The two big ceramic caps up top are each in parallel with one ASIC. So measuring across each one gives each ASIC's voltage, and measuring between GND and the bottom cap gives total string voltage. That's where I'm probing during bench testing.
|
|
|
The old 55nm chips were designed to string without an external level shifter. Maybe some passives between chips, but nothing serious. Take a look at the old OneString boards, they're pretty barebones.
|
|
|
That could be fun, just remember active PFC is also a boost converter so your output voltage would be more like 350.
|
|
|
On your stick I think I used different resistors that'll go to a bit lower voltage than the release model, probably 600mV flat instead of 630. I had trouble getting sticks to initialize stably in that low range so I decided to nix it.
I'll run my test stick at those underclock numbers tomorrow and add to the efficiency chart.
|
|
|
2Pac PCBs arrived today, actually earlier than expected instead of the other thing. I got ahold of them because I was trying to order solder paste stencils as well, got that taken care of but I don't expect them before Monday. I should have time tomorrow to get the robot set up to run a full panel so Monday we can dive right in. One minion drilled out about 300 heatsinks today, and another helped me strip about 1200 chips from S5 boards, so that's pretty well on pace.
Proof 2Pacs will be sent out tomorrow, to CrazyGuy, HolyScott and MacEntyre (the big resellers) and one to philipma1957 since he offered to do a review.
I need to get the sales queue updated, since a few more orders have come in this week.
|
|
|
I don't know if it's in there yet, but yesterday I talked to VH about adding frequency support below 100MHz. I tested my stick at 3.8W at 100MHz, so about 75MHz and below should run on a 2.5W port. It is still definitely recommended to use a good powered hub, especially if you want to push higher frequencies.
|
|
|
Anything in Germany will likely want to talk to MacEntyre, the proprietor of bitshopper.de - he'll be stocking 2Pacs as soon as they're ready, and has already expressed interest in licensing BF16 designs for manufacture and sale in Europe.
|
|
|
Yeah, no doubt. The current generation of Bitfury and Bitmain chips are literally 100 times better than the first-gen Avalon and ASICMiner chips from less than four years ago.
|
|
|
It should also be encouraging that the BF16 stickminer will use the same power system (different voltage range of course) and formfactor, just with upgraded controls. So if this guy is comfortable up above 13W, think what'll happen when it's got 0.1J/GH chips instead of 0.4J/GH chips. The BF16 stick could be most of an S1 by itself.
|
|
|
No, I don't have good tools for that. What I do have is a cheap case fan capable of cooling a couple dozen sticks.
The BF16 stick will have an onboard temp sensor, but not this one.
|
|
|
Finally got around to building a test 2Pac with good chips and ran a quick efficiency test, only up to 300MHz because my USB power meter is only rated for 3A. Freq | GH | Vcore | Power | J/GH | 100 | 11.00 | 1.26 | 3.67 | 0.33 | 125 | 13.75 | 1.26 | 4.48 | 0.33 | 150 | 16.50 | 1.26 | 5.27 | 0.32 | 175 | 19.25 | 1.26 | 6.12 | 0.32 | 200 | 22.00 | 1.32 | 7.43 | 0.34 | 225 | 24.75 | 1.36 | 8.73 | 0.35 | 250 | 27.50 | 1.40 | 10.11 | 0.37 | 275 | 30.25 | 1.46 | 12.25 | 0.40 | 300 | 33.00 | 1.50 | 13.98 | 0.42 |
At 300MHz, the total device efficiency is only about 15% worse than the datasheet chip-level efficiency, which I think is pretty good for a stick miner. It also puts up over 30GH, which I think is pretty darn good for a stickminer. Note this is about the same power draw the original Compac expected to see at 22GH, so we're seeing a 50% improvement in hashrate. If the stick could be pushed to 400MHz you'd see it drawing about 23W for 44GH, but I don't want to test this until I have a bit stouter measurement tool. I'll copy this table into the OP for reference.
|
|
|
Finally got around to building a test 2Pac with good chips. The first 3 test sticks (used by VH for driver dev) were built with out-of-spec rejects from Compac manufacture, meaning they ran hot or poorly from the get-go and wouldn't be representative of the performance of a stock manufactured stick. Yesterday I got the robot set up to run out 2Pacs, today I did a bit of dabbling with voltage ranges on the buck. I had noted that these guys don't particularly like to initialize below about 630mV per node even though at low frequencies (100MHz) they should be good to 600mV, but at the bottom threshold of voltage, chip power draw balance and node-level storage matters a whole heck of a lot. So I figured up buck voltage adjustment resistors that'll give you 630mV to about 830mV, which means it should be good for at least 400MHz. I ran a quick efficiency test, only up to 300MHz because my USB power meter is only rated for 3A. Freq | GH | Vcore | Power | J/GH | 100 | 11.00 | 1.26 | 3.67 | 0.33 | 125 | 13.75 | 1.26 | 4.48 | 0.33 | 150 | 16.50 | 1.26 | 5.27 | 0.32 | 175 | 19.25 | 1.26 | 6.12 | 0.32 | 200 | 22.00 | 1.32 | 7.43 | 0.34 | 225 | 24.75 | 1.36 | 8.73 | 0.35 | 250 | 27.50 | 1.40 | 10.11 | 0.37 | 275 | 30.25 | 1.46 | 12.25 | 0.40 | 300 | 33.00 | 1.50 | 13.98 | 0.42 |
As you can see, efficiency across the bottom end of frequencies is pretty flat. That's because they're all at the same voltage level. Efficiency actually improves slightly with an increase in frequency because the maintenance current (buck controller, LDOs, CP2102 etc) is pretty constant so a higher proportion of power is used for mining. At 300MHz, the total device efficiency is only about 15% worse than the datasheet chip-level efficiency, which I think is pretty good for a stick miner. It also puts up over 30GH, which I think is pretty darn good for a stickminer. Note this is about the same power draw the original Compac expected to see at 22GH, so we're seeing a 50% improvement in hashrate. If the stick could be pushed to 400MHz you'd see it drawing about 23W for 44GH, but I don't want to test this until I have a bit stouter measurement tool.
|
|
|
Most of the additional tools I needed to work up heatsinks have arrived. The heatsinks are only 6 fins instead of 8, but should still work well enough. I still need to build a jig for drilling, but my power tapping head arrived today so a bit more testing to dial it in and we should be in business.
I finally got around to checking out the robot program, making sure all the feeder alignments and parts rotations were correct. Tomorrow I'll have to run through it to check the actual parts placement calibrations but that takes a lot less time. Tomorrow I should get all of the prototypes built and start testing. Hopefully I'll be getting sample sticks out Friday, including one for philipma1957 to review.
I'm assuming that the PCBs will be delayed a bit, partly because they were expected to ship today but I just confirmed the panelized gerber layout yesterday so I doubt they've been made yet. Shouldn't affect much in the long run.
I still haven't gotten out BW sticks, parttly because they're entirely manual placement and that kinda sucks and partly because I've been juggling a lot of other stuff. Yesterday's workload didn't finish until 1AM. I took Sunday off instead of making progress but that's because I'd worked the previous 13 days. Anyways, once 2Pacs are rolling off the line I'll be giving a free 2Pac to the people who bought the BW sticks to help make up for the delay.
Once Chinese New Year ends and the cheap-but-easy prototype PCB house gets back to work, I should get some Biggie prototype PCBs to play with. I probably won't get around to building them until at least the middle of February because the robot and I will be quite busy with 2Pacs. There won't be a lot of retooling for parts feeders, but it will take a full day to get the placement program made and calibrated. I don't have the numbers offhand, but there are at least 40 parts per node and 5 nodes in the chip field, plus the fan control, USB interface and core power. Looking at around 250-300 components to verify placement and rotation on.
|
|
|
Just lit up a pair of T9 on DPS1200 PSUs. Meter estimates 2850W for the pair and one-hour reading of 23.43TH between 'em for an average of 0.121J/GH.
|
|
|
That's why I'd just use hot air to replace the PIC. Less likely to hit the wrong thing. Course it also requires more complex tools.
|
|
|
If you were previously running Novak's cgminer-gekko, it's the same ramp code it always was. His code jumps the frequency up in steps to ramp up the power use, which is actually pretty twitchy if you look at the current draw - but still better than just dumping into full speed from the get-go.
VH's new code, which will be available soon for 2Pac releases (exactly when is up to him), mimics the smooth ramp routine the S5 uses. From a hardware standpoint I'm not sure what's going on since the ramp itself is implemented inside the ASICs when given some specific data, but I suspect it's actually sequentially enabling hashing cores so the current draw increases in a clean linear ramp. This was essential to make work out for proper 2Pac functionality, and is one reason why VH's driver will be better overall for the original Compac than Novak's original code.
I haven't tested the new cgminer to see how it actually behaves, maybe I'll do that today.
|
|
|
Nifty, I'll have to check that out and see how well it works. If it's Novak's code, it'll work okay for Compacs. If it's VH's code, it'll work well for Compacs, 2Pacs and the Biggie.
|
|
|
Is Compac support based on Novak's Icarus rework from 2015 or so, or VH's new code from the last couple months?
|
|
|
Far as I know, Bitmain Warranty is not Bitmain at all but a separate entity that got good at fixing Bitmain's stuff on their own and decided to make a business of it.
Additionally, Bitmain never sold a 335MH stick miner.
|
|
|
That's a new one. According to member lists there are at least 4 fake sidehacks, I hadn't seen sidehack hyphen yet. Good catch.
Please note that no Bitfury-based hardware is for sale yet, and the only miner I'm currently selling is the 2Pac and the only payment address for it is posted in the sales thread. When BF16 gear is available, I will post that publicly and provide pictures and info to back it up.
Also please note that, unless you're one of the big resellers like CrazyGuy or HolyScott, or someone often specifically sought out like philipma1957, I will probably never come to you and say "hey that thing you wanted is for sale now". I just don't have the time to follow people around like that. The only time I'll take your money is if you come to me and offer it first.
|
|
|
|