GT 755M vs GT 740M
#1
Hello everybody!
I'm a little confused.
What we have:
GT 755M on GK107:
[Image: cat32.jpg]
[Image: proxy.php?image=http%3A%2F%2Fimg.techpow...ff704f52a7]

And
GT 740M on GK208
[Image: hJNtyvyYu3kpcAaf-hgvg3rhu1qVeeDs5Q0o4Haw...a0dc413b7e]
[Image: proxy.php?image=http%3A%2F%2F1.downloade...8158783fd7]

So, what's the catch? Why the speed of GT 740M (GK 208) is twice faster than GT 755 M (GK 107)?
Reply
#2
And i really don't understand why does it writes: "Old CUDA chipset 3.0 detected, OpenCL performance is reduced. For ideal hashcat performance on NVIDIA GPU you need Shader Model 5.0 or higher"
[Image: shaders.jpg]
Reply
#3
The GPU is bad for cracking if the NVIDIA OpenCL runtime returns it as Shader Model 3.0
Reply
#4
Ok great so im not the only one getting this error. Is it because our graphic card is too old? I have a gtx780
Reply
#5
Look at the screenshot in the second post. Shader model version is 5.0 ! There is some bug in OpenCL drivers or hashcat itself
And in the first post I had shown that at almost same GPU (GK208 even worse) there is no such error
Reply
#6
Then OpenCL runtime reports it back as SM 3.0. If you want to complain you need to go to nvidia forum. We have nothing to do with it. You can also use clinfo to retrieve that info, just to make sure it's not somehow wronly interpreted by hashcat.
Reply
#7
I guess I understand. I have just looked the source code and have seen next:
Code:
CL_rc = hc_clGetDeviceInfo (hashcat_ctx, device_param->device, CL_DEVICE_COMPUTE_CAPABILITY_MAJOR_NV, sizeof (sm_major), &sm_major, NULL);
This parameter is NOT a shader model. It is called Cuda Compute Capability. And, yes, in my adapter on chip GK107 it version is 3.0
For chip GK208 Cuda compute capability version is 3.5, and it is obviously the reason why GK208 shows better performance
There is a table here where are shown various chips and their compute capabilities.
So I think the warning message should be changed.

P. S. Thank you for attention and sorry for my English Smile
Reply
#8
Added with https://github.com/hashcat/hashcat/commi...f84728e540
Reply