It will be interesting to see if the smaller OCZ dual controller (and even quad controller) drives perform equally as well, considering the price these units will fetch they could be a very real option for application and OS installation.
Originally Posted by felix the cat Third graph (page 3 I believe) the Y-Axis Legend (data description labels or whatever you want to call it) are a wee bit messed up.
Otherwise very nice review indeed. Anyone want to lend me 500 bucks??
If it's the response time graph, there's not a lot I can do about that - the SSDs are so much quicker than the mechanical drives that their bars can't fit the data in on the same graph. I'll work with our developer to see if we can find a solution.
Correct me if I'm wrong but looking at the charts there doesn't seem to be any real reason to buy an SSD or even the more affordable Raptor over the Baracuda and Samsung, at least in terms of all-round performance, storage capacity and price. Or have I misunderstood the charts?
Originally Posted by rustor Correct me if I'm wrong but looking at the charts there doesn't seem to be any real reason to buy an SSD or even the more affordable Raptor over the Baracuda and Samsung, at least in terms of all-round performance, storage capacity and price. Or have I misunderstood the charts?
As useful as the theoretical and read/write/copy tests are, it's the real world tests that you see the biggest jump from using a mechanical to using a decent SSD. Booting from off to working desktop in just 35 seconds is so much faster than a mechanical drive is capable of, and you'll notice the nippier hard drive performance in almost every aspect of OS use. yes, the price is steep and the storage limited, but this is the only reliable way to get performance on this sort of level in a single drive.
You say you could build a RAID 5 array with some Samsung terabyte drives for the same price, so why didn't you include such a setup in the comparisons?
Meh, so even with a dual controller there are still situations where SSD's are ~50% slower than mechanical drives, if they would atleast be faster than a mechanical drive in every situation, one could say it's ok that they are smaller and more expensive, but as it is I just dont see the point.
Well, maybe the OCZ Vertex which will come with Cache and a non Jmicron Controller will be able to beat the mechanical drives in every situation.
SSD are norm more relable then HDDs and if bad sectors happen the disk should start to get smaller (if the hdd maker adds that option not likely on any jmicron SSDs) the bigger the SSD more likely the longer it will last as the wear leveling can work over an larger space of the flash, turning off any auto defrag (win7 off by default for SSD) is very important as it likey to shorting the life by a alot (as well as that pointless to defrag on SSD due to data is the same speed where ever it is accessed from and wear leveling)
its the access time on SSD that makes the system respond faster, the dual Jmicron is an intresitng idea seems to work well one side can be writeing other side can be reading
Originally Posted by mauvecloud You say you could build a RAID 5 array with some Samsung terabyte drives for the same price, so why didn't you include such a setup in the comparisons?
My guess would be that g.skill provided the SSD for testing, but they don't have 3-4 1tb samsung's and a high end raid controller sitting around in the office.
Originally Posted by teamtd11 My guess would be that g.skill provided the SSD for testing, but they don't have 3-4 1tb samsung's and a high end raid controller sitting around in the office.
We do have 4 Seagate 1Tb 7200.11s, but two or three of them are dead actually...
The RAID controller is another option: what do you go for, crappy CPU controlled onboard RAID or high end and expensive controller?
Also, if you want a 2.5" SSD the size, temperature, noise and energy consumption difference is also CONSIDERABLE between this and 3/4 1Tb 3.5" disks.
Originally Posted by leexgx the dual Jmicron is an intresitng idea seems to work well one side can be writeing other side can be reading
I was wondering about this - what happens if the data you want to read is on the same side as you're writing to? would you occasionally get the same slowdown as the 128GB?
Originally Posted by Bindibadgi Also, if you want a 2.5" SSD the size, temperature, noise and energy consumption difference is also CONSIDERABLE between this and 3/4 1Tb 3.5" disks.
Also, SSDs are more rugged than HDDs. In that sense, I'd expect they're a better bet for computers that will go through a lot, such as car PCs, some laptops depending on how they're used, and other computers that have to put up with a lot of jostling.
I'm guessing that it's the next generation of SSDs that are created that start to make them make sense for the mainstream.
Originally Posted by Baz If it's the response time graph, there's not a lot I can do about that - the SSDs are so much quicker than the mechanical drives that their bars can't fit the data in on the same graph. I'll work with our developer to see if we can find a solution.
for the 'higher is better' graphs make the starting value 'x' instead of zero, 50 looks like a good number. As for 'lower is better' only thing i can thinl of is to make it a relative(%) graph instead of an absolute one.
once these babies hit the $0.90/GB mark I'm all over these like a fat man at a buffet.
Two more questions: Why wasn't the WD Velociraptor included? and what drive adapter would be needed?
Originally Posted by Baz If it's the response time graph, there's not a lot I can do about that - the SSDs are so much quicker than the mechanical drives that their bars can't fit the data in on the same graph. I'll work with our developer to see if we can find a solution.
for the 'higher is better' graphs make the starting value 'x' instead of zero, 50 looks like a good number. As for 'lower is better' only thing i can thinl of is to make it a relative(%) graph instead of an absolute one.
once these babies hit the $0.90/GB mark I'm all over these like a fat man at a buffet.
Two more questions: Why wasn't the WD Velociraptor included? and what drive adapter would be needed?
Percentage wouldn't help much for the response time graph - what they really need is a bar graph generator smart enough to put the caption to the right if the bar is too small. I disagree with the idea of starting at non-zero for the "higher is better" graphs like the speed in MB/s, because that tends to exaggerate the differences. If there's really that big a difference, it'll still be visible in a zero-based graph.
Originally Posted by mauvecloud I disagree with the idea of starting at non-zero for the "higher is better" graphs like the speed in MB/s, because that tends to exaggerate the differences.
+1
Graphs that start at non-zero are evil, especially if mixed with start at zero graphs in the same review, it just defeats the points of having graphs since you are going to have to read it very carefully then.
Graphs that start at non-zero are evil, especially if mixed with start at zero graphs in the same review, it just defeats the points of having graphs since you are going to have to read it very carefully then.
They're also mathematically incorrect and, as a degree-level mathematician/scientist/engineer, I'm not going to publish graphs that aren't actually representative of the data.
SSDs are approaching a stage where they are being used on Enterprise hardware, so reliability is improving (unsure on G.Skill, but I know a manufacturer)
You didn't have the test I really wanted to see, compile an application, like (for example) the Linux kernel.
Actually, what I really want to see is how long it takes to load up the test suite (something that happens before every test is run) in a Ruby on Rails application. I found that having to wait several seconds for it to load up on a slow 5400rpm drive really slowed down my development, especially since I develop in a test-first manner. But I can't expect you to do something so specialised. But compiling a Linux kernel is similar, and much more relavent to most Linux users.
It took long enough, but someone finally answered my long-standing suggestion - a flash drive of RAIDed cells presented as a single "Disk".
I wonder just how this thing would perform being a 4x64 split, rather than a 2x128. And, indeed, when someone's going to make an MLC or SLC chip with integrated controller, so SSD makers can use RAID parallelism to split loads, and mitigate flash media's limited read/write bandwidth.
I wonder just how this thing would perform being a 4x64 split, rather than a 2x128.
It can technically be done - there's a JMicron controller out there that will handle four, but to be honest, I think it's a combination of diminishing returns and squeezing everything through one SATA 3Gbps port :|
Originally Posted by Bindibadgi It can technically be done - there's a JMicron controller out there that will handle four, but to be honest, I think it's a combination of diminishing returns and squeezing everything through one SATA 3Gbps port :|
Well, read speeds would run into the limitations of SATA2, but write speeds (the weak point of ssd's) should still profit. But what I think will help more is adding cache like the OCZ vertex will that will be out soon.
Comments 1 to 25 of 30
ReplyOtherwise very nice review indeed. Anyone want to lend me 500 bucks??
If it's the response time graph, there's not a lot I can do about that - the SSDs are so much quicker than the mechanical drives that their bars can't fit the data in on the same graph. I'll work with our developer to see if we can find a solution.
As useful as the theoretical and read/write/copy tests are, it's the real world tests that you see the biggest jump from using a mechanical to using a decent SSD. Booting from off to working desktop in just 35 seconds is so much faster than a mechanical drive is capable of, and you'll notice the nippier hard drive performance in almost every aspect of OS use. yes, the price is steep and the storage limited, but this is the only reliable way to get performance on this sort of level in a single drive.
Well, maybe the OCZ Vertex which will come with Cache and a non Jmicron Controller will be able to beat the mechanical drives in every situation.
its the access time on SSD that makes the system respond faster, the dual Jmicron is an intresitng idea seems to work well one side can be writeing other side can be reading
My guess would be that g.skill provided the SSD for testing, but they don't have 3-4 1tb samsung's and a high end raid controller sitting around in the office.
We do have 4 Seagate 1Tb 7200.11s, but two or three of them are dead actually...
The RAID controller is another option: what do you go for, crappy CPU controlled onboard RAID or high end and expensive controller?
Also, if you want a 2.5" SSD the size, temperature, noise and energy consumption difference is also CONSIDERABLE between this and 3/4 1Tb 3.5" disks.
Dreaming - we just don't know yet to be honest.
I was wondering about this - what happens if the data you want to read is on the same side as you're writing to? would you occasionally get the same slowdown as the 128GB?
Also, SSDs are more rugged than HDDs. In that sense, I'd expect they're a better bet for computers that will go through a lot, such as car PCs, some laptops depending on how they're used, and other computers that have to put up with a lot of jostling.
I'm guessing that it's the next generation of SSDs that are created that start to make them make sense for the mainstream.
for the 'higher is better' graphs make the starting value 'x' instead of zero, 50 looks like a good number. As for 'lower is better' only thing i can thinl of is to make it a relative(%) graph instead of an absolute one.
once these babies hit the $0.90/GB mark I'm all over these like a fat man at a buffet.
Two more questions: Why wasn't the WD Velociraptor included? and what drive adapter would be needed?
Percentage wouldn't help much for the response time graph - what they really need is a bar graph generator smart enough to put the caption to the right if the bar is too small. I disagree with the idea of starting at non-zero for the "higher is better" graphs like the speed in MB/s, because that tends to exaggerate the differences. If there's really that big a difference, it'll still be visible in a zero-based graph.
If you want an adapter to mount one or two 2.5 inch drives in a desktop computer (whether SSD or mechanical), you'll probably want one of these:
http://www.startech.com/item/SATA35252X-35-Bay-to-Dual-25-SATA-HDD-Adapter.aspx
+1
Graphs that start at non-zero are evil, especially if mixed with start at zero graphs in the same review, it just defeats the points of having graphs since you are going to have to read it very carefully then.
They're also mathematically incorrect and, as a degree-level mathematician/scientist/engineer, I'm not going to publish graphs that aren't actually representative of the data.
Actually, what I really want to see is how long it takes to load up the test suite (something that happens before every test is run) in a Ruby on Rails application. I found that having to wait several seconds for it to load up on a slow 5400rpm drive really slowed down my development, especially since I develop in a test-first manner. But I can't expect you to do something so specialised. But compiling a Linux kernel is similar, and much more relavent to most Linux users.
I wonder just how this thing would perform being a 4x64 split, rather than a 2x128. And, indeed, when someone's going to make an MLC or SLC chip with integrated controller, so SSD makers can use RAID parallelism to split loads, and mitigate flash media's limited read/write bandwidth.
It can technically be done - there's a JMicron controller out there that will handle four, but to be honest, I think it's a combination of diminishing returns and squeezing everything through one SATA 3Gbps port :|
Well, read speeds would run into the limitations of SATA2, but write speeds (the weak point of ssd's) should still profit. But what I think will help more is adding cache like the OCZ vertex will that will be out soon.
-
« Previous
-
1
-
2
-
Next »
Discuss in the forums