cuModuleLoad 301 error GTX 750 ti only on some hash types

[RainingStars]

Getting the GTX 750 Ti to work is not really the issue, I can run through most of the benchmarks, it's just certain hash types that cause this failure. Primarily the unix based ones.
Like:
500 = md5crypt, MD5(Unix), FreeBSD MD5, Cisco-IOS MD5
1800 = sha512crypt, SHA512(Unix)

However
1500 = descrypt, DES(Unix), Traditional DES
Does work, but seems really slow for being so old
Speed.GPU.#1.: 14306.9 kH/s

This is my result with just MD5
Speed.GPU.#1.: 596.5 MH/s

Anyway here is my output on trying to do a sha512crypt (unix)

Code:

cudaHashcat v1.20 starting...

Device #1: GeForce GTX 750 Ti, 2047MB, 1254Mhz, 5MCU
Device #1: WARNING! Kernel exec timeout is not disabled, it might cause you errors of code 702

Hashes: 1 hashes; 1 unique digests, 1 unique salts
Bitmaps: 8 bits, 256 entries, 0x000000ff mask, 1024 bytes
Applicable Optimizers:
* Zero-Byte
* Single-Hash
* Single-Salt
* Brute-Force
Watchdog: Temperature abort trigger set to 90c
Watchdog: Temperature retain trigger set to 80c
Device #1: Kernel ./kernels/4318/m1800.sm_50.64.ptx


ERROR: cuModuleLoad() 301

[3kings]

What is your driver version?

[RainingStars]

Driver version : 334.21

[atom]

descrypt is slower than md5, around 90 times.

[atom]

on nv

[RainingStars]

Any thoughts on the failure to load module on only certain hash types?

[atom]

I'd say it's missing the bzero or markov kernels. Did you remove some of the kernels in the kernels/4318/ directory? It's part of the archieve so it must be something like this. If unsure, just unpack it again

[atom]

we've worked on that issue on irc and found out a bug related to sm_50. If you have the same problem, come to IRC and tell me or wait for v1.21 release.