Card benchmark - what's the optimal write buffer size?

[squig]

The Lexar 1000x and Toshiba 1066x are the top performers.

A1ex have you done the funky graph on the Toshiba benchmarks I submitted?

[quantumrider]

hey guys, so what is the fastest CF card for 5D2 shooting? I have no clue how to read all these numbers here...

[quantumrider]

ridiculous, I just asked the question the same moment you answered it... didn't see your post yet

[quantumrider]

just one quick note, the Lexar 1000x 16GB is much slower!

[KMA_WWC]

Don't forget Hoodman Steel is the top performer too!

[ptunstall]

The Lexar 1000x and Toshiba 1066x are the top performers.

A1ex have you done the funky graph on the Toshiba benchmarks I submitted?

Do you ever get dropped frames in 1080p with the Toshiba 1066 64gb card?

[squig]

@ 1080p sometimes the buffer fills after a couple of seconds on the first recording but after that it's rock solid. I'm shooting mostly 1920x840.

[ptunstall]

@ 1080p sometimes the buffer fills after a couple of seconds on the first recording but after that it's rock solid. I'm shooting mostly 1920x840.

Yea, I figure that initial recording frame drop happens regardless of card speeds. Good to know its solid thereafter though! Thanks for the info! I'll pick one up!

[squig]

There's been at least one build in the last week that didn't drop frames on the first recording, damned if I know which one.  Shooting is going really smoothly for me, I wish I could say the same about post. Adobe camera raw just wasn't designed for moving pictures.

[mvejerslev]

5D Mark II, Sandisk Extreme 60MB/s (will follow up with Sandisk Extreme Pro 90MB/s card in a minute)

Buffer size experiment
ML v2.3.NEXT.2013May24.5D2212, bff47c520333+ (unified) tip
Mode: movie 1920x1080 24p, Global Draw: OFF
10312704 446
15885312 510
80896 123
8949760 495
28661760 463
16264192 488
24521728 465
16424960 457
13148160 457
15619072 498
1658880 382
21001216 494
8548352 429
19760128 481
5965824 429
24906752 516
13968384 461
16770048 493
4542464 421
13063168 484
33263616 442
30312448 508
16713728 457
3887104 453
21841920 484
12846080 495
16192512 441
3333120 454
6719488 444
17313792 508
23374848 434
2865152 457
32488448 422
18588672 504
4186112 426
821248 365
3661824 413
18359296 505
19875840 420
22754304 498
13665280 450
24957952 504
28555264 452
33442816 491
10708992 433
20242432 478
8774656 479
11216896 494
23156736 447
12750848 488
12369920 436
2157568 408
15279104 455
29943808 501
30321664 442
19004416 469
268288 222
23778304 486
467968 276
1905664 431
32368640 452
19829760 493
26281984 436
24153088 480
11364352 450
32160768 502
5944320 439
24118272 458
28511232 505
31526912 460
15956992 454
1732608 413
13881344 451
6106112 471
17367040 462
21693440 484
30835712 431
1139712 373
13698048 454
19388416 493
1532928 365
16333824 474
15963136 432
23540736 505
23770112 474
7317504 474
6019072 446
13303808 475
32039936 452
13137920 496
24311808 478
25442304 493
8986624 437
25421824 490
2720768 439
18526208 490
12448768 509
6514688 465
6632448 472
17897472 502
26569728 457
22568960 508
11912192 469
22007808 503
30651392 489
29615104 489
18367488 477
21049344 471
15778816 502
24951808 498
5359616 466
23476224 489
17125376 478
3318784 444
12782592 499
30965760 499
1974272 423
32021504 485
24463360 484
28246016 480
23352320 495
27161600 504
25191424 443
10841088 480
22846464 458
19111936 468
26234880 461
32610304 487
14385152 459
18029568 469
11080704 456
3933184 436
22414336 466
8314880 488
27988992 459
20073472 490

[mvejerslev]

5D Mark II, Sandisk Extreme Pro 90MB/s

Buffer size experiment
ML v2.3.NEXT.2013May24.5D2212, bff47c520333+ (unified) tip
Mode: movie 1920x1080 24p, Global Draw: OFF
10312704 603
15885312 596
80896 133
8949760 613
28661760 618
16264192 586
24521728 618
16424960 617
13148160 612
15619072 480
1658880 500
21001216 618
8548352 594
19760128 510
5965824 594
24906752 627
13968384 610
16770048 611
4542464 574
13063168 612
33263616 629
30312448 605
16713728 605
3887104 566
21841920 624
12846080 504
16192512 501
3333120 552
6719488 594
17313792 607
23374848 620
2865152 544
32488448 629
18588672 610
4186112 542
821248 444
3661824 559
18359296 615
19875840 505
22754304 621
13665280 609
24957952 618
28555264 627
33442816 622
10708992 604
20242432 613
8774656 582
11216896 620
23156736 608
12750848 511
12369920 616
2157568 432
15279104 620
29943808 619
30321664 600
19004416 606
268288 270
23778304 617
467968 342
1905664 526
32368640 628
19829760 623
26281984 508
24153088 604
11364352 605
32160768 626
5944320 572
24118272 507
28511232 630
31526912 622
15956992 605
1732608 497
13881344 611
6106112 591
17367040 612
21693440 604
30835712 627
1139712 442
13698048 510
19388416 502
1532928 485
16333824 623
15963136 606
23540736 615
23770112 616
7317504 599
6019072 493
13303808 606
32039936 623
13137920 609
24311808 506
25442304 623
8986624 594
25421824 632
2720768 554
18526208 617
12448768 620
6514688 583
6632448 589
17897472 511
26569728 630
22568960 619
11912192 610
22007808 509
30651392 630
29615104 615
18367488 618
21049344 613
15778816 515
24951808 628
5359616 585
23476224 620
17125376 566
3318784 554
12782592 614
30965760 634
1974272 533
32021504 606
24463360 580
28246016 633
23352320 614
27161600 614
25191424 589
10841088 602
22846464 629
19111936 626
26234880 616
32610304 512
14385152 615
18029568 614
11080704 609
3933184 574
22414336 621
8314880 601
27988992 621
20073472 628
23487488 620
30233600 621
25299968 517
2406400 450
28059648 619
2145280 531
27503616 567
20705280 579
28731392 593
2458624 530
19662848 588
6198272 730
21655552 790
32225280 793
21788672 558
16669696 480
12007424 568
4321280 521
14878720 519
10118144 399

[xvince1]

Hey a1ex and g3gg0, there's something that maybe you're missing. All flash cards (SD/CF/SSD/etc) have a fixed erase block size.

If there's no partition and the card is just formatted with filesystem then it is ok. But for those who have a partition table on them, the partition must be aligned with EBS to have optimal performance (See articles on partition alignment for SSD)

All partitions of the cards I've tested were misaligned out-of-the-box. Besides Dane-elec and Sandisk I couldn't recall the other brands. I couldn't recall the numbers but I did see performance improvements.

Instead of trying to find EBS for all cards, I suggest aligning the partition at 1MB, so it should fit almost (if not all) EBS.

An utility to check alignment: http://www.tcpdump.com/kb/os/windows/disk-alignment/all-pages.html
(Or using MS System Information - msinfo32.exe) http://whirlpool.net.au/wiki/checking_ssd_alignment

Instructions for Windows XP here: http://support.microsoft.com/kb/929491
Linux users should set fdisk to show entry/units to sectors and create a partition with starting sector 2048.
Of course you can use any partition manager to move/resize the beginning of the partition.

Regards,

Rodrigo

Sorry for this noob Question, but I do not understand how you can create an align partition that can be usable in the camera, as every time the 5D read the new partition (with a good alignement), it detect that partition wasn't create in the DSLR, so I'm obliged to format the CF inside and the alignement goes out... I'v try many time, using the freeware "active partition manager" and I very don't understand what's wrong ...

I'm very disapointed as I have 2 Komputerbay x600 wich max up at 30Mb/s with bad alignement. Hoped to gain some biterate with alignement... 

[xvince1]

There's something I don't understand, I try my Komputerbay 600x (just formated in the 5Dmk2) on cristal diskmark on an CF reader USB3, and I got 61MB/s with 4000MB test files but 0.5 MB/s with 512k files. When I try benchmarking in ML menu, I maxed 25-30MB/s

My question : Is there difference or inconsistency between caching RAM / controller on the 5D hardware as I see many people with mk2 maxing at 60/70 MB/s with some KB, or is it only the CF ?

really Sorry for this noobesque questions 

[togg]

Is there no test with Komputerbay 32GB 1000X?

[ashtrai]

Is there no test with Komputerbay 32GB 1000X?

Interested in this as well.

[squig]

Komputerbay write speeds aren't consistent, it's a lottery.

[ashtrai]

I just ordered a Komputerbay 32GB 1000x card. Will update with benchmarks with my 50D once it arrives (should be within a week).

[seanturco]

5dmark II
Sandisk Extreme, 8go 60mb/s ( very short bench)

Buffer size experiment
ML v2.3.NEXT.2013May31.5D2212, d6786d675069+ (unified) tip
Mode: movie 1920x1080 24p, Global Draw: OFF
10312704 487
15885312 498
80896 123
8949760 491
28661760 478
16264192 506
24521728 456
16424960 501
13148160 476
15619072 493
1658880 416
21001216 503
8548352 451
19760128 502
5965824 460
24906752 494
13968384 502
16770048 496
4542464 451
13063168 470
33263616 492

[togg]

Komputerbay write speeds aren't consistent, it's a lottery.

( I hope to win 

I just ordered a Komputerbay 32GB 1000x card. Will update with benchmarks with my 50D once it arrives (should be within a week).

Yup!

[mihaii]

from a amazon comment on 128 GB Komputerbay card


 "The tech support person was great to talk to, here is a bit of inside info he gave me. They use micron/crucial chips on the 128GB card and a MSI (I think it's LSI) controller if I remember correctly. He said it the same as the Lexar 1000x cards. The Komputerbay smaller CF cards use samsung chips."

So maybe the 128GB is not a lottery and the price is between the Lexar 32 and Lexar 64.

What do you think?

I don't know if I should go for a Komputerbay 128GB or a Lexar 32GB 1000x

[odlan]

from a amazon comment on 128 GB Komputerbay card


 "The tech support person was great to talk to, here is a bit of inside info he gave me. They use micron/crucial chips on the 128GB card and a MSI (I think it's LSI) controller if I remember correctly. He said it the same as the Lexar 1000x cards. The Komputerbay smaller CF cards use samsung chips."

So maybe the 128GB is not a lottery and the price is between the Lexar 32 and Lexar 64.

What do you think?

I don't know if I should go for a Komputerbay 128GB or a Lexar 32GB 1000x

Hey dude, it is better for you forget about Komputerbay 128Gb
I have bought a new one from Amazon and this is the score



YOU CAN'T record RAW 1920x1080 16:9 25 Fps

I'll go with refund and I'll use my Lexar 32Gb (record flawlessly raw 19820x1080) until CF-to-SSD will be ready

[Jonit]

I made a quick benchmark of a SanDisk Extreme 45MB/s. After formatting it to exFAT file system, it gives stable 20+MB write speeds.

https://www.youtube.com/watch?v=HI4XZPAMoZk

[driftwood]

Code: [Select]

Buffer size experiment
ML v2.3.NEXT.2013Jun20.5D3113, 1b42291d5085+ (unified) tip
Mode: movie 1920x1080 25p, Global Draw: ON
CF LEXAR ATA FLASH CARD LEXAR ATA FLASH CARD                   
10312704 852
15885312 812
80896 238
8949760 900
28661760 821
16264192 872
24521728 858
16424960 819
13148160 831
15619072 799
1658880 694
21001216 924
8548352 749
19760128 881
5965824 815
24906752 893
13968384 757
16770048 818
4542464 801
13063168 860
33263616 829
30312448 866
16713728 808
3887104 793
21841920 876
12846080 785
16192512 870
3333120 726
6719488 876
17313792 871
23374848 865
2865152 744
32488448 896
18588672 847
4186112 736
821248 672
3661824 845
18359296 814
19875840 922
22754304 814
13665280 775
24957952 863
28555264 826
33442816 824
10708992 833
20242432 794
8774656 853
11216896 809
23156736 873
12750848 840
12369920 824
2157568 805
15279104 879
29943808 872
30321664 857
19004416 803
268288 408
23778304 926
467968 496
1905664 740
32368640 843
19829760 761
 

[grooveminister]

@ all the people using Transcend cards operating near 83 MB/s:
Have you already tried the "AutoFormat" tool that Transcend provides to low-level format the cards?
http://de.transcend-info.com/support/DlCenter/dllogin_t.asp?Link=|support|dlcenter|software|AutoFormat.zip
I´ve read one report of a ML user on amazon.de that autoformat did solve his problems and he was able to record 1080p afterwards.

Best wishes,
Andreas

[Danne]

I never had to use any special format-tool. I worked straight away