APFS vs HFS+: benchmarks on 2017 MacBook Pro with macOS High Sierra
<– Page 1: dd, ioping, sysbench
blogbench
Blogbench is a filesystem stress tool that imitates a hosting environment. While it sounds like it may not be an ideal benchmarking utility for a laptop or workstation, Blogbench creates multiple write / read / rewrite operations, uses 8K chunks of data and was used primarily because of multi-threading.
HFS+
$ blogbench -d /tmp
Frequency = 10 secs
Scratch dir = [/tmp]
Spawning 3 writers...
Spawning 1 rewriters...
Spawning 5 commenters...
Spawning 100 readers...
Benchmarking for 30 iterations.
The test will run during 5 minutes.
Nb blogs R articles W articles R pictures W pictures R comments W comments
173 1151009 9724 714923 8232 591264 24967
295 1248120 6876 820262 7104 653722 17051
423 1165883 6565 780699 6566 602302 17566
549 1217983 6678 819287 6930 627100 17641
674 1231432 6814 824949 6106 643147 17867
789 1238335 6328 825122 6593 633810 17028
918 1142464 6889 758554 5575 586111 15823
1049 1218096 6452 811547 6313 631171 16172
1178 1236903 6433 815218 7004 632066 16580
1302 1233998 6673 818920 6080 640355 16687
1424 1213624 6134 806283 5911 630331 17029
1555 1198290 6217 795245 6848 608787 16082
1679 1214499 6622 814224 6504 634264 17132
1804 1213402 6118 818276 6604 625071 16453
1925 1145822 6224 768443 6179 596363 16692
2043 1011932 6256 681732 6054 524392 15582
2149 903087 5426 608006 5470 471889 14687
2224 699150 4501 473189 4431 364103 10130
2313 698974 5453 480588 4956 362468 9693
2391 703301 4836 476964 3826 361573 11082
2459 634829 4385 429054 3872 327426 9732
2521 617930 4079 419608 3547 326351 11306
2606 601020 4897 407524 4231 312970 9817
2691 547121 4999 374539 4648 286723 10793
2755 537186 4393 366589 3528 278197 9229
2826 522027 4095 352830 3984 276855 10006
2889 515149 4171 348546 3526 269746 8865
2952 499819 4375 342504 3692 262521 8369
3039 479147 4854 325683 4750 245609 9198
3109 460991 4309 312898 3977 243007 8359
Final score for writes: 3109
Final score for reads : 549437
HFS+ encrypted
$ blogbench -d /tmp
Frequency = 10 secs
Scratch dir = [/tmp]
Spawning 3 writers...
Spawning 1 rewriters...
Spawning 5 commenters...
Spawning 100 readers...
Benchmarking for 30 iterations.
The test will run during 5 minutes.
Nb blogs R articles W articles R pictures W pictures R comments W comments
159 1047319 8503 785307 8823 562227 21606
288 1136847 7540 852135 6789 627755 20764
409 1142448 7228 834397 7053 643790 18797
514 1154866 7182 838339 6003 664550 17390
651 1128184 8180 808199 7666 620338 18508
762 1108561 7189 789626 6235 599676 16711
875 1126680 7495 799168 6253 618659 16988
971 1186674 5945 840671 5337 647478 14431
1069 1168613 6392 828585 6050 635756 15852
1165 1182315 6370 839453 5218 629151 16283
1256 1135763 6520 802254 5377 621340 16424
1374 1122660 6592 786225 6689 607902 17142
1496 1154108 7554 809104 6411 630031 17020
1615 1136312 7191 800905 7023 613196 17841
1742 1151949 7046 813946 6327 613785 16960
1865 1049526 6584 735034 5776 563548 16341
1975 941379 6274 654847 5422 490101 15014
2075 819056 5604 564476 5038 426068 13126
2137 701153 3484 483396 3515 365671 7565
2212 686552 4553 477357 4228 350334 9791
2291 597985 5008 413336 4683 309456 10302
2361 559614 4263 386959 3745 281652 8128
2436 548433 4863 380552 4450 281132 9163
2518 524520 4388 362262 4043 261152 10281
2602 483096 4857 332363 4523 248484 9148
2654 490778 3647 339266 3155 253620 7711
2721 469509 4633 323205 3686 237544 8368
2778 443888 3317 308761 3594 227099 8384
2836 430795 3909 296151 3183 217926 8773
2900 428126 3777 298828 3891 213226 8244
Final score for writes: 2900
Final score for reads : 529108
APFS
$ blogbench -d /tmp
Frequency = 10 secs
Scratch dir = [/tmp]
Spawning 3 writers...
Spawning 1 rewriters...
Spawning 5 commenters...
Spawning 100 readers...
Benchmarking for 30 iterations.
The test will run during 5 minutes.
Nb blogs R articles W articles R pictures W pictures R comments W comments
102 1047160 5502 753670 5561 538026 14129
159 1135784 3373 800554 3489 621399 9167
220 1155966 3630 821857 3506 627210 9222
283 1134478 3793 822116 3948 629208 9539
350 1141677 4197 809858 3640 603920 8624
421 1148273 4258 817907 4437 594403 10256
492 1129645 3940 800718 4025 592163 11259
565 1079623 3852 768173 3878 581369 11097
642 1107088 4356 788464 3835 607439 11147
719 1035475 4265 724601 4251 547627 10377
797 1046436 3782 739908 4038 565335 10028
878 1109830 4370 774863 4257 588692 10859
955 1098560 3975 775496 4208 600607 11249
1031 1076059 4174 761763 4037 569403 10482
1111 1124256 4690 784115 3899 569054 10825
1186 1118244 3970 780563 4064 580944 10537
1257 1110296 3875 776137 3776 565402 10295
1335 1070674 3806 742652 4127 560730 10386
1400 1044126 3603 728794 3830 531727 10536
1467 1104249 3747 773696 3687 586241 10196
1545 1118511 3963 781478 3940 581093 10103
1618 1109166 3695 778530 3984 576598 9981
1692 1077301 3597 756706 4188 554409 10332
1758 1083163 3567 760868 3893 564309 9836
1813 1097803 4039 774478 3470 584914 8526
1857 1054846 3283 736910 2572 555452 7479
1921 975166 3862 687565 3320 509402 6764
1973 895691 3528 633044 2968 470482 8184
2026 775250 3746 545057 3061 414864 9067
2085 734539 3925 518748 3319 390159 8332
Final score for writes: 2085
Final score for reads : 652794
APFS encrypted
$ blogbench -d /tmp
Frequency = 10 secs
Scratch dir = [/tmp]
Spawning 3 writers...
Spawning 1 rewriters...
Spawning 5 commenters...
Spawning 100 readers...
Benchmarking for 30 iterations.
The test will run during 5 minutes.
Nb blogs R articles W articles R pictures W pictures R comments W comments
94 1045218 5710 745903 5084 479978 12955
151 1137694 3745 797434 3796 548218 8757
219 1147899 4158 813481 3360 577646 10298
283 1173336 3980 798788 3448 604826 9641
339 1171412 3436 810892 3788 604117 9378
402 1155112 3986 811343 3266 591223 9593
461 1139333 3929 799848 3392 588281 9770
517 1147593 4287 793375 3303 596903 9621
592 1136911 4485 771825 3459 593174 10187
671 1118111 3870 759065 3936 579052 10878
742 1125614 4106 767965 4026 595456 10544
808 1117016 3611 765990 3545 590469 10809
879 1042473 3803 717562 3515 550728 10203
947 1100384 3989 751434 3447 571741 10531
1023 1120981 3854 763309 3578 569113 10683
1092 1111106 3829 758346 3802 584764 10209
1167 1121487 4234 766284 3374 573690 10278
1234 1126331 3654 760581 3796 578727 10309
1300 1109761 3596 745995 3345 567933 10492
1378 1105059 3913 743056 3597 560403 9865
1432 1081010 3798 726357 3119 546768 7988
1486 1100699 3996 742941 3093 565180 9482
1546 1094152 3772 738002 3717 565265 8525
1595 1125944 3560 764444 3360 585776 8295
1653 1117273 4189 762856 2957 582396 9131
1719 1107814 4499 749517 3531 566665 9195
1779 1112700 3851 748007 3364 572693 8394
1837 1050906 3601 713795 3403 540644 8052
1890 1002112 3155 677631 3038 520734 7153
1949 845934 3202 573127 3176 444848 8378
Final score for writes: 1949
Final score for reads : 635140
First – overall scores of blogs written and read for both filesystems
And now the fun part – Blogbench creates timely statistics, which helps to visualise the transfer over time:
Articles written
Articles read
Pictures written
Pictures read
Comments written
Comments read
Bonnie++
HFS+
$ bonnie++ -d ~/tmp/ -m hfs
Version 1.97 ------Sequential Output------ --Sequential Input- --Random-
Concurrency 1 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
fs:hfs 32G 547 98 1290032 86 889147 69 1549 99 2275523 90 +++++ +++
Latency 33249us 20326us 19939us 13613us 4438us 2135us
Version 1.97 ------Sequential Create------ --------Random Create--------
fs:hfs -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
files /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP
16 22798 88 +++++ +++ 26146 91 25061 94 +++++ +++ 22166 87
Latency 4638us 1019us 1669us 1126us 123us 11011us
HFS+ encrypted
$ bonnie++ -d ~/tmp/ -m hfs
Version 1.97 ------Sequential Output------ --Sequential Input- --Random-
Concurrency 1 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
fs:hfs 32G 752 98 1238481 77 885117 59 1949 99 2111772 83 +++++ +++
Latency 22162us 10980us 18607us 10515us 5660us 1766us
Version 1.97 ------Sequential Create------ --------Random Create--------
fs:hfs -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
files /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP
16 22771 90 +++++ +++ 26532 90 24540 92 +++++ +++ 21315 84
Latency 1964us 81us 2035us 539us 47us 6507us
APFS
$ bonnie++ -d ~/tmp/ -m apfs
Version 1.97 ------Sequential Output------ --Sequential Input- --Random-
Concurrency 1 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
fs:apfs 32G 184 99 888304 95 886564 84 1090 99 2014824 87 +++++ +++
Latency 56710us 156ms 13898us 8551us 17769us 1978us
Version 1.97 ------Sequential Create------ --------Random Create--------
fs:apfs -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
files /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP
16 17052 96 +++++ +++ 11066 84 17192 97 +++++ +++ 11912 92
Latency 355us 163us 1590us 252us 51us 10109us
APFS encrypted
$ bonnie++ -d ~/tmp/ -m apfs
Version 1.97 ------Sequential Output------ --Sequential Input- --Random-
Concurrency 1 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
fs:apfs 32G 130 99 842978 82 769741 67 1424 99 870519 45 7153 296
Latency 81616us 417ms 22196us 20405us 96709us 102ms
Version 1.97 ------Sequential Create------ --------Random Create--------
fs:apfs -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
files /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP
16 17962 94 +++++ +++ 14766 96 19228 94 +++++ +++ 15020 97
Latency 7433us 200us 1199us 190us 73us 517us
Blackmagic Disk Speed Test
HFS+ encrypted
APFS encrypted
APFS vs HFS+?
There isn’t really a ‘versus’ here – it’s hard to fully compare two filesystems that were created 2 decades apart. While modern filesystems may have some disadvantages, they offer a whole bunch of new features and improvements that a normal user is even unaware of.
Older filesystems are fast – all of the widely used ones can saturate SATA hardware with no issues. But compared to modern ones, they offer next to no traits other than “you can store objects here”. Over time, additional layers of software improvements were added to some filesystems, but also new ones were created from scratch, with a variety of new and old, but now built-in capabilities. And while one can think that the later should be faster – it sometimes isn’t because of how things work. In short: faster doesn’t necessarily mean better and all of the above charts only show how fast (or slow) things are. If I were to compare NTFS to ReFS or EXT4 to ZFS, some new bottlenecks would also arise in specific workflows and specific hardware configurations.
Apple didn’t officially said much about APFS and most websites are just quoting Apple’s developer documentation: “Apple File System (…) features strong encryption, copy-on-write metadata, space sharing, cloning for files and directories, snapshots, fast directory sizing, atomic safe-save primitives, and improved file system fundamentals”.
I could speculate on why some tests shown a decreased performance but my knowledge on the subject is thin and that would just be some maybes and probablys (for example: I have no idea why encrypted HFS+ achieved better results than plain HFS+ in some tests but I repeated these tests with a comparable outcome) so instead, here are some observations:
- in some cases APFS appears to be slightly slower than HFS+ when it comes to writing data, especially in smaller chunks
- APFS appears to be a lot faster than HFS+ when it comes to reading data
- AFPS’ built-in encryption shows decreased throughput speeds in comparison to HFS+ with FileVault2
- seek rates seem a tiny bit higher with APFS but that would be unnoticable on most workflows
- apart from speeds, APFS introduces a lot of under-the-hood improvements
And again, all these benchmarks were performed on a really fast NVMe SSD disk and even on this disk, the differences aren’t big. On SATA SSD there should’t be any visible problems with the filesystem upgrade.
APFS documentation
To learn about APFS advantages, check Apple File System Guide on developer.apple.com. Here’s a short features comparison:
Thanks for reading!
< Page 1: dd, ioping, sysbench
Related:
How to upgrade macOS to High Sierra without filesystem change (HFS+ to APFS)
APFS and HFS+ benchmarks on 2017 MacBook Pro with macOS High Sierra beta (July 2017)
13 thoughts on “APFS vs HFS+: benchmarks on 2017 MacBook Pro with macOS High Sierra”
Thanks for writing this article! Been wondering how the file systems actually perform against each other. My first impression is that my Mac seems a little snappier, maybe because of the improved read speeds.
Impressive post! Thanks!
Thanks for doing the legwork on this, it saves a lot of people a lot of time and some wouldn’t even know where to start on doing benchmarks like these. if nothing else, for me personally, this actually reinforces my interest in the new system. I’ve been researching which way to go while I’m getting all set up to install on High Sierra on Skylake with NvMe boot and a couple Samsung PM851’s in raid. I know how silly I am getting with wanting my drives to get every bit of mbps squeezed out of them, like I would really even notice the difference between 700 and 800 reads. But, that’s what makes this a hobby for some and work for others I guess. anyway………………….. sorry for the rant………. THANKS!!!!!
I keep my 500GB music library on an external 1T spinning disk.
iTunes has always been slow — when it became possible with the betas to format at disk with APFS, I formatted my music disk with APFS.
iTunes became even more sluggish (unusably slow, much more than twice as slow), so I reverted to HFS+.
. . .
When High Sierra came out, I installed it on my late-2013 MB Pro (16G ram, 250G ssd). I kept it on through 10.13.1.
After a couple of weeks it became obvious that my computer was sluggish (except for duplicating files).
It was evident everywhere, but it was especially obvious when emuxing video files (almost twice as long).
Remembering my experience with iTunes, I decided to reformat my HD as HFS+ and see if it made a difference.
The computer did become much more responsive, so for now I am keeping my Macbook on HFS+.
Thanks!
My MacBook Air (Early 2015) dropped write speeds to about 150MB/s but was very quick before installing High Sierra.
Is the low powered CPU the bottleneck coding/compressing files to the SSD.
Nice work but no one will be noticing the differences really, they just are too small either way. No one uses a setup like the tester did in a real world working environment. Other loaded apps and various background processes running take their toll on system performance.
An issue I ran into with APFS encrypted is file corruption. Running MacBook Pro (late 2015). Even after 4 formats, and re-install (one done by tech at the Apple store), I always ended up with errors. Could not be repaired by tools, even in boot mode. Switched back to HFS+ and all is running smoothly. When at the Apple store they ran a diagnostics on my 1TB SSD and it passed with flying colors.
Awesome. Thanks! On the other hand, a conclusion section would be handy. Have you tested the latest macOS versions?
Note: this web form reports:
Error: Time limit is exhausted. Please enter a CAPTCHA value again. Click the BACK button on your browser, and try again.
Just wanted to say thanks and more importantly pointing out that while speed may be important but it’s not the complete picture. All of the benefits of a modern filing system, while may be milliseconds slower, you are getting benefits; A,B,C,D,E,F,G,etc… etc… I highly doubt that any “AVERAGE” human could be able to tell the speed difference but would see it if a drive crashes because it didn’t have X-feature.
I think this isn’t comparing Apples to Apples as it were, when it comes to disk encryption — the encryption scheme in APFS is to my knowledge a lot stronger by default. I couldn’t find you saying what cipher suites you used for each.
Upgraded early 2013 15″ MacBook Pro Retina 16GB/500GB to Mojave APFS encrypted, and Apple Mail (particularly smart mailboxes with over 100K messages) became dog slow — almost unusable. Then did a clean install same configuration — no difference. Then cloned drive to convert drive to HFS+ (no encryption), and same configuration became snappy again. Conclusion: APFS encrypted on this configuration is *not* a good choice (haven’t yet tried APFS unencrypted)
Mail has some issues on its own. I have recently had to switch to Thunderbird, literally mid-work.