cbuchner1 (OP)
|
|
February 10, 2014, 11:10:27 PM |
|
|
|
|
|
IronEagle
Newbie
Offline
Activity: 49
Merit: 0
|
|
February 10, 2014, 11:19:45 PM |
|
My GTX570 has 480 Cuda-Cores, same MHz (732 MHz) as yours and the same amout of RAM (1280 MB). Which would be the best configuration?
try -l F960x16 with keccak, the differences between various configurations are rather small. I suppose you won't be getting much extra speed. Some people reported that keeping a web browser window with some flash playing in it open to boosted the GPU speeds somewhat. Christian What stands this "-L" for? Is this the amount of RAM?
|
|
|
|
cbuchner1 (OP)
|
|
February 10, 2014, 11:22:45 PM |
|
What stands this "-L" for? Is this the amount of RAM?
-L makes the scrypt scratchpad buffers smaller. These buffers aren't actually needed for keccak hashing, and to prevent an out of memory situation we have to shrink them to fit the card at least. This keccak feature was implemented in a hurry and I haven't cleaned up the initialization of the memory buffers and the autotune yet. Christian
|
|
|
|
IronEagle
Newbie
Offline
Activity: 49
Merit: 0
|
|
February 10, 2014, 11:27:34 PM |
|
What stands this "-L" for? Is this the amount of RAM?
-L makes the scrypt scratchpad buffers smaller. These buffers aren't actually needed for keccak hashing, and to prevent an out of memory situation we have to shrink them to fit the card at least. This keccak feature was implemented in a hurry and I haven't cleaned up the initialization of the memory buffers and the autotune yet. Christian Ok, thx. And what does the number behind this -L mean? cudaminer.exe --algo=keccak -d gtx560ti -i 0 -m 1 -l F896 x16 -L 1024 -o 192.168.2.2:9999 -O cbuchner1.1:x "F" is for the Fermi chip "896" are the cuda-cores doubled "x16" is what? "1024" is what?
|
|
|
|
cbuchner1 (OP)
|
|
February 10, 2014, 11:31:21 PM |
|
"x16" is what?
"1024" is what?
16 warps (one warp is a group of 32 threads) per thread block. And to understand this, you need to know the CUDA programming model in some detail. 16 is the upper limit for Fermi. the -L 1024 means to shrink the scrypt scratchpad buffer by a factor of 1024.
|
|
|
|
IronEagle
Newbie
Offline
Activity: 49
Merit: 0
|
|
February 10, 2014, 11:52:22 PM |
|
"x16" is what?
"1024" is what?
16 warps (one warp is a group of 32 threads) per thread block. And to understand this, you need to know the CUDA programming model in some detail. 16 is the upper limit for Fermi. the -L 1024 means to shrink the scrypt scratchpad buffer by a factor of 1024. Thx cbuchner1 How do I setup CUDAMINER for Solomining?
|
|
|
|
cbuchner1 (OP)
|
|
February 11, 2014, 12:07:44 AM |
|
How do I setup CUDAMINER for Solomining?
Pretty much like every mining software is configured for solo mining. I will have to refer you to some other online sources or other forum members for help. Christian
|
|
|
|
cbuchner1 (OP)
|
|
February 11, 2014, 12:54:04 AM |
|
I have set up my desktop PC for VertCoin mining using p2pool today here's one advanced command line for you note how I chose T12x20 instead of T12x24 for my gtx780 because somehow performance dropped sharply for any number of warps above x20 the +0.00029103 behind my wallet address is a difficulty modifier for the p2pool. @echo off cd ..\cudaminer-2014-02-09\x86 cudaminer.exe -q --algo=scrypt:2048 -d gtx780,gtx660ti,gt640 -l T12x20,K7x32,T4x24 -C 0,1,0 -b 512,2048,2048 -i 1,0,0 -m 1 -H 2 -o http://q30.qhor.net:9171 -u ViTWB96tJXSR4YwUdJCURiPtKpegtVRFif+0.00009155 -p x
Christian
|
|
|
|
djm34
Legendary
Offline
Activity: 1400
Merit: 1050
|
|
February 11, 2014, 12:59:05 AM |
|
I have set up my desktop PC for VertCoin mining using p2pool today here's one advanced command line for you note how I chose T12x20 instead of T12x24 for my gtx780 because somehow performance dropped sharply for any number of warps above x20 the +0.00029103 behind my wallet address is a difficulty modifier for the p2pool. @echo off cd ..\cudaminer-2014-02-09\x86 cudaminer.exe -q --algo=scrypt:2048 -d gtx780,gtx660ti,gt640 -l T12x20,K7x32,T4x24 -C 0,1,0 -b 512,2048,2048 -i 1,0,0 -m 1 -H 2 -o http://q30.qhor.net:9171 -u ViTWB96tJXSR4YwUdJCURiPtKpegtVRFif+0.00029103 -p x
Christian Can you explain a bit the difficulty modifier ? Is this something that p2pool requires or is this something you can set up ?
|
djm34 facebook pageBTC: 1NENYmxwZGHsKFmyjTc5WferTn5VTFb7Ze Pledge for neoscrypt ccminer to that address: 16UoC4DmTz2pvhFvcfTQrzkPTrXkWijzXw
|
|
|
cbuchner1 (OP)
|
|
February 11, 2014, 01:00:04 AM |
|
the p2pool's stats page suggested the modifier. https://bitcointalk.org/index.php?topic=438194.0Worker Tracker
Hash Rate Worker Rate: 393kH/s (Stale: 1.19%) Effective Rate: 388kH/s
Estimated Payouts Next Block: 0 VTC Day Average: 0.0478 VTC Day Total: 3.19 VTC
Suggested Difficulties Local Share: +0.00009155
|
|
|
|
ken-ray
Newbie
Offline
Activity: 24
Merit: 0
|
|
February 11, 2014, 01:29:56 AM |
|
I had trouble getting good performance from Ubuntu 64 bit and problems with cuda 5.5 I switched to Debian Jessie 32 bit which had Cuda 5.5 in non-free repository and now get hash rate equal to google spreadsheet. I wish I could compile 32 cudaminer on debian 64 bit system. Does anyone have experience with this? Best regards, Ken
|
|
|
|
DeltaQuebec
|
|
February 11, 2014, 02:05:18 AM |
|
Is any one mining VERT On a GTX 650 TI and getting more than 95 kh/s K4x24 in the bat GPU Settings: Cudaminer 2014.2.9 Results: Thanks in advance!
|
|
|
|
djm34
Legendary
Offline
Activity: 1400
Merit: 1050
|
|
February 11, 2014, 02:13:16 AM |
|
I had trouble getting good performance from Ubuntu 64 bit and problems with cuda 5.5 I switched to Debian Jessie 32 bit which had Cuda 5.5 in non-free repository and now get hash rate equal to google spreadsheet. I wish I could compile 32 cudaminer on debian 64 bit system. Does anyone have experience with this? Best regards, Ken The best way to install cuda 5.5, is to download the source from nvidia and compile it on the computer. I did it on ubuntu without any problem (had more problem with the driver though...) and even on fedora (on a laptop) Even though, I have it installed right now on ubuntu, I wish I had installed fedora... I think I hate ubuntu... hugly, and a completely clueless package management system... (have to search for hours to find where is hidden xterm)
|
djm34 facebook pageBTC: 1NENYmxwZGHsKFmyjTc5WferTn5VTFb7Ze Pledge for neoscrypt ccminer to that address: 16UoC4DmTz2pvhFvcfTQrzkPTrXkWijzXw
|
|
|
ivanlabrie
|
|
February 11, 2014, 02:16:10 AM Last edit: February 11, 2014, 02:34:16 AM by ivanlabrie |
|
I have set up my desktop PC for VertCoin mining using p2pool today here's one advanced command line for you note how I chose T12x20 instead of T12x24 for my gtx780 because somehow performance dropped sharply for any number of warps above x20 the +0.00029103 behind my wallet address is a difficulty modifier for the p2pool. @echo off cd ..\cudaminer-2014-02-09\x86 cudaminer.exe -q --algo=scrypt:2048 -d gtx780,gtx660ti,gt640 -l T12x20,K7x32,T4x24 -C 0,1,0 -b 512,2048,2048 -i 1,0,0 -m 1 -H 2 -o http://q30.qhor.net:9171 -u ViTWB96tJXSR4YwUdJCURiPtKpegtVRFif+0.00009155 -p x
Christian I got to 318kh/s with those settings...seems stable so far. 80% tdp 1267mhz core and 1100mv :p (+325 mem offset for 3301mhz) Crossing fingers...If I can get to 400 I'll be a happy camper :p (264w...skynet bios) EDIT: Memory seems to affect speed more than core speed...core still at 1267mhz but raised mem to +351 offset and got to 337kh/s stable (so far). Edit2: and it also seems to make my miner crash xD Downclocked mem by 13mhz...let's see how that goes.
|
|
|
|
debido666
Member
Offline
Activity: 68
Merit: 10
|
|
February 11, 2014, 02:35:14 AM |
|
Doesn't work, only booo... with the same configuration i mine doge, i can't mine VTC, why?
did it ever occur to you that Vertcoin isn't doge? No paws, no fur, no wow! seriously, you need --algo=scrypt:2048 and later on 4096, 8192, etc... I am not entirely sure about their N-factor schedule. The original wallet had this schedule: | N | Memory | Timestamp | Date/Time | | |----------|--------|-------------|-------------------------------|---| | 2048 | 256 kB | 1389430800 | Sat, 11 Jan 2014 09:00:00 GMT | | | 4096 | 512 kB | 1422985232 | Tue, 03 Feb 2015 17:40:32 GMT | | | 8192 | 1 MB | 1448151056 | Sat, 22 Nov 2015 00:10:56 GMT | | ... but apparently they have decided to relax the aggressiveness and change it to this more relaxed schedule: | N | Memory | Timestamp | Date/Time | | |----------|--------|-------------|-------------------------------|---| | 2048 | 256 kB | 1389306217 | Thu, 09 Jan 2014 22:23:37 GMT | | | 4096 | 512 kB | 1456415081 | Thu, 25 Feb 2016 15:44:41 GMT | | | 8192 | 1 MB | 1506746729 | Sat, 30 Sep 2017 04:45:29 GMT | | ... but I am not aware of any wallet update having been pushed out recently, so the code may still run on the old schedule at the moment. https://docs.google.com/spreadsheet/ccc?key=0Aj3vcsuY-JFNdC1ITWJrSG9VeWp6QXppbVgxcm0tbGc&usp=sharing#gid=0
|
|
|
|
ivanlabrie
|
|
February 11, 2014, 03:00:11 AM |
|
Doesn't work, only booo... with the same configuration i mine doge, i can't mine VTC, why?
did it ever occur to you that Vertcoin isn't doge? No paws, no fur, no wow! seriously, you need --algo=scrypt:2048 and later on 4096, 8192, etc... I am not entirely sure about their N-factor schedule. The original wallet had this schedule: | N | Memory | Timestamp | Date/Time | | |----------|--------|-------------|-------------------------------|---| | 2048 | 256 kB | 1389430800 | Sat, 11 Jan 2014 09:00:00 GMT | | | 4096 | 512 kB | 1422985232 | Tue, 03 Feb 2015 17:40:32 GMT | | | 8192 | 1 MB | 1448151056 | Sat, 22 Nov 2015 00:10:56 GMT | | ... but apparently they have decided to relax the aggressiveness and change it to this more relaxed schedule: | N | Memory | Timestamp | Date/Time | | |----------|--------|-------------|-------------------------------|---| | 2048 | 256 kB | 1389306217 | Thu, 09 Jan 2014 22:23:37 GMT | | | 4096 | 512 kB | 1456415081 | Thu, 25 Feb 2016 15:44:41 GMT | | | 8192 | 1 MB | 1506746729 | Sat, 30 Sep 2017 04:45:29 GMT | | ... but I am not aware of any wallet update having been pushed out recently, so the code may still run on the old schedule at the moment. https://docs.google.com/spreadsheet/ccc?key=0Aj3vcsuY-JFNdC1ITWJrSG9VeWp6QXppbVgxcm0tbGc&usp=sharing#gid=0Huh? So we gotta use 4096 now?? Christian: Can it be that --benchmark shows the correct hashrate and whilst mining I see lower figures, yet the pool perceives it as the hash the bench showed? I mine at 333kh/s now, stable, when using --benchmark with your settings and my clocks. When mining at the pool I see cudaminer reports 183-190-174kh/s and numbers all over the place but shares appear one after the other, pretty fast, faster than before. Pool reports 318kh/s. Edit: dammit, no, pool reports 160kh/s (confused shares with kh/s) hmmm
|
|
|
|
69charger
|
|
February 11, 2014, 04:07:11 AM |
|
I am loving me some Vertcoin
|
|
|
|
|
patoberli
Member
Offline
Activity: 106
Merit: 10
|
|
February 11, 2014, 07:45:41 AM Last edit: February 11, 2014, 08:07:10 AM by patoberli |
|
Ah here, why bitcoin went down yesterday: http://www.businessinsider.com/mtgox-statement-on-withdrawals-2014-2[edit] I see a LOT of wallet updates coming in the next weeks.
|
YAC: YA86YiWSvWEGSSSerPTMy4kwndabRUNftf BTC: 16NqvkYbKMnonVEf7jHbuWURFsLeuTRidX LTC: LTKCoiDwqEjaRCoNXfFhDm9EeWbGWouZjE
|
|
|
Legionnairez
|
|
February 11, 2014, 08:09:13 AM |
|
I'm trying to figure out best settings for scrypt-jane MRC, with GTX 680. So far I've got about 330 kh/s with: cudaminer.exe --algo=scrypt-jane:MRC -i 0 -H 2 -l K8x32 Is 330 low, average or high for MRC?
I'm especially interested in the use of -s and -b commands. What should I base these values on?
Also I'm interested how much the N-value effects the hashrate. The readme mentions low-count kernels instead of high (e.g. "k" instead of "K") for high N-values, but are there any means to increase hashrate for low N-value?
|
|
|
|
|