SSD oR raptor

[cdeez]

Ok so im about to build a new i7 system , i will be using windows 7 when it comes out but for the time being using a copy of vista 64 bit . but to get down to business

i am thinking of getting 1x 300 gig raptor drive , or

Patriot Torqx PFZ128GS25SSDR 2.5" 128GB SATA II Internal Solid state disk (SSD) - Retail

Patriot Torqx PFZ128GS25SSDR 2.5" 128GB


i want to make sure this ssd is going to out preform the 10000 rpm drive ,


also if there is something better then the patriot torqx let me know

 

[r_manic]

Yes it will. Keep in mind though that if you're copying data to and from the SSD to, say, a hard disk or usb drive, you'll be limited by the speed of the non-SSD device.

 

[wuzy]

Torqx performs the same as every other SSD based on Indilinx controller + same graded MLC NAND out there.
If you want to go faster then there's SLC NAND like the uber-expensive Intel X25-E series.

The main advantage SSD has over any (with no exception) HDD is multi disk-tasking performance. In the factor of over 200% difference. It's night and day.

 

[sub mesa]

The newer X25-M G2 beats the X25-E in several important benchmarks (random write notably). So i would recommend X25-M G2 as its argueably the fastest MLC SSD available at the moment. And at the price of OCZ Vertex/Patriot Torqx. These two share an Indilinx controller, but the Intel controller is superior.

 

[outlw6669]

+1 for a SSD.
The difference is night and day between a good SSD and even the fastest hard drives.

Just be sure and avoid an of the current JMicron controlled SSD's.
They have an issue where, during small file wrights, the drive will stop responding for 0.5-2 seconds.
You can locate a good explanation in This Article (good read but quite long).

Your chosen SSD is an excellent choice!
I would, however, recommend changing it to This OCZ Vertex 120Gb drive and saving a bit of cash ($40 after the $30 MIR).
While both drives are completely identical inside the case, Patriot is advertising like Hd manufacturers do (calculating space in base 10) where OCZ gives you the formatted size.

Which ever you end up getting, make sure and update to the latest firmware before installing your final OS.
This will ensure your drive operates reliably and at top speed the longest.
Also note, when you update the firmware, your drive will be formatted loosing all data.

 

[randomizer]

And don't, whatever you do, stop the firmware update if it appears to hang. OCZ Vertex firmware updates can take 20 mins to 4+ hours.

 

[rcpratt]

Intel X25-M G2 kthx.

 

[randomizer]

I just got a WD Black 500gb 32mb for $59. A couple of those in Raid would save you some big bucks.

The 640GB is better though as it has 320GB platters instead of 250GB.

 

[cdeez]

+1 for a SSD.
The difference is night and day between a good SSD and even the fastest hard drives.

Just be sure and avoid an of the current JMicron controlled SSD's.
They have an issue where, during small file wrights, the drive will stop responding for 0.5-2 seconds.
You can locate a good explanation in This Article (good read but quite long).

Your chosen SSD is an excellent choice!
I would, however, recommend changing it to This OCZ Vertex 120Gb drive and saving a bit of cash ($40 after the $30 MIR).
While both drives are completely identical inside the case, Patriot is advertising like Hd manufacturers do (calculating space in base 10) where OCZ gives you the formatted size.

Which ever you end up getting, make sure and update to the latest firmware before installing your final OS.
This will ensure your drive operates reliably and at top speed the longest.
Also note, when you update the firmware, your drive will be formatted loosing all data.

Thanks for all the info , i just want to make sure that me spending 160$ more then a raptor is worth it ,also if i were to change to the ocz vertex is it as reliable as the patriot torqx.Finally im seeing some talk about Sequential Access - Write and Sustained Write . whats the difference ?? and are sustained writes on these SSD faster then the 10000 rpm hardrives ?

 

[outlw6669]

If you can afford an extra $160, the SSD is well worth it.
As the Vertex and Torqx are identical inside the box (same controller, same flash), they should both have the same reliability.

Short answer to the rest, yes, a SSD (except for the JMicron controlled drives) will be substantially faster in practice than a 10k RPM drive.
It has more to do with the random performance than anything else.
In real life usage, you are not going to be making very many large file sequential wrights.
You will, however, be making tons of random reads/wrights anytime you do basically anything on your system.

Nearly every SSD available is capable of massively outperforming the best performing hard drives in Sequential Access'.
A good, non enterprise, SSD is not terrible with Sequential Wrights but falls behind the faster hard drives.
When you move to Random Read performance, SSD's again are king with a massive advantage.

Random Wright performance is a little more complicated.
With some non JMicron based SSD's you are again looking at massive performance gains, others perform a little more lackluster.
According to Anandtech's update on the OCZ Vertex, it performs 297% faster than the Raptor (6.47MB/s vs 1.63MB/s).
The current Samsung based SSD's give you performance a bit slower still somewhat acceptable, about like a notebook drive (0.55-077MB/s).

The JMicron based drives, on the other hand, perform pathetically bad with Random Wrights.
A single JMicron JMF602B controller gets you 0.02MB/s and the 'new improved' drives with a pair running internal RAID 0 gets you a whopping 0.03MB/s

Here is a quick performance chart from Anandtech's SSD Anthology:

And this is from the OCZ Vertex Performance/Firmware Update

If you have not already done so, I would still recomend reading Anandtech's: The SSD Anthology (or at least skipping ahead to the Benchmarks).
Although a long read, it gives you a great background on how SSD's work and why they perform so much better than standard hard drives.

 

[sub mesa]

OCZ Vertex and Patriot Torqx only differ in naming; its the same controller and product.

Both are cheap alternatives to the best SSD on the market: the Intel X25-M. However, now that Intel lowered its prices with the new "G2" or Generation 2 34nm drives, they may even be cheaper than the Vertex. So i would recommend taking the Intel drive, as its alot of faster in random write benchmarks and generally is superior to the Vertex. If the difference in cost is less than 10-40 dollars i would consider the Intel SSD.

 

[sub mesa]

Oh outlw6669 beat me too it, with nice pics.

 

[cdeez]

asesome info , thanks guys !!!

 

[cdeez]

ok , new game plan . I got a pay check in and am gunna drop some more on a ssd , im now going to go for the best , and ima go with a intel x25-m g2 when they start shipping again guys . Thanks for the great guide and info !!


cant wait

 

[RushRocks]

i dunno guys...SSD's will be the only game in town sometime in the future, but I'm not sure they are ready for prime-time.

I just built a Core i7 Vista64 machine with 4 Velociraptors in a RAID 0+1, and it is incredibly fast (it's my Photoshop machine) For the money, I can't see buying SSD's (yet) for the extra little performance in read only (the write times still suck in most of them) I'm not a gamer, so they may work better for that application.

Before someone asks, I did look at performance (and price/performance ratio) beforehand. Adobe told me that they are not ready to roll with SSD's just yet. Apparently, all that will change with Windows 7...it is built to handle SSD's in the future.

Laptops on the other hand......

 

[cjl]

10.2? Only if he set it up horribly wrong. A velociraptor array with properly done caching has roughly a 7ms read access and 4-5ms write access. Not like an SSD, but FAR more capacity for quite a bit cheaper. It still has way better access times than standard hard drives too. I have a pair of velociraptors in RAID 0, and it works great. I might get an X25-m as an OS drive when Win7 comes out though (and keep apps and programs on the velociraptors - I have >200GB of apps). Keep in mind that not everyone can get away with a 100GB or so main drive.

 

[sminlal]

A velociraptor array with properly done caching has roughly a 7ms read access and 4-5ms write access. Not like an SSD, but FAR more capacity for quite a bit cheaper.

SSDs vs. hard drives are like cache memory vs. DRAM. One is a LOT faster, but smaller and more expensive. The trick is to put the fast, expensive drive to work on the relatively small number of files that get hit the hardest.

For a lot of users, using an SSD for the OS disk is a great fit. The OS disk typically gets beat up a lot more than a data drive does, it typically gets hit with a lot of random reads which play to the biggest advantage for an SSD, and it typically needs only a relatively small amount of storage (ie, ~80GB is very reasonable for for most OS disks).

IMHO, folks that are spending a lot of money on high-end CPUs and overclocking solutions are kind of missing the boat if they're not seriously thinking about how to improve their disk performance as well. And for no-compromises disk performance then it's awfully hard to beat an SSD.

 

[Onus]

I want one, but the price is still too high for me. When 80GB non-jmicron SSDs are <$200, I'll wait maybe another month to see if 120GB drives also drop into that range, then either way I'll get one.

 

[cjl]

CJ,

It was just a number i threw out there. What are the read's like with Vrap's in R0?

250MB/s roughly. It certainly feels faster than a standard 7200rpm drive (or even any RAID array of 7200s that I've used).

 

[sub mesa]

If you would get 250MB/s (2x125) at the outer tracks, you would get around (2x75) at the inner tracks, or 150MB/s. Pair that with slight fragmentation and layout of the filesystem (who does not keep files large contiguous on the disk, so it requires seeks) this would be lowered down to 120-130MB/s realistically.

Whereas an SSD with "250MB/s reads" would in this case still provide 250MB/s, even with slight fragmentation, and the location on the disk does not matter.

Though this argument is more relevant to the issue of Intel X25-M SSDs having relatively low sequential write speeds; only 75MB/s. However, saying the velociraptor does 125MB/s and Intel SSD does 75MB/s write, would be kind of unfair and also incorrect in realistic conditions.

Numbers may lie, just not enough to be considered illegal in court.

 

[randomizer]

Sequential reads are almost completely meaningless. Throw some file fragmentation into the mix and test random reads.

 

[cjl]

If you would get 250MB/s (2x125) at the outer tracks, you would get around (2x75) at the inner tracks, or 150MB/s. Pair that with slight fragmentation and layout of the filesystem (who does not keep files large contiguous on the disk, so it requires seeks) this would be lowered down to 120-130MB/s realistically.

Whereas an SSD with "250MB/s reads" would in this case still provide 250MB/s, even with slight fragmentation, and the location on the disk does not matter.

Though this argument is more relevant to the issue of Intel X25-M SSDs having relatively low sequential write speeds; only 75MB/s. However, saying the velociraptor does 125MB/s and Intel SSD does 75MB/s write, would be kind of unfair and also incorrect in realistic conditions.

Numbers may lie, just not enough to be considered illegal in court.

150 at the inner tracks? Nope - try 210. The velociraptors don't taper off as much as my 7200rpm Caviar Black does. They really do hit those speeds too - when copying larger files up to about 2GB from the Velociraptors onto my Caviar Black, I get roughly 230MB/s real copy speed (since the write is being cached by the controller, so the read is the limiting factor). It's true that the velociraptors can't match good SSDs as an OS drive, they certainly aren't bad.

As for randoms, here are some results from PassMark's disk test (for comparison, I've included numbers from both my velociraptor array and my 1TB Caviar Black storage drive):
Test 1, 50% read, 50% write, 12GB file accessed in 8k blocks via Win32 api uncached, 100% random, 60s duration: 3.0MB/s (vraptor), 1.4MB/s (Caviar Black)
Test 2, 100% read, 12GB file in 8k blocks, win32 api uncached, 100% random, 60s: 1.8 MB/s (vraptor), 1.3MB/s (Caviar Black)
Test 3, 100% write, 12GB file in 8k blocks, win32 api uncached, 100% random, 60s: 6.5 MB/s (vraptor), 3.6MB/s (Caviar Black)
Test 4, preset workstation benchmark (70/30 R/W, 16k blocks, 20/80 seq/random): 4.2 MB/s (vraptor), 2.7 MB/s (Caviar Black)

I'm sure that a good SSD would absolutely destroy these numbers, but it does clearly show the benefit of the Velociraptors compared to the Caviar Blacks.


EDIT: on an interesting side note, I can get 300k iops on my vraptor array if I do cached 128 byte writes, 100% random. I know it's meaningless, but it's kind of entertaining watching the numbers rack up that quickly

 

[randomizer]

Don't put an X25-M in there, the Velociraptor will refuse to spin up in shame. :lol:

 

[cjl]

Nah. Actually, given the size of the apps I use, the vraptors would remain as my app drives. If I tried to install programs to the x25, I would run out of space quite quickly. The x25 would be OS only. It would actually end up as a fairly nice tiered setup from the fastest, smallest drive up to the slowest, largest drive.

Keep in mind though - I intentionally chose those benchmarks to make the hard drives look as bad as possible - you'll note that all of them except the preset workstation benchmark are 8k chunks with 100% random access - an absolute worst case scenario for a hard drive, and one that is quite unlikely in any normal use. Those are also all uncached numbers, which isn't quite accurate (especially in random write performance). For example, if I set everything to the same settings as the random write test above, but use the win32 cached api, I get 1750 MB/s from the velociraptor array instead of only 6.5 MB/s. Admittedly, that's aggressive caching on the part of the RAID controller, but it does show the other extreme end of what can happen in my storage setup.

 

[randomizer]

And just because you "cheat" (cache) to get higher speeds doesn't mean someone can dismiss the fact that you are getting higher speeds.