CCminer(SP-MOD) Modded GPU kernels.

[thevictimofuktyranny]

Anyone, doing any work on the new X11evo algo at moment?

[Stein]

tpruvot, has made a command to benchmark all the algos. I don't have it in my fork. (1.5.80)

That would be a usefull thing to have. Would help send tests to you.

[tbearhere]

no problem SP!
tpruvot 1.6.7 cuda 8.0 build
Results :
970+960 Stock clock:
blake 2370
Blake2s 1700
blakecoin 4120
bmw 1077
c11 11.53
decred 2271
deep 32.80
dmd-gr 35.12
fresh 13.66
fugue256 404
groestl 35.35
heavy 42.16
jackpot 20.50
luffa 262
lyra2 2600
lyra2v2 14.33
mjollnir 42.21
myr-gr 55.50
neoscrypt 515
nist5 38.60
penta 91.50
quark 20.45
qubit 17.25
sib 2568
scrypt -jane 558.17
skein 349
skein2 235
s3 33.28
x11 11.36
x13 9055
x15 7900
x17 7900
vanilla 3964
whirlcoin 21.72
whirlpool 21.56
zr5 4725

That quark ... is that on the 970?  thx  I use that for a reference... sp did good on quark for one.

[sp_]

No. That is quark on 2 cards total. Gtx 970+gtx 960. I just tested my latest private. It does over 30mhash on standard clocks.

[Velgelm]

sp, which miner i should use ? Private or Public ?
now i am at home. tomorrow when i will be in work, i can run tests and compare it vs cuda 8.0 results.
here is my small work/project and results in some optimization ccminer in Nicehashminer`s (960+970 standard clocks) - you can get this results for basis in cuda 6.5 and 7.5

Could you run the same test with the cuda 6.5 and 7.5 build please?

This results in mixed cuda 6.5 and 7.5 depend on algorithm
tomorrow i compare all algos in my project`s program and write there results

[sp_]

Priv6 is slower than priv5 on compute 5.2 devices exept keccak.

Why don't you compare release 74(cuda 6.5) and release 80(cuda 7.5)

https://github.com/sp-hash/ccminer/releases/

And then priv #5

[Velgelm]

ok tomorrow test it

[sp_]

And here is sp-mod #6 tuned for compute 5.2(unreleased)



Since the shadercount  is almost the same as a gtx 970+gtx 960 I would expect it to do almost the same results

[restless]

Are you saying that priv build #6 makes 30MH on quark with single 750Ti ?

Is there a link with ... something like table of these priv builds
build #X - optimized for Y, price Z ?

[sp_]

Not 30mhash.

32 mhash on the 980ti (single card)
 
But the latest#6 is tuned for compute 5.0.( I have decided not to release #6 compute 5.2)

have 6 private miners for the people who donate (donators will get updates).
 
1. 0.1BTC: Pentablake +100-120% (3 releases)
2. 0.1BTC: Cryptonight +10% (one release)
3. 0.1BTC: Spreadcoin +10-20% (with full sourcecode / linux compatible) (9 releases)
4. 0.1BTC: All nicehash algos optimized. 0-10% (6 releases)(x11,x13,x14,x15,x17,nist5,quark,skein,lyra2v2,neoscrypt)
5. 0.1BTC: decred +18-25% (9 releases) (Full sourcecode(linux) 0.4BTC)
6. 0.2BTC: Vcash(+13%+decred(+18-25%) (0.1 btc discount for the decred buyers) (6 releases)

[bensam1231]

no problem SP!
tpruvot 1.6.7 cuda 8.0 build
Results :/snip

Did you just run each algo one after another manually to get those results ? Wow

Pretty sure Nicehash miner has a built in benchmark mode and you can drop in exe's to use with it.

[tbearhere]

Anyone, doing any work on the new X11evo algo at moment?

What coin is that for?   Thx

[Amph]

Anyone, doing any work on the new X11evo algo at moment?

What coin is that for?   Thx

revolver, i also missed the launch, apparently it was launched at a bad hour

[Velgelm]

sp, which miner i should use ? Private or Public ?
now i am at home. tomorrow when i will be in work, i can run tests and compare it vs cuda 8.0 results.
here is my small work/project and results in some optimization ccminer in Nicehashminer`s (960+970 standard clocks) - you can get this results for basis in cuda 6.5 and 7.5

Could you run the same test with the cuda 6.5 and 7.5 build please?

This results in mixed cuda 6.5 and 7.5 depend on algorithm
tomorrow i compare all algos in my project`s program and write there results

as promised, here are the results of the comparison Cuda 8.0
Now I'm very busy, so scanning a piece by hand
80-x86 CUDA 8.0
80-X64 - Cuda 8.0
1.7.6X86 - cuda 6.5
1.7.6X64 - cuda 7.5

[sp_]

as promised, here are the results of the comparison Cuda 8.0
Now I'm very busy, so scanning a piece by hand
80-x86 CUDA 8.0
80-X64 - Cuda 8.0
1.7.6X86 - cuda 6.5
1.7.6X64 - cuda 7.5

Good job.

Why do you use sp#6?. On compute 5.2 you should compare with sp#5. Some algos are missing. Like The Keccak algo.

[Velgelm]

today i tested it and add to my compare

[restless]

So these results are from 960+970 and could be assumed that single 980 will get similar results, right?

[sp_]

So these results are from 960+970 and could be assumed that single 980 will get similar results, right?

970+960 should get similar results as the 980ti.  For neoscypt I get 1 MHASH in sp-mod #5 on the 980ti. The cuda 8.0 (1.7.6) build is only around 0.5MHASH. (the DJM34/sp-modded code is 100% faster)
The optimized decred/blake/blakecoin/Vcash is moved to the decred sp-mod. (I was hoping to do a sp-mod ultimate wich was a refork of 1.7.6 with all the kernals optimized, but I don't have time now)

[Amph]

mining cryptonight with more than 4 gpu result in "failed to allocate device memory for long state" on gpu 0, something to do with what memory exactly gpu?

edit it was the page memory too low, fixed with 6144

[sp_]

Here are my results from page one of this thread:

Sp-Mod #5 ( All nicehash algos optimized) default intensities hashrates in MHASH/s. (980ti 1290 coreclock/1650 memclock) faster than ccminer 1.7.5 cuda 7.5 32 build

nist5: 43.80 (+17%)
quark: 27.85 (+61,91%)
lyra2v2 16.10 (+16.6%)
x11:   12.85 (+18%)
x13:   10.27 (+19.97)
x15:    8.83 (+18.52 )
neoscrypt: 0.99 (+73.68%)