[ANN] sgminer v5 - optimized X11/X13/NeoScrypt/Lyra2RE/etc. kernel-switch miner

[DStefanov]

Its fake  hashrate. Adding a few arguments to sgminer we can get +5MH, but our WU will be same as with 10MH per card

If you mean the image that I have and that fake say, i will tell you that talking nonsense !!! Hashrate is completely true and I can prove it with other photos from Nicehash !!! I would not argue with any nonsense!
Cheers!  

p.s. Would you showed your "real" values of hashrate and WU

[1714760008]

1714760008

[1714760008]

1714760008

[1714760008]

1714760008

[BitmoreCoin]

Its fake  hashrate. Adding a few arguments to sgminer we can get +5MH, but our WU will be same as with 10MH per card

If you mean the image that I have and that fake say, i will tell you that talking nonsense !!! Hashrate is completely true and I can prove it with other photos from Nicehash !!! I would not argue with any nonsense!
Cheers!  

p.s. Would you showed your "real" values of hashrate and WU

It is not fake.

When I use xI=512 instead of xI=2048, the accepted hash rate is the same at the pool, but locally, hash rate for xI=2048 is higher than xI=512, and use more power.

[drr0ss]

Its fake  hashrate. Adding a few arguments to sgminer we can get +5MH, but our WU will be same as with 10MH per card

If you mean the image that I have and that fake say, i will tell you that talking nonsense !!! Hashrate is completely true and I can prove it with other photos from Nicehash !!! I would not argue with any nonsense!
Cheers!  

p.s. Would you showed your "real" values of hashrate and WU

It is not fake.

When I use xI=512 instead of xI=2048, the accepted hash rate is the same at the pool, but locally, hash rate for xI=2048 is higher than xI=512, and use more power.

+1
payment of the pool corresponds to this hashrate

[sp_]

Just got the modded qubit and tahiti-bin kernal(russian?) to work on the hawaii chipset.(390x etc)

with -X 512:

On the 290:

quark: 13.45 MHASH
qubit:  13.12 MHASH

[Eastwind]

Just got the modded qubit and tahiti-bin kernal(russian?) to work on the hawaii chipset.(390x etc)

with -X 512:

On the 290:

quark: 13.45 MHASH
qubit:  13.12 MHASH

So the speed is not faster than 280X?

[sp_]

I don't have the sourcecode. The precompiled binaries are compiled and optimized for the 280x..

Without the sourcecode of the fastest miners, the R9 furyx is performing a bit worse than a R9 280x.

Read more here:

https://bitcointalk.org/index.php?topic=1091591.0

first results with the opensource miner (source cryptomining blog)

AMD Radeon R9 Fury X Results (sgminer opensource):
– X11 default: 6.778 MHS
– X13 default: 5.614 MHS
– X15 default: 4.69 MHS
– Neoscrypt default: 147 KHS
– Lyra2RE default: 287 KHS
– Lyra2RE Pallas Mod: 450 KHS

[pallas]

I don't have the sourcecode. The precompiled binaries are compiled and optimized for the 280x..

Without the sourcecode of the fastest miners, the R9 furyx is performing a bit worse than a R9 280x.

Read more here:

https://bitcointalk.org/index.php?topic=1091591.0

first results with the opensource miner (source cryptomining blog)

AMD Radeon R9 Fury X Results:
– X11 default: 6.778 MHS
– X13 default: 5.614 MHS
– X15 default: 4.69 MHS
– Neoscrypt default: 147 KHS
– Lyra2RE default: 287 KHS
– Lyra2RE Pallas Mod: 450 KHS

Can't wait for my Fury X.

please run whirlpoolx, groestl and sia (maybe keccak) on them and let us know :-)

[sp_]

please run whirlpoolx, groestl and sia (maybe keccak) on them and let us know :-)

Look on the bright side.

The pallas moded lyra2re was 57% faster on the furyx.  

From 287 KHS 450 KHS  


Djm34's private NVIDIA kernal does 3.2MHASH on the 780ti (11 times faster than the sgminer opensource furyx kernal)

[pallas]

please run whirlpoolx, groestl and sia (maybe keccak) on them and let us know :-)

Look on the bright side.

The pallas moded lyra2re was 57% faster on the furyx.  

From 287 KHS 450 KHS  


Djm34's private NVIDIA kernal does 3.2MHASH on the 780ti (11 times faster than the sgminer opensource furyx kernal)

Cause it doesn't use global.

I've tried hard to avoid using global ram (actually most of it is probably cached) but failed.
Tried using local ram but it was slower (you need a very little worksize in order to fit).
Probably there is a way to reduce memory usage and fit it in the registers or in local ram, but I haven't dedicated any more time to it (there was very little interest back then).
How does the nvidia version avoid using global ram?

[MaxDZ8]

How does the nvidia version avoid using global ram?

There are two possibilities. At the bottom, they use proper SIMD programming. Just open some random Cuda kernel and you'll notice their scan hash is sometimes 2D. This allows them to sum the amount of registers and fill the worksize more effectively as multiple WIs are used on the same hash.

The other point is that NV exposes SIMD shuffle in Cuda (being device-specific). OpenCL tries to be device-independent and therefore has no SIMD shuffle (CL2 has some basic shuffle functionality but still lacks full shuffle). AMD driver team is either too busy fixing bugs or improving perf on the game of the month so they couldn't be bothered in providing CL_AMD_GCN_SIMD_shuffle.

[Monopoly]

ANY prepared profit switching CONFIG FILE FOR miningpoolhub.com ? i couldn't write these command lines .

[ivcelmik]

Hi guys can someone tell me how much a 7950 OC Vapor-X hash in quark and x11 wolf's version.

Thx in advance

[Eastwind]

Hi guys can someone tell me how much a 7950 OC Vapor-X hash in quark and x11 wolf's version.

Thx in advance

clock for clock, that will be 7/8 of 7970/280X.

[serg_25]

Hi Wolf, as far as i can see new generation of AMD cards (R9 380) already in all shops. Looking around the forum i could see that hashing result for this model is ~3 mh\s (x11) seems like it's too low.
do you have any results for this card of your private miner in x11 algo?

just interesting of buying this cards and (if you have good results) your miner as well.

Thanks.

[pallas]

Hi Wolf, as far as i can see new generation of AMD cards (R9 380) already in all shops. Looking around the forum i could see that hashing result for this model is ~3 mh\s (x11) seems like it's too low.
do you have any results for this card of your private miner in x11 algo?

just interesting of buying this cards and (if you have good results) your miner as well.

Thanks.

R9 380 is a rebrand of the R9 285. As far as I can see, the stock clocks for the 380 are 1000/1550. Test results incoming.

EDIT:

(NSFW): https://ottrbutt.com/miner/x11wolf-07082015.png

slower than a lower clocked 7950... disappointing
maybe there is some kind of specific optimization for gcn 1.2 to make it worth mining with.

[MaxDZ8]

slower than a lower clocked 7950... disappointing
maybe there is some kind of specific optimization for gcn 1.2 to make it worth mining with.

According to an user on the AMD forums which also hangs out here every once in a while, Tonga has something ... which I understood to be independent per-byte LUT lookup. This could be very beneficial for AES based algos but I don't think there's much chance the compiler emits the proper code - assuming my understanding of his statements is correct.

[sp_]

I don't think Wolf0's AES implementation uses precalctables for the sbox, but I might be wrong. Perhaps he jsut reduced the size of the tables.

[Ram_x2]

Hello,

I've been out of for a while, but still my rig operational the whole time without any changes for over couple months. I can see Wolf0 made gr8 job on the moded bin files.

I have 2x R-290X Saphire Trix-OC (Hynix Ram) reached 7.948 MH on x11 by using Wolf0 bin, but I saw several posts that I could reach 10mh..



I tried xI 64 wasnt even close to those numbers...

Here's my conf  file..

],
"profiles": [
{
      "name" : "x11",
      "algorithm" : "darkcoin-mod",
      "intensity" : "19,19",
      "gpu-threads" : "1",
      "gpu-powertune" : "30",
      "worksize": "64",
      "gpu-engine": "1040,1040",
      "gpu-memclock": "1300,1300",
      "gpu-fan" : "20-100"
   }

],
"devices" : "0,1",
"hamsi-expand-big" : "1",
"failover-only" : true,
"shaders" : "2816",
"gpu-vddc" : "0.900",
"temp-cutoff" : "95",
"temp-overheat" : "90",
"temp-target" : "70",
"auto-fan" : true,
"auto-gpu" : true,
"api-port" : "4028",
"expiry" : "1",
"gpu-dyninterval" : "7",
"hotplug" : "5",
"log" : "5",
"queue" : "0",
"scan-time" : "1",
"temp-hysteresis" : "2",
"shares" : "0",
"no-restart" : true,
"failover-switch-delay" : "30",
"show-coindiff" : true,
"remove-disabled" : true,
"extranonce-subscription" : true
}


Could someone point how to do it?
Help appreciated..

[pallas]

Hello,

I've been out of for a while, but still my rig operational the whole time without any changes for over couple months. I can see Wolf0 made gr8 job on the moded bin files.

I have 2x R-290X Saphire Trix-OC (Hynix Ram) reached 7.948 MH on x11 by using Wolf0 bin, but I saw several posts that I could reach 10mh..



I tried xI 64 wasnt even close to those numbers...

Here's my conf  file..

Could someone point how to do it?
Help appreciated..

try "gpu-threads" : "2" or "gpu-threads" : "4"

[dashminer]

Hello AMD GPU miners!

I know you are all looking for better income and faster kernels...

Our algorithm switching multipool dashminer.com with BTC and DASH payouts makes it easy for you to get the most income from your hardware.

We already have an optimized kernel built in (it is faster than Wolf's leaked BINs). So you no longer need to search for a faster driver or download binaries.
Our custom open source miner (based on sph-sgminer) is very simple to tweak (as we removed most parameters and you just need to set intensity and overclocking params) and it even shows your daily income in DASH. It also shows speed not in megahashes but rather in % of R9 290. This way we can keep this value constant when we switch to a different algorithm (as all algos have different speeds in megahashes)

With our current promotion you would be earning 0.4 DASH per day or approximately 0.004 BTC per day (if you are paid in BTC) per each AMD R9 290 (R9 280X is about 85% of that, R9 270X is 50% and so on).
If this is a good price for you, please join DashMiner.com asap!
Note: Our pool only works with AMD GPUs. Nvidia or CPUs are not supported.

* Excellent profits: we quickly add new coins, we have a fast mining kernel, we can switch between coins and algorithms.
* Custom miner software optimized for speed and simplicity. No need to install special drivers, search for faster kernels or fiddle with parameters for sgminer. Just install and start mining. Our miner also shows you base rate payout in DASH so you know how much to expect.
* We pay miners once a day. Payouts are in BTD and DASH.
* No registration is required, just your wallet address.
* We have a unique payouts scheme: there is a base rate plus bonus (bonus depends on pool luck, coin prices, etc) . You will never receive a payout which is less than the base rate, unlike other multipools where sometimes you’re getting extra low payment (because of bad luck, wrong fork or something like that).
* We have 3 high performance stratum servers: San Francisco, New York and Amsterdam. We will add more servers if there are enough miners.
* We support DASH community by paying 1% of DASH miner profits to DASH donation address (XpAy7r5RVdGLnnjWNKuB9EUDiJ5Tje9GZ8).

This is DashMiner.com BitcoinTalk thread. Feel free to post there directly instead of polluting this thread.

PS: Kernel developers... If you have something interesting that is not publicly available - please contact us and we will make you an offer that you cannot refuse :-)

Thank you
dashminer.com