Comparison of compression type performance

[Cregrant]

Did someone else compare the performance of different compression methods? It will be interesting if you share some numbers in this thread.

Deprecated speed test

7-Zip CRC-32 hashing + copy to RAM disk
3 x 1.2 GB files

Source type	Compression	CRC-32 hashing, MB/s	Copy to RAM disk, MB/s
RAM cached	1.00	1800	580
SSD	1.00	450	370
SSD + XPRS4	0.63	240	310
SSD + XPRS8	0.59	260	340
SSD + XPRS16	0.57	270	350
SSD + LZX	0.49	230	290

---

Simple script in Java to measure latency and linear read speed (JVM warm-up included)
300 x 317 KB files for a latency check
6 x 322 MB files for a speed check

Source type	Mean latency, ms	Min latency, ms	Linear read, MB/s
RAM cached	0.011	0.010	2520
SSD	0.220	0.120	440
SSD + XPRS4	0.170	0.120	310
SSD + XPRS8	0.230	0.180	330
SSD + XPRS16	0.270	0.220	330
SSD + LZX	0.450	0.410	260

My PC specs:
Intel Core i5-3570K, 4.4 GHz
DDR3 2000 MHz, 32 GB
Crucial BX500 240GB, SATA-III

[AFaustini]

Seems to me that XPRESS16 is the better method with higher compression and faster reading from disk

[Cregrant]

Our sibling CompactGUI is able to compress RAM disks, so here is the real comparison of 'fast' disk reading speed using my script

Game asset 1 (nice compression) (4x1.4Gb)

Type	Linear read, MB/s	Compression
RAM	4300+	1.00
RAM + XPRS4	1170	0.27
RAM + XPRS8	1410	0.23
RAM + XPRS16	1590	0.21
RAM + LZX	790	0.16

Game asset 2 (bad compression) (4x1.5Gb)

Type	Linear read, MB/s	Compression
RAM	4300+	1.00
RAM + XPRS4	1170	0.94
RAM + XPRS8	1170	0.94
RAM + XPRS16	1170	0.94
RAM + LZX	1170	0.94

Latency

Type	Min latency, ms
RAM	0.021
RAM + XPRS4	0.042
RAM + XPRS8	0.054
RAM + XPRS16	0.077
RAM + LZX	0.211
SSD access	0.147
HDD access	12.600

My PC specs:
Intel Core i5-3570K, 4.4 GHz
DDR3 1800 MHz, 32 GB

[AFaustini]

XPRS16 continues to be the better compression method in this new batch of tests