it 148? Okay, I'm already up to 129GH/s. 0.95 could never get pass 100GH/s so time to give it a test. ---------------- Okay, no good. Lots of cores being disabled. No way. Ah, too bad. My last two tantalum caps and it took nerves of steel to put stands of silver solder to the board. |
|
|
|
Okay, up and running with a couple of good sized tantalum caps on the module 12v rail. Will catch up on posts while it settles out on 0.94 then see if 0.95 is any better. Opps, forgot to enablecores as a few were down on 1 die.
Now running 0.94, that's .91-.94volts on the core, all cores 100% but I suspect I'll see cores dropping on die 3 as it's drawing 50.6amps while the others are drawing under 50amps. But the real test isn't if die 3 drops 1-3 cores or not, it's if the cores will stay almost as good on 0.95 with .7volts on the cores. Big difference, 270watts at the wall versus what was it 148? Okay, I'm already up to 129GH/s. 0.95 could never get pass 100GH/s so time to give it a test. ---------------- Okay, no good. Lots of cores being disabled. No way. |
|
|
|
|
Show Posts
