Replacing 4x30GB Vertex SSDs With?

[Elfear]

I'm running out of capacity on my Vertex Raid array and I'd like to replace it with something faster. I've been pouring over reviews for the past couple days and I'm a little confused as to what would be the best solution. According to a lot of sites, the Vertex 3 seems to be the best thing since sliced bread. Anandtech and a bunch of other sites show it being on the top of the heap by a good margin. However, Hardware France did a pretty in-depth review and Crucial's drives seem to offer much faster random read/write speeds (cheaper too for the C300).

The drives will be used for multi-tasking (lots of browser tabs, MS Office, etc.) and gaming. Random 4k performance seems to be where regular users would notice the difference so I'm rather confused by different reviews. Any recommendations?

 

[LokutusofBorg]

You can't really go wrong with any of these newer drives, although the anti-OCZ crowd will surely chime in with their dire warnings of anything OCZ. I agree they have had some issues, but take all the negativity with a grain of salt if it does hit your thread here.

If you're keen on Intel reliability then they've given you two choices: the 300 line is the actual Intel controller, and the 500 series is their move to a third-party controller (Marvell, same as Crucial has used all along). If you are a fan of Crucial (I personally always buy Crucial RAM) then their SSDs are very solid and fast. If you've had good luck with your Vertexes then the V3 is neck and neck for benchmark/review top spots, as well.

It really comes down to brand loyalty/preference more than saying "this one is better than the others".

 

[noblemo]

The charts posted in this review provide comparative performance info. It's a good place to start.

Looks like your EVGA E758 mobo is SATA II. Unless you already have or plan to add a SATA 6Gbps controller card, you will not benefit as much from the Vertex 3, Crucial M4, Intel 510, etc.

 

[Elfear]

The charts posted in this review provide comparative performance info. It's a good place to start.

Looks like your EVGA E758 mobo is SATA II. Unless you already have or plan to add a SATA 6Gbps controller card, you will not benefit as much from the Vertex 3, Crucial M4, Intel 510, etc.

Thanks for the link. Those Anandtech charts are what was swaying me towards the Vertex 3 drives but the Hardware France review seems to cast the C300/M4 in a better light (in terms of random 4K performance). My plans for now are to upgrade to Bulldozer or SB-E this summer so I will have SATA 3 eventually.

I'm leaning towards a pair of 128GB C300s. That would save me ~$200 over a pair of Vertex 3s. My current SSDs have treated me well so I have nothing against OCZ. I just don't want to spend big bucks for the V3s only to find out they feel about the same in a typical user environment.

 

[noblemo]

Sorry for the redundant link; got distracted and should have reread the OP before submitting.

The Hardware.FR and AnandTech reviews both seem to show:

- C300 128GB is faster than Vertex 3 120GB performing 4KB random reads
- Vertex 3 120GB is faster than C300 128GB performing compressible 4KB random writes
- Vertex 3 120GB write speed drops to be about the same as C300 128GB when data is incompressible

Vertex 3 120GB vs C300 128GB

4KB Random Read

- AnandTech, 6Gb/s, QD=3: 37.0 vs 83.8 MB/s
- AnandTech, 3Gb/s, QD=3: 35.4 vs 78.8 MB/s

- Hardware.FR, 6Gb/s, QD=4: 44.1 vs 109.8 MB/s
- Hardware.FR, 3Gb/s, QD=4: 43.6 vs 94.5 MB/s

- Hardware.FR, 6Gb/s, QD=4, incompressible: 41.4 vs 109.8 MB/s
- Hardware.FR, 3Gb/s, QD=4, incompressible: 39.9 vs 94.5 MB/s


4KB Random Write

- AnandTech, 6Gb/s, QD=3: 214.3 vs 110.4 MB/s
- AnandTech, 3Gb/s, QD=3: 162.1 vs 110.4 MB/s

- Hardware.FR, 6Gb/s, QD=4: 241.7 vs 124.7 MB/s
- Hardware.FR, 3Gb/s, QD=4: 158.1 vs 110.5 MB/s

- Hardware.FR, 6Gb/s, QD=4, incompressible: 123.3 vs 124.7 MB/s
- Hardware.FR, 3Gb/s, QD=4, incompressible: 123.4 vs 110.5 MB/s

AnandTech Vertex 3 120GB Review
AnandTech C300 Review

 

[Zap]

Random 4k performance seems to be where regular users would notice the difference so I'm rather confused by different reviews.

Correction. Random 4k performance to a certain extent. If you note the AnandTech articles show the Intel drives as having poor random 4k performance compared to the latest/greatest, but the Intel drives often have better "real world" performance. I think we're hitting the saturation point in IOPS. For instance the Sandforce 1500 controller has three times the random 4K of the Sandforce 1200, but in "real world" performance the drives perform almost the same.

 

[Elfear]

Sorry for the redundant link; got distracted and should have reread the OP before submitting.

The Hardware.FR and AnandTech reviews both seem to show:

- C300 128GB is faster than Vertex 3 120GB performing 4KB random reads
- Vertex 3 120GB is faster than C300 128GB performing compressible 4KB random writes
- Vertex 3 120GB write speed drops to be about the same as C300 128GB when data is incompressible

Vertex 3 120GB vs C300 128GB

4KB Random Read

- AnandTech, 6Gb/s, QD=3: 37.0 vs 83.8 MB/s
- AnandTech, 3Gb/s, QD=3: 35.4 vs 78.8 MB/s

- Hardware.FR, 6Gb/s, QD=4: 44.1 vs 109.8 MB/s
- Hardware.FR, 3Gb/s, QD=4: 43.6 vs 94.5 MB/s

- Hardware.FR, 6Gb/s, QD=4, incompressible: 41.4 vs 109.8 MB/s
- Hardware.FR, 3Gb/s, QD=4, incompressible: 39.9 vs 94.5 MB/s


4KB Random Write

- AnandTech, 6Gb/s, QD=3: 214.3 vs 110.4 MB/s
- AnandTech, 3Gb/s, QD=3: 162.1 vs 110.4 MB/s

- Hardware.FR, 6Gb/s, QD=4: 241.7 vs 124.7 MB/s
- Hardware.FR, 3Gb/s, QD=4: 158.1 vs 110.5 MB/s

- Hardware.FR, 6Gb/s, QD=4, incompressible: 123.3 vs 124.7 MB/s
- Hardware.FR, 3Gb/s, QD=4, incompressible: 123.4 vs 110.5 MB/s

AnandTech Vertex 3 120GB Review
AnandTech C300 Review

Thanks for the breakdown. I didn't realize they were saying the same thing.

So for the average user/gamer, I assume the read speeds are more important since most people create content on their drives much less frequently than they use the content. Is that a reasonable assumption?

Correction. Random 4k performance to a certain extent. If you note the AnandTech articles show the Intel drives as having poor random 4k performance compared to the latest/greatest, but the Intel drives often have better "real world" performance. I think we're hitting the saturation point in IOPS. For instance the Sandforce 1500 controller has three times the random 4K of the Sandforce 1200, but in "real world" performance the drives perform almost the same.

Interesting. So what would you attribute the better real world feel to? Is there a way to quantify it or do you have to rely on review sites and user experience?

I could really care less about benchmarking the drives but I do want something that feels very fast in everyday use. My Vertex drives feel decently quick but I figured there would be some benefit to be gained by modern drives.

 

[MarkLuvsCS]

Interesting. So what would you attribute the better real world feel to? Is there a way to quantify it or do you have to rely on review sites and user experience?

I could really care less about benchmarking the drives but I do want something that feels very fast in everyday use. My Vertex drives feel decently quick but I figured there would be some benefit to be gained by modern drives.

If i'm not mistaken, one of Anand's recent reviews showed that on the newer intel drives garbage collection occured more regularly and thus kept a much higher "min speeds" due to performance degradation as the drives filled. Although the Intel drives weren't getting as high max speeds, they didn't deviate very much at all from their maxes because of pretty frequent garbage collections.

 

[groberts101]

Only thing that's really missing here is the fact that the tests quoted there are with incompressible data which is worst case on a Sandforce drive. Once you get into more easily compressed data(such as an OS) the Sandforce will pull far ahead of the C300's capability especially when writing(even moreso in reads AND writes when moved to 6G ports).

They are WYSIWYG on all benchmarks whereas Sandforce can speed up as the data becomes more easily compressible. Kind of like having Turbo speed available with much of the data that would be typically used for such a machine.

PS. I went from a 6 x 30GB Vertex array to a 6 x 50GB Vertex 2 array and the snappiness and multitasking ability is on a whole other level. In fact even 1 Sandforce drive will beat Indilinx controlled drives in R0 due to the ultra low latency and excelent small file performance they offer.

 

[Elfear]

If i'm not mistaken, one of Anand's recent reviews showed that on the newer intel drives garbage collection occured more regularly and thus kept a much higher "min speeds" due to performance degradation as the drives filled. Although the Intel drives weren't getting as high max speeds, they didn't deviate very much at all from their maxes because of pretty frequent garbage collections.

That makes sense. I'll have to look into it some more.

Only thing that's really missing here is the fact that the tests quoted there are with incompressible data which is worst case on a Sandforce drive. Once you get into more easily compressed data(such as an OS) the Sandforce will pull far ahead of the C300's capability especially when writing(even moreso in reads AND writes when moved to 6G ports).

They are WYSIWYG on all benchmarks whereas Sandforce can speed up as the data becomes more easily compressible. Kind of like having Turbo speed available with much of the data that would be typically used for such a machine.

PS. I went from a 6 x 30GB Vertex array to a 6 x 50GB Vertex 2 array and the snappiness and multitasking ability is on a whole other level. In fact even 1 Sandforce drive will beat Indilinx controlled drives in R0 due to the ultra low latency and excelent small file performance they offer.

Glad to hear you noticed a speed difference between the generations. Have you had a chance to compare V3's in R0?

 

[Zap]

Interesting. So what would you attribute the better real world feel to? Is there a way to quantify it or do you have to rely on review sites and user experience?

It isn't that one thing is contributing to better real world feel, but rather that after some point some things no longer matter as much. BITD because lack of IOPS was holding back performance, more IOPS meant better performance. These days SSDs have such high IOPS that more IOPS no longer matters.

It's like this... let's say the highway you take commuting to work has a 65MPH speed limit that is very strictly enforced so there's no chance of speeding. Let's say that you have been commuting to work riding a Segway. Your commute takes an hour, because you were being limited by the top speed of that device. Let's say you upgrade to a 50cc gas scooter. Now your commute gets done in around 25 minutes because it is much faster by the Segway, but you are still limited by the top speed of the scooter. Let's say you upgrade again, to a Honda Civic. Your commute now takes around 11 minutes because you are no longer limited by your top speed. Now, how much shorter would your commute be if you got a Porsche 911 Turbo? It wouldn't be any shorter, because you were no longer limited by the top speed of the device. Well, maybe a few seconds here and there due to better acceleration away from a traffic light so technically it would benchmark faster, but at the end of the 12 mile commute the difference may not be that great by this point.

Think of the Segway as a 5400RPM HDD, the scooter as a VelociRaptor 600GB, the Civic as your old Vertex SSDs and the Porsche as a Vertex 3.

Anyways, the point of that analogy was to illustrate the concept of something no longer holding you back, no longer being a bottleneck.

I could really care less about benchmarking the drives but I do want something that feels very fast in everyday use. My Vertex drives feel decently quick but I figured there would be some benefit to be gained by modern drives.

You could do like what groberts101 did and just swap with a newer generation, though you can skip over to Vertex 3 at this point.

A couple thoughts...

You might want to target fewer but larger drives. The thing is that SSDs internally are very parallel in operation already, meaning larger capacity drives with more NAND chips outperform smaller drives. This means your gains with RAID0 are lost with using the slowest drives. Of course you can also target the lowest cost/GB as well.

Do you have everything you run on your computer (other than pure data) already on your SSDs? If not, I'd recommend more capacity, maybe going as far as keeping your original Vertex array along with anything new you get (maybe get a PCIe RAID card).