qasdfdsaq 2 Report post Posted February 19, 2009 1) Your offset (where the partition starts) must be a common multiple (divisible by without remainder) of number of drives minus one and your stripe width. Number of drives = 3. 3-1 = 2. So offset must be divisible by 2 and your stripe size (32k), i.e. a common multiple of both. So your offset could be 2*32, i.e 64, 2*2*32, i.e. 128, 3*2*32, i.e. 192 (probably a dumb idea), 4*2*32, i.e. 256, etc. etc. I'd recommend leaving it to where Vista/2008 puts it by default - 1024KB. Windows 2003 puts it in the same place as XP as it is based on XP (sector 63) Windows 2008 puts it in the same place as Vista as it is based on Vista (sector 2048) 2) OK 3) Windows 2003 is the same as XP as it is based on XP (64KB) Windows 2008 is the same as Vista as it is based on Vista (1-8MB) 4) You might be better off with 32KB to be honest - let us know how it goes. Share this post Link to post Share on other sites
master blue 0 Report post Posted February 25, 2009 (edited) ok, i understand. my config now: windows server 2008 sp1 stripe size: 32k cluster size: 32k this are the atto results: it's better than before with win2003, but i think there's something wrong. i expected "a bit" more. i have installed win2008 on a partition (exactly 30720 mb...this looks better ) of the raid (cluster size: win standard) and then i have partitioned a second one with cluster size 32k (partition D). Edited February 25, 2009 by master blue Share this post Link to post Share on other sites
master blue 0 Report post Posted February 26, 2009 ok, i have some news: after activating this option (red marked): (i dont know what it means or what it does. the first option turns on the disk cache, the second -> ? maybe RAM caching, but this wouldn't speed up the raid itself i think.) http://members.inode.at/599459/device%20manager.jpg the results look like this: partition C (system partition, standard cluster size): http://members.inode.at/599459/atto%20c.jpg partition D (data partition, 32k cluster size): http://members.inode.at/599459/atto%20d.jpg why 32k is so slow and 64k so fast? Share this post Link to post Share on other sites
Munkee915 0 Report post Posted February 26, 2009 Hi guys. I found this thread when i was looking for advise on setting up my raid 5 on my nforce 570 board. I setup the raid 5 w/ 3 1tb discs in win Vista. Set stripe to 32k and Vista picks it up ok as 1.81 tb. Formated using 32k cluster. As far as i know vista takes care of the rest. also disabled NCQ on the channels. However after all this i test it and still get about 40mb write and 150mb read. Is this average or should i be seeing better? I was considering trying again we 32k strip and 64k cluster, or 64/64, but from what i've read here 32/32 would be the best for my setup. Any suggestions? Share this post Link to post Share on other sites
master blue 0 Report post Posted February 26, 2009 Hi guys. I found this thread when i was looking for advise on setting up my raid 5 on my nforce 570 board. I setup the raid 5 w/ 3 1tb discs in win Vista. Set stripe to 32k and Vista picks it up ok as 1.81 tb. Formated using 32k cluster. As far as i know vista takes care of the rest. also disabled NCQ on the channels. However after all this i test it and still get about 40mb write and 150mb read. Is this average or should i be seeing better? I was considering trying again we 32k strip and 64k cluster, or 64/64, but from what i've read here 32/32 would be the best for my setup. Any suggestions? maybe you have to turn on the option, which i described in the posting above. you find it in the properties of your array in the device manager. if you have an english os installed, please tell me, how this option is named and described there. (the german version of this text does not explain what this option is doing.) Share this post Link to post Share on other sites
Munkee915 0 Report post Posted February 27, 2009 (edited) The option is to "enable advanced performance". Properties -> Policies -> Check the box. Tried it and reran atto and got similar numbers as before (almost exactly the same as your first run). I wish i could pull out the numbers like your 2nd D drive test Edited February 27, 2009 by Munkee915 Share this post Link to post Share on other sites
Munkee915 0 Report post Posted February 27, 2009 Well, reformated in 64k clusters, left at 32k stripe, and now my 2nd run at atto looks like yours . Same low preformance 32k and below but 140mb + after that. Find it odd, but hey, it works! Share this post Link to post Share on other sites
sedernota 0 Report post Posted March 1, 2009 I found this thread after having problems with my nforce raid 5 array. It is very slow. I ran HD Tune and got very erratic results. Huge spikes with lows in the 7s up to the 120s. I read that HD Tune isn't great with arrays, but I couldn't find the ATTO Benchmark that some of you seem to use. Anyways, I have 4 500GB Hitachi drives forming a RAID 5 array. This motherboard, ASUS P5N-E SLI with nforce 650i SLI and 430i, has 4 channels Serial ATA with RAID 0, 1, 0+1, 5 support. I have the latest drivers installed. Vista 32. If I understand this all okay, I am not going to be able to get much speed out of this current configuration. I don't game, I am a professional photographer who set this up about 14 months ago for 1.5GB storage with redundancy. Since my mobo only supports 4 drives, what is the best thing to do at this point? Buy two 1.5GB drives and set up a raid 1? Or buy another 500GB drive and a RAID card (which one) for five disk RAID 5 array? Any advice would be appreciated. Thanks, Seder Share this post Link to post Share on other sites
dx1 0 Report post Posted March 2, 2009 Hello, I run a software RAID 5 on a patched WinXP-SP3. HDDs involved are 3 Hitachi 7k1000.b 1TB. I have one RAID 0 for Games with around 3x81.5 GB, a RAID 5 had performed very bad (5 to 6 mb per second write speed) and I was looking for an solution. The disk were replacement for 3 Samsung 250 GB, which had wiriting speeds around 20 to 30 mb per sec. The starting offset for the new RAID 5 was around 8xxxxxx somewhat. I would like to set the offset to 8388608. This should be one of the "magic offsets" if I did not miscalculate the values. Is there a way to do it? btw: o&o partitionmanager2 showed something like 850 gb total space, 1200 free (139 %) when it show sth. like 1700 gb totaL; 1200 FRee, . I deleted the volume by trying to create a new one with better starting offset by using the PM2 boot disk. Didn't work. I guess because of the limited feature set of "native XP". Creating a volume with diskpart failed with a paramter error (create volume raid disk=2,3,4 offset=8388608 size=870400) Please ignore my word-bumpyness, English is not my native language. (PMs in German words will be welcome. ) Share this post Link to post Share on other sites
TheFatKid 0 Report post Posted March 6, 2009 Does this same protocol hold w/ new nVidia 700-series chipsets? Share this post Link to post Share on other sites
master blue 0 Report post Posted March 7, 2009 Does this same protocol hold w/ new nVidia 700-series chipsets? on page 2 the user Wiinter is doing this procedure on a 780i and its working too. Share this post Link to post Share on other sites
steve8 0 Report post Posted March 25, 2009 http://www.alternativerecursion.info/?p=31 check this out for raid performance analysis w/ ICH9r Share this post Link to post Share on other sites
HachavBanav 0 Report post Posted May 14, 2009 http://www.alternativerecursion.info/?p=31check this out for raid performance analysis w/ ICH9r First time, I read a comparison of HDD/RAID/OS cache modes. Very nice article. Share this post Link to post Share on other sites
Executron 0 Report post Posted August 16, 2009 Hi! This is an extraordinary topic, very interesting stuff on RAID 5. But i have some problems The basic data is : i am using 3 x 250 HDD which i want to set up as RAID 5 on a Silicon Image chip which is incorporated into my DFI SLI-DR Expert nforce4 motherboard. Even if i set up the RAID in BIOS using its special software or if i use the Windows based software of the chip it won't let me choose anything else but "chunk" size and drives. Things like drive offset, cluster size, stripe size, I/O block size i can't seem to choose. Currently the benchmarks tell me that the RAID5 array has an averange rate transfer speed of 15mb/sec which is very low. Thanks and salutations from Romania! Share this post Link to post Share on other sites
Migelo 0 Report post Posted May 7, 2011 I registered on this forum just to say this: I can't thank you enough qasdfdsaq, thanks to you, I'm finally able to use my RAID5 array properly! Really THX. And since I'm already here, let me ask you if I'll still be able to recover from a drive failure using your settings. (I don't see a reason why, just wondering) Share this post Link to post Share on other sites
Samapico 0 Report post Posted July 11, 2011 (edited) HOLY COW! I registered just to post this... I'm using a nForce 980a SLI motherboard, with 3x 1.5TB Caviar Green drives in RAID 5. I use it mainly for large files like movies, and copying back my movies to the array was VERY slow (~15MB/s). I thought I just had to deal with it since RAID 5 had a slow write (or so I read), especially with these Green drives which are not meant for performance... until I did a desperate search about RAID 5 on nForce and found this topic. My partition was properly aligned already (Win7 did this for me), but all the rest, I left to default. My stripe size was set to 64, so I recreated the array with 32. Cluster size was 512 bytes, changed it to 64k. I just copied a large file at 90-110 MB/s!!! This is awesome! Edit: After just a few GB's of write, it dropped to 75MB/s, but that's still way better than 15MB/s Edited July 11, 2011 by Samapico Share this post Link to post Share on other sites