Which motherboard(s) are you using in your FM2 rigs?
|
|
|
Using a laptop to mine with is a VERY BAD IDEA - they are not designed to handle the heat of running your CPU (or the GPU in them) 24/7 at the kind of power level that mining puts on them.
|
|
|
Here in the Philippines we have so expensive rate for electricity because of old infrastructures to produce electricty. Sometimes we really experience power interruptions because it cannot sustain the power needed especially during summer season.
You also have to import most if not all of your fuel, that helps drive your rates up. Underground is NOT generally a good idea if you have to worry about high heat dissippation (like in cryptocoin mining), it might be inherently cool but all that rock/dirt/etc acts like insulation making it hard to get a LOT of heat out of the space. Might be OK if you're using some sort of active cooling, like a geothermal loop or some such, or if you set it up like a quanat (sp?) with lots of airflow driven by height difference between the input and the output.
|
|
|
Ethereium is slated to go POS a while before *3* GB cards become unviable for mining - if the timeframe doesn't get pushed back again.
The real limit is likely to be "when will the diff increase to the point it overcomes the price and mining become uneconomic if your power cost is higher then VERY VERY cheap".
|
|
|
You want to mine with a 9400?
*ROFLMAO*
Looks like you're the one that's been smoking the weed - that GPU is so old it would have a hard time mining SHA256, much less anything that is currently considered to be a GPU-mined coin.
In THEORY, it might be able to do something - it did technically have CUDA units - but the performance would be worse than any semi-current CPU.
For the record - CUDA showed up sometime during the GeForce 8xxx series, I forget offhand if ALL the 8xxx were CUDA capable though. They would be VERY VERY POOR performing CUDA units though, very slow clocks by current standards AND small amounts of very slow memory. Forget Ethereum entirely - no way those ever had a 2 GB option, or even a 1GB option - I'm not sure if they had a 256 MB option offhand.
Also, expecting an answer to a rather OBSCURE question like this in a half hour is just DUMB. This is a forum, NOT an IRC channel with hundreds of folks active at the time you asked.
|
|
|
If they'll clock to even 4Gh ballpark, they shouldn't be even close to CPU limiting the GPUs.
Problem appears to be something else.
If we were talking Foldcoin, then I could easily see the issue being CPU limiting, there's a lot more interaction between the CPU and any GPU(s) there. Ethereum though pretty much only uses the CPU for DAG file genration (and not always then depending on the miner software), and there is only a lot of data transfer TO the GPU when loading a DAG file.
AMD hasn't even announced an RX 490 yet - it won't be soon, though in THEORY it could show up late this year.
|
|
|
All temperatures are arbitrary based on the sensor location, its not a great idea to compare meltTemps between miners.
Absolutely! SP20 sensors were on the CHIP, thus the high "rated" temps - Antminer sensors IME have been board-level, and need to stay a LOT lower for the actual silicon to not overheat. Personally I VASTLY prefer the sensor to be on the chip, as there are way too many variables involved in "board-level" sensors that can distort their readings badly.
|
|
|
BTW - the 8086 didn't have enough memory space to even THINK about trying to process Bitcoin - not sure if the 80*3*86 did for that matter.
Pentium - perhaps, SHA256 isn't that much harder to process than RC5 depending on the RC5 keylength - but it would be incredably SLOW doing so. If I had to use something out of that generation I'd go with the AMD K5 over the Pentium.
|
|
|
because they are equal to a 970 in gaming or slightly worse, usually this mean that they will be there in hashing
Hashing varies - sometimes NVidia can match AMD on "comparable" cards, sometimes it's not even in the same ballpark - depends on the algorythm and how it works. Scrypt, for example, "comparable" AMD cards blew NVidia completely out of the water - but X11 Nvidia was the ruler for a long time, even the last of the X11 improvements to the AMD side were only able to get them into the same ballpark per the benchs I've seen. Foldcoin, NVidia blows AMD away - the GTX 970 is competative with ANYTHING AMD on that despite being a much "lower end" card than the R9 390x or the Nano/Fury line. Gaming performance is NOT a good indicator of mining performance except PERHAPS within the SAME archetecture and generation of cards. With that said, I still suspect that the RX 480 will be a MUCH better $/hash performer than any Pascal card, and probably will compete well on W/hash as well - but we won't know for SURE 'till late June at best when/if someone with a review RX 480 benches it on mining algorythms.
|
|
|
The history of large monolithic mining ASIC's like the Minion (Hashfast?) or BFL's Monarch chips with lots of cores/pipelines/engines/ whatever ya want to call them, let's stick to 'cores', is a travesty of wasted and stolen money (pre-order$).
Not ALWAYS the case. Consider the Spondoolies "Rockerbox" chips in the SP20 and bigger same-gen miners. There are tradeoffs to make on the chip size debate - smaller chips = higher yield and less heat per chip but you need a LOT MORE CHIPS to achieve comparable miner performance, which leads to more-complex BOARD level design and a lot more components to go bad, as well as more complexity on heatsink design / air path routing. Also consider the Gridseed GC3355 - small chip, yet the miners it went into MOSTLY ended up being very poor reliabilidy due to BAD board-level designs (the orb EVENTUALLY managed to achieve decent reliability, but the early versions had a TON of issues. The BLADES never overcame their crap buck design issues).
|
|
|
It actually applied to ALL coins at some point to a degree - some of the CPU-specific coins might run better on Intel, but Intel's presense in graphics is a sad joke for the most part (even the Haswell and Skylake APUs don't match my 2 generation old A10-5700s on graphic performance overall).
Doesn't apply CURRENTLY to SHA256/Scrypt/X11 coins as the ASICs have taken those over, but that leaves a very large number of coins that work well on AMD hardware (and a FEW that work better on Nvidia).
|
|
|
Do keep in mind how MANY S5/S7/SP20 generation miners are out there - the S7s might keep mining after the halfing, with the recent Bitcoin price runup they should still be profitable - but that's a LOT of S5/SP20 generation gear that probably WILL become unprofitable.
I'm not guessing a 30% hashrate drop due to the halfing, but I AM guessing more than 10% and perhaps 20% ballpark short-term.
30% would need a significant number of S7/A6/B11 units to be taken offline - and I don't see that happening right around the halfing.
|
|
|
1 or 2 weeks?
You were a pessimist.
|
|
|
Hey BlackSpidy, are you kicking yourself yet?
|
|
|
Difficulty should show a drop after the halfing - older miners going unprofitable and being taken offline - but with the S9 and quite likely other 14/16nm full custom gen miners showing up around that time or shortly after, we'll most likely have a difficulty that IS higher than right at the halfing before October 2016.
|
|
|
I believe the R9 390 was basically a rebadged R9 290 with an updated BIOS - try 1100 core / 1500 clock as a start point, *IF* you can keep the thing cool enough.
|
|
|
from everything I have read so far (Phil @ NFW), the winning combo will probably be S9LN, IF they make it. Should be 6.75-8.6 Th at 700-850W-would be a best seller by far. Something like B2, but with less premium should be OK as well.
8.6ish at more like 900w - just take 2/3ds of the TH and watt spec of the S9, then add a few watts 'cause you still have the controller and fans eating the same power. Anyone remember what process the 4004 used?
|
|
|
I've been using mostly Seasonic for a long time - did get a couple of EVGA 1300 G2 units when I had the S5 farm, as they were on sale at the time and lots of "they're good" recommendations - they worked fine but haven't had them long enough or pushed them long enough to have a solid opinion yet.
A couple of my A2 units got X1250 transplants - the EVGA flat out didn't fit right, the Seasonics did fit fairly well, partly they're a bit shorter and partly the modular connections were better placed to clear the fan/control board.
I wouldn't hesitate to trust Rosewill for cases and keyboards and the like (not mice - I HATE mice, nothing but Logitech Trackmen around here!) but their lack of transparency in who makes the PS makes me inclined to not trust them there.
Oh for those that say "you'll never repurpose those ATX power supplies, just go server" - of my current 4x X1250 and pair of EVGA 1300s, ALL of the Seasonics are currently in machines (2 A2s, 2 computers) and one of the EVGA's is my current "bench setup" power supply (the other one is currently a spare but it might get hooked to a Batch 2 S9 when the price gets reasonable).
Not knocking the server supplies - just prefer the flexability of "ack, dead PS, grab whatever I've got as a spare 'till I can get a real replacement!" for almost anything in house.
I can see the S5 still being a good option in some cases for a FREE ELECTRIC miner, they've come down enough that you can match the hashrate of a S7 with enough S5s at less cost with a LITTLE digging - abet the power usage and heat are higher, the flexability of being able to use much smaller PS and the lower power draw per unit might be a major plus if they have to share circuits with something else. I wouldn't bother with anything older though.
|
|
|
There's a pic floating around somewhere about one of those units mining at very close to (bit OVER as I recall, 151-152Mh I think) the spec 150 MH/s - but eating about 53 watts.
STILL very nice efficiency, best in the X11 ASIC market - if valid.
|
|
|
My triple-GPU Nvidia Win7 rig (2x 950 1x 960) runs a 4-core APU - AND also runs the DNet client on the APU's GPU part, AND Prime95 on all 4 CPU cores I DO notice that one of the CPU cores seems to run Prime a LOT slower than the others - obviously the one that's actively running the system and various software.
I don't notice ANY hashrate difference between when I'm runnnig all of it and just running the miner by itself.
The only difference I've seen on multi-core CPU vs single-core is during DAG file creation - I sometimes get 2 DAG file "threads" running at the same time, and the system doesn't seem to bog down as much.
The Big Rig (triple AMD R9-290) is quite happy with it's Semperon 145 - though when I shift it over to something else after ETH gets unprofitable I'll probably swap one of my X240 dual-cores in it's place - planning to build a new file server and that 145 would be overkill for it.
Might it be the g3240 is just too slow clocked, rather than being single-core? I don't know Intel low-end CPUs well enough to rememeber even appx specs on their G-series stuff. Semperon 145 is the same series as the X240 (Regor) except it's got one core disabled, I forget which clocks higher but might be the same - 2,8 Ghz comes to mind.
|
|
|
|