Post your HDTach 3 Results!

Judas

yah ahahaa..

now that's a steller disk preformance..


just to bad that nearing the ned.. it realy starts to drop off.

PangingJr

well i think that was the best i can do with this onboard Intel firmware/hardware raid controller, it's just a cheap raid solution, no addition cost just some hard drives. i'm sure i should see a big improvement in real-world raid performance with these seagate low-cost drives under a good SATA raid card that have its own onboard CPU and memory.

Judas

imo, aside from the cpu usage being roughly 10%.... i still think those are excellent results reguardless of a dedicated raid controller or not.. (isn't it using the southbridge which includes the raid controller?)

IMO Panging, i still think the best raid controller is the Highpoint tech.

PangingJr

the Intel onboard SATA raid controller is not a hardware raid, it's a firmware raid, well it's more like partly software-based and partly hardware-based raid solutions.
but i really don't need to have a real good raid performance, just wanted to add some disk performance to all the systems that i have.
the idea was began after a friend of mine have asked me to make some OS image files for him. as i see that i can get these drives at a very low price, so i just add 1 or 2 more hard drives to set up a raid volume on the systems that already have an onboard SATA raid motherboards, i plan to use just maybe one or two systems that still use AHCI with no raid, since these two systems have been set up from the beginning without a complete raid-ready but AHCI only.

about the Highpoint tech, you did not tell me what model?
Areca Technology ARC-12xx, AMCC/3ware 9500S/9550SX, HighPoint RocketRAID 23xx...
these are a good ones but it cost a lot more just one of this raid card alone...

Judas

well i'm really used to the IDE ones actually. But i've seen some pretty steller results from the Highpoint SATA raid controllers..

I just miss having the highpoints onboard raid controllers..

It sat aside byitself on the motherboards with all the little components (example would be the Abit boards that were not only the first to feature raid but ATA100 spec and ATA133 via the Highpoint tech controllers)

Such boards include the very first HT360 ATA100 Abit KA7-100 and following some of the athlon and intel socket boards from ATI that were considered "more high end" that included the via/intel/sis ide solutions and also the 2 extra IDE channels from highpoint... (up to 8 IDE devices per board, which i ran quite a setup at the time)

I've email Abit and several other board makers about the chipsets that highpoint have produces and asked if they had ever considered them and what they figure needs attention before they would consider using them. Abit responded stating that they had been in talks with highpoint with there sata controllers... but never got any further then that, Asus however hadn't really heard much about them and thanked me for providing links and information on them (which i find highly unlikely or untrue).

I just remember with 2 30gb maxtor 2mb cache 7200rpm drives that were poor performers (cheap though) pulled of one hell of a steller result... with sub 1% cpu usage with the onboard Highpoint. (which was unheard of) and sustained very excellent bursts and high read/write results.

PangingJr

i think those highpoint raid card was also a firmware raid. i used to use and have many PCI raid cards in the past years.

as for SATA raid, the better SATA raid the ones that is consider as a real hardware-based SATA raid must have its own CPU and memeory.
a good hardware-based SATA raid is supposed to give you the overall system performance, not just disk performance but sometimes you still see the system is lacking in some tasks...

Judas

i'm waiting on a true PCI-ex set of cards to be released (sound/network/raid/various others) as the PCI bus just is to damn slow imo.

zanzabae

u guys need to run the long test so u cant cheat with small blocks and get results closser to real use

deathman20

Had sign up and post my results after seeing this thread.



Top Imaged are without Write-Back Cache Enabled and the bottome 2 Images are with it enabled. The smaller voulme of couse is my OS/Apps and the larger is my data storage.

Gotta love 3x 320Gig Perp drives.

temeteus82

Nice

Ferret_Style

Not sure how mine stacks up but using the intel based raid with 2 x 80gb SATA2 Seagate drives.

Ultrasonic2

after some testing i have decided that HD tach and h2benchw aren't good benchmarks for raid configs. at all ! As the best setup for the benchmarks ( 4-8k stripe size ) produced the worst real life performance. PCMark05 and MS bootvis seems to give a better real life measurement.

MS Bootvis records my boot to the desktop time of 4.65 seconds ( at 128k stripe size ) at the moment
PCMark05 xp startup is 16 mb ish

i however get 253mb average with 4 seagate 7200.9 80gb drives .but to get that i have to turn off the cache, NCQ and have a stripe size of 4-8 k.. where as with bootvis, pcmark and battlefield 2 load times are best to have the cache and NCQ on with a stripe size of 128k . i might point out that the pc is faster in windows ( opening and close apps ) in this mode too

Judas

can you select 256kb striple size? that should produce even better results.

ATM i'm still looking at getting a new motherboard and hoping to hell that perhaps 6 drives in raid 0 will be support with stripe sizes of at LEAST 256kb if not 1mb.

Ultrasonic2

no the largest i can use is 128k on my Asus A8N-SLI SE i also want to go to a AM2 cpu so i can use the 570 chipset on a Asus M2N-E board to run 6 disks . i dont know if i'll know notice much difference in performance over 4 disks though . ive looked into it and it doesn't look like i can run a larger than 128k stripe size on the M2N-E board

i have considered buying an after maket raid controller which i think would produce better performce but it's alot more money i could buy three Asus M2N-E boards for the price of the RocketRAID 2320 though that would give you 8 disks

im not sure increasing the strip size would make to much difference. i could have tested this on my old raid setup but not now with the largest size being 128k . the stripe size ( 4-128 ) does seem to make a bigger difference to performance with the more disks you have. i would love to see some PCmark xp startup with a bigger strip size though.


i must say my disks are so quite and cool it's exellent

i found this interesting, though it's only on 2 disks which doesn't make it relevant. as i also found with 2 disk the stripe sizes didn't make to much difference
http://www.overclockers.com/articles1063/index02.asp

Judas

Well the more disks you throw into the mix, the larger your bandwidth increase substaially.

It would also make more differences more obvious.

PangingJr

if what deathman20 has posted on #39 is also using a 128K stripe size, then this is the 2nd best example of a Intel onboard RAID's striped set that using a 128K with all of RAID array member drives were working evenly together and across the array which meets the basic idea of setting up a striped set. the more higher score that i can easily get it from this controller and drives all means nothing... when comparing to this value,
and as far as i am concern this is the best striped set setups i can have from my existing hardwares/softwares which i find it's working and serving me quite well in the real-world applications and their performances.
however, thing is i can only get this from using 4 drives which i'll have to leave a lot of free drive space that i'll ever need to use it, so as of now i reduce the member drives to 3 and will also checking with only 2 drives and may be with other hard drives models or firmwares when i have more free time to try it.

Ultrasonic2

PangingJr when I also test my 4 disk arry with a 128k stripe size I get about half what I should in HD tach.with a 4k strip i get a 243mb average
At the end of the day it’s a bad test it doesn’t read files on file system . Which is obvious as it can do a read test of the disk that has no data on it. Well that’s my opinion

I have been looking into what the cluster size of the disk makes to performance.

Ok this is my understanding of how strip / cluster size on 4 disks works.

If the strip size is 128k that means that the total strip size of all disks is 512 ( 128 x512 ) not that this matters till we include the below
Windows default cluster size is 512
Any file smaller than 512k will take up one cluster of 512

If we accept the above statements that means if we were to create a disk based on the above .
And read a 1k file it would take up 512k on the disk(s) and therefore access all disk to retrieve the data .

Where as if we used a cluster size of 64k ( 128 isn’t available ) when reading the 1k file it would only access one disk which would slightly decrease the access time and freeing up the other 3 disk for other i/o’s


I haven’t tested small cluster sizes. has anyone else ?

Judas

every test i've ran with modified Windows NTFS cluster size variances has proven zero impact on the transfer rate or on raid systems mostly.

Now Stripe(block) size while setting up the raid array is what we are looking at..

i beleive i explained this above, but

4 drives, Stripe size of 128kb, = 512kb of data across all 4 drives = 128kb block size per drive.

256kb file = first 2 drives blocks filled, 2 other drives not, bandwidth used is 1/2 of it's potential, and leaving extra space on the other 2 drives (this is mostly fact + theory).

1mb file = all 4 drives get 2 sets of 128kb blocks filled, full drive potential forfilled.

The larger the file, the better, the larger the block/stripe size set (say 2mb vs 128kb) for large files, the better (as far as my tests have gone).

Basically, the moment the block size is larger then the file being read/writen across all the drives correctly, that's when the performance gets seriously cut down.

And being that we live in a world in which not all files are nice even 128kb/512kb/6mb files, the majority of the file will be evenly processed, while the last bits will leave voids in the raids drive setup.

Plus this makes a HUGE impact if your raid controller doesn't handle things the most efficent way.

Ultrasonic2

actually may last post was kind of wrong the default size is 512 bytes not 512k which means more than one cluster should fit in the stripe size .

thanks for your posts

Judas

sorry i ment that as well.... i just leave windows to auto/default, i've just learnt to leave it the hell alone lol

PangingJr

Ultrasonic2,
file system driver accesses files by their cluster number and is therefore unaware of any other information about the physical drive.
file system is a different layer from physical drive access. increase NTFS cluster size... may affect (but not necessary increasing) "file system" performance and data accessibility at the operating system level, Not at the RAID controller's and physical hard drive device's drivers level.

That said. and it sounds like the RAID stripe size and file system cluster size could hardly be matched.
however, i'd say they will never be matched. i suggest you don't waste the time trying to match them,
and just set your RAID's stripe size to controller's default/recommended value setting, leave file system's cluster size at the Windows default.
Hopefully all systems will work at their optimum, try that, and i hope this helps.

Ultrasonic2

These tests were done with PCMark05. Each stripe size was tested multiple times. The drives are Seagate 80 7200.9 SATA2 drive on a Asus A8N-SLI –SE ( Nvida Raid )

4 DRIVES

128k
Score 7411
Test set: HDD - XP Startup
HDD - XP Startup: 16.917234MB/s
Test set: HDD - Application Loading
HDD - Application Loading: 6.744542MB/s
Test set: HDD - General Usage
HDD - General Usage: 7.417566MB/s
Test set: HDD - Virus Scan
HDD - Virus Scan: 72.196426MB/s
Test set: HDD - File Write
HDD - File Write: 150.519882MB/s

64k
Score 7299
Test set: HDD - XP Startup
HDD - XP Startup: 15.316092MB/s
Test set: HDD - Application Loading
HDD - Application Loading: 6.382443MB/s
Test set: HDD - General Usage
HDD - General Usage: 6.926970MB/s
Test set: HDD - Virus Scan
HDD - Virus Scan: 72.399605MB/s
Test set: HDD - File Write
HDD - File Write: 173.936325MB/s

32k
Score 4687
Test set: HDD - XP Startup
HDD - XP Startup: 13.641394MB/s
Test set: HDD - Application Loading
HDD - Application Loading: 5.688534MB/s
Test set: HDD - General Usage
HDD - General Usage: 6.260658MB/s
Test set: HDD - Virus Scan
HDD - Virus Scan: 10.369295MB/s
Test set: HDD - File Write
HDD - File Write: 184.745605MB/s

16k
Score 7808
Test set: HDD - XP Startup
HDD - XP Startup: 11.991607MB/s
Test set: HDD - Application Loading
HDD - Application Loading: 5.822765MB/s
Test set: HDD - General Usage
HDD - General Usage: 5.870546MB/s
Test set: HDD - Virus Scan
HDD - Virus Scan: 129.009308MB/s
Test set: HDD - File Write
HDD - File Write: 225.797699MB/s

8k
Score 7555
Test set: HDD - XP Startup
HDD - XP Startup: 10.950196MB/s
Test set: HDD - Application Loading
HDD - Application Loading: 5.872395MB/s
Test set: HDD - General Usage
HDD - General Usage: 5.423572MB/s
Test set: HDD - Virus Scan
HDD - Virus Scan: 128.713074MB/s
Test set: HDD - File Write
HDD - File Write: 225.606461MB/s

4k
Score 7368
Test set: HDD - XP Startup
HDD - XP Startup: 10.121690MB/s
Test set: HDD - Application Loading
HDD - Application Loading: 5.795400MB/s
Test set: HDD - General Usage
HDD - General Usage: 5.176408MB/s
Test set: HDD - Virus Scan
HDD - Virus Scan: 135.871506MB/s
Test set: HDD - File Write
HDD - File Write: 216.524231MB/s

Windows software raid
Score 7515
HDD - XP Startup: 17.463575MB/s
Test set: HDD - Application Loading
HDD - Application Loading: 6.236611MB/s
Test set: HDD - General Usage
HDD - General Usage: 6.940607MB/s
Test set: HDD - Virus Scan
HDD - Virus Scan: 84.266342MB/s
Test set: HDD - File Write
HDD - File Write: 154.892548MB/s

2 DRIVES

128k
Score 6357
Test set: HDD - XP Startup
HDD - XP Startup: 11.625346MB/s
Test set: HDD - Application Loading
HDD - Application Loading: 6.822273MB/s
Test set: HDD - General Usage
HDD - General Usage: 6.300878MB/s
Test set: HDD - Virus Scan
HDD - Virus Scan: 81.015060MB/s
Test set: HDD - File Write
HDD - File Write: 105.548340MB/s

16k
Score 6480
Test set: HDD - XP Startup
HDD - XP Startup: 9.711584MB/s
Test set: HDD - Application Loading
HDD - Application Loading: 6.084504MB/s
Test set: HDD - General Usage
HDD - General Usage: 5.423079MB/s
Test set: HDD - Virus Scan
HDD - Virus Scan: 129.852493MB/s
Test set: HDD - File Write
HDD - File Write: 113.021835MB/s

4k
Score 6419
Test set: HDD - XP Startup
HDD - XP Startup: 8.877170MB/s
Test set: HDD - Application Loading
HDD - Application Loading: 6.233236MB/s
Test set: HDD - General Usage
HDD - General Usage: 5.121596MB/s
Test set: HDD - Virus Scan
HDD - Virus Scan: 139.451736MB/s
Test set: HDD - File Write
HDD - File Write: 113.449753MB/s

1 Disk

Score 5615
Test set: HDD - XP Startup
HDD - XP Startup: 9.030895MB/s
Test set: HDD - Application Loading
HDD - Application Loading: 6.645724MB/s
Test set: HDD - General Usage
HDD - General Usage: 5.169858MB/s
Test set: HDD - Virus Scan
HDD - Virus Scan: 103.591942MB/s
Test set: HDD - File Write
HDD - File Write: 71.458069MB/s


it's a shame 128k is the max size as id love to test larger sizes
Michael Dixon