I am interested to this, while stumbled into good idea to deploy.
What exactly are the performance research you mentioned, and hi-precision libraries?
If we look for at how "hash" is computed, we can see many functions mpf_something. Those come from the high-precision library GMP. Hi-precision is critical to many applications in scientific computing. Google leads to several documents about this. I cannot quite tell if those can work as a drop-in replacement for GMP but it's quite obvious there's a lot of research going into some GMP-like GPU-accelerated processing (mostly cuda, as standard in high performance computing). So we can say that Magi, by leaving the realm of pure cryptography, is fostering competition in something that has real world value.
|
|
|
I was thinking... about that "wasted pow" power and using hashing for useful purposes.
By using GMP, XMG is possibly positioning itself to be a promoter for performance research which would translate to real world uses. Leaving aside I have been told big number crunching is already possible in Cuda (unclear if also on GPU), how would you react on knowing your fav coin is promoting development of hi-precision libraries?
|
|
|
Great idea, GBSTY now much more accessible!
|
|
|
linearIntensity 512. I'm a bit uncomfortable providing more details right now.
|
|
|
You were right! I missed it! Now my own build is doing well, 4378khs in Qubit
Pretty good for 512 cores.. (The GTX 750 Maxwell 1gb card retails for around $120) A AMD radeon 280x R9 does 5,5MHASH with 2048 cores. (optimized opensource miner with 5 times the power usage). What is this "optimized" miner you're talking about? This number is very odd as R7750 can push out >2000khs with ported kernels.
|
|
|
Why not use most recent drivers? They improved hashrate by a lot over the last year. Can confirm. GRS-MYR went from ~1500 to 3000 khs if memory serves, but maybe that was just me. Worth a try I'd say.
|
|
|
What is going on here?
|
|
|
Sounds like a good thing to me.
|
|
|
100% smart move with mface.
|
|
|
Yes, it's the pool. I really meant to say DBG qubit on blocksfactory.
|
|
|
Hi I'm mining at 52 mh/s and your pool is showing 27mh/s. After 1 hour. What is happening ? I can reply to that. The server knows nothing about your miner. There's no way for it to know how much hashpower you have. The server tries to guess your hashpower by looking at how often you send results and how good the results are. Both are random so it's fairly usual to have half the estimate, or twice as much. I mined at your pool in the past and everything was ok..the usual displayed pay per mh/s 25% above what it really is...and worker difficulty set to 1 when it should be .02 for gpu's. Whats going on plz. Only showing 27mh/s throwing in 52mh/s.dgb-qubit This. DBG qubit requires --difficulty-multiplier 1 as far as I can tell.
|
|
|
What is the state of qubit? Surprisingly, while the above mentioned version of sgminer does not hash correctly, sgminer 5 does work for me! No idea what's going on, have they been updated?
|
|
|
I'm not well aware of the state of DigiHash (I see discussed quite some time ago) but just to let you know, it does not seem to hash qubit correctly on my system (radeon 7750, 14.12 omega).
|
|
|
Interesting. Include your driver version - drivers are involved in the process. I can tell for sure sgminer4.1 (bundled with digihash) does not work anymore on omega drivers.
|
|
|
Considering the hardware/drivers you use I'm not surprised. Previous message relates to an old build of sgminer-sph only.
|
|
|
how much is the hashrate with a 290X in myrgroestl now ? thx 25 Mh/s with stock kernel
|
|
|
When I used to mine qubit using sph-sgminer performance fluctuated +-5% if memory serves. I would have to go great lengths to exceed 1MH on my hardware. I couldn't observe any such variation in the profiler (~1%, perhaps lower), I'd say there's a good chance this is due to stop-n-wait work dispatch. It would be interesting to see if those are somehow chipset related as it is my understanding transfer latency is higher on AMx chips.
|
|
|
So basically the misconception eventually has a basis of truth which is now largely obsolete.
Thanks!
|
|
|
I agree. 7000 series while still very relevant for gaming (especially on budget) have a serious problem: their model number is ... two years old? Maybe even three. Do not buy. Sell asap if you already have one as a new model will be out in a couple of months (actually sooner AFAIK, but availability is a big question).
7970 is exactly the same card as R9 280x. Probably the most widely used card for mining. Their model number is still two years old. Explain this when trying to resell (albeit I would admit they're probably recognized as R something at driver level). Sticker is sticker.
|
|
|
|