[ANN] cudaMiner & ccMiner CUDA based mining applications [Windows/Linux/MacOSX] - page 986.

SystmHash

newbie

Activity: 37

Merit: 0

Quote from: cbuchner1 on January 24, 2014, 09:18:38 AM

Quote from: SystmHash on January 24, 2014, 09:09:54 AM

Does the TI justify the additional $140? Roll Eyes

So sorry to post this again, but maybe someone could give me some advice, I would highly appreciate it!

My opinion is No. The 780Ti's have virtually no TDP headroom for overclocking (106% is the maximum with the stock BIOS), whereas the 780 cards can be sent to 120% (or 125%?) TDP and much higher overclock.

In Windows then the 780 and 780Ti get about the same output with OC. On Linux your overclocking options are more limited however.

I am currently doing 1.88 MHash/s on 3x 780Ti on Linux with a slightly modified BIOS. But now I wish I'd bought 780's instead...

Christian

Thanks Christian! Keep up the good work buddy! Kiss

(no homo)

bathrobehero

legendary

Activity: 2002

Merit: 1051

ICO? Not even once.

The single memory switch turned on (-m 1) reduces my max total warps from 181 to 118 (for N 14 SJ, L 3), but it's needed for any texture caching (-C).
Is it supposed to work like this?

cbuchner1

hero member

Activity: 756

Merit: 502

Quote from: SystmHash on January 24, 2014, 09:09:54 AM

Does the TI justify the additional $140? Roll Eyes

So sorry to post this again, but maybe someone could give me some advice, I would highly appreciate it!

My opinion is No. The 780Ti's have virtually no TDP headroom for overclocking (106% is the maximum with the stock BIOS), whereas the 780 cards can be sent to 120% (or 125%?) TDP and much higher overclock.

In Windows then the 780 and 780Ti get about the same output with OC. On Linux your overclocking options are more limited however.

I am currently doing 1.88 MHash/s on 3x 780Ti on Linux with a slightly modified BIOS. But now I wish I'd bought 780's instead...

Christian

BK124

newbie

Activity: 23

Merit: 0

Quote from: cbuchner1 on January 24, 2014, 03:45:57 AM

Quote from: BK124 on January 24, 2014, 01:00:54 AM

I have two 560 ti 1GB and from basing your stats off the numbers Ive seen with my cards, it seems your about maxing out the performance you can get. I cant push either of my 560 TIs past .44kh/s on scrypt jane and about 175 k/hs on scrypt usually with -l F7x1. A thing to remember with the flags, is that it doesnt matter where or what order you put them in as long as the syntax is right.

the X kernel was meant to push Fermi a bit further with scrypt-jane. Together with -L 2 or -L 3 you might see some 1.x kHash/s for Jane.

The next release version will likely have kernel names like FS, FJ for the respective scrypt and scrypt-jane versions.

Thanks Christian, I had read about the X kernel earlier in the the thread, didnt realize it was in the recent 122 commit that I already had Tongue

. I saw my scrypt-jane numbers double to about .88k/h from before so good improvement there. I've got two 4GB 760s arriving today to replace these old 560 ti 1GB so it was a good send off to see their numbers double Cheesy

Thanks again for all the work on this.

SystmHash

newbie

Activity: 37

Merit: 0

Quote from: SystmHash on January 23, 2014, 09:17:51 PM

Hi Guys,

Planning to get a new card, thought maybe you could give some advice? Huh

GIGABYTE GV-N780GHZ-3GD GeForce GTX 780 3GB $540
http://www.newegg.com/Product/Product.aspx?Item=N82E16814125488

GIGABYTE GV-N78TOC-3GD GeForce GTX 780 Ti 3GB $680
http://www.newegg.com/Product/Product.aspx?Item=N82E16814125489

Does the TI justify the additional $140? Roll Eyes

Thanks!

So sorry to post this again, but maybe someone could give me some advice, I would highly appreciate it!

chup

sr. member

Activity: 737

Merit: 262

Me, Myself & I

Quote from: taltamir on January 23, 2014, 10:41:52 PM

that window is unsightly and i wish you could minimize it to tray. there are methods to hide it entirely but if you do that, then you can't do things like turning it off and running it again on another pool.
Anyone knows a way to minimize it to tray?

Try cryptominers in a tray: https://bitcointalksearch.org/topic/crypto-miners-in-tray-a-lightweight-front-end-for-unattended-mining-149442
Highly customizable, so You have homework (xml-configuration file) to do.
Works for me.

Lacan82

sr. member

Activity: 247

Merit: 250

Quote from: taltamir on January 23, 2014, 10:41:52 PM

Quote from: Jayjay04 on January 23, 2014, 06:50:09 PM

But will it be without any windows opening ?

that window is unsightly and i wish you could minimize it to tray. there are methods to hide it entirely but if you do that, then you can't do things like turning it off and running it again on another pool.

Anyone knows a way to minimize it to tray?

put Start /B cudaminer.exe blah blah blah

in your batch file

knox203

member

Activity: 79

Merit: 10

Quote from: cbuchner1 on January 24, 2014, 04:52:48 AM

Quote from: knox203 on January 24, 2014, 04:02:53 AM

I just built the newest revision from github and am getting an error launching with: "cudaminer.exe -D --benchmark". I have two GTX 770's, so I'm not sure why it's selecting the "salsa" kernel. It should be using "Kepler", shouldn't it? Thanks!

pass the -d 0 flag to make it use just one device.

There seem to be problems using two at a time. These problems started when I moved to CUDA 5.5

Christian

I appreciate the help, Christian. That seemed to have done the trick! I noticed that after cudaMiner would crash, my main video card would mine about 100kh/s slower than my secondary back on the 12-18 release. A reboot would fix it, but then I found that enabling SLI then disabling SLI would take my main video card back up to the normal has rate. Dunno if this provides any insight, just something I noticed!

dgc81

newbie

Activity: 1

Merit: 0

Quote from: Zenitur on January 23, 2014, 11:13:11 PM

Code:

/home/zenitur/cudaminer-2013-12-18/CudaMiner-master/salsa_kernel.cu:166: undefined reference to `NV2Kernel::NV2Kernel()'
/home/zenitur/cudaminer-2013-12-18/CudaMiner-master/salsa_kernel.cu:165: undefined reference to `NVKernel::NVKernel()'

Run ./autogen.sh to regenerate the configure file, then rerun ./configure

cbuchner1

hero member

Activity: 756

Merit: 502

Quote from: knox203 on January 24, 2014, 04:02:53 AM

I just built the newest revision from github and am getting an error launching with: "cudaminer.exe -D --benchmark". I have two GTX 770's, so I'm not sure why it's selecting the "salsa" kernel. It should be using "Kepler", shouldn't it? Thanks!

pass the -d 0 flag to make it use just one device.

There seem to be problems using two at a time. These problems started when I moved to CUDA 5.5

Christian

knox203

member

Activity: 79

Merit: 10

I just built the newest revision from github and am getting an error launching with: "cudaminer.exe -D --benchmark". I have two GTX 770's, so I'm not sure why it's selecting the "salsa" kernel. It should be using "Kepler", shouldn't it? Thanks!

cbuchner1

hero member

Activity: 756

Merit: 502

Quote from: BK124 on January 24, 2014, 01:00:54 AM

I have two 560 ti 1GB and from basing your stats off the numbers Ive seen with my cards, it seems your about maxing out the performance you can get. I cant push either of my 560 TIs past .44kh/s on scrypt jane and about 175 k/hs on scrypt usually with -l F7x1. A thing to remember with the flags, is that it doesnt matter where or what order you put them in as long as the syntax is right.

the X kernel was meant to push Fermi a bit further with scrypt-jane. Together with -L 2 or -L 3 you might see some 1.x kHash/s for Jane.

The next release version will likely have kernel names like FS, FJ for the respective scrypt and scrypt-jane versions.

BK124

newbie

Activity: 23

Merit: 0

Quote from: cuoficr on January 23, 2014, 11:31:39 PM

Hey folks, relatively new to this, but I'm trying to squeeze the most I can out of my GTX 550Ti..

Just letting it auto-tune I'm gettting low 90kh/s on scrypt coins, and .44kh/s on scrypt-jane (was hoping for > .65kh/s) I've tried changing some of the commands, but it usually ends up crashing my driver. For example, I don't know where to put the kernel command in the code Huh

. I've noticed some are definitely better than others. (F32x4 > F83x2) Part of the problem is I dont really understand what the commands do... I'm sure at some point in these 150+ pages this has been addressed, but, my search is to broad Sad

I appreciate any help I can get.

I have two 560 ti 1GB and from basing your stats off the numbers Ive seen with my cards, it seems your about maxing out the performance you can get. I cant push either of my 560 TIs past .44kh/s on scrypt jane and about 175 k/hs on scrypt usually with -l F7x1. A thing to remember with the flags, is that it doesnt matter where or what order you put them in as long as the syntax is right.

For a better understand of the flags visit the Readme of cudaminer.
https://github.com/cbuchner1/CudaMiner/blob/master/README.txt

cuoficr

newbie

Activity: 2

Merit: 0

Hey folks, relatively new to this, but I'm trying to squeeze the most I can out of my GTX 550Ti..

Just letting it auto-tune I'm gettting low 90kh/s on scrypt coins, and .44kh/s on scrypt-jane (was hoping for > .65kh/s) I've tried changing some of the commands, but it usually ends up crashing my driver. For example, I don't know where to put the kernel command in the code Huh

. I've noticed some are definitely better than others. (F32x4 > F83x2) Part of the problem is I dont really understand what the commands do... I'm sure at some point in these 150+ pages this has been addressed, but, my search is to broad Sad

I appreciate any help I can get.

Zenitur

sr. member

Activity: 362

Merit: 250

Code:

g++ -g -O2 -pthread -L/usr/local/cuda-5.5/lib64 -o cudaminer cudaminer-cpu-miner.o cudaminer-util.o cudaminer-sha2.o cudaminer-scrypt.o cudaminer-scrypt-jane.o salsa_kernel.o spinlock_kernel.o legacy_kernel.o fermi_kernel.o kepler_kernel.o test_kernel.o titan_kernel.o -lcurl compat/jansson/libjansson.a -lpthread -lcudart -fopenmp
salsa_kernel.o: In function `validate_config(char*, int&, int&, KernelInterface**, cudaDeviceProp*)':
/home/zenitur/cudaminer-2013-12-18/CudaMiner-master/salsa_kernel.cu:166: undefined reference to `NV2Kernel::NV2Kernel()'
/home/zenitur/cudaminer-2013-12-18/CudaMiner-master/salsa_kernel.cu:165: undefined reference to `NVKernel::NVKernel()'
salsa_kernel.o: In function `find_optimal_blockcount(int, KernelInterface*&, bool&, int&)':
/home/zenitur/cudaminer-2013-12-18/CudaMiner-master/salsa_kernel.cu:381: undefined reference to `NV2Kernel::NV2Kernel()'
/home/zenitur/cudaminer-2013-12-18/CudaMiner-master/salsa_kernel.cu:379: undefined reference to `NVKernel::NVKernel()'
collect2: error: ld returned 1 exit status
make[2]: *** [cudaminer] Oшибкa 1
make[2]: Bыxoд из кaтaлoгa `/home/zenitur/cudaminer-2013-12-18/CudaMiner-master'
make[1]: *** [all-recursive] Oшибкa 1
make[1]: Bыxoд из кaтaлoгa `/home/zenitur/cudaminer-2013-12-18/CudaMiner-master'
make: *** [all] Oшибкa 2
zenitur@zenithar:~/cudaminer-2013-12-18/CudaMiner-master>

taltamir

full member

Activity: 196

Merit: 100

Quote from: chapmanjw on January 23, 2014, 11:00:14 PM

Quote from: taltamir on January 23, 2014, 10:41:52 PM

Quote from: Jayjay04 on January 23, 2014, 06:50:09 PM

But will it be without any windows opening ?

that window is unsightly and i wish you could minimize it to tray. there are methods to hide it entirely but if you do that, then you can't do things like turning it off and running it again on another pool.

Anyone knows a way to minimize it to tray?

This article has a number of suggestions for minimizing to the tray: http://www.actualtools.com/articles/detail.php?ID=888.

Every single one of them boils down to "buy our program for 20$"
I tested the trial version and it works just fine. Not sure if its worth 20$ though
I appreciate the help though

chapmanjw

newbie

Activity: 53

Merit: 0

Quote from: taltamir on January 23, 2014, 10:41:52 PM

Quote from: Jayjay04 on January 23, 2014, 06:50:09 PM

But will it be without any windows opening ?

that window is unsightly and i wish you could minimize it to tray. there are methods to hide it entirely but if you do that, then you can't do things like turning it off and running it again on another pool.

Anyone knows a way to minimize it to tray?

This article has a number of suggestions for minimizing to the tray: http://www.actualtools.com/articles/detail.php?ID=888.

taltamir

full member

Activity: 196

Merit: 100

Quote from: Jayjay04 on January 23, 2014, 06:50:09 PM

But will it be without any windows opening ?

that window is unsightly and i wish you could minimize it to tray. there are methods to hide it entirely but if you do that, then you can't do things like turning it off and running it again on another pool.

Anyone knows a way to minimize it to tray?

flatlanded

newbie

Activity: 7

Merit: 0

I've been lurking for a while now so I figured I'd post. I just wanted to say a big thank you to Christian for all of your hard work on this. The results have been awesome.

Just picked up a second GTX760 but this one is a 4gb model so it looks like I might have to dual-boot Ubuntu to get the most out of it.

3.85~ khash/s - MSI GTX760 2gb (has to go in for an RMA, the fan has thrown oil all over my mobo and case)
4.00~ khash/s - MSI GTX760 4gb (might need some tweaking, side note: the 4gb model is much better built card)

Thanks again!

SystmHash

newbie

Activity: 37

Merit: 0

Hi Guys,

Planning to get a new card, thought maybe you could give some advice? Huh

GIGABYTE GV-N780GHZ-3GD GeForce GTX 780 3GB $540
http://www.newegg.com/Product/Product.aspx?Item=N82E16814125488

GIGABYTE GV-N78TOC-3GD GeForce GTX 780 Ti 3GB $680
http://www.newegg.com/Product/Product.aspx?Item=N82E16814125489

Does the TI justify the additional $140? Roll Eyes

Thanks!

Topic: [ANN] cudaMiner & ccMiner CUDA based mining applications [Windows/Linux/MacOSX] - page 986. (Read 3426989 times)