RAID0 Benchmarks (e.g. Samsung F1 HD/HE103UJ)

Discussion in 'Mac Pro' started by Mac Husky, Jul 15, 2009.

  1. Mac Husky macrumors regular

    Joined:
    Mar 28, 2009
    Location:
    Bavaria, Germany
    #1
    As Tesselator did (see this thread) I am thinking about going RAID0 with three of Samsungs EcoGreen F2 (154UI).

    I allready ordered some of them for using them in the new QNAP NAS TS-439 Pro.
    I think they should work fine with it after reading a lot of reviews concerning the Samsung F2 1500MB.

    Now thinking about going for a RAID0 with 3 of them in my Mac Pro Nehalem Quad 2,66 6GB RAM that just arrived yesterday.

    Not shure, whether the performance might be noticeable better with a RAID0 using three Samsung Spinpoint F1 1000MB (HD103UJ)
    or its RAID Class Version (HE103UJ), due to their single performance is better than that of the F2 EcoGreen (see comparison here including WD Caviar Black).

    Any suggestions?
    Does anyone use the Samsungs in a RAID0 and may have some benchmarks?

    Benchmarks of a RAID0 of 3 WD Caviar Black are welcome, also.
     
  2. Tesselator macrumors 601

    Tesselator

    Joined:
    Jan 9, 2008
    Location:
    Japan
    #2
    Yeah, I looked at those too. UI vs. UJ... and the only important difference was the "Data Transfer Rate / Media to/from Buffer (Max.)": 175 MB/sec for the UJ vs. 166 MB/sec for the UI. - besides of course the lower power consumption. I looked into the UJ prices but at the time as is still the case, no one in all of Japan had any. No reader reviews, no shop sales, no nothing - it's almost like it doesn't actually exist for the children in the land of the rising sun. :D

    The UJ is only 1TB. Both are 3-Platter (perpendicular recording) drives - which means that only the UI is 500GB/platter. Both are RAID Class spec drives except for the spindle speed of the UI (if that's a RAID Class issue).

    That's all I got. ;)

    Mine are still working and the RAID now at ~ 25% full, still benches at exactly the same speeds. :)

    EDIT: There was someone here who posted a 3-Drive WD Black RAID0 spec. Have a search. It was faster than mine by a fair bit too!
     
  3. Mac Husky thread starter macrumors regular

    Joined:
    Mar 28, 2009
    Location:
    Bavaria, Germany
    #3
    Thanks Tess!

    Prices right now here in Germany

    Samsung EcoGreen F2 1500MB (154UI): € 98,-
    Samsung Spinpoint F1 1000MB (HD103UJ): € 67,-
    Samsung Spinpoint F1 1000MB Raid Class (HE103UJ): € 108,-

    Western Digital Caviar Black 1000MB (WD1001FALS): € 80,-
    Western Digital RE3 1000MB (WD1002FYBS): € 118,-
    Western Digital RE4 2000MB (WD2002FYPS): € 243,- :eek:

    Having good experiences with Samsung I would tend to go
    for the 154UI or the HD103UJ. The price for the RAID class
    version is kind of high. And as I am no hardcore user of the
    RAID system (24/7) I would not go for it. Or would you?

    Did not found the benachmark of the caviar black so far.
    But I am still surching for it. Samsungs F1 should be a little
    bit better than?!

    You are talking about the values presented on the Samsung homepage?!
    The comparison I linked abough (klick) shows some more speed differences, doesn´t it?!
     
  4. Tesselator macrumors 601

    Tesselator

    Joined:
    Jan 9, 2008
    Location:
    Japan
    #4
    Well I didn't so I guess I wouldn't. :D I wonder what the differences are between the HE103UJ and the HD103UJ though? Samsung posts the exact same specs it looks like. No? Did I miss something? There's one extra spec for the HE: MTBF 1,200,000 POH but that's the same as for both HDxxx drives in your list too (when you can find that spec). I'm confused. :)


    I think it was in this thread: http://forums.macrumors.com/showthread.php?t=726879 Pretty sure anyway. :)


    Yeah, I was just talking about the samsung pages. I see Tommy added the HD154UI to his list since I looked last. It doesn't look so hot by those marks, huh? :D But I knew that going in. My plan succeeded. I wanted to use a rugged (RAID spec), huge (1.5TB ~ 2TB), cheap (<$100) drive and RAID0 them to make up for the differences. It worked. :) It's certainly NOT the fastest RAID set around but they ended up doing better than I expected too. If you're going for the high performance thing you probably don't want the F2 green drives. I keep running out of space all the time so size was important to me. I leave my system on 24/7 so KW/h profiles and UBE was of a concern, and I'm a cheap bastard so much over $100 and they can kiss my grits! :D The F2's just kinda fell right in place to meet all those goals. Had I been going for the performance thing I would have been willing to spend more, make sure they were RAID class and probably bought myself a caching controller with user installable DDR2/3 RAM (maybe even yummy SAS ;)). I also would have considered a higher spindle speed even though it might mean I'd have to hang my Mac From the ceiling or mount a shelf on the wall for it.
     
  5. Mac Husky thread starter macrumors regular

    Joined:
    Mar 28, 2009
    Location:
    Bavaria, Germany
    #5
    What I found about the differences between HD and HE is, that the HE shell have a sensor measuring the vibration and balances it by influencing the control of the drive (however?!). The reliability of the HE has been doubled looking at the hours of usage. Tests say that the HE is a little bit louder and a little bit slower than the HD version (german article).

    So I am not shure, whether it is really a good idea to spend that money into the HE instead of the HD one?!

    Found your summing up here.
    Found results for 8 of them ;)

    I am no speed junky and all the other goals you mentioned are right and I agree with your decision. And I was also like going for 3x 154UI until I found the data of the HD/HE103UJ. Following the recommendations here in MR a RAID0 demands RAID class drives for reliability. The more money may be worth the more reliability they promise.

    So finally I would have to decide between these two:

    - Samsung EcoGreen F2 1500MB (154UI): € 98,-
    - Samsung Spinpoint F1 1000MB Raid Class (HE103UJ): € 108,- (really worth +€ 41,- instead of taking the HD version???)

    If I knew, that the performance jump between the F1 and the F2 would be no more than 10% or some more in RAID0, I would tend go for the F2 for sure due to noise reduction and lower power consumption. And some more space at least.

    The performance of the single drives let me expect a bigger jump between F1 and F2 in a RAID0 system?! What do you think?
    Would be glad someone might be running a F1 RAID0 and could send some benches.
     
  6. Tesselator macrumors 601

    Tesselator

    Joined:
    Jan 9, 2008
    Location:
    Japan
    #6
    According to Samgung's page they both have that.

    HD154UI Features
    1. • MAX.500 GB Formatted Capacity Per Disk
    2. • Environment friendly product with RoHS compliance
    3. • Serial ATA 3.0 Gbps Interface Support
    4. • Improved performance with dual-ARM based firmware
    5. • Improved recording stability over temperature with PMR
    6. • ATA S.M.A.R.T. Compliant
    7. • Advanced dynamic FOD control for best data integrity
    8. • ATA Automatic Acoustic Management Feature
    9. • Intelligent compensation of external disturbance
    10. • ATA 48-bit Address Feature
    11. • SATA Native Command Queuing Feature
    12. • ATA Device Configuration Overlay Feature
    13. • Device Initiated SATA Power Management
    14. • NoiseGuard™
    15. Rotational vibration sensor
    16. • SilentSeek™


    HE103UJ Features
    1. • Formatted capacity : 1 TB
    2. • Serial ATA 3.0Gbps Interface Support
    3. • Improved recording stability over temperature with PMR
    4. • Advanced dynamic FOD control for best data integrity
    5. • Intelligent compensation of external disturbance
    6. • SATA Native Command Queuing Feature
    7. • Device Initiated SATA Power Management
    8. Staggerd Spin-up Support
    9. Rotational vibration sensor
    10. • Environment friendly product with RoHS compliance
    11. • Improved performance with dual-ARM based firmware
    12. • ATA S.M.A.R.T. Compliant
    13. • ATA Automatic Acoustic Management Feature
    14. • ATA 48-bit Address Feature
    15. • ATA Device Configuration Overlay Feature
    16. • NoiseGuard™
    17. • SilentSeek™

    <shrug> :confused:

    Also I read a review somewhere that said the HD154UI was also 1,200,000 MTBF. You read one that said it's not? Or just one that said the HE103UJ "doubled typical ratings"? Maybe ONLY Samsung knows... LOL



    :D


    I dunno what to say. Draw straws?

    Here some relevant XB specs:


    Code:
    *** One WD 1TB Black  ***
    Results	98.60	
    	System Info		
    		Xbench Version		1.3
    		System Version		10.5.7 (9J61)
    		Physical RAM		6144 MB
    		Model		MacPro4,1
    		Drive Type		WDC WD1001FALS-41K1B0
    	Disk Test	98.60	
    		Sequential	168.08	
    			Uncached Write	175.09	107.50 MB/sec [4K blocks]
    			Uncached Write	172.29	97.48 MB/sec [256K blocks]
    			Uncached Read	132.69	38.83 MB/sec [4K blocks]
    			Uncached Read	210.66	105.88 MB/sec [256K blocks]
    		Random	69.77	
    			Uncached Write	24.85	2.63 MB/sec [4K blocks]
    			Uncached Write	318.09	101.83 MB/sec [256K blocks]
    			Uncached Read	115.63	0.82 MB/sec [4K blocks]
    			Uncached Read	188.95	35.06 MB/sec [256K blocks]
    
    Code:
    *** Two WD 1TB Black setup as RAID 0 ***
    
    Results		168.19	
    System Info	
    Xbench Version	 1.3
    System Version	 10.5.7 (9J61)
    Physical RAM	 16384 MB
    Model	 	MacPro3,1
    Drive Type	 Me
    Disk Test	168.19	
    Sequential	206.01	
    Uncached Write	320.12	196.55 MB/sec [4K blocks]
    Uncached Write	309.90	175.34 MB/sec [256K blocks]
    Uncached Read	95.45	27.93 MB/sec [4K blocks]
    Uncached Read	386.20	194.10 MB/sec [256K blocks]
    Random		142.09	
    Uncached Write	58.82	6.23 MB/sec [4K blocks]
    Uncached Write	769.30	246.28 MB/sec [256K blocks]
    Uncached Read	176.15	1.25 MB/sec [4K blocks]
    Uncached Read	239.75	44.49 MB/sec [256K blocks]
    

    Hmmm, can't seem to find the 3-Drive one. I even used http://mroogle.*************/ MR's search thingy. ;)
     
  7. Mac Husky thread starter macrumors regular

    Joined:
    Mar 28, 2009
    Location:
    Bavaria, Germany
    #7
    I was talking about the differences between HD and HE - not UI.
    Just quoting the linked article, that said: HE 1,200,000 MTBF and HD 600,000 MTBF. Don´t know about the UI but heared about 1,200,000 also. Maybe from one of your postings :confused::D

    Having a look at the/some disk test data:

    1 TB WD Caviar Green: 75,72 (but was an older EACS)
    1 TB WD Caviar Black: 98,60
    2 TB WD Caviar Black in RAID0: 168,19
    3 TB Samsung 154UI: 176,78

    Doesn´t that mean, that a RAID0 of 2 Caviar Black is nearly as
    fast as a RAID0 of 3 Samsung 154UI?! And does this suggest,
    that a RAID0 of 3 Caviar Black (resp Samsung F1) would be
    in the disk test range of 250-260 and sowith kind of more
    than 40% faster?
     
  8. Tesselator macrumors 601

    Tesselator

    Joined:
    Jan 9, 2008
    Location:
    Japan
    #8
    Hmm, got a link to that article?


    Yup! That's what it implies. That's what I was trying to say about this green RAID. It's much faster than I expected but it's not high performance by comparison. Blacks are faster for sure - no doubt about that! If there were 1.5TB or 2TB blacks I would have considered them for sure.
     
  9. Mac Husky thread starter macrumors regular

    Joined:
    Mar 28, 2009
    Location:
    Bavaria, Germany
    #9
    Sure - linked it abough allready. Here you are!

    I have read it again and they point it out as one of the advantages of the HE103UJ that makes it labeled as a RAID class model for 24/7 use.
    But you are right: they didn´t say, that die HD103UJ has a smaller MTBF.

    I did not thought it would be that much difference between them in a RAID0?!
    40% or more is a bigger jump than I expected...

    Question to test the disks in a RAID0 as you did:

    What software did you use? Where to download?
    Has the software to be installed within the RAID0?
    Or maybe on a fourth disk with OSX?
    Or has OSX to be installed on the RAID disks?

    I ask that, because I could test 3 of the Samsung 154UI on my system
    and than 3 of the Samsung F1 disks - if the expenditure isn´t to much?!
    I would do that before setting up the system and everything else.

    Could return the F1 disks if I don´t wont them after that.
    Have the 154UI ordered allready and they will definetily stay for my NAS.

    Would you go for the RAID class model of the 103UJ in the Mac Pro???
     
  10. Tesselator macrumors 601

    Tesselator

    Joined:
    Jan 9, 2008
    Location:
    Japan
    #10
    I think your cut&paste is broken. That's twice that you think you already posted something that you haven't. :D No biggy. I have it now tho. :) Here it is in semi-English for those not knowing German:
    http://66.249.89.132/translate_c?hl...le.com&usg=ALkJrhhVmY8bjvqvKTJdi2K9gsgm_pv0hg


    Right. I think the HD103UJ and the HE103UJ are identical except maybe for the "Staggerd Spin-up Support" in the HE model.


    Right. It's like every other test for hardware. While 40% may be attainable in some instances that won't be the average performance difference. So it's more accurate to say that "it's up to 40% faster". The average difference between three HD154UI (green) in RAID0 and three HE103UJ in RAID0 will probably be more like 2% to 10% as is typical. Still, when that 40% increase happens, it's pretty nice. ;)


      1. SpeedTools - http://www.speedtools.com/
      2. IO Guage - http://www.SANtools.com -- http://mac.softpedia.com/get/Network-Admin/IO-Gauge.shtml
      3. AJA System Test - http://www.aja.com/ajashare/AJA_System_Test_v601.zip
      4. XBench - http://www.xbench.com/ <-- Lame
      5. Disk Speed Bench X - http://christian-hornung.de/binary/DiskSpeedBenchX-0.1.zip <--- (also kinda lame)
    1. Either way.
    2. Either way.
    3. Try both.


    I think they are identical and in a desktop or workstation class system staggered spin-up isn't really useful, so whichever is the cheapest. :)
     
  11. Mac Husky thread starter macrumors regular

    Joined:
    Mar 28, 2009
    Location:
    Bavaria, Germany
    #11
    No, it isn´t :D See posting #5 line 2: (german article) ;)
    Thx - I guess I will try soon.
    Does the system RAM influence these tests in any way?
    That would leed me to go for the HD103UJ :)
    That - on the other hand - would be the best argument for the 154UI at all :rolleyes:
     
  12. Tesselator macrumors 601

    Tesselator

    Joined:
    Jan 9, 2008
    Location:
    Japan
    #12
    Damn - there it is. :D


    Yes, of course. So does CPU Speed, and bus speed. How critical that is hasn't been measured by me personally but I would imagine not more than 2% to 8% overall between say the 2009 and the 2006 2.66 octads - and it'll vary from test to test and from software suit to software suit.


    Hehehe, If it's any help, in your specific shoes I would get the WB 1TB Blacks for inside the Mac and the F2 1.5TB Green drives for the NAS. I'd have to check the data on the blacks again but I'm pretty sure of that - yes.

    Remember also that even 500,000 hrs. MTBF is an awfully long time!

    500,000 ÷ 24 = 20,833 Days
    20,833 ÷ 350 = 59.5 Years.

    And that's if you run them 24/7 :D

    Even an unrecoverable bit error these days usually just means a sector or block has to be relocated - which is usually transparent to the user except for a tiny speed hit right when it actually happens.
     
  13. Mac Husky thread starter macrumors regular

    Joined:
    Mar 28, 2009
    Location:
    Bavaria, Germany
    #13
    I know it has been recommended here a lot of times.
    But the data of the newer 103UJ some tests are pro Samsung F1 vs WD black.
    See the link I set in my introducing post: klick

    Or are the Database and Workstation Benchmark Pattern (average I/O operations per second for all queue depth) Score in I/O most important?
    That would - indeed - lead to the "good old" Caviar Black for usage in the Mac Pro - even in RAID0 (RE3 to think about again).

    But in the end they may/will both not be presenting noticable differences at all in daily work :D
     
  14. seisend macrumors 6502a

    seisend

    Joined:
    Feb 20, 2009
    Location:
    Switzerland, ZG
    #14
    Well, I saw in some benches that the caviar black is the loudest 1TB disc. So that's the only reason against a Caviar Black, so I think. The RE3 would be the BEST decision for a Mac Pro RAID0 and isn't loud, but pricy !
     
  15. Mac Husky thread starter macrumors regular

    Joined:
    Mar 28, 2009
    Location:
    Bavaria, Germany
    #15
    Right. The test says:
    Samsungs F1 idle: dBA 42,42
    WD Blacks idle: dBA 46,67
    Plus 40% in price. But same noise. Even louder in rest. Found on WD´s page.
     
  16. Tesselator macrumors 601

    Tesselator

    Joined:
    Jan 9, 2008
    Location:
    Japan
    #16
    Yeah, I dunno. I think at this point the spec differences are anal at best and meaningless at worst.

    I mean if you look at tom's results the HE is 8MB/s or something, faster at one and then in the very next test testing virtually the same thing, the other one is 8MB/s faster. o0 Kinda goofy. And a db test? Hmm, K. Trust me you're not going to hear either dive over the you're own breath except maybe when the machine powers up. And then from what I understand the Blacks go WEeeshhh... And the Samsungs go Click, tock, tick Weeeeefffuuu Click, click. and after that all is silent. Unless your ear is right on top of the drive.

    And what tests? At what, in what? For example I just finished reading this article:http://www.tomshardware.com/reviews/1.5tb-hdd-caviar,2331.html

    Where they test two different 3-Drive RAID0's against each other in a mix of some others. Some of the results are a little strange! For example on this page they show IOPs in ranges from 50 to 150. I couldn't get mine to go below 500 in the worst case and with a decent data size more like 500,000 IOPs (no cache, RAW I/O). Of the 30 or 40 test results there I'm guessing only 2 or 3 of them actually apply to anything you might be interested in. The question is which ones. :p

    Pages like this might make me feel good about my drives seeing that Samsung won but I'd be willing to bet that on a mac with different benchmarking software the results would come out somewhere between all together different and the same. Meaning that just temperature, the order the tests are run in and, the platform they're being performed on, and other unknown conditions could easily change the results more than the differences graphed here in most cases.

    It looks pretty much the same case with tom's tests on the Black vrs. the HE103UJ to me. You really need someone with a mac pro to bench the two setups in whatever software you're primarily interested in using them with - or do it yourself. Synthetic benchmarks can be great if you have and know the software being used - which is why it's best to always use freely available benchmarking software. But on a totally different platform with software we have no access to? They would only be useful if the results were drastically different. 10 or 20 IOPs or 5 to 10 MB/s at up around 120 MB/s isn't significant enough to make a reasonable judgement call from - IMHO.

    Sorry if this lands you back to the level of guessing but in all seriousness I can't see it any other way. We have many many testimonials on the Mac platform claiming that the Blacks are FAST and work well in RAID0. We have nothing to tell us what the HE103UJs are like. So either you go with the pack and get the Blacks or you get the HE103UJs and benchmark them for us for the first time. I don't believe Tom's results are any other than a very general indication of how they will perform on a mac - and so general as to not be any indication at all of which is faster the Blacks or the Sammy HE...
     
  17. Tesselator macrumors 601

    Tesselator

    Joined:
    Jan 9, 2008
    Location:
    Japan
    #17
    BTW, just to give Tom credit where credit is due they say the same things I just said except they don't also have to consider software and platform differences. Tom concludes:


    "Still, the characteristics of these two hard drives are significantly different. Both are low at idle power and stay really cool, but performance and efficiency varies. WD adjusted its 1.5 TB Caviar Green to deliver quick access time and high I/O performance, which results in a respectable performance per watt result for intensive I/O workloads. Samsung’s Spinpoint F2 EcoGreen is different. It delivers higher maximum transfer rates and lower power consumption in all application scenarios except when idle. As a result, Samsung does better in performance per watt for sequential streaming.

    However, the differences will only be slightly noticeable, and they will only be noticeable if you actually use these units as system drives or for applications that require high performance. If you intend to use the drives for plain storage, backup, and archiving, you will probably not be able to see many differences between them in everyday operation. Both drives rest at a nice performance level considering their low spindle rotation speed, and it will be hard to further decrease power consumption without making fundamental design changes."​
     
  18. Mac Husky thread starter macrumors regular

    Joined:
    Mar 28, 2009
    Location:
    Bavaria, Germany
    #18
    Thx. That´s what I ment when i wrote:
    A lot of testing and as much different results - even more :D

    You are right: just go for one of these. You won´t be disappointed at all.
    They probably all work on a high level of speed and reliability.

    I had a look at some more conclusions like that yesterday night and thought that the Caviar Black would be the better
    decision due to faster access time and i/o performance using it as a system drive RAID0.
     

Share This Page