WD Raptor vs SSD

I just had a friend install an Intel X25-V 40Gb on an old i865 motherboard with a P4 2.8C.

That southbridge is ancient. In fact, it doesn't even support NCQ as its 2 SATA ports are basically just bridged ATA133 ports.

The read speed maxes out at 134Mb/s b/c of the older 1.5Gbps motherboard, BUT at least he can still run the Intel toolbox weekly to keep the SSD trimmed. He's very happy with the boot times and resposiveness of the system. Its made him want to upgrade his other PC to an Intel X25-M 80Gb
 
Be aware that SSDs like Intel use NCQ to process multiple I/O at once, you'll lose that benefit if you run with a controller that doesn't support NCQ or when the mode isn't set to AHCI.

But even a castrated Intel SSD is still a very fast drive, vastly superior to any HDD for the function of the system disk (random I/O).
 
sub.mesa said:
Thanks Pete, but i run something much more advanced already, take a look at this topic if you like:
http://hardforum.com/showthread.php?t=1500505

However, i would not recommend any Indilinx-SSD at the moment, its price is only slightly lower than the Intel based SSDs while being factor 5-10 less fast when it comes to random I/O. So that means that 10 Indilinx SSDs in RAID0 would give you about the same random IOps as one Intel. If you can't afford the 80GB version, the Intel X25-V 40GB is probably the best buy.
Click to expand...

I find this all very interesting. Not to say this isn't true, but are there tests that confirm this?

The reason I find it interesting is that from my understanding, with these SSD drives, random I/O patterns should not hold back the drives performance very much, what with the ultra low seek times. Why would the Intel drives be 5-10x faster (!!) than the OCZ drives when access times are so low for all MLC SSD's? (and the Vertex are all MLC) I don't get it.

And I just have to say, that many of the people (not criticizing anybody here) who used to make the argument that sequential transfer rates via things such as RAID-0 are more important than access times on hard drives have suddenly switched their position and now argue the latter after the release of SSD's, which as far as I can tell is just the opposite of what is really going on...Again, very interesting. I would love to hear people's input / discussion on this.

*EDIT* This article has shed some light on the subject for me (I hadn't done my homework :p) It seems the Intel SSD's are the ones to beat, with OCZ's drives coming in a close second (according to the real-world performance tests in the article, not synthetic benchmarks, which are a slaughter).
 
Last edited:
Well Intel makes its own controller; the other brands have to 'shop' for Indilinx, JMicron, Samsung, Sandforce, Micron controllers; which are less good especially the first three brands.

Secondly, the reading performance is quite similar, though Intel uses NCQ as command buffering which the cheap SSDs don't use. But the real performance differences are when doing random writes; or actual I/O patterns with mixed read+write most of it highly random. That would be the kind of I/O where Intel SSDs severly outclass the competition; at least the Indilinx/JMicron/Samsung generation. Compare the two benchmarks below:

Random Read benchmark (Intel on top but nothing too extreme)
http://images.anandtech.com/graphs/intelx25mg2perfpreview_072209165207/19506.png

Random Write benchmark (extremely high differences between SSDs that do not vary much in price)
http://images.anandtech.com/graphs/intelx25mg2perfpreview_072209165207/19508.png

Compare the last benchmark, especially the bad JMicron/Samsung controller will get a good beating; and this shows those controllers suck. Indilinx is pretty decent, but looks mediocre compared to Intel.

Looking at the price levels of Intel SSDs, they are only marginally more expensive per GB than its competitors, while providing much better I/O performance and likely to last much longer than the competition as well. Intel is one of the few SSDs with very low write amplification, which is a key factor for looking at how advanced your SSD is.

The reason behind these differences is in write remapping; in order to change 2 bytes on an SSD, you would have to read a whole data block first, then erase the whole block, then reprogram. So to process 2 bytes you actually may need to write 1 megabyte to the SSD. This is called write amplification and can be a major problem for an SSD. To cope with this, intelligent controllers like the Intel is using write remapping; it actually lies to Windows and intentionally writes to a different location where free flash cells are; then it only has to write the 2 bytes not do the whole read-erase-program cycle which is very slow and wears your drive excessively.

So besides offering superior performance, the endurance and longetivity of Intel SSDs in actual use should be much better than many of the mediocre SSDs.
 
Last edited by a moderator:
sub.mesa said:
The reason behind these differences is in write remapping; in order to change 2 bytes on an SSD, you would have to read a whole data block first, then erase the whole block, then reprogram. So to process 2 bytes you actually may need to write 1 megabyte to the SSD. This is called write amplification and can be a major problem for an SSD. To cope with this, intelligent controllers like the Intel is using write remapping; it actually lies to Windows and intentionally writes to a different location where free flash cells are; then it only has to write the 2 bytes not do the whole read-erase-program cycle which is very slow and wears your drive excessively..
Click to expand...

Very informative, thanks! Still glad I got the OCZ for the price I paid.
 
I would also like to thank you for the info. The only question I have left is what happens to the write capabilities of the Intel when the entire drive is encrypted. I have heard good reports of no major differences in speed and others that their drive slowed down considerably after encrypting their drives.
 
Encrypting it how? If on the filesystem level, that would mean it's just one big (encrypted) file that resides on an NTFS filesystem. For the SSD/filesystem it is just a file; while in essence it is a filesystem of its own that is encrypted.

If this is about full-disk encryption, encrypting the whole disk, then you need to be more careful. If possible, use partitions so the encryption-driver does not write to the end of the drive. That means you make a smaller partition than the full capacity. So a 80GB drive gets 70GB for example. That means 10GB is unused.

So it doesn't matter if its encrypted or not; to the SSD its just binary data anyway. But if the encryption-thing causes every flash cell on the SSD to be written to, it doesn't have enough free space to remap writes to.

Intel SSDs use the difference between GiB and GB as internal reserved data already. So the 80GB Intel SSD is actually 80GiB ("real gigabyte") but uses about 8% for internal usage and exposes itself as a 80GB drive, which is like 73GiB or something.

So you already got about 7GB of free space, but you might need more - 10-15%. That would make sure your advanced SSD stays fast, even after time. It would also mean you don't need trim.
 
Thanks for more info again. The company I work for requires that the entire drive be encrypted. We use TrueCrypt for the encryption. No one here has been allowed to use an SSD because of the companies own unresolved certainty of long term reliability and a sustainable speed with TrueCrypt or any other encryption program. We use Raptors or VeloRaptors currently depending on the age of the machines.
 
Don't buy an old SSD. I don't know what it is, so I would rather buy a new Oynx.

If you want to see speed, check out the OCZ Vertex 100G LE. It uses the new sandforce controller that whips the Intel
 
Rembrandt said:
Thanks for more info again. The company I work for requires that the entire drive be encrypted. We use TrueCrypt for the encryption. No one here has been allowed to use an SSD because of the companies own unresolved certainty of long term reliability and a sustainable speed with TrueCrypt or any other encryption program. We use Raptors or VeloRaptors currently depending on the age of the machines.
Click to expand...
So your company prevents you from using SSDs because they think SSDs are less reliable and less fast than HDDs? Makes no sense to me.

I could argue that from a security point of view, most intelligent SSDs that do write remapping may cause deleted/zero-written data to still be accessible, if you have direct/internal access to the SSD. This would fall into the "NSA-class" of "heroic measures to recover data"; don't expect this to actually work unless you spend some big money on it.

So i don't see any issue. As far as reliability is concerned; HDDs die all the time so even if you got an 'unproven' SSD, it's likely to be more reliable than a 'proven and failed' HDD. With the HDD you know it is going to fail within 10 years, with the SSD there's no particular reason it would.
 
You must log in or register to reply here.
Back
Top