cRARk 5.0

Fastest RAR password recovery

New 5.0 version with RAR 5.0 support!



User's opinions

GUI & distributed RAR password recovery

Version history
5.5b - Intel/AMD Ryzen SHA extensions support, up to 4 times faster. For the first time RAR 5.0 password are testing faster than RAR 3.0! 5.5 - PDL 4.0 with Unicode support in definition and dictionaries - '*' now can be not only at the end and has explicit length - other password definition language features, including generators 5.2b - NVIDIA Volta and AMD RX Vega support, 50% speed-up RAR 3.0 testing on RX Vega card - zero length files are now skipped 5.2a - serious bug with stored files fixed (password was not found) 5.1 - 64-bit version, 32-bit is now obsolete. Up to 5-20% speed-up on different archive types. 5.0 - WinXP support fixed. But future WinXP support is not guaranteed. 5.0 BETA 2 - Haswell AVX2 support, 2 times faster, CPU rate up to 4.80! - Linking CUDA 5.5 surprisingly resulted in 30%-50% speed-up on Fermi and Kepler cards - Intel HD grapichs OpenCL support 5.0 BETA - RAR 5.0 support and optimization for modern Intel, AMD and NVIDIA hardware - no more crark and crark-hp files, all features is now in one program 3.45 - -n option to specify GPU device number - correct recognition of AMD APUs - slightly faster XOP code - old Pentium III/4 code removed 3.43 - Kepler tuning - GPU rate and grid size are calculated more accurately - due to a lot of OpenCL compiler errors, the self test is included. (OpenCL version only) 3.42 - better RAR 4.x compatibility - OpenCL Mac OS X version 3.41 - -d option is now available in OpenCL version - fixed few (serious enough) bugs in cRARk-HP 3.4e - NVIDIA Kepler support (no optimization now, in fact GTX 680 is very bad in hash calculation) - AMD GCN (Radeon 7xxx) speed-up about 25% - yet more accurate CPU rate and SHA-1 clocks. Turn Turbo Boost off when to get correct results! 3.4d - Bulldozer code fixed, now it works fine and it's the fastest code I ever seen (CPU rate > 3.0)! - GPU grid sizes slightly changed - seriuos bug in AMD OpenCL code fixed - crash when running AVX code on non-AVX aware OS 3.4c - hang with Catalyst 11.12 fixed 3.4b - at least 5-7% speed-up on Core 2/iX - dictionary attack speed increased up to 20-40% - more accurate and detailed CPU rate calculation 3.4a - AVX, XOP and AES-NI instructions support, at least 20% speed-up on Sandy Bridge and more on Bulldozer (expected) 3.4 - OpenCL BETA version, works on both NVIDIA and AMD/ATI GPU - up to 15% speed-up on G80 NVIDIA cards (CUDA version) - stored files testing speed highly increased - most stored files are now supported by GPU 3.3d - -d option fixed - CUDA 4.0 compatibility added 3.3c - "File rarcrypt.dll not found" bug fixed - Upper/lower conversion for european languages changed, now is working for Turkish alphabet (spanish.def is also changed) 3.3a,b - Fermi support and optimization. 3.3 - CUDA engine totally rewritten, up to 5% faster; it is also running more smoothly - new -d option 3.25 - CUDA 3.0 linked which should fix some CUDA problems - Core 2, Core iX slight optimizations, up to 5% faster - RAR 2.0 few bugs fixed, -o option 3.2c-d - password was not found on GPU if charset > 128 chars 3.2b - support of SFX archives with encrypted headers 3.2a - 3-5% speed-up on both CPU and GPU code 3.2 - 50% speed-up on GeForce 8xxx and 9xxx, more than 100% on GTX 2xx serie - CUDA engine added in crark-hp - several CUDA improvements and fixes 3.2 BETA - NVIDIA CUDA support added (only in brute-force mode) - -c option to disable CUDA 3.1c - significantly sped up on very large archives - minor bug-fix of multivolume and encrypted headers archives processing - documentation corrected and converted to html - Mac OS X version compiled - new domain name 3.1b - Bugfix of false passwords in crark-hp again - Linux version available 3.1a - Bugfix in crark-hp (false passwords found) - Limitation for password length (28 chars) due to SHA implementation error in RAR 3.x - New -f option for choosing crypto functions 3.1 - Support archives with encrypted headers 3.0a - Core 2 optimization 3.0 - RAR 3.0 support with Pentium III/4/AMD optimizations 2.4 - Although RAR 2.x password cracking is not an actual problem, this upgrade is "just for fun". The brute-force rate is increased up to 25% on 3-chars (and longer) passwords and it is not depending of password length anymore. Thanks to Eugene D. Shelwien for his outstanding ideas. 2.3b - Just recompiling for P4. RAR 3.0 support is not expected now. 2.3a - Minor changes, about 5% speed-up. Documentation is improved. 2.3 - Pentium II optimized (2 times faster), CPU type identification, -5, -6 options added. PCL updated, including printing password in hex and "not defined .c(1)" bug fixed (thanks to Jan Polonsky). -v option improved. 2.2 - stored files support added. Linux (ELF) executable file added - maybe 5-10% faster. 2.1a - "5th line" bug in PCL 2.0 fixed. Only password definitions greather than 4 lines have been affected. Thanks to Dmitry Lisiy. 2.1 - PCL v. 2.0 added. Thus, appeared: support of different languages and encodings, charset definition, new modifiers and their parameters, timing and benchmarking functions, maximal password length is now 255. 2.0 - Password indication added (about 1 of 200000). Multimedia compressed files (-mm) bug fixed. 1.99 - RAR 2.0 support is finally included. Older RAR versions are not supported. 1.5a - Library Password Cracking Library (PCL) v. 1.1 added, al- lowing multifunctional dictionary attacks, brute force attack with known symbols, recovery of the incorrectly typed password and much more. An error, in connection with small (less than 10 bytes) files, is corrected. Documentation is complemented and corrected. 1.02 - Error, causing the result faulty in 0.5 % of the cases, is corrected. The speed is increased by 5 %. 1.01 - Insignificant error in an initial code UNRAR 1.01 (see what's new in RAR 2.0) is corrected 1.00: - First version released.

  Last updated: Dec, 2006 © Pavel Semjanov, 2000-06