Compression Software: PeaZip, NanoZip, 7-Zip, WinRAR, WinZip, Tar
Abridged Results (see original)
| Format | Software | Time (s) | Ratio (%) ^ | Memory (MB) |
| PAQ8O | PeaZip | 16,502.748 | 58 | 900.0 |
| NZ (CM) | NanoZip | 290.880 | 60 | 105.0 |
| LPAQ8 | PeaZip | 338.960 | 60 | 102.0 |
| 7Z (LZMA) | PeaZip | 85.059 | 61 | 1,331.2 |
| NZ (Optimum1) | NanoZip | 91.770 | 61 | 461.0 |
| LPAQ8 | PeaZip | 273.290 | 61 | 13.0 |
| PAQ8O | PeaZip | 2,623.902 | 61 | 40.0 |
| 7Z (LZMA) | PeaZip | 38.592 | 62 | 56.0 |
| 7Z (LZMA) | 7-Zip | 46.000 | 62 | 200.0 |
| RAR | WinRAR | 56.000 | 63 | 42.0 |
| QUAD | PeaZip | 58.935 | 63 | 34.5 |
| BALZ | PeaZip | 91.168 | 63 | 67.0 |
| ARC | PeaZip | 74.153 | 64 | 80.0 |
| ZIP (Deflate64) | PeaZip | 21.380 | 65 | 4.0 |
| 7Z (LZMA) | PeaZip | 23.124 | 65 | 3.0 |
| BZ2 | PeaZip | 27.702 | 65 | 22.0 |
| BZ2 | 7-Zip | 248.000 | 65 | 48.0 |
| ZIP (BZ2) | 7-Zip | 249.000 | 65 | 26.0 |
| RAR | WinRAR | 8.000 | 66 | 42.0 |
| GZ | 7-Zip | 18.000 | 66 | 6.0 |
| ZIP (Deflate) | 7-Zip | 20.000 | 66 | 5.5 |
| GZ | PeaZip | 20.140 | 66 | 4.0 |
| ZIP | WinZip | 23.000 | 66 | 13.0 |
| PEA | PeaZip | 25.848 | 66 | 9.0 |
| BZ2 | 7-Zip | 27.000 | 66 | 44.0 |
| BZ2 | Tar | 40.000 | 66 | 9.0 |
| ARC | PeaZip | 4.828 | 67 | 27.0 |
| ZIP | WinZip | 9.000 | 67 | 13.0 |
| GZ | PeaZip | 9.570 | 67 | 4.0 |
| GZ | Tar | 10.000 | 67 | 1.5 |
| ZIP (Deflate64) | PeaZip | 10.595 | 67 | 4.0 |
| ZIP | WinRAR | 12.000 | 67 | 11.0 |
| BZ2 | 7-Zip | 21.000 | 67 | 10.0 |
| NZ (LZPF) | NanoZip | 1.400 | 68 | 164.0 |
Analysis
I want to point out a quick caveat here: my computer didn’t have enough RAM to run PAQ8O (PeaZip) at level 9/9 compression or ARC (PeaZip) at levels 8 or 9/9. Therefore, PAQ8O and ARC could have (theoretically) compressed at ratios better than 58% and 64%, respectively. Since they’re both in PeaZip and it already ranked #1, I think this is just overkill anyhow.
One interesting thing to note here is that NanoZip managed to pound this 143MB group of files down to 68% of its original size in only 1.4 seconds (using LZPF)! Now notice how long it took PeaZip to compress with PAQ8O – over 4.5 hours @900MB! The results, however, were very impressive. Even only running at level 8/9 it still beat out the others by a long shot. The problem is that no one else out there really supports the PAQ8O archive format or algorithm, and so anyone who wants to open this file up will have to use PeaZip to do it. Also, if PeaZip ever bites the dust someday, you might not be able to decompress the files ever again (without digging through archives to find an archaic version of PeaZip). The same goes for NanoZip (NZ), QUAD, BALZ and PEA files, though. Before RAR files caught on, they had this same issue as well (and still do, somewhat).
Fortunately for PeaZip, they also support LPAQ8 (a faster implementation, which sacrifices ratio). LPAQ8 managed to squeeze the file down to 60% of its original, while taking less than 6 minutes and 102MB to run. NanoZip was able to match this ratio using their CM algorithm in ’modest mode’. Interestingly enough, using ‘modest mode’ had the same results as ’standard mode’ with my data, but used 1/4 the RAM and was ~5 seconds faster. Notice that this also took ~5 minutes to run (compared to 4.5 hours for PAQ8O) but compressed 2% less than PAQ8O.
The very first time we see the 7Z format on the list is, interestingly, in PeaZip. PeaZip outperformed 7-Zip’s own LZMA algorithm by 1% (61% vs 62%), however it took almost twice as long and required 1.3GB of RAM. Using different settings, PeaZip also matched (62%) 7-Zip’s compression ratio but in less time (~38sec vs 46sec) and almost 1/4 the RAM usage (56MB vs 200MB).
At 2% below the best 7Z implementation (63%) finally comes the best RAR implementation (WinRAR is the only program I know of that offers RAR compression). Compressing into RAR format took longer than with 7Z (almost 20 seconds longer than PeaZip’s 62% 7Z result), and used a comparable amount of RAM.
Quite a ways down the list comes the best ZIP implementation (again, by PeaZip). This managed to compress down to 65% and in only about 21 seconds (using only 4MB of RAM). 7-Zip matched that ratio, coming in at 65% but taking over 4.5 minutes (the 20 second result from 7-Zip only compressed to 66%). Finally, WinZip’s own implementation compressed the file down to 66% and took 23 seconds. WinRAR seemed to have the worst ZIP implementation, only compressing to 67% but in 12 seconds (compare this to the other WinZip result, which also compressed to 67% but only in 9 seconds).
Notice that the BZ2 format (BZip2) was a close rival to ZIP. It compressed to the same ratio as the best ZIP implementation (PeaZip – 65%), but took about 6 seconds longer in PeaZip (and used slightly more RAM). When compared to the best WinZip ZIP implementation in particular (66%), PeaZip and 7-Zip both compressed with BZ2 at 66% but in less time.
Not surprisingly, GZ (GZip) came in at the bottom of the list. The best implementation (by 7-Zip this time) cranked out the file at 66% the original size and in 18 seconds. PeaZip also compressed it to 66% and in 20 seconds. Tar’s best implementation compressed it to 67% in 10 seconds.
Love the article , thanks a million man !
I personally prefer 7zip.
Comment by rohin — April 25, 2009 @ 9:59 pm
Yeah I can’t deny that … 7zip has always been my favorite. I’m at least starting to use peazip to see how much I can get used to it. The *nix version of peazip looks like it’s better than the windows one I tested here (the GUI aspects of it, at least). I’d give it a try.
Comment by QuadCEM — April 29, 2009 @ 11:02 pm
Squeez have full rar support.
http://www.speedproject.de/enu/squeez/index.html
Comment by Ivica — September 3, 2009 @ 7:17 am
Nice catch, it looks like WinRAR and Squeez can write in RAR format. Both are still shareware, though. I wonder if Squeez licensed RAR out or if they’re affiliated in some way.
Comment by QuadCEM — September 3, 2009 @ 8:49 am
While achieving better compression than ZIP, WinRAR is by far NOT the most powerful compression.
You will find that NanoZip will beat RAR file size by huge margins (sometimes even 20%).
But I agree that it is very experimental and maybe not suitable for everyday office use. Yet.
Comment by Huffman — February 15, 2010 @ 5:11 pm