Dudes, I found some blocks (with my 7750) but they were apparently all 0 tx blocks. I'm not sure this (I mean, your tx-fees only policy) is working as expected. The last reported blocks by hashlink, look at the dates! Last Found Blocks - hashlink.eu| Block Finder | Time | Actual Shares | | 492370 | 06/25/2015 22:01:37 | 17,920 | | 492350 | 06/25/2015 21:38:59 | 30,720 | | 492319 | 06/25/2015 21:02:04 | 768 | | 492312 | 06/25/2015 20:52:39 | 90,427 | | 455775 | 05/30/2015 06:41:37 | 39,846 | | 455772 | 05/30/2015 06:40:11 | 59,730 |
Where are you going? |
|
|
|
Hey dudes, how are you? I'm running some testing and I hope suprnova is off with their estimations because that's how they report on the dashboard:  Anyway, suprnova diff is way too high for me. It should be cut in half. This is hashlink:  Again, ~22MHs reported. I have issues with this pool BTW. Somehow I detect wrong difficulty ... but still get no rejects  I will have to look into it. On the "my workers" page, it is signaled my worker has diff 256, but I get diff 1 from stratum. What other pools are available for testing? |
|
|
|
|
I think it's an interesting idea, from a purely educational experience. I don't really see it going anywhere else.
At that point, why not to just 3D print computer cases? Enthusiasts are going to shell out an arm and a leg for a pretty case.
And anyway, how do you 3D-print metal with decent electrical properties? 3D printing is mostly plastic AFAIK.
|
|
|
|
honestly you may not get much better than the work Wolf has done with these until new generations of cards come out with more ram.... I don't know how you guys keep beating the "moar RAM is better" horse. As a start, do you have an idea of how much RAM hashing algorithms consumes? X 11 has basically no memory requirement; it takes more or less as a couple of (diffuse) textures from an AAA video game. A couple of them. |
|
|
|
So if this is legit, then 1 user is having close to 70% of the network hashrate daily mining ~15 BTC worth of coins. That is either thousands of GPUs or FPGAs even with 200-300% faster private miners. Or ASICs or an exploit. Either way, it's ugly.
The currently available Lyra2RE kernel runs at about ~1% GPU efficiency. The authors themselves wrote on their paper (note that I had to copy this by hand as the authors are douchebags and didn't let me copy paste from the pdf) - 60: such performance penalty is most likely due to latency ... since GPUs are optimized for delivering high throughput rather than low latency.
This is false in general. GPUs have worked on hiding latency since bumpy-shiny shader was demonstrated back on NV10 days. - 60: in practice they later elaborate on it by writing
latency ... can usually be masked by the GPU ... - 61: ... the throughput provided in our tests is still 4.5 times lower than ... CPU
Only? It should be at least double digits. - 61: confirming again they cut it short
... hindering an attacker's ability to take full advantage of ... latency hiding capabilities of commercial GPUs.
There was also another statement about latency, very indicative. But due to the PDF being full of rubbish I haven't been able to find it. You know the drill. GPUs right now are slower... but not even all that much. And they're running at 1%. The private miner could as well be 10x-20x (1000%). Your observations about FPGA bandwidth is correct but your conclusion is not. That bandwidth is peak and most likely block write/read. |
|
|
|
|
I've stopped following NV a couple of years ago so take it with some salt...
As far as I've understood, the ALU-to-memory resource ratios in small maxwell are much more appropriate to efficient memory usage. Approximated: there are more "memory ports" compared to ALU count. What could be happening is that big GPUs ALUs are tripping on each other. Very approximated: every ALU has "more chance" to get to memory.
It is also possible to reproduce this on GCN1.0 (which has quite smaller buffers BTW).
Try measure the amount of memory effectively moved. I'd expect it to be much bigger for big maxwell and in both cases far higher than it should.
As a side note: "memory load" is unlikely to be what you think it is. It's most likely "memory controller active"; it could be active but stalled (~ shut down but burning energy).
|
|
|
|
Just two things to point out: - GPUs are basically better at everything (not just mining) they are a proper compute-oriented architecture, vector processing as it always have been. The "CPU for compute" mindset is borderline Intel propaganda. HPC switched to GPUs years ago and the only points where CPUs are competitive is when shit is written on purpose.
- AMD does not suck at efficiency. My 7750 hashes half as fast as a 750Ti, it also costed me half as much, a few months before the 750Ti launched...
- BUT! Maxwell consumes half as much! Maybe. I estimate room for 150% in hashrate with perhaps a 10% power increase. I consider Maxwell a good chip but not to the point to make GCN obsolete; leaving aside there's no low-end Maxwell yet.
You guys keep talking about efficiency but it's obvious to me you should talk efficiency of the algorithm implementations instead of drawing a connection to the hardware. That's fairly obvious considering how much talk about ah-hoc kernels around here. |
|
|
|
Why a cpucoin mined with coprocessor that teorically give you the oppurtunity to reach 5mh/s with 1 card?
It's a cpu coin, not other.
If the idea of m7m xmg cpucoin is to be asic and gpu resistant maybe GPU and other similar is not the good route to take.
But maybe i'm wrong.
You might be wrong, but you ask good questions. Apparently somebody does not have notion of "computing devices". Apparently somebody thinks some devices are "good" while some others are "bad". Lowering operative costs increases margins. |
|
|
|
And to be completely honest I wish more people would advertise their private miners because as much as I appreciate sp_, tpruvot, djm34 and other's work their optimizations are being used by everyone because they are free and in reality if everyone has fast miners then nobody really have an advantage.
It might be a good example (or just a coincidence) that since sp_ announced a week ago that he "pushed Quark on the AMD 280x from 1.7 MHASH to 4,5MHASH on standard clocks" the weekly average hashrate on yaamp for quark increased from 572 mh/s to 852 mh/s even though the profit decreased from 0.051 to 0.0398 mbtc/mh/d. And apparently there's a private russian miner which does 11mh/s (280x) and if that was free, quark mining would turn unprofitable very fast. So really, you either have very cheap electiricty, a fast private miner or tons of time to research coins nobody is looking at.
Miners mine mainly for profit so why shouldn't devs prioritize their profits as well? Assuming sp_ haven't changed his donation address he only received 4.6 BTC in the last 9 months meanwhile on yaamp alone currently there are 98 miners (with who knows how many cards per miner) using different SP-MOD ccminer forks. How many of those people ever donated or even knew this thread existed? Sure, private miners have plenty of cons on both sides but I'd certainly like to know the options. Quoted for emphasis. Everyone should truly consider some sort of crowdfunding. I have a long running joke about people wanting more performance and sort but don't tell anyone (LOL)! |
|
|
|
|
Yes indeed, I think I understand the connection you're trying to draw here.
Pool difficulty / share difficulty does indeed cause a different amount of magic numbers to be sent, in theory it does not change the amount of shares generated across time which is always a function of difficulty.
Nonetheless, changing diff does not generate new work and indeed block size does not affect the size of the results.
|
|
|
|
Increasing block size would not increase bandwidth for people mining in pools. Pools distribute work based on the worker hashrate and difficulty - the blocksize has nothing to do with the distribution of work in the pool. I'd like you to elaborate. As far as I know they just send some new work every time they hook up a new header/block or a new transaction / merkle tree. I cannot see any connection with difficulty. |
|
|
|
AMD-A6 5200 APU w/ Radeon 8800 HD graphics
That chip is at the lower end so I think 24 Kh/s would be the maximum hashrate you could expect from CPU mining. I do not know if the graphics section (are you sure it is a HD 8800? My references say it is a HD 8400) will support GPU mining but if it does... Kabini yay! As far as I can recall, GPU here is Hawaii-derived. Given proper kernels it will be a powerful GPU... for its price. I hope to see results! |
|
|
|
|
Interesting algo variations.
CPUs better mine this fast. As soon as a GPU miner is released the difficulty will go up up up!
|
|
|
|
You may be right here, I am always mixed which is which between CryptoNote and CryptoNight.
Sorry about the confusion. Indeed, the names are very badly chosen. I definitely won't blame you! I remember seeing a site that list CPU mine-able cryptocurrency. The thread with the site link should be around this section. Not more than a month ago. Check it out. I would also suggest you to check out OrbitCoin. I'm not sure if it was a dead coin or not but the difficulty seem low enough for CPU to mine with.
You're probably talking about cpucoinlist.com. It has not been updated in a while. A recent discussion about that. |
|
|
|
Afaik, Magi is similar to CryptoNote algo Absolutely not. They are very different things. Magi ANN post tells: M7M is based on the M7 PoW algorithm introduced by XCN.So I'd say they are not that different. But XCN is Cryptonite, a CryptoNote coin. There's no such thing as CryptoNote algo - that's probably CryptoNight. You can thank CryptoNote developers for being total douches at choosing names. M7 is definetely not CryptoNight unless you get very flexible in your mind. At that point all algos will be the same except maybe Quark derivatives. |
|
|
|
|
That's some truly slick updates!
I was indeed planning to do full AES round without t-tables as the amount of masks are nonsensical.
I had the impression the SALU was immensely updated for Tonga given it takes much more VGPRs on the analyzer.
I wonder how to trick the CL compiler in emitting this code.
But most importantly, what are they waiting to just make an AMD_GCN_swizzle extension!?
|
|
|
|
As a start, use sgminer. Last version is 5.something, you can download it from various sources (such as here). Be warned mining scrypt with GPU is useless now. Thread concurrency and intensity are way too high for your card. Start low and go up from there. |
|
|
|
Afaik, Magi is similar to CryptoNote algo Absolutely not. They are very different things. The only truly GPU-resistant algo around right now is yescrypt (GBST) but coin value is nonsensically low. The globalboost team sponsored creation of a GPU miner instead of barring themselves in the "CPU ivory tower". GPUs right now don't have any advantage. |
|
|
|
I have fresh support and I need testers. Hit github! Be warned current version is a bit... quirky but if it works it's interesting. |
|
|
|
|
Speaking of the wallet... I have been fiddling with it a few days and I cannot find the generate new address button... I'm pretty sure it used to be... I don't remember. Maybe I'm just dumb those days. Can someone give me a pointer?
|
|
|
|
|
Show Posts
