|
Micon: me too, I won our 200 BTC bet (150+ Mh/J), which was on the same terms as SgtSpike. For the exact terms, see links in post
#1 of this thread.
|
|
|
|
You can probably set all these to 1 since you don't use GBT/getwork with an Avalon.
You mean Arvicco specifically did not configure it to use GBT/getwork, or cgminer on Avalon is by default is configured to not use it (ie. use stratum?) |
|
|
|
I finally got altium viewer installed and looked at the board designs. Removing J1 would remove all power to the USB. But you would need to power the tp-link by some other means
You are reading it wrong  F1 powers USB-A1, nothing is connected there. The tp-link is connected to USB-B1. |
|
|
|
Thank you for the explanation. Another question that I haven't been able to find an answer for. Why does removing the fuse matter if there is nothing plugged into the port?
Because the fuse supplies 5V to both USB-A1 (unused) and to the USB hub (which is supposed to get 3.3V). Removing it makes USB-A1 unpowered (we don't care) and stops the overvoltage of the USB hub (which is okay with 0V/GND on this pin). From the link to github I read BitSyncom answer where he says: "Thanks, this is true, will cause the chip to be about ~20C higher than avg, which is still fine [...]"
So, adding a small heat sink could be a "simpler" solution, isn't it?
With all due respect to Yifu, I think he is too negligent here. I don't think it is "fine". We all saw senseless' picture of a melting chip package. And nobody has enough data to proves whether a heatsink is sufficient or not. Even if it helps, the chip's pin input is still being overvolted, so it is still reducing its operating life... |
|
|
|
|
I am not saying that disabling wifi won't help. There is definitely what seems to be a second problem, that is taken care of by disabling wifi, which is a good thing to do in addition to removing the fuse F1. I have a hunch that it is because there is no decoupling capacitor close to USB-B1 (C44 is too distant), and the power consumption of the TP-LINK with wifi enabled is very variable, causing ripple across the internal 5V rail of the control unit...
|
|
|
|
The description that the TP-LINK is "drawing too much power" is actually incorrect. The USB hub chip does not overheat because the TP-LINK draws too much power. The TP-LINK is powered via a 5V line that does not even come from the USB hub chip. Instead, the hub chip overheats because of a flaw in the control unit v1.5 design which I found while poring over the schematics: they accidentally feed 5V to one of the pins of the chip instead of 3.3V. This overvoltage is what causes the chip to overheat. I pinged Yifu who confirmed it, and I filed a bug quoting his response: https://github.com/BitSyncom/avalon-ref/issues/5Removing the fuse F1 merely cuts voltage to this pin. It leaves the USB type B port (which powers the TP-LINK) completely unaffected. |
|
|
|
|
abracadabra: shunting J1 does precisely that: it powers the USB type B port on the control unit. Remove the J1 jumper when connecting the control unit to a host computer.
As to J2, I don't know its purpose.
|
|
|
|
|
At Bitcoin 2013, Yifu said they will ship the sample chips (30 pieces per order of 10k) by the end of May.
|
|
|
|
YMMV, but I happened to get it to work, too. This photo's from SlaveInDebt. Goto link for thread -- he prescribes 67.5ohm. Or actually go to the source for this photo which gives more information: http://blog.zorinaq.com/?e=11 |
|
|
|
If they where more transparent, maybe i would have invested something.
The business plan alone https://picostocks.com/businessplan/19 makes them more transparent than, say, ASICMINER. Personally, I think that this scam because bitfury admitted in a public conversation that he has no higher (university) education but it is going to make the chips.
And yet, out of all the Spartan6-LX150 miner designs, he is the person who made the fastest bitstream (300 Mh/s per FPGA, compared to the "professional" competition at 200-220 Mh/s). Sometimes autodidacts are better than university-taught people. I see that quite often in my professional field as well. (Disclosure: I think the BitFury ASIC is real. I have shares in 100Th-Mine) |
|
|
|
The missing component is a capacitor. However I have no idea if replacing it will fix it.
Actually, that's an inductor, not a capacitor. Actually (really, this time), it is a ferrite bead. Confirmed by reading the reference design. |
|
|
|
Weird, on the hash unit, the SS34 diode is missing from the schematic (PDF).
It turns out that what Yifu released is a new revision of the hash unit PCB that does not need this diode. Most likely this is the revision that batch 2 and 3 will be built on. |
|
|
|
|
Weird, on the hash unit, the SS34 diode is missing from the schematic (PDF).
|
|
|
|
They'd be spending a few million, bare minimum, for marginally (relatively speaking compared to the leaps of cpu->gpu->fpga->asic(any nanometer-node)) little return.
A die shrink from Avalon's 110nm to 90nm or even 65nm would not cost millions. http://blog.zorinaq.com/?e=73 (that said I am certain batch 2 and 3 will be on the same 110nm). |
|
|
|
Interesting that each ASIC has its own reset.
But is not connected. Tied high. I think this was for testing. There is a capacitor on the pin. Obviously that's for delaying the power up of the chip. |
|
|
|
Thank you so much Yifu! I was really looking forward to these specs. You never replied to the 1 PM I sent you, but I know you are a busy man See you at the conference. |
|
|
|
Probably made the assumption that the device shipped met the criteria, and thus the outcome known at this point. I'm presuming what shipped doesn't meet the criteria thus the bet should remain open (as there could be et another product shipped before the deadline that does meet the criteria).
Exactly. The bet should have remained open. coinjedi, any answer on that? We do not always close bets when the outcome is absolutely certain. For example we close a bet on a product shipping date, when the company announces an official shipping date, but we do not payout until they actually do. We judged that in this case starting of the initial batch releases significant enough information for the purposes of closing the bets. Initial reports tell that the performance does not meet the criteria, so we will wait until the announced event date in case they ship another product. If you are the submitter and think that our decision is unreasonable, please contact in private and we will reimburse your submission fee. If you would like to continue betting on this event please submit a new statement so that the betting can continue under the new conditions. What you missed is that the outcome of the BFL bet is not certain. They seem to be merely 46% off the bet's 350 Mh/J limit (239 Mhash/J at 12V == 191 Mhash/J at the wall) and they are currently working on redesigning the power supply circuitry to make it more efficient... The fact that the amounts bet on each side did not vary much in the last weeks before you closed it proves my point: bettors were not certain which side was going to win. Beside, even if the outcome was certain, why are you closing it anyway? Why not let people bet on the side that is "so certain" to win? I am upset you closed the bet  |
|
|
|
So powered extenders are:
a) unnecessary for 7xxx cards
b) unnecessary if adequate power supplies that have double-spring plugs are utilized
c) unnecessary for non-scrypt mining
I concede that there are applications that can warrant powered extenders, but stand by the "waste of time/money" designation. There is a clear profit motive for the product to be promoted the way it has been, which seems to be based on misinformation.
a) No, I only stated I tested the 7970. In fact lewlzminer above claims 3 Amp per 7950. So it seems my model of 7970 (a reference model) is not representative of the whole 7xxx series. b) Yes. However I also suspect you need good quality mini-fit jr male terminals (on the motherboard's 24-pin connector) to benefit from the low electrical resistance of double-spring mini-fit jr female terminals (on the psu's 24-pin connector). c) No. Again, many Bitcoin miners certainly do need powered extenders. I needed them even with my 4x5970 and 3x6990 running with downclocked memory (everything else at stock settings). As to the profit motive of vendors: I don't know. I am not associated with any of them. I certainly did not start a conspiracy theory to entice users to buy powered extenders |
|
|
|
Time to clear up the many wrong statements that people are making in this thread. I am the person who made the first powered PCIe extender and who documented why this is necessary: http://blog.zorinaq.com/?e=44Firstly, the current drawn through the PCIe slot vary a lot depending on the card. Dual-GPU cards usually draw more than single-GPU ones (contrary to what someone said). Here are some of my own measurements with a clamp meter around the 12V lines of a PCIe extender, while Bitcoin mining: HD 6990: 4.2 A HD 5970: 4.1 A HD 5870: 3.2 A HD 6950: 2.5 A HD 7970: 0.9 A So for example the 7970 (and probably most other 7xxx series cards, but I have not tested them) draws so little power that they pretty much don't need powered extenders. Also, eroxors, most of your cards are pretty low-end/mid-end. They likely all draw less than 3 Amp or so because your most power-consuming one seems to be the HD 5870 (3.2 A through the slot). Even if you had 4 of these cards, that's only 12.8 Amp total, so only 6.4 Amp per 12V wire of your 24-pin ATX adapter. But most importantly, whether the 24-pin ATX connector overheats or not depends a lot on whether the Molex Mini-Fit Jr. pins in the PSU 24-pin plug have a single or double spring. You can tell by removing the pin from the plastic housing and seeing if it has one or two pairs of "dots" at the end. This one has two: http://www.alliedelec.com/images/products/Small/70090646.jpg The single spring pins (typically used by inexpensive PSUs) have a higher (edit: electrical) resistance and are more prone to causing the connector to burn out over time. For example, in my farm of 4x5970 and 3x6990 machines from the old days, I noticed that upgrading from single-spring pins to double-spring pins pretty much made powered extenders unnecessary. Finally, eroxors, you are right that the memory subsystem of Radeon cards (at least the 5xxx I tested years ago) is powered through the slot, not through the 6-pin or 8-pin power extenders. So Litecoin miners need powered extenders more than Bitcoin miners (at least for 5xxx and 6xxx cards, again the 7xxx series seems to draw so little from the slot, that powered extenders may not be necessary). |
|
|
|
|
Show Posts
