hard drives & raid - benchmark and compare!

[suspekt]

I can disable command queing but where do you disable read cache?

Should be in Windows Device Manager, under IDE ATA/ATAPI controllers. There you should see atleast two Serial ATA controllers. Right click them and go to Properties. Under Primary and Secondary Channel you can change Transfers Mode (if possible), Read Caching, Write Caching and Command Queing.

 

I don't know if you see this with the Windows default driver, but I think you have to use the SW IDE driver to see this (or have used F6 floppy with the Windows install).

[suspekt]

ic. but even my older 160Gb WD SATA1 can reach about 135Mb/s burst and 95mb/s average.

 

how come 10,000rpm doesnt help at all?

suspekt> Ok, everybody with me.. ?

suspekt> 1

suspekt> 2

suspekt> 3:

everybody> ACCESS TIME!

 

 

I think you should theoretically get better sustained read and write speed with higher rpm drives, along with the tighter access, seek and positioning time. You will benifit from tighter access, seek and positioning time more with smaller files than you will if you primary use larger files. Windows is a good example of many smaller files scattered in many different places. But I bet you will see best results in a server enviroment than on workstations. (Yes I know.. SCSI is better for servers etc etc but for small companies and lesser demands Raptors is a good alternative).

[blinker0]

Its under IDE ATA/ATAPI CONTROLLER Double click NVIDIA nForce4 ADMA controller

select primary and secondary channel. If you are using three or more drives select both ADMA Controller channels. And do the same to both. I dont have the IDE SW Drivers installed. So it shouldnt matter. I may try to install The IDE SW driver later just to see what effect it has,

 

I am running O/S on that aray, And have actualy (should say in my case) have seem a major improvement in Windows speed.

 

Dont Dissable write Caching It has very adverse affects

[Tecklord]

Want to thank both you guys for the info. I have 3 raptors and HD Tach kept giving me speeds of 110 MB's a second. Now with read and que disabled I am hitting 160 to 165 MB's a second. Might be worth posting this at the top for people with 3 or more drives on sata? Is this an sata problem only? I have seen 2 other people here with slow raptors in raid. Again thanks. :-) :nod:

[suspekt]

My 4 raptors in RAID 0 (64/4/4) with Write Cache enabled, Read Cache enabled and Command Queing Enabled:

 

 

 

 

113Mb/s in sequential read and 227MB/s burst read. Good ATTO write and read speed.

 

My 4 raptors in RAID 0 (64/4/4) with Write Cache enabled, Read Cache disabled and Command Queing disabled:

 

 

8mb HDTach:

 

 

32mb HDTach:

 

 

156.3Mb/s in sequential read and 245.8MB/s burst read. Great ATTO write speed and poor read speed.

 

I don't know which benchmark type to trust more, but my ATTO read speed is greatly affected in a negative manner with read cache and command queing disabled. The great ATTO write speed that was acheived I almost can get by modifying the IoPageLockLimit in the Windows registry but if I do that my read speed is greatly affected in a negative manner... and I really don't "need" 350-360mb/s write speed. High read speed is of more use to me.

 

Maybe this time next year I will re-do my array for 16kb stripe and 4kb cluster. The earlier Windows crashes and I need to do a re-install, the sooner the RAID gets rebuilt hehe :nod:

 

As you all can see 64/4 is not optimal for my system. That 8mb HDTach is some crazy .. By the looks of things, I think I should be able to get ~200mb in sequential read if I get the stripe and cluster size right.

[SoyoJoe]

I came to almost the same conclusion with different drives. With read and write off and cq off my write speed was severly crippled. With just write on and read off as well as cq off the write speeds improved greatly and read stayed about the same.

 

I would trust atto more over HDtach, but niether are what i would call accurate tests.

 

Those HDtach benchies look funky, its almost like it was trying to draw something for you.

[Redbeaver]

My 4 raptors in RAID 0 (64/4/4) with Write Cache enabled, Read Cache enabled and Command Queing Enabled:

 

113Mb/s in sequential read and 227MB/s burst read. Good ATTO write and read speed.

 

My 4 raptors in RAID 0 (64/4/4) with Write Cache enabled, Read Cache disabled and Command Queing disabled:

 

156.3Mb/s in sequential read and 245.8MB/s burst read. Great ATTO write speed and poor read speed.

 

I don't know which benchmark type to trust more, but my ATTO read speed is greatly affected in a negative manner with read cache and command queing disabled. The great ATTO write speed that was acheived I almost can get by modifying the IoPageLockLimit in the Windows registry but if I do that my read speed is greatly affected in a negative manner... and I really don't "need" 350-360mb/s write speed. High read speed is of more use to me.

 

Maybe this time next year I will re-do my array for 16kb stripe and 4kb cluster. The earlier Windows crashes and I need to do a re-install, the sooner the RAID gets rebuilt hehe :nod:

 

As you all can see 64/4 is not optimal for my system. That 8mb HDTach is some crazy .. By the looks of things, I think I should be able to get ~200mb in sequential read if I get the stripe and cluster size right.

 

those Hdtach are just totally out of this world............................ im talkin about the shape of it.... wtf lol

[Trevor]

What are the speed and noise differences between these two?

 

Western Digital Caviar SE16 320GB 3.5" SATA 3.0Gb/s Hard Drive - OEM

 

and

 

HITACHI Deskstar T7K250 250GB 3.5" SATA 3.0Gb/s Hard Drive - OEM

 

I noticed that the 74gb raptor was very quiet compared to the 36gb raptor (when idle) What kind of noise do these ones make compared to that?

 

I want to get another drive or two so I can have more space and use my computer as a pvr.

[red930]

Search for those models on www.storagereview.com.

[suspekt]

I came to almost the same conclusion with different drives. With read and write off and cq off my write speed was severly crippled. With just write on and read off as well as cq off the write speeds improved greatly and read stayed about the same.

 

I would trust atto more over HDtach, but niether are what i would call accurate tests.

 

Those HDtach benchies look funky, its almost like it was trying to draw something for you.

That's what I'm saying. I think the 8mb HDTach picture is saying "I need balance... perfect balance.. look at my center ".

 

And the 32mb HDTach picture is saying "Look at these tops.. look at these tops... this is how high I wanna go" :shake:

 

First feeling I have with read cache and cq disabled is that I get more reads to the disk... but I have not noticed any negative "real world" effects to speak about.

[Xak]

Here are my results with software RAID5 with the Silicon Image 3114 chip on my SLI-DR with 4x Western Digital WD3200KS 320GB 16MB SATA2 3.0Gb/sec, operating at SATA 1.5Gb/sec (controller limitation). They are configured as 800GB RAID5 16K/64K and 63GB RAID 0+1 4K/64K.

 

As far as real-world results, which I personally am most interested in, I get about 20MB/sec writing to the array and about 60MB/sec reading. So basically, a good evening of sorting/copying/benchmarking the array and I had it over a quarter of the way full with over 250GB of data =)

 

To me, this makes RAID5 much more attainable than a $400 controller card that does it right. Write speeds are actually tolerable for a storage-only volume. I would NOT recommend this for booting, or even a high-volume write array, but for mass storage, it fits the bill perfectly.

 

To test the controller, I have even done a CRC compare of my entire music collection (70GB), and it all matched perfectly, as well as simulated a 'drive failure' by hot-pulling one of the SATA cables during an active file copy, which hesitated, then finished completely. I am very pleased with it's performance, moreso than I would have thought, but that may possibly be due to reading horror stories here at DFI Street about RAID5 performance on this chip.

 

Without further adieu, here is the HD Tach for the RAID5 array (if anyone would like other benchmarks just ask, I didn't know how much interest there would be on this topic):

 

[PÿRÓ]

^ Maybe I should install the Nvidia IDE SW driver? Gurus here say not to install it if I'm not running RAID, but maybe it will make a difference. Right now I'm pretty sure Windows sees my Raptor as an IDE drive.

 

I would need to image my hard drive first in case it messes something up. I've read a few horror stories about this driver, but that was a little while ago.

 

burst speed means 0 ..go by avg read/write (STR) speeds only