Sign in with
Sign up | Sign in
Your question

Intel X25 160 GB or two 80 GB in RAID 0?

Last response: in Storage
November 28, 2009 10:56:15 AM

I couldn't seem to find any topic adressing this directly through searching, so hopefully someone can give me an answer or point me towards another thread.

I'm simply wondering, would two 80 GB Intel X25m G2 drives in RAID 0 be faster than one single 160 GB drive?

And if so, are there any drawbacks, or any particular requirements to get the best out of the two drives in RAID 0?

Thanks in advance for any input.

More about : intel x25 160 raid

a b å Intel
a c 415 G Storage
November 28, 2009 3:31:53 PM

RAID 0 will be faster for sequential transfer rates (ie, copying large files, reading or writing large files such as RAW camera images or movies), assuming you haven't saturated the RAID controller so that it becomes a bottleneck.

RAID 0 will not be all that much faster for random transfers (ie, booting the OS or starting application programs).
a b å Intel
a c 353 G Storage
November 28, 2009 6:49:54 PM

Do not believe win 7 TRIM cmd works when the drives are connected as a Raid 0, or other Raid modes. Also firmware upgrade can not be applied in raid mode (I think this is still correct).
Related resources
December 2, 2009 1:23:45 AM

No Win7 Trim for Raid.
December 20, 2009 9:17:30 PM

Raid-0 can be better for random accesses, if your OS issues several access requests in parallel, which is the case beginning with XP and its prefetch.

Then, the sectors requested by the software (requested in advance by XP's prefetch) will, with a little luck, reside on different disks which may respond in parallel, even if these requests belong to different files.

Now, about one or two X25M in particular, only someone having tried it can answer. My single X25E (faster than X25M at random requests) is limited by the Cpu (E8600 @4GHz), as any Cpu frequency change reflects at 100% on boot speed.
a b å Intel
a c 415 G Storage
December 21, 2009 3:30:40 AM

Pointertovoid said:
Raid-0 can be better for random accesses, if your OS issues several access requests in parallel, which is the case beginning with XP and its prefetch.
But in the real world of booting and starting applications, RAID 0 only increases performance by about 10% compared with the 50% increase you'd expect from being able to do twice as many concurrent I/O operations. This is because a lot of the work done during booting and loading is serial: read one file, then read a registry entry, then read the next file, etc.
December 23, 2009 5:10:11 PM

Experiments tell quite the opposite. Enabling Ahci accelerates boot time and applications load time quite effectively by serving simultaneous requests. Ahci doesn't improve a bit throughput nor random access time.

Such simultaneous requests are produced by Xp's prefetch mechanism, whether Windows or the application were designed to issue them or not.

Again, this is experimental, so theory shall adapt.
a b å Intel
a c 415 G Storage
December 23, 2009 7:20:41 PM

The 10% figure I quoted is what my experiments told me, by measuring boot and application startup times first with Windows 7 freshly installed on a single drive in AHCI mode, then repeated using ICH10R RAID 0 on two identical drives.
a b å Intel
a c 353 G Storage
December 23, 2009 9:16:55 PM

sminlal - agree with your end result.

one minor point. While the events are serial, the HDD reads during load of operating system are by inlarge random reads of small files. Raid0 does not boost random read/writes very much over non-raid0 - Supporting your 10% boost during booting. The biggest performance benifit is in seq read and writes of larger files. I've used raid0 for many years and prefer it, But my new SSD,s are so much faster - Do to Random reads/writes on 4k files.

This is also why SSD totally blow away HDD raid0 - Random read/writes and very-very low access time.
December 30, 2009 3:26:09 PM

Access requests are parallel, thanks to the prefetch! What happens with SSD is that they aren't the limiting factor any more. With my X25-E, any Cpu clock change scales boot time 1:1. That's why two SSD aren't much faster than one.

As Ahci, which improves only Ncq, accelerates booting a lot, both on a mechanical disk and an SSD, the OS does issue parallel requests.
a b å Intel
a c 353 G Storage
December 30, 2009 5:05:54 PM

But the execution (HDD/SSD read/writes) are serial: Interface is SATA for serial.
The old IDE interface was parallel, but execution was still serial. A head can only be at one point at a time, for SSD while a head is not involved it is also serial in reading/accessing a cell.
December 30, 2009 8:47:35 PM

Mamma mia!

Next one please!
December 31, 2009 4:04:43 AM

Well, what really matters with SSDs is TRIM.

And Win7 does not support TRIM on RAID. :( 

Go with the 160GB SSD. The drive itself is probably >10% faster than the 80GB X-25M G2. :p