Topic: [ANN][BLC] Blakecoin Blake-256 for GPU/FPGA With Merged Mined Pools Stable Net - page 188. (Read 409641 times)

I've seen two very different targets reported by Blakecoin wallet during reaper debug. (Or think I've seen, didn't made logs, unfortunately ) Visually these values differed by 256 factor, extra "00" in hexadecimal representation. I've turned off target checking and made reaper to submit all the shares, just to be sure that no block will be lost. Well, downside of this decision is high stale shares count on the screen and I can live with that.
Could you re-check difficulty calculations in the wallet code, may be some old code wasn't upgraded?

And now I'm going to be linched by Smelter users waiting for the fresh version
Anyone testing OpenCL miner, please report GPU temperature during BLC and BTC mining, the difference will give me an idea how much the kernel can be improved.

I have an idle 5870 to test with. Since I'm down from having 2x 5870's and a 5850, a 5870 and the 5850 died on me. I hardly gpu mine now because on just one card and XPM mine at the moment.

To clear up some potential confusion, I didn't write the OpenCL miner, I paid smolen a bit to modify reaper to support blakecoin.

Closed beta (open to anyone who has some hardware not already covered by testers who currently applied) will be going out in a few hours (Let's say, target of 5PM forum time, so about 2.5 hours). After that, sometime Monday (probably monday afternoon for me, might be early Tuesday forum time) I'll gather all the data (best settings for GPUs, GPU speed, temps, etc.), put together an official launch thread. However, since I only funded the project and didn't write the miner myself, I won't be able to provide any new builds based on code changes, etc. unless they are fairly straight-forward. My hope is that the community will pick the software development up.

Here's some current stats:
7950: >2GH/s at a decent overclock
7970: >1.6GH/s at a decent overclock

Also, I'd like to note that, since the blakecoin algorithm (blake[8,10]) is pretty simple, it seems that GPUs spend a bit of time idle between each round, meaning that higher clock rates are stable. For example, my 7950s at 1100MHz are running great. They came stock at 960.

Most graphics cards support OpenCL these days including some of the Arm SoC but it is best to check with the manufactures website and Check for OpenCL on your device info page.

but as for the hash rate on any of these devices only testing the OpenCL miner will give a hash rate that you can compare with others results, so until it has been release only Vorksholk knows the performance  

I am happy to do some testing I have a couple of 5870, 6850 and 6950  

Many thanks to Vorksholk for working on the OpenCL miner it does save me quite a bit of work  

still working on pool software atm current results are:

p2p has quite a few bugs that still need to be fixed before I could do a release
pushpool has a huge amount of bugs and I am not going to continue working on that

currently I am working with slush0 stratum pool software and creating a new version that works with Blakecoin and will release on Github once it is working and tested.

my plan was to get some pool software ready for this weekend but it has been quite a disappointing week with lots of bugs  

as soon as I get something working I will release same day

Edit:
Blakecoin is 2 weeks old today and overall I am happy with the progress we could use a few more developers working on supporting software but kramble and now Vorksholk have been very helpful in finding solutions 

when the GPU miner

Yeah, sorry to nitpick, but your 16 day figure just "felt" wrong to me, so I thought I'd put another calculation out there (I wasn't entirely certain I had it right though). Good luck with the OpenCL, even though that's not going to help with my mining (FPGA only, I've not got an even halfway decent GPU rig).

Ahh, appears my napkin math was wrong, never mind

At 10Mh/s it takes (2^32)/10000000 = 429 seconds (7.15 minutes) to scan the nonce range for a difficulty one share. Now my statistics knowledge is a bit rusty, but I reckon this should be the same as the long term average to find a share. So that gives 900 * 429 seconds = approx 4.5 days rather than 16 days, which chimes in rather better with my current experience of a roughly 2-4 blocks per day on around 89MHash/s.

Diff is now at 1168 (first time above 1000?), are the GPU's coming online already?

R9 280X it will be compatible?
R9 280X has the same statis as 7970.

OpenCL works on all platforms  

its just that ATI cards are faster at mining and most use them  

Are NVIDIA cards completely unsupported with no chance of even working?

From my observations, it seems that an average high-end processor (i7-3770k, AMD FX-8320, AMD FX-8350) gets around 10MH to 12MH.

To make the math easy, at 10MH, the CPU will get (from test results) somewhere around 0.06 difficulty-one shares per minute, or around one difficulty-one share every 24.6666667 minutes. At a difficulty of 900, you need 900 difficulty-one shares. Therefore, it would take nearly 16 days to get a block at current difficulty with one of those CPUs.


The OpenCL miner is coming soon, I want the launch to go real smooth, so if someone with one or more of the following cards can PM me so we can get performance statistics together and figure out optimal settings, that would be great:
-7990
-78xx
-7770
-6990
-6970
-6950
-68xx
-5970
-5850

Also, I'm gonna try to time the GPU miner release with some pool software from BlueDragon(747), but that is a huge job, so we'll see how timetables work out...

Even better!!!

I'm hopeful that BLC will hit an exchange in the near future.  Who knows, maybe it will turn up on "Cryptsy 2.0" later today.  However, the timing might not be the greatest against a soring BTC.

that's not the reduced round version!

blake256.cl from my fork is and should be faster
https://github.com/BlueDragon747/clblake

Sweeeet!

I compiled and ran the clblake ( https://github.com/wfr/clblake ) experiment code and tested it on an HIS 7950 with the following result:


I'm not sure how MiB/s equate to hashes per second but I would guess it would be rather good.

Right on. Now just need a mining pool.

Gonna try to release an OpenCL miner in the next 24-48 hours, keep your eyes peeled

selling some ks pm me with an offer

Cool. Thanks for putting effort into it. It's just to hard to solo mine at the moment.