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

[cbuchner1]

Speed scrypt-jane went from 0.46 kh/s (non-interactive) to 0.59 kh/s (interactive) or 0.63 (non-interactive)
Scrypt went from 84 kh/s (interactive) down to 63 kh/s (interactive), including some HW errors:
With scrypt and -C 0 speed is down to 57 kh/s.

At the moment I am primarily interested in boosting the jane speed for Fermi (for purely selfish motives) - and it seems I have succeeded.

My list of Fermi devices is long:
GTX 560 Ti 448 core edition
GTX 660 (OEM label "V660-4098B") <-- what a shameful product mislabeling!
GT 630

However I find it rather bizarre that scrypt performance dropped quite a bit. That is somewhat unexpected.

I have seen some validation errors even on Kepler myself when used with the -C option. I have to investigate this a bit more.

Christian

[1714239537]

1714239537

[1714239537]

1714239537

[1714239537]

1714239537

[1714239537]

1714239537

[patoberli]

As I guess that my Edit could go forgotten, just discovered that my CPU load went from ~1% to 11% with this new version with scrypt-jane.
11% = 1 CPU core fully loaded.

[cbuchner1]

As I guess that my Edit could go forgotten, just discovered that my CPU load went from ~1% to 11% with this new version with scrypt-jane.
11% = 1 CPU core fully loaded.

also getting this with -b 32768  and / or -i 0 ?

[patoberli]

-b 32768 it's even a little bit higher (13 %) and interactive mode is "gone", means it's laggy again.
Both options is laggy and 4 % CPU load.
-i is laggy and 4-8 % load.

Can do more tests tomorrow.

[sambiohazard]

-b 32768 it's even a little bit higher (13 %) and interactive mode is "gone", means it's laggy again.
Both options is laggy and 4 % CPU load.
-i is laggy and 4-8 % load.

Can do more tests tomorrow.

Now I am no expert but I have observed that lag can increase with increase in use of graphic card RAM. e.g. on my gtx 275 896 MB

L60x3 uses 840 MB and PC is laggy

L80x1 uses 650 MB and PC is less laggy. I can watch youtube video at least.

also 80x1 gives more khashes. Just my observation.

[ManIkWeet]

I just found another YACoin block 
I say when Christian finds 2 in a row on a frequent basis...

[Snard]

I am confused by the GTX 670 cudaminer scrypt-jane hash rates as they seem quite low for certain coins. For example, in cache using cgminer with scrypt jane support I am getting in the mhps range compared to cudaminer, which gives me around 20 khps.

Is this simply a difference in how cgminer vs cudaminer are reporting the hash rate or something else?

[Treggar]

I am confused by the cudaminer scrypt-jane hash rates as they seem quite low for certain coins. For example, in cache using cgminer with scrypt jane support I am getting in the mhps range compared to cudaminer, which gives me around 20 khps.

Is this simply a difference in how cgminer vs cudaminer are reporting the hash rate or something else?

Depends on the coin... the older a coin is the lower the hashrates you'll get

[cbuchner1]

I am confused by the cudaminer scrypt-jane hash rates as they seem quite low for certain coins. For example, in cache using cgminer with scrypt jane support I am getting in the mhps range compared to cudaminer, which gives me around 20 khps.

Is this simply a difference in how cgminer vs cudaminer are reporting the hash rate or something else?

cudaminer is not yet suitable for low Nfactor coins, as the Kekkac part is done on the CPU with unoptimized code

for such coins cgminer for scrypt-jane will beat cudaminer.

[Snard]

I see...I thought i had read that elsewhere, but couldn't find the post. Thanks for clarifying.

[ozie]

This is due to the fact that with scrypt-jane you also need to calculate a KECCAK hash. This is done by the cpu singlethreaded atm.
I hope in the future there will be an option to offload it to the GPU like the SHA256 hash for scrypt.

But right now support for old deprecated Fermi Devices is more important 

Maxwell btw. is looking very promising for mining. If the rumors are true even Mobile GPU gets 6 Gig RAM.

[cbuchner1]

But right now support for old deprecated Fermi Devices is more important  

My GTX 560 Ti 448 core edition is anything but deprecated. It does something in the region of 280 kHash/s for scrypt mining.
Before Dave Andersen came along, Fermis were hashing stronger than Keplers.

Christian

[Lionel]

Guys let's try playing games while mining with cuda!
Try adding this options and let me know if the game lags

-i 1 -l S2x1

If the game lags try with

-i 1 -l S1x1

Lemme know

Lol that hashrate... 3.8kh/s while my 670 does 240kh/s on my own settings

too low you're right

try with these ones
-i 1 -l S3x3
-i 1 -l S5x5
-i 1 -l S6x6

should give a decent Kh/s while gaming,
for me
20Kh/s
40Kh/s
50Kh/s

and you? I have to try more aggressive configs

[Tweek]

Guys let's try playing games while mining with cuda!
Try adding this options and let me know if the game lags

-i 1 -l S2x1

If the game lags try with

-i 1 -l S1x1

Lemme know

Lol that hashrate... 3.8kh/s while my 670 does 240kh/s on my own settings

too low you're right

try with these ones
-i 1 -l S3x3
-i 1 -l S5x5
-i 1 -l S6x6

should give a decent Kh/s while gaming,
for me
20Kh/s
40Kh/s
50Kh/s

and you? I have to try more aggressive configs

For gaming I use
-i 1 -l K8x2 = 55kh/s and game (guild wars 2) feels a little laggy with ~15fps
-i 1 -l K6x2 = 40kh/s and game runs normal with ~20fps
Tested on GTX 670 btw

[cbuchner1]

A bit of research on Fermi performance with the new X-perimental kernel (Dave Andersen's work ported over to Fermi)

GTX 560Ti 1280MB: 0.83 kHash/s with -X 8x1    <--- the low VRAM is really hurting
GT 630 4GB VRAM: 0.72 kHash/s                      <--- low cost, low performance. The RAM doesn't help
GTX 660 OEM 4GB VRAM: 1.25 kHash/s             <--- that is one strange OEM part, I must say.

The new code is about 50% faster then the existing Fermi kernel for scrypt-jane. But I do get occasional
validation errors on Fermi + Kepler when I enable -C 1 or -C 2. Strange. Hence avoid Fermi parts like the
plague if you intend to do scrypt-jane.

In comparison.

A GT 640 (GK107) 4GB at stock clocks will do something in the range of 1.65 kHash/s. This is somewhat less than I expected
because my GT750M laptop part (same chip) delivers 2.1 kHash/s.

A GT 640 (GK208) 1GB GDDR5 manages to do 1.25 kHash/s with mild overclock. Again the low VRAM is hurting.

I will be getting more Kepler parts for comparison. GTX 650, GTX 650Ti (not the Boost version) with 2GB each.

Christian

[beachking2000]

Hey Christian. Is there anyway to get the texture cache feature to work in the 12-18 build? I haven't had success compiling off the github with windows so cant get the extra hash rate.

[Lionel]

Guys let's try playing games while mining with cuda!
Try adding this options and let me know if the game lags

-i 1 -l S2x1

If the game lags try with

-i 1 -l S1x1

Lemme know

Lol that hashrate... 3.8kh/s while my 670 does 240kh/s on my own settings

too low you're right

try with these ones
-i 1 -l S3x3
-i 1 -l S5x5
-i 1 -l S6x6

should give a decent Kh/s while gaming,
for me
20Kh/s
40Kh/s
50Kh/s

and you? I have to try more aggressive configs

For gaming I use
-i 1 -l K8x2 = 55kh/s and game (guild wars 2) feels a little laggy with ~15fps
-i 1 -l K6x2 = 40kh/s and game runs normal with ~20fps
Tested on GTX 670 btw

20fps is not the best experience, try with -i 1 -l K4x2

[sairon]

Hi there! Can you please plug your figures into this table for nvidia performance?
http://yacoinwiki.tk/index.php/Mining_Hardware_Comparison

[RbelMonstr]

Hi there! Can you please plug your figures into this table for nvidia performance?
http://yacoinwiki.tk/index.php/Mining_Hardware_Comparison

There is a survey/spreadsheet combo for Scrypt-Jane a few pages back, you could get your data from there.
And in the OP there are also 2 spreadsheets.

[bathrobehero]

A GT 640 (GK107) 4GB at stock clocks will do something in the range of 1.65 kHash/s.

I wonder how this one performs with an N factor of 15.