Fastest Hard drives for OS...

[Inigmatics]

Are what?

So now you can get like an 8 or 16 gig flash mem hard drive for a large chunk of change or whatever I think my bro was showing me.

Then there is that new card where you have it powered externally and you can slide in like 8 slots of ddr mem for I think a total of 8 gigs and you can use that as a drive, but it has to have a continuous power source I think or it'll lose all the info, so might not be a good idea for an operating system.

Then there is the whole raid vs. non raid and what would be best for the small operating system file and what drive would be best for them.

I'm thinking a single raptor would be most reliable and fastest to have xp on and a raid 0 array of raptors would be thae fastest for anything large like encoding.

What do you think?

[cmay119]

Well there's extremely fast, and then there's fast but cost efficient. What is your budget?

Personally I think Fast and cost efficient is the best way go. I.E. 2x SataII 3.0GB/s drives in RAID 0. Faster would be 2x Raptors in RAID 0, and most expensive would be those flash drives you were talking about. SCSI is also another expensive solution.

[nn_step]

A 20Gb Flash drive with a SATA interface would be the fastest you can buy but it is extremely expensive and hard to come by.
a 4GB I-Ram drive, is relatively cheap compared to flash.
and a 150GB raptor would be even cheaper but with even lower performance..

[Inigmatics]

I guess what I was really wondering was how fast is really fast? But I guess that can't be easily answered in terms I'd understand.

Also, what I was getting at is- would xp installed on a raid 0 of raptors be even noticably faster than xp installed on a single raptor?

[K404]

I-RAM is the fastest for O.S.- 8 seconds from power to desktop but its not got enough space to be worth it.

2 raptors in RAID0 have a higher burst speed.

I think 2 raptors in a stripe would give a boost over a single, but you`d need a 3rd disk separate for moving files to really notice a difference.

[Order]

15k SCSI.

[[cTx]Philosophy]

I was wondering this as well
I currently have a raptor 76 partitioned 10G for os rest for storage
80G as my download drive
And 2X250G sata II raid 0 16k for more storage and installing :banana::banana::banana::banana: and such, is this the correct way you guys would do it? or sell of the raptor and put os on SATA II raid0 16k?
raid is alot fasted on benchies and such but the raptors 10k rpm makes up for that im guessing, plus one more thing, I read somewhere about the NCQ being a killer to raptors on windows XP, how do I desable this does anyone know?
I really wanna et the most performance out fo this pc that I can..

[DAK1640]

I am running 74GB Raptors in a Raid0 and they are very very fast. I am ALWAYS 1st into a new CSS server...NCQ is in your BIOS options.

[uOpt]

And 2X250G sata II raid 0 16k for more storage and installing :banana::banana::banana::banana: and such, is this the correct way you guys would do it? or sell of the raptor and put os on SATA II raid0 16k?


Standard lecture:

when your goal is to speed up program start time, raid-0 is not the right thing to do. Program start is almost exclusively random seeks and small reads in the random locations. This pattern is not sped up by raid-0.

Raid-1 can speed this pattern up, and pretty substancially, but only with clever raid-1 software. Raid software found with SATA Raid controllers is not clever.

[ibby]

scsi 15k

[ibby]

Standard lecture:

when your goal is to speed up program start time, raid-0 is not the right thing to do. Program start is almost exclusively random seeks and small reads in the random locations. This pattern is not sped up by raid-0.

Raid-1 can speed this pattern up, and pretty substancially, but only with clever raid-1 software. Raid software found with SATA Raid controllers is not clever.

Care to prove this theroy ?
Maybe some links i can read .
thanks
:D

[uOpt]

Care to prove this theroy ?
Maybe some links i can read .
thanks
:D

I actually have a huge set of results from different RAID policies.

Most raw data here:
http://forum.useless-microoptimizations.com/forum/raid.html

Discussion here:
http://forums.2cpu.com/showthread.php?t=69817&highlight=linux+software

I also gave more technical details on this thread, which is identical to this one:
http://www.xtremesystems.org/forums/showthread.php?t=95707

[ibby]

Ok.. good read.

However i still have some questions.

The results on the links you gave me where
linux and freebosd.

Do you have any data on windows loaded software.

Ive got an adaptec 1200a raid controller.
IDE, but works like a treat. I have 2 old maxtor hdds in raid 0.

I get resonable results when i run a bench mark util

[uOpt]

Ok.. good read.

However i still have some questions.

The results on the links you gave me where
linux and freebosd.

Do you have any data on windows loaded software.

Ive got an adaptec 1200a raid controller.
IDE, but works like a treat. I have 2 old maxtor hdds in raid 0.

I get resonable results when i run a bench mark util

Well, the big thing missing from my benchmarks is hardware raid. I only compared software RAID.

Software RAID works great for me but most people want a real RAID hardware controller.

To my knowledge there is no Windows software RAID in wide use, so Windows installations are usually hardware controller based.

The SATA builtin RAID stuff is good only for raid-0 anyway, I don't like it much.

[ibby]

Well, the big thing missing from my benchmarks is hardware raid. I only compared software RAID.

Software RAID works great for me but most people want a real RAID hardware controller.

To my knowledge there is no Windows software RAID in wide use, so Windows installations are usually hardware controller based.

The SATA builtin RAID stuff is good only for raid-0 anyway, I don't like it much.

Windows has raid, 1 + 0
but it is rather naff.

well some good info their,
thanks for the input

[[cTx]Philosophy]

I dont want to disable NCQ totally just for the raptor is this available in the bios as well?

[Nanometer]

RAID 1 does not improve any program performance period.

[uOpt]

RAID 1 does not improve any program performance period.

Random seeks can be greatly sped up with a smart raid-1 solution, for reasons I outlined on this forum earlier.

If you had bothered to read the benchmarks I posted you would have known, but for your benefit I can repeat what you need:
http://forum.useless-microoptimizations.com/forum/raid.html






First, the speed for the disks solo:
------------------------------------

Single disk, partition at the beginning of the disk:
49.23 MB/s (51622036 B/s) (15.9% CPU, 3164.6 KB/sec/CPU)
64.43 MB/s (67554626 B/s) (6.4% CPU, 10372.8 KB/sec/CPU)
-Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks---
2047 51795 67.8 54641 13.5 25679 3.8 43896 60.0 66375 4.8 140.0 0.3

Raid 1 (three disks):
---------------------------------

Doing level 1, chunk 1024, at beginning of disk
65.09 MB/s (68250723 B/s) (11.1% CPU, 5983.0 KB/sec/CPU)
65.53 MB/s (68714278 B/s) (6.7% CPU, 10045.5 KB/sec/CPU)
-Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks---
2047 70108 83.6 67393 6.8 28096 4.0 55067 76.8 67643 4.8 464.8 2.3

Raid 0 (three disks):
----------------------------

Doing level 0, chunk 256, at beginning of disk
182.59 MB/s (191457784 B/s) (43.8% CPU, 4267.8 KB/sec/CPU)
159.23 MB/s (166967535 B/s) (43.1% CPU, 3779.7 KB/sec/CPU)
-Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks---
2047 79491 93.4 204622 19.3 68458 9.7 60790 86.6 167182 10.5 289.3 1.0

[nn_step]

uOpt, you always have good benchmarks but Random seeks really shouldn't be extremely common in day to day use.
At the same Time I do know that varies per person and what they are doing.

[uOpt]

uOpt, you always have good benchmarks but Random seeks really shouldn't be extremely common in day to day use.


I'm sorry but that is untrue.

Random seeks are precisely what happens when you start a program.

It has to
Map the executable
Resolve a first set of shared libraries
Read lots of fifty GUI stuff (icons etc)
Read config files
Read registry entries
Read program data (in games textures, map files, 3D models)
Read more shared libraries (plugins)
Load all the virusses that infected the program


All of that is synchronous random seeks. And it is exactly what is not sped up by raid-0 but sped up by my raid-1 software.

The only thing going against that is writing temporary files, which is writes on program startup. But that is only a tiny amount compared to all that random reading that is going on.

[Butcher_]

Everytime your disk "clicks", that's a seek. Disks click a lot when you start stuff.

[crackhead2k]

4x150gb raptors would perform better without using all your pci slots on Raid 0 and ms is low and is great for 3d work fast loading on games. On a good controller it will reach speeds faster than a pair of iram maxed out on Riad 0 + you get alot more storage.

Go check storagereview.com for details.

[[XC] hipno650]

the only reason they beat i ram is because i ram is limited to sata 150 (150mbs) and it totaly saturates the bus and uses all 150mbs so raiding just gives you more space no performance increase. but 150GB raptors use sata 3b so they can go faster with enough of them in raid. but when i ram2 comes out nothing will touch it at all. it will use all of the sata 3b bus (300mbs). i ram2 is supposed to use ddr2 sat 3b and up to like 16GB of storage!!!! if you wanted to get increased performance raiding ram ram's you would need a sata conection of over 3.2GBS and with iram2 over 5.6GBS so raiding ram disks wil give you no benifits in speed. i would say the best bang for your buck would be the iram2 when it comes out or 2 raptors in raid. 4 raptors is a waste because iram is cheaper even with ram. it's like $150 plus 4gb of value ram for like $300. so lets say $500 for iram and over $1000 for 4 150gb raptors in raid. but you could use i ram for the os and 2 74gb raptors in raid for the games and storage. now that would be fast:) :) :). i hope people understand now:). for a good scsi card it costs about $200 at least but you can get 15k drives for like $80. so i guess scsi is an option as well.

[[XC] hipno650]

oh yeah a pair of good sata 2 drives with 16mb caches would be good as well if you where on a budget. or mabye 4 for for the price of say $400.

[[XC] hipno650]

4 of these in raid would give you 1tb of storage and awsome performance.
Western Digital Caviar SE16 250GB 3.5" SATA 3.0Gb/s Hard Drive - OEM $88.00 (newegg.com) i have one of these drives and it performes awsome.please read my long post for performance drive details. READ IT> IT TOOK A LONG TIME TO WRITE!!!!!!!!:stick:

[uOpt]

The advantage of the iram over any raid-0 would of course be random seek :)

[[XC] hipno650]

yes the random seek is 0.0ms according to max pc compared to the raptors:8.2ms.

[[XC] hipno650]

in canada it is cheaper to buy the iram with 4gb mem than it is to buy 2 150gb raptors by like $100.

[crackhead2k]

You wont feel the difference except in superpi you will see the difference and slightly faster load times. Ive tried iram on raid and I hated the lack of storage space... with a pair raptors (150gb) i didnt notice a difference.

Pi scores are onlyfaster by like 1 second. (average through 5 tests consistantly)

And id rather not sacrifice too much pci slots for pair of iram either or even a single.
Consider:
Compare 150GB vs tiny 4GB
Noticable system performance
Price per GB
No backup battery needed
Extra Ram for your computer etc etc

Their has been other reivews like at tomshardware.com if you like to read it. I am sure if your try both you ll wont notice much difference... ull be more satisfied with the raptor i think personally......