[ANN] cudaMiner & ccMiner CUDA based mining applications [Windows/Linux/MacOSX]

[bathrobehero]

GTX 670Ti: 7 SMX  (not enough memory for optimal config K7x8)
GTX 760: 6 SMX     (not enough memory for optimal config K6x8)
GTX 660: 4 SMX (optimal config probably K4x8, but because of some CUDA memory overhead you have to use K31x1)
GTX 440: 2 SMX (optimal config: K2x8)

Christian

My GTX 660 has 5 SMX units (2GB), and I can't find any near-decent config with -C 2. With -C 0, 14x1 works best.

[cbuchner1]

cbuchner1 do you have any 640s running currently? I'm curious what scrypt-jane hash rate you're able to hit in Linux. With your most recent texture cache update I'm able to get 1.65 kH/s with -C 2 (up from 1.5 yesterday). I only picked up one to try out, but I'm considering buying a few more for a new rig, which would be running Linux (my main box is a hackintosh)

Also, would there be any sense in attempting to OC it? I've heard low end cards don't get much out of it, but even 0.2 kH/s would add up given enough cards.

I will try this during the weekend. I am also going to check out a Fermi based GT630 for comparison, and a GTX 660 OEM card (sold in bulk)

EDIT: my 750M Laptop does 2.15 kHash/s now with the -C 2 option, using 2GB of its 4 GB video memory. But it gets hot and the fan jumps to 100%, which sounds like a turbine engine. This should be similar to what I expect from a GT 640.

Christian

[cbuchner1]

GTX 670Ti: 7 SMX  (not enough memory for optimal config K7x8)
GTX 760: 6 SMX     (not enough memory for optimal config K6x8)
GTX 660: 4 SMX (optimal config probably K4x8, but because of some CUDA memory overhead you have to use K31x1)
GTX 440: 2 SMX (optimal config: K2x8)

Christian

My GTX 660 has 5 SMX units (2GB), and I can't find any near-decent config with -C 2. With -C 0, 14x1 works the best.

You're right. http://www.geforce.com/hardware/desktop-gpus/geforce-gtx-660/specifications
So it's 5 then.

[cbuchner1]

Screenshots and proof are here:

http://forums.evga.com/tm.aspx?m=2089066

And here:

http://forums.evga.com/tm.aspx?m=2091247#2091885

these links belong onto reddit's /r/litecoinmining. Who's going to post?

Christian

[cbuchner1]

Yay! 3000 Yacoins mined. 3 million more to go.

Christian

[bathrobehero]

Yay! 3000 Yacoins mined. 3 million more to go.

Christian

Congrats! I'm waaaay behind you  ...but I'm enjoying tinkering around


The highest I can go with -C 2 is K10x1, K11x1 triggers out of memory error.

Does it mean that -C 2 has no use to GTX 660's with 2GB?

[eduncan911]

Screenshots and proof are here:

http://forums.evga.com/tm.aspx?m=2089066

And here:

http://forums.evga.com/tm.aspx?m=2091247#2091885

these links belong onto reddit's /r/litecoinmining. Who's going to post?

Christian

Ha, go for it!

[Schleicher]

Memory speed is a confusing one, most cards today have GDDR5 memory on them which means their effective clock speed is 4 times the default.
So for example my car has 1502 Mhz memory clock, which is 6008 Mhz effective speed. So when I let's say downclock it by 2000 Mhz in OC Guru, it will end up being 1000 Mhz instead of 1502 due to the multiplier. To further confuse the matter, some OC tools are using a 2 times multiplier so for example in EVGA Precision a -502 Mhz downclock (max) results in ~1251 Mhz and some use no multiplier, just the default speed. Which is yet another reason to use GPU-Z when it comes to overclocking.

Yeah, the clocks are confusing.
The CUDA driver is saying this:

Code:

Device 0: GeForce GTX 660
   1032.5 MHz peak clock rate
   3004 MHz peak memory clock rate

and GPU-Z 0.7.5 is saying 1033 and 1502 MHz.

[bathrobehero]

Is there any way to tell a scrypt-jane crypto's N factor other than trial and error if it's not in getmininginfo?

[cbuchner1]

Is there any way to tell a scrypt-jane crypto's N factor other than trial and error if it's not in getmininginfo?

the latest cudaminer prints it...

[bathrobehero]

I don't mean for YaCoin, but other coins, for example VelocityCoin, which has an N factor of 7 and cudaMiner still says 14.
I compiled multiple instances with different fixed N factors, it's just if there's a coin without any pools whatsoever, I can't tell if I'm hashing in the right gear or not.

[cbuchner1]

I don't mean for YaCoin, but other coins, for example VelocityCoin, which has an N factor of 7 and cudaMiner still says 14.
I compiled multiple instances with different fixed N factors, it's just if there's a coin without any pools whatsoever, I can't tell if I'm hashing in the right gear or not.

erm, really? I thought the N factor computation takes the currently mined block's timestamp and derives the N factor from that. But maybe the parameters for the coin are hardcoded. I have to check that.

Christian

[bathrobehero]

Well, you're using a fixed timestamp, 1367991200 which points to 2013-05-08, which I guess is the start of YaCoin, I don't know.

Edit: yeah, that date seems to be the launch date for YaCoin.

[manofcolombia]

scrypt-jane hasnt been merged into anything on the OP right? so everyone running jane has compiled it themselves from the git correct? Or have I been wasting time thinking that it hasnt been in the 12/18 xD

[Sarg338]

Gotcha, thanks!

Finally got around to installing Visual Studio and compiling the newer version, and I got a drop of about 20 khash/s on my GTX 760  (~290 with the 12/18 build down to ~270). Is this normal? From what I read, there was suppose to be a bump up in hash rate if you used -H 2 now instead of -H 1. I built it in Release mode for x64.

the added flexibility for scrypt-jane indeed costs some performance. I may later decide to roll out dedicated scrypt kernels and separate out the jane code. But for now, I am focused on getting the scrypt-jane feature completed first (there's more mining profits in it)

Alright, got it!

[jots]

Hi-ya Christian,

My 660Ti registers at ~270 khash/s peak (scrypt) for a single instance of cudaminer (which appears to start a single mining thread) - By starting two instances, I'd have expected that rate to halve.  Instead, each instance reports ~200khash/s peak.

This is surprising. What is your launch configuration?
What does GPU-z show for GPU utilization when running just a single instance?
How's GPU utilization and memory usage with 1 and 2 instances?

I used the the same launch config (-l K7x32 -C 1 -i 0 -m 0 -H 1 a product of not realizing it was already running in another terminal).  After messing around with it for the last day, I can tell you the reported numbers on the second cudaminer instance are lies.

Juggling the launch config around (7x32, and the other at 7x14), I can get two instances to submit results correctly, but they don't exceed my 270 khash/s (one settles at 180 something khash/s and the other at 80 something).  There's a +500MiB GPU memory usage increase in this scenario (Don't know about utilisation... nvidia-smi won't report it).  There doesn't appear to be any operational efficiency to gain here.

Apologies for wasting your time.  

[bigjme]

Yay! 3000 Yacoins mined. 3 million more to go.

Christian

Congrats. I have 150 in 3 days. Havent ran it the full time though. Heck I didnt run it for over 12 hours yesterday abd most of today it has been off. Guess I should really leave it on more.

Hmm just found some gtx 640 4gb cards for £87 and the guy has over 10 of them. If I had the money! Maybe if yacoin actually starts getting me coins I might invest but the return is very low right now

Has anyone tried the 4gb gtx 630? Their about £63 so very cheap. And if they run faster then my 780 you could be seeing almost 5khash/s each for 4gb memory. And 3 of those for £190 for 15khash/s isnt bad

[bathrobehero]

Has anyone tried the 4gb gtx 630? Their about £63 so very cheap. And if they run faster then my 780 you could be seeing almost 5khash/s each for 4gb memory. And 3 of those for £190 for 15khash/s isnt bad

I'm not completely sure, but I really doubt the GT 630's would be anywhere near 5kH/s. Sure, those have 4gb memory but they also have 96 cuda cores and direct compute 2.1.

[cbuchner1]

Has anyone tried the 4gb gtx 630?

I get 1.6 kHash out of a GT640 on Windows right now.

GT630 is expected to be worse...

[bigjme]

I was guessing :-) from Christian's new rig idea of 15khash, using 4 640's and a 660. Which is abiut 3khash/s each.

But windows is getting very bad results. I gained over 1khash/s by switching to linux