Show Posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.


Messages - Darkbane

Pages: 1 2 3 4 5 6 [7] 8
91
BitShares PTS / Re: fast AMD OpenCL PTS miner released
« on: February 09, 2014, 09:24:26 pm »
Quote from: NaN on February 09, 2014, 02:08:51 pm
I released a new version with large performance improvements. The syntax of the command line options of v0.1.x has changed and one has to specify the device IDs like in v0.2.x (see README). Furthermore, the command line option -a was introduced to select an algorithm. I guess that -a 0 or -a 2 should be fastest on R9 290(X) and -a 1 or -a 3 should be best for non-R9 290(X) cards.

clpts v0.1.4 win 64bit
Windows 8.1 64bit
AMD driver 13.251.0.0
two R9 270 2GB (975mhz / 1400mhz stock)
two R9 270X 4GB (1100mhz / 1400mhz stock)

30min test for each (previous cpm was on v0.1.2)

4463 to 5590 (-a 0)
4463 to 5827 (-a 1) -- update: after 2 hours I am getting 5840, rejections increased to 1.26%
4463 to 5579 (-a 2)
4463 to 5750 (-a 3)

so looks like a solid 25% increase across the board, 30% for option 1 on the R9 270 series cards works best as mentioned... now that I've done the 30min test for each option... I'll burn in for a couple hours on option 1 and check back in a couple hours...

do you think it would be a fair test if I ran 4 copies of clpts at the same time, one on each card? with each option? (thinking I could try all 4 options at once for several hours at the same time) or is it better to run one instance and check hours later to get a real read of the program? I'm not sure how its internals function if it would make any difference...

I did notice a roughly 20 watt increase over 4 cards, seems to be using slightly more memory on the cards.

92
BitShares PTS / Re: fast AMD OpenCL PTS miner released
« on: February 06, 2014, 12:58:05 am »
Quote from: NaN on February 05, 2014, 11:59:33 pm
Thanks for your feedback. I also reduced the number of workgroups, perhaps this causes the performance regression on non R9 290(X) cards. I'm working on improving the performance of non R9 290(X) cards, too.
EDIT: Could someone test this? You have to take e.g. version 0.1.2 and have to overwrite gpuhash_gcn with the gpuhash_gcn-file of v0.1.3.

Okay I copied the one file from 0.1.3 into 0.1.2 and its been running a half hour with almost identical results as 0.1.2 stock...

4460 c/m 17.38 sh/m (0.1.2 stock)
4467 c/m 17.04 sh/m (0.1.2 one file changed) -- 30min later
4449 c/m 17.17 sh/m (0.1.2 one file changed) -- 90min later
4445 c/m 16.97 sh/m (0.1.2 one file changed) -- 120min later
4449 c/m 17.09 sh/m (0.1.2 one file changed) -- 160min later

93
BitShares PTS / Re: fast AMD OpenCL PTS miner released
« on: February 05, 2014, 10:24:13 pm »
I downloaded the windows v0.1.3 64bit file, but it says v0.1.2 when I start the program up... and I do have a roughly 2% drop in performance for c/m and sh/m with the same system after letting it run for a couple hours... copied the v.0.1.2 files back in, and its within 0.3% where it was before after just a few minutes.

windows 8.1 64bit
amd driver 13.251.0.0
R9 270 2GB & R9 270X 4GB

94
BitShares PTS / Re: [ANN] ptsweb.beeeeer.org - Protoshares mining sub-pool
« on: February 04, 2014, 07:34:10 pm »
Quote from: xolokram on February 04, 2014, 07:17:14 pm
Quote from: Darkbane on February 04, 2014, 07:02:50 pm
Quote from: xolokram on February 04, 2014, 06:39:34 pm
...
in theory, more cpm for the entire pool, means more blocks will be found, however they will each be a lower payout as its now spread out among more cpm... but averaging out, more blocks despite being lower pay, would ultimately equate the same level of pay as before... in theory...
this would be true, when the overall power of the network wouldn't increase, but just the pool

but

that's not the case. with the new implementations for GPUs the overall network power has increased.
thus the difficulty was increasing to let the network stay at the desired block rate (there are not more blocks found!!)
thus the number of shares needed to find a block has increased too
thus you proportion of the overall network, which is the most important indicator for your payout, is decreasing (if you're still using the same miner/hardware)

i hope that was helpful. :)

- xolokram

it helps some thank you... when I say find more blocks, I do not mean faster than intended by design (which if I remember is every 5min?)... I mean the pool with more cpm has a higher better statistical chance to find that block than if say we there were half the cpm total for the pool...

100,000,000 cpm = more likely to find block and 50,000,000 cpm = less likely to find the same block... still chance for either, just more likely with more cpm...

95
BitShares PTS / Re: [ANN] ptsweb.beeeeer.org - Protoshares mining sub-pool
« on: February 04, 2014, 07:02:50 pm »
Quote from: xolokram on February 04, 2014, 06:39:34 pm
Quote from: xolokram on February 04, 2014, 05:48:15 pm
i don't know what hardware you're using, but there's opencl & cuda optimized gpu mining software for pts out there.
you have to compete with GPUs and different implementations (OpenCL/CUDA/etc.pp.)
it's likely that more efficient miners or software are "eating" your "missing" part of the payout
increased # of miners or found blocks won't help you with that

i guess you're missing the connected problem of increased difficulty that arise with the new efficient gpu implementations

But what I am asking... has the number of "cpm" on the pool, almost doubled in the past 5 days... I am trying to understand the information displayed in the line...

from what I understand...

46833 2014-01-31 16:07:17 30838 / 2546117 -> 0.18334 PTS / 15.14 PTS -> PAID

means on 2014-01-31 I did 30838 shares out of 2546117 total pool shares for block 46833 correct? so I was roughly 1.2% of total pool cpm, so I get 1.2% of the 15.14 PTS so .18334 pts...

now looking today at....

48086 2014-02-04 17:40:20 965 / 131346 -> 0.11180 PTS / 15.22 PTS -> block waiting for confirmations

means on 2014-02-04 I did 965 shares out of 131346 total pool shares for block 48086 correct? so I was roughly 0.07% of the total pool cpm, so I get 0.07% of the 15.22 PTS so .11180 pts...

so am I correctly reading this, and understand, that over the past 5 days the number of people using this pool, and total pool cpm rate, has almost doubled? (irregardless of which method they are using to create cpm, just the overall cpm rate has increased that much?)

in theory, more cpm for the entire pool, means more blocks will be found, however they will each be a lower payout as its now spread out among more cpm... but averaging out, more blocks despite being lower pay, would ultimately equate the same level of pay as before... in theory...

english as a second language sometimes makes words used different than my language, so I am trying to make sure I understand things correctly.

96
BitShares PTS / Re: [ANN] ptsweb.beeeeer.org - Protoshares mining sub-pool
« on: February 04, 2014, 06:31:28 pm »
Quote from: xolokram on February 04, 2014, 05:48:15 pm
?

your payout won't be affected by the shares-target-changes i've made recently! and the pool hasn't found a block since, so you can't judge this by now anyway.

the cause for changes in payouts over last days/weeks is the effect of competition regarding the efficiency of the miner software out there. i don't know what hardware you're using, but there's opencl & cuda optimized gpu mining software for pts out there.

- xolokram

going from this...

46602 2014-01-30 22:03:09 13253 / 1084416 -> 0.18500 PTS / 15.14 PTS -> added

to this...

47967 2014-02-04 09:47:39 15645 / 2137928 -> 0.11087 PTS / 15.15 PTS -> PAID

same cpm in both instances... but the average PTS per block has dropped almost in half over the last 5 days... I was attributing it to a large increase of people joining the pool... hence, more blocks, but lower payouts per block? but it averages out with more blocks found? or am I not reading any of this correctly?

so basically where I was getting .18500 pts per found block, now with the same cpm, I only get .11087 pts per block? so half the pay for the same work, but in theory we'll find more blocks so it averages out?

97
BitShares PTS / Re: [ANN] ptsweb.beeeeer.org - Protoshares mining sub-pool
« on: February 04, 2014, 05:23:41 pm »
Quote from: xolokram on February 04, 2014, 04:14:23 pm
server under maintenance atm
sorry for the inconvenvience

done, i changed the share target to reduce the server workload.

shares/min will decrease, col/min & payout/day won't! just fyi.

- xolokram

Wow you weren't kidding... 1/5th the sh/m now... I'm guessing a lot of folks have come over recently as my payouts have shrunk per share, granted more blocks found so it seems to even out?

98
BitShares PTS / Re: fast AMD OpenCL PTS miner released
« on: February 03, 2014, 02:47:45 pm »
Quote from: dongjin0 on February 03, 2014, 01:59:35 pm
Quote from: Schwede65 on January 28, 2014, 08:48:33 pm

Here are some numbers of my cards:
HD 7850:  930 cpm
HD 7950: 1815 cpm
R9  270: 1115 cpm
R9 270X: 1265 cpm
R9 290X: 2535 cpm
It's very strange that my 3x Asus 7850 could just run with about 700 cpm, even less than your one card. I tried new driver, it still doesn't work normally. Could you please help me find the solution? My English is not very good, hope you could understand me!

are all three cards actually being used... you have to tell the program which gpu's to use on the command line... example...

clpts_x86-64.exe yourptsaddressheresoyougetpaid 0 1 2

the 0 1 2 tells it to use three graphics cards

0 = gpu 1
1 = gpu 2
2 = gpu 3


HOWEVER... if your video card is only 1GB, it will not work from what I understand, you need 2GB... is only one of your cards 2GB?

99
BitShares PTS / Re: fast AMD OpenCL PTS miner released
« on: February 02, 2014, 03:02:55 pm »
Quote from: Bigboppa on February 02, 2014, 02:35:18 pm
New AMD driver 14.1 beta1.6 gives ~ 4150 cpm @ 2x 280x !!!
Some "new" warnings appearing..

what were you getting with the clpts miner before updating the driver?

100
BitShares PTS / Re: fast AMD OpenCL PTS miner released
« on: February 01, 2014, 10:37:49 pm »
mining 35s for author... I hope this actually lands on the found blocks once in a while heh... so what does it do anyhow, just switch the user address on the beeeeer.org pool and then convert back for the 30min? looks like today has been a good day so far over there... been running v0.1.2 for I think 48 hours straight now without any issues...

whats on the to-do list for your next version?

101
BitShares PTS / Re: fast AMD OpenCL PTS miner released
« on: February 01, 2014, 01:55:33 am »
Quote from: agc on January 31, 2014, 11:48:36 pm
Quote from: Darkbane on January 31, 2014, 11:36:02 pm
so we roughly know... using NaN's pts miner...

R9 270X = about 100 watt usage 1000 cpm
R9 290X = about 200 watt usage 3000 cpm

Anyone have a watt usage on the R9 280X?

is it as simple as splitting the difference at about 150 watts and 2000cpm?

Hi there,

my rig is drawing ~ 500 Watts according to Corsair Link when shuffling 5500 c/m.
GPU: 3 x 280x
CPU: AMD Sempron 145

How do you measure Watts per GPU??

Well the way I do it is with a digital display on my UPS device... I start the system with no cards, get a base wattage, then I add 1 card and start using it, check the wattage... then I repeat this with 2 3 4 cards... and check to make sure the increase in-between each card is relatively the same...

alternatively if you don't want to rip your rig apart... start it up without mining... you'll get your base while the cards are in low power mode since they aren't being used... then start mining with just 1 card, take a reading, then start mining with 2 cards, take a reading, then mine with all 3 cards and take a reading... then check the wattage between each one and you can get a relatively accurate idea what its using when mining... just be sure to wait until you reach maximum operating temperature so the fan is going as fast as its going to go... should be able to see a pattern between each card...

looking at your total wattage... looks like a fair guess would probably be around 150 watts each...

102
BitShares PTS / Re: fast AMD OpenCL PTS miner released
« on: February 01, 2014, 01:49:06 am »
Quote from: solar1 on January 31, 2014, 11:57:10 pm
Hi NaN

want to mine on beeer but
still having problems with v0.1.2 (R9 270, Win64, Driver 13.12)

Vendor of platform #1 / 2: Intel(R) Corporation
error in gpuhash.h, line 131: Method clGetDeviceIDs(...) failed (errorcode -1)

The AMD Card is not recognized.

The v0.2 Version is running perfect with 1070 cpm



any idea ?

Try updating your AMD driver, that seems to be helping most people who run into that issue... I'm running R9 270 and R9 270X together using v0.1.2 with the AMD 13.251.0.0 (12/6/2013 version) and have been running solid for 48 hours now... on windows 8 64bit for me...

103
BitShares PTS / Re: fast AMD OpenCL PTS miner released
« on: January 31, 2014, 11:36:02 pm »
so we roughly know... using NaN's pts miner...

R9 270   = about 100 watt usage 1000 cpm
R9 270X = about 110 watt usage 1100 cpm
R9 280X = about 150 watt usage 1800 cpm
R9 290X = about 200 watt usage 2500 cpm

your results will vary as people change engine/memory speeds, as well as different manufacturers having different core settings with the cards they ship... your results will vary...

(edited to add the 280X)

104
BitShares PTS / Re: fast AMD OpenCL PTS miner released
« on: January 30, 2014, 10:42:03 pm »
Quote from: NaN on January 30, 2014, 10:33:50 pm
My GPU miner uses a completely different algorithm than CPU miners and more memory is not a big advantage. The work for the GPU is produced locally until a new block comes in. Therefore the GPU should never idle. But the current kernels are executed in less than 0.05s or even faster, so there should be some noticeable CPU overhead. Additionally, the current kernels aren't 100% optimized so that the GPU is not doing any significant work for about 0.05s per round. This explains the low power usage compared to scrypt miners. Furthermore, the r9 290(X) has a different memory interface than the other GPUs. This might explain why it is better to use two threads per GPU on these GPUs.

it is quite amazing how much higher the cpm rate is with your work compared to other miners... and with less power! (less power is my favorite part, the fan noise is a blessing, I've actually got mine down to 30% now) would more difficult work (or larger blocks, not sure what to call it really) benefit the program or does it all end up being relatively the same with the method being used?

EDIT... sound is also a huge reason why I use the CORSAIR AX760 (the non-i version)... the fan doesn't even spin up until well after 50% load... so when I use these R9 270 cards, it never spins up in hybrid mode...

105
BitShares PTS / Re: fast AMD OpenCL PTS miner released
« on: January 30, 2014, 10:15:53 pm »
Quote from: duckproto on January 30, 2014, 08:16:42 pm
270X shows 99% GPU Load (930/1500) @ 50% Fan Speed but when I monitor the 290X I see 930 Core and Memory changes from 1500 to 300 every 10-15 seconds and the GPU load shows as 0% (while putting out 2300-2400c/m) - go figure.  The 290X shows between 90W and 143W.

crazy idea, but is it possible your 290x is working so quick, you're simply running out of work to do... which might explain why it has a burst and then backs down as the work is done... on other coins I've been able to increase packet difficulty in the pools, but I've never seen any options in any of these pts pools to increase or decrease difficulty... could it be as simple as you ran out of work for that brief moment heh... I mean none of the miners seem to take full advantage of the memory cards have (it just seems to be a minimum required to start), in other miners we're able to set the threads and intensity to fine tune that sort of thing... some of the pts cpu miners seem to have the ability to increase memory use which in turn boosts performance a little... we could just be leaving a lot on the table in this early stage of gpu mining for it?

Pages: 1 2 3 4 5 6 [7] 8

Page created in 0.187 seconds with 23 queries.