|
My main farm area with 16 s7's in it in summer runs 85-95F ambient and the miners routinely have run at 70-75C running full speed 24x7x365. Zero problems and they batch-1's through batch 16's.
|
|
|
|
I was meant to update this but as usual I forgot. I bought a cheap 120mm filter and took the filter element out and placed it over the intake fan. The suction held it in place. Within 10mins the temps had risen 10degrees although they settled at around 65-66c with the heat we've had the last few days (30c) I'm not happy leaving it on. As you guys have said already, it's not a major issue. I've tried it and it wasn't suitable for me.
Um, don't do that. Typical small fan filters, even simple wire screens, slow down the airflow way too much. Ref what I do using open-cell filter foam https://bitcointalk.org/index.php?topic=1512186.msg15348138#msg15348138That stuff give me only a 1-2 deg temp rise and also greatly reduces and even sometimes eliminates the siren effect. |
|
|
|
Anyone have the upper and lower operating temperatures for the Temp(PCB) and Temp(chip) for the S9?
Only critical one is chip temp. Anything under 100C is fine, 95 or less ideal but you need reasonable ambient to do that. Not great but I have a b1 running at highest die temp of 100-102C, works fine. Danger point starts around 110, Bitmain usually specs an absolute max junction (die) temp of 125C |
|
|
|
Arrg... Got the PICkit3 programmer,, dl'd the latest software for it http://pic-microcontroller.com/download-mplab-ide/ All well there, installed to Win7just fine. Problem is with "You'll want to set the IPE "Device" drop-down to PIC12F1572". Guess what device is not in the drop down list? Closest match the sw will let me do is PIC12F1501 and that throws a mismatched device ID# error when trying to connect.... Possibly that device was removed in a recent 'update"? Ideas on getting maybe an older device lib for the programmer to use? Also details: you mentioned disconnecting the 18pin cables to control board. Done that. Using the PICkit to power the device with 3.375v. So, do the hash boards need their normal +12v power to them or is the MCU for each strictly powered from the programmer? EDIT: Looking at microchips site for the PIC12F1572, http://www.microchip.com/wwwproducts/en/PIC12F1572 its data sheet starts with: "PIC12(L)F157X microcontrollers combine the capabilities of 16-bit PWMs with Analog to suit a variety of applications" I'll try that tomorrow as I recall seeing the LF version listed... Really wanna get my my 22 online s7's knocked down in power. Looking to run using 625MHz vs default 700 for clock as that produces 4.1THs vs 4.7THs, should be a good target for when I can get the Vcore reduced  |
|
|
|
Fan dynamics...
I was thinking and I don't like doing that these days...
scenario
intake fan max RPM is 4500 rpm and is spinning at 3400 rpm
exhaust fans max rpm is 2400 rpm and is spinning at 2600 rpm
a> is the slower exhaust fan actually slowing down the intake fan as it is being forced to spin the exhaust one faster than it can electrically spin itself?
b> is the exhaust fan even electrically doing anything since its exceeding its max rpm due to the intake fans higher rpm
I am guessing the exhaust fan being that much slower isnt doing anything but causing impedance and should be removed. correct?
I can match it but don't like the extra noise but if its not actually doing anything then I dont need an exhaust fan anyway....
While I am not an "airflow engineer", here's another possibility: c> The exhaust fan is lowering the pressure that the intake fan "sees", and hence the intake fan speed can be reduced without having the miner overheat. Bingo. It's all about static pressure. I haven't actually checked their specs but I'd wager that despite the exhaust fan spinning slower it is still rated with same CFM when working against a light resistance as the intake fan is when the intake is working against high back pressure. Difference is that that intake fan can move that air against the high resistance all those heatsink fins produce. The unloaded CFM of the intake fan is much higher. Put the 2 fans together and it's a win-win. side note: The 1-fan s7 proves that the 'help' the exhaust provides is not necessarily needed. I have 3 of them running full stock speed in 85F ambient and they are perfectly happy. |
|
|
|
|
Why would want to pay almost $700 more than what Bitmain charges for them? "free shipping" from the Amazon seller sounds nice but - shipping from Shenzhen to the USA is only around $140 so buying direct is still cheaper.
|
|
|
|
Likewise my one recommendation would be to put a 30 amp automotive fuse in line to each Titan from that big supply. That way if a Titan shorts a capacitor it will blow the fuse before catching fire.
Actually a super idea would be to put a 10 amp fuse on each one of the three lines that feed each titan's PCI plug. That way if a plug line went bad the other two fuses would blow, saving the titan from the ground fault fail.
I think someone looked at those once and they blew often due to heat lol 😄 The auto fuse anyway idea Then make sure the fuses are in a reasonably cool location or derate fusing as needed (increase fuse size). That said - look around for thermal derating charts from Bussman/LittleFuse/etc so it is done right. The only downside to derating is that under cold starts the permitted current will be quite a bit higher... |
|
|
|
Its clear that bitmain is only out to make a quick buck since its introduction of the S5. I've read nothing but bad news with all the S7's and now S9. As a owner of S1,S3 and S5 and can say that their attention to detail after the S5 is questionable. Ofcourse what happened after the S3 was they didn't care about the noise.
I now have an Avalon 6 and think people should switch.
Er? I have 22 s7's starting from batch-1 of up through b18, not a bit of trouble with any of them. Ever. So far have 3 s9's running, all perform as advertised, no problems so far. There. You have now read GOOD news on the s7 and 9's |
|
|
|
|
Just popping in since power and its distribution is part of my thing... "The thinnest trace in the bridge board" Perhaps KnC's idea of a fusible link vs using a picofuse?
Options when repairing: Insert a 1 or 2a pico fuse to what feeds the RasPi. Quick, cheap and dirty version, use a piece of 30ga wire-wrapping wire perhaps
|
|
|
|
Sure, if there's dev money and a guaranteed market. I don't want to say it's not a good idea, but you're also doubling the cost of something that's already a race to the bottom.
No doubt about higher price vs an unprotected breakout board but -- when dealing with any supply capable of putting out 3-4kw or more I'd definitely pay the additional cost for peace of mind. To me it would be a nice selling point and considering there are pics a bit back in the thread to show the consequences of not having current limits methinks it should be an easy sell. Don't suppose there's a way to monitor a few loads and externally trip a PSU's internal current limiter is there? IF there is a limit sense input that could shut down the PSU's internal DC on/off FET's that would certainly help with cost but then also becomes an all or none on situation. |
|
|
|
Hmm, a thought... Problem: If using a (really) big psu or a bank of 1+n load share ones there is a need to limit power feeding each load. Fuses not good choice due to the voltage drop they introduce. DC breakers in the >100A range give very low vdrop but are very expensive. So... how about incorporating electronic fusing into these high current breakout boards like the ones for 4kw IBM supplies. Say 2-3 banks of mosfets driven by adjustable current monitors (1 bank per-miner attached)? Just dial in the desired current trip point and problem solved. Near zero vdrop and response times from almost instantaneous to whatever is desired. Oh Sidehack and friends, up for a new project? |
|
|
|
Hmm, sorta looking like 16nm production at TSMC is at full speed again no?  Now the huge question is what the Typhon due to hit Taiwan in a couple days will do... My b5 is showing as arriving Thurs but methinks will be the last one I get for a bit. Then again, I said that, um, a couple weeks ago with the 2nd b3 I got  |
|
|
|
I am really impressed that neither miner next to the burned up one were affected. In fact, I am impressed the whole house did not burn down.
Generally the materials used won't support combustion so unless something really catastrophic happens, the worst case should be as pictured. I don't remember what the certification series is called but I think boards are meant to self extinguish within 10 seconds or so. True, I am aware of those facts and I agree with you. I think you are referring to the FR-4 materials grade designation for PCB construction. However, that does not preclude surrounding items, i.e. shelving or other objects depending on material composition, from catching fire. The self extinguishing ratings only apply after the ignition source/condition is removed. In this case after the mains breaker feeding the PSU tripped shutting off power to the (massive) short. Odds are the PCIe power leads were the only thing setting a limit to the power being fed into the short circuit. This miner was tied to a 4kw psu no? If safety is a concern, um, putting a couple 1200-1300w loads on a single high energy source demands fusing EACH of loads to just over whatever current they take, for an s7 that equates to 110-115A to prevent exactly this sort of thing from happening. Yes because of how they work fuses introduce a voltage drop, if that is too high then using DC circuit breakers is the only option. Now as what started the meltdown by apparently shorting out the incoming 12V power planes... Would certainly be nice to know. This also points to one of the few advantages to feeding each miner from a PSU more closely matched to its power needs. Certainly could still have a catastrophic failure similar to this but at least the maximum power that can feed the failure point is far less and more in line with reasonable worst case design considerations. |
|
|
|
I know for a fact that the IBM 2kw psu's will comfortably run 3x s5's which is around 1,800w total. From before I updated to s7's https://i.imgur.com/XRlk5Tq.jpgdon't even need fans for them in this config  |
|
|
|
Might this whole grand adventure be better suited to an Alt Coin sub-forum?
I see mention of a new protocol and infrastructure. None of that sounds particularly Bitcoin Mining centric to me.
As I suggested when OP first showed up, he might find a more receptive audience in another sub-forum. The title doesn't really seem very accurate to the overall scope and desires he has.
Just my $.02.
Exactly. We get that blockchain technology has a myriad of uses, contract verification being just one of them. Hell going way out, the major lighting makers are working with Public electric utilities to incorporate local blockchains to identify and verify 'smart' LED streetlight installations eg what is it, where is it (from built in GPS), when installed, etc. So far installation ID time has dropped from ~15min scanning the unit, filling in contractor information as needed. to less than 2 most of which is the lights finding/setting up a local network link. Point is -- this is a HARDWARE section. Wrong crowd to pitch to. |
|
|
|
Yes I expected "ready to ship" to mean a shipping notification within 24 hours. They demand that you pay in just one!
I'm willing to bet that in the fine print the 'within 24hrs' is after it clears their orders system and shows up in your account orders as 'paid'. My b3 took a full 28hrs before I got the Paid email from them. Based on the Paid email date/time technically it took another 26-30hrs before I got the Shipped email but they got it close enough for me. Now as to the Scenic route that UPS decided to send it on which added another day. Plus the import fees kerfuffle... |
|
|
|
|
Show Posts
