[ANN] Kryptohash | Brand new PoW algo | 320bit hash | ed25519 | PID algo for dif - page 40.

WORE

sr. member

Activity: 350

Merit: 250

Quote from: mrmanager on December 19, 2014, 12:54:23 PM

Quote from: WhyMe on December 18, 2014, 12:38:13 PM

Quote from: ?? on ??

Quote from: WhyMe on December 17, 2014, 04:02:01 PM

Quote from: SpeedDemon13 on December 17, 2014, 02:38:40 PM

Anyone test the power consumption of the algo versus X11 and other algos? Just curious about the results...

Full config with one 6950@6970 :
x11 = 195w
khc = 240w

stop killing your gpus!!! running full voltage???

280x
x11 6.5mhs 144W
khc 151khs 137W
Neoscrypt 325khs 157W

Lol.
You have values with your hardware, i have values with mine.

On MY config with 1 6950 with shaders unlocked, I have :
x11 : @850/1300, ~1.9MHs, ~190w
khc : @825/1200, ~110kHs, ~240w
neoscrypt : @875/1300, ~6khs, ~165w

Compare what it can be !

EDIT : on idle, full config is at ~90w ...

You can likely lower the power usage and definitely the temps by underclocking the mem clock. With this algo it doesn't seem to need a high mem clock like scrypt. I have the mem clock set to under 400mhz on all of my rigs with no loss of kH.

Latest omega driver will tolerate no lower than 750 memclock in CCC on a 280x, noted that in GPU-Z the clock goes right to 1500mhz as soon as cgminer is launched even if overridden in CCC overdrive. How are you doing this?

WhyMe

sr. member

Activity: 661

Merit: 250

Quote from: mrmanager on December 19, 2014, 12:54:23 PM

You can likely lower the power usage and definitely the temps by underclocking the mem clock. With this algo it doesn't seem to need a high mem clock like scrypt. I have the mem clock set to under 400mhz on all of my rigs with no loss of kH.

On this card, lower memclock have an effect on hashrate.
On my 6950 1Go ( with faster memory than 2Go models ), memclock @800 have same result than @1300 ( stock value )
If i lower more than 800, GPUZ show me memclock is forced @ stock value.

Are you sure your memclock is really < 400 ?

And my GPU actually serve for heating my basement, so i don't really search heat performance Wink

wr104

sr. member

Activity: 329

Merit: 250

Quote from: ?? on ??

the fee is insane
100kch fee when sending
there should not be such a case that fee is greater than blockreward
not likely to trade this one
maybe needs some fine tuning..

IMO, zero fee coins are not sustainable in the long run. This is why KHC will always carry a fixed fee per transaction.

However, we can debate whether or not Half of 1% is the right value for the fee. The end goal here is to keep enough honest miners interested in KHC which in turn will ensure the security of the network over time.

Also, if you think about it, a fixed fee changes people's behavior regarding how the coins get spent. Knowing there is a fee, you'd likely think twice before you spend all your KHCs. And if you think about selling them in the exchanges, you'd likely wait for the right price or select the right amount of coins before you hit the submit button, which IMO will help with KHC price over time.

mrmanager

newbie

Activity: 50

Merit: 0

Quote from: WhyMe on December 18, 2014, 12:38:13 PM

Quote from: ?? on ??

Quote from: WhyMe on December 17, 2014, 04:02:01 PM

Quote from: SpeedDemon13 on December 17, 2014, 02:38:40 PM

Anyone test the power consumption of the algo versus X11 and other algos? Just curious about the results...

Full config with one 6950@6970 :
x11 = 195w
khc = 240w

stop killing your gpus!!! running full voltage???

280x
x11 6.5mhs 144W
khc 151khs 137W
Neoscrypt 325khs 157W

Lol.
You have values with your hardware, i have values with mine.

On MY config with 1 6950 with shaders unlocked, I have :
x11 : @850/1300, ~1.9MHs, ~190w
khc : @825/1200, ~110kHs, ~240w
neoscrypt : @875/1300, ~6khs, ~165w

Compare what it can be !

EDIT : on idle, full config is at ~90w ...

You can likely lower the power usage and definitely the temps by underclocking the mem clock. With this algo it doesn't seem to need a high mem clock like scrypt. I have the mem clock set to under 400mhz on all of my rigs with no loss of kH.

WORE

sr. member

Activity: 350

Merit: 250

Quote from: ?? on ??

the fee is insane
100kch fee when sending
there should not be such a case that fee is greater than blockreward
not likely to trade this one
maybe needs some fine tuning..

1) Every Kryptohash Coin (KHC) transaction carries a fixed fee of 0.5% that is paid to the miner who solves the block. Minimum fee is 5 Kryptoha-toshis or 0.00005 KHC.

•Blocks: 0 to 99 -> A fixed subsidy of 50 KHCs.
•Blocks: 100 to 999 -> A fixed subsidy of 50 KHCs plus a random subsidy of 0 to 399 KHCs.
•Blocks: 1,000 to 9,999 -> A fixed subsidy of 50 KHCs plus a random subsidy of 0 to 199 KHCs.
•Blocks: 10,000 to 99,999 -> A fixed subsidy of 50 KHCs plus a random subsidy of 0 to 99 KHCs.
•Blocks: 100,000 to 124,999 -> A fixed subsidy of 50 KHCs plus a random subsidy of 0 to 49 KHCs.
•Blocks: 125,000+ -> A fixed subsidy of 50 KHCs.

So;

Block Reward = TXN fee + 50KHC + Blockheight subsidy RND.

Block reward can't be lower than the reward for the block the TXN is stored in. Maybe you should try and solve blocks that contain your TXNs!

wr104

sr. member

Activity: 329

Merit: 250

The Kernel change was just one line where the FOUND value was increased from 0xF to 0xFF.

In other words, with the old Kernel there could potentially be up to 15 FOUND results in one OpenCL Kernel execution. With current GPU generations executing around 22K threads simultaneously and a low Difficulty in the KHC network, there was a chance that some people was getting more than 15 FOUND results and this caused the HW errors.

Now this limit has been increased to 255 so, the chances of this happening again are slim. Also, none of the other Kernels I saw would use a value lower than 0xFF so, I decided to go with what it seems to be a standard value for it.

The only real impact this value change could have caused was an increase of the OpenCL Output buffer size from 64 bytes to 1024 bytes.

SpeedDemon13

hero member

Activity: 518

Merit: 500

Quote from: WhyMe on December 19, 2014, 08:11:40 AM

Is solo mining OK ?
I found some blocks in solo, nothing in my wallet after 10 hours ...

They might be rejected or the miner might not be submitting shares.

Also, the new miner seems to be a little unstable. The previous miner would get me a solid 144kh/s, but the new one bounces around from 129kh/s to 140kh/s on a HD7970/280x. Maybe the new kernel is having an issue.

WhyMe

sr. member

Activity: 661

Merit: 250

Is solo mining OK ?
I found some blocks in solo, nothing in my wallet after 10 hours ...

SpeedDemon13

hero member

Activity: 518

Merit: 500

Quote from: wr104 on December 18, 2014, 10:05:43 PM

Here is a new 3.7.6 BETA version of cgminer-khc

https://github.com/kryptohash/cgminer-khc/releases

Changes since version 3.7.5

- Fixes for the stratum pool submitted by nonce-pool (thanks for the patch).
- Fixed the functions that calculate Diff which caused wrong diff values to be displayed and bogus found block messages.
- Fixed one of the causes of HW errors on secondary GPUs (more fixes might be needed).

Note: This version comes with an updated version of the OpenCL Kernel called kshake320-546.cl file. To avoid conflict with the old Kernel, place these new files in a brand new directory

When will the pool start using stratum? Is it possible to increase the intensity of the miner while still able to mine stable and increase hashing?

berbip

member

Activity: 143

Merit: 10

Quote from: wr104 on December 18, 2014, 10:05:43 PM

Here is a new 3.7.6 BETA version of cgminer-khc

https://github.com/kryptohash/cgminer-khc/releases

Changes since version 3.7.5

- Fixes for the stratum pool submitted by nonce-pool (thanks for the patch).
- Fixed the functions that calculate Diff which caused wrong diff values to be displayed and bogus found block messages.
- Fixed one of the causes of HW errors on secondary GPUs (more fixes might be needed).

Note: This version comes with an updated version of the OpenCL Kernel called kshake320-546.cl file. To avoid conflict with the old Kernel, place these new files in a brand new directory

Great work on the miner, had to lower --shaders-mul to 7 to keep the second card below 90c, lost some hash there.
my 2 280x are hashing @ 125 each

digital_money

newbie

Activity: 56

Merit: 0

Quote from: wr104 on December 18, 2014, 10:05:43 PM

Here is a new 3.7.6 BETA version of cgminer-khc

https://github.com/kryptohash/cgminer-khc/releases

Changes since version 3.7.5

- Fixes for the stratum pool submitted by nonce-pool (thanks for the patch).
- Fixed the functions that calculate Diff which caused wrong diff values to be displayed and bogus found block messages.
- Fixed one of the causes of HW errors on secondary GPUs (more fixes might be needed).

Note: This version comes with an updated version of the OpenCL Kernel called kshake320-546.cl file. To avoid conflict with the old Kernel, place these new files in a brand new directory

Does this sgminer sopport 6 cards?

MemoryShock

hero member

Activity: 728

Merit: 500

The continuing dev involvement with this thread is pretty cool...I wish I had the tools to mine...

WORE

sr. member

Activity: 350

Merit: 250

Quote from: wr104 on December 19, 2014, 12:35:20 AM

Quote from: WORE on December 18, 2014, 11:09:08 PM

Quote from: wr104 on December 18, 2014, 10:24:02 PM

No change in speed. I made a small fix in there that could have been the cause of HW errors. It shouldn't impact performance.

Device0 running about 10 degrees hotter with Intensity dynamic. Should we be using the load-balance with this kernel?

Temp cutoff at 90 not working but is not going over 95 so either it knows the card is a Tahiti and has it's own temp safe zones coded or there could be trouble with that part.

Size of share found only reports 0/1 but the block diff up on top is reported correctly now. Haven't seen Block! yet but I'd bet that is why block diff got your attention.

So far so good.

Intensity Dynamic setting doesn't have an effect in this version. I can try adding something similar to Dynamic in the next build. For the mean time, I would suggest you to lower the 'shaders-mul' parameter in order to allow your GPU0 some room to breathe. It seems setting 'shaders-mul' to 8 would be equivalent to the Max Intensity value in Scrypt.

cgminer uses dynamic if it is not specified in conf or bat file but doesn't appear to be changing the clock as designed. Just strange that it is running hotter on this kernel than the last one but it is not going over around 92 degress. If it becomes problematic I'll just lower gpu-engine as it is running around 1080 and powertune is already -20 and I could undervolt it but I don't want to starve the hardware beyond what is allowed in CCC overdrive. Gpu-engine 1050 seems to be a sweet spot, not going over 90 but still delivers +140k, shaders-mul is already at 8.

-Correction - It is kicking the clock down, dropped to 1000mhz in range from 850-1050, it is just taking it's time about doing it.

What driver version did you drop in? Is this the newest omega driver?

wr104

sr. member

Activity: 329

Merit: 250

Quote from: WORE on December 18, 2014, 11:09:08 PM

Quote from: wr104 on December 18, 2014, 10:24:02 PM

No change in speed. I made a small fix in there that could have been the cause of HW errors. It shouldn't impact performance.

Device0 running about 10 degrees hotter with Intensity dynamic. Should we be using the load-balance with this kernel?

Temp cutoff at 90 not working but is not going over 95 so either it knows the card is a Tahiti and has it's own temp safe zones coded or there could be trouble with that part.

Size of share found only reports 0/1 but the block diff up on top is reported correctly now. Haven't seen Block! yet but I'd bet that is why block diff got your attention.

So far so good.

Intensity Dynamic setting doesn't have an effect in this version. I can try adding something similar to Dynamic in the next build. For the mean time, I would suggest you to lower the 'shaders-mul' parameter in order to allow your GPU0 some room to breathe. It seems setting 'shaders-mul' to 8 would be equivalent to the Max Intensity value in Scrypt.

Quote from: antonio8 on December 19, 2014, 12:15:57 AM

I take it the new miner will not work for Nvidia cards?

Also I am not the brightest. Is putting the file in a new directory as simple as just putting the miner in a separate folder?

Nvidia cards should work but, you're only going to get a fraction of the Kh/s you'd get with any AMD card. This is because KSHAKE320 and other PoW algos that calculate hashes, take advantage of the high number of very simple Arithmetic/Logic Units (or ALUs) found in AMD GPUs. nVidia on the other hand, have fewer but more complex ALUs in their GPUs.

Copy all the files into a new folder and you'll be ready to go.

antonio8

legendary

Activity: 1400

Merit: 1000

I take it the new miner will not work for Nvidia cards?

Also I am not the brightest. Is putting the file in a new directory as simple as just putting the miner in a separate folder?

WORE

sr. member

Activity: 350

Merit: 250

Quote from: wr104 on December 18, 2014, 10:24:02 PM

No change in speed. I made a small fix in there that could have been the cause of HW errors. It shouldn't impact performance.

Device0 running about 10 degrees hotter with Intensity dynamic. Should we be using the load-balance with this kernel?

Temp cutoff at 90 not working but is not going over 95 so either it knows the card is a Tahiti and has it's own temp safe zones coded or there could be trouble with that part.

Size of share found only reports 0/1 but the block diff up on top is reported correctly now. Haven't seen Block! yet but I'd bet that is why block diff got your attention.

So far so good.

wr104

sr. member

Activity: 329

Merit: 250

No change in speed. I made a small fix in there that could have been the cause of HW errors. It shouldn't impact performance.

Youghoor

sr. member

Activity: 1036

Merit: 275

Quote from: wr104 on December 18, 2014, 10:05:43 PM

Here is a new 3.7.6 BETA version of cgminer-khc

https://github.com/kryptohash/cgminer-khc/releases

Changes since version 3.7.5

- Fixes for the stratum pool submitted by nonce-pool (thanks for the patch).
- Fixed the functions that calculate Diff which caused wrong diff values to be displayed and bogus found block messages.
- Fixed one of the causes of HW errors on secondary GPUs (more fixes might be needed).

Note: This version comes with an updated version of the OpenCL Kernel called kshake320-546.cl file. To avoid conflict with the old Kernel, place these new files in a brand new directory

So what are the difference in speeds in this Kernel.

wr104

sr. member

Activity: 329

Merit: 250

Here is a new 3.7.6 BETA version of cgminer-khc

https://github.com/kryptohash/cgminer-khc/releases

Changes since version 3.7.5

- Fixes for the stratum pool submitted by nonce-pool (thanks for the patch).
- Fixed the functions that calculate Diff which caused wrong diff values to be displayed and bogus found block messages.
- Fixed one of the causes of HW errors on secondary GPUs (more fixes might be needed).

Note: This version comes with an updated version of the OpenCL Kernel called kshake320-546.cl file. To avoid conflict with the old Kernel, place these new files in a brand new directory

digital_money

newbie

Activity: 56

Merit: 0

Quote from: WORE on December 18, 2014, 09:18:04 PM

Quote from: digital_money on December 18, 2014, 09:10:23 PM

Dev,what's status of your developmnet of the sgminer that supports 6 cards?? Huh

Pretty sure it is not done yet, but you can read through the pages in this thread to get an idea maybe when.

maybe 2 weeks later,I'll wait. Wink

Topic: [ANN] Kryptohash | Brand new PoW algo | 320bit hash | ed25519 | PID algo for dif - page 40. (Read 149454 times)