Samsung 840 EVO 1TB x2 + 840 EVO mSATA 1TB in RAID 0 Benchmarks

[G73S]

Very disappointing scores I must say, I was expecting 3X the performance but it only seems slightly higher than a 2X RAID setup which I had with my 840 EVO 1TBs before adding the mSATA 1TB to the equation

Write-Back Cache was enabled for maximum performance and disable Windows Write-Cache Buffer flushing.

Without doing the above 2 steps, the performance was no better than a single SSD but it went higher once I disabled the Write-Cache Buffer flushing and Enabled Write-Back Cache

I OPed by 10% (that's the max I can as I need the other space for storing my data)

AS SSD Benchmark with IRST 12.8.0.1016 in RAID 0 RDx3 (W8.1)

CrystalDiskMark with IRST 12.8.0.1016 in RAID 0 RDx3 (W8.1)

 

[smakme7757]

Aren't the drives rated for around 500MB/s read/write.

You get 1547MB/s Seq Read and 1315MB/s Seq write. Doesn't seem too far off a 3x increase to me. Taking into account normal system "anomalies" and that you are most likely not using the best RAID controller on the planet.

 

[G73S]

Aren't the drives rated for around 500MB/s read/write.

You get 1547MB/s Seq Read and 1315MB/s Seq write. Doesn't seem too far off a 3x increase to me. Taking into account normal system "anomalies" and that you are most likely not using the best RAID controller on the planet.

yeah it makes sense I guess.

I just don't feel my system any much more snappier with this setup than when it was with a single SSD

 

[smakme7757]

yeah it makes sense I guess.

I just don't feel my system any much more snappier with this setup than when it was with a single SSD

The snappyness comes from faster access times. When you RAID the drives you are increasing throughput, but not access times. Actually, if I'm not mistaken access times will go up slightly when you put the drives in RAID.

Transferring files will go faster, especially big sequential files like ISOs and Disk images. But the "Snappyness" will be the same.

 

[Galatian]

RAID should introduce some Latency, shouldn't it?

 

[G73S]

The snappyness comes from faster access times. When you RAID the drives you are increasing throughput, but not access times. Actually, if I'm not mistaken access times will go up slightly when you put the drives in RAID.

Transferring files will go faster, especially big sequential files like ISOs and Disk images. But the "Snappyness" will be the same.

I see, with that said, since I rarely transfer files at all, RAID is not ideal for my situation.

I am breaking the RAID array and reinstalling Windows 7 and enabling RAPID mode with the new Magician 4.4

 

[BrightCandle]

Even with increases in 4k throughput (which is where the OS tends to be locking up) you often find that you are reaching a CPU limitation.

One of the tests I did a few years ago was with Java compilation with a big project. I tested it with an old and slow 80GB hard drive, tested it with a more modern 1TB drive and then an Intel G1 80GB SSD and finally a Vertex 2 120GB. While there was a definite improvement going from the hard drives to the Intel SSD going from that very early SSD to a much faster one made no difference in performance. But when I varied the CPU clock speed I saw linear improvements. But the overclock didn't help as much on the hard drives, it was far from linear although there were still improvements. Basically the SSD had improved 4k performance so much that it was no longer the problem and compilation was dominated entirely by clock speed and CPU performance (single threaded unfortunately).

So it might very well be the case that the snappiness improvement people see from SSDs is relieving the bottleneck from the hard drive but then more SSD performance doesn't directly translate into actual application performance. You can see that in the non artificial benchmarks most sites do. If you look at the pc application test suite the differences are minimal between the different SSDs, so while file copy improvements OS startup and application startup doesn't, and its because the SSD isn't the problem, its the CPU.

I see people RAIDing SSDs and I often wonder whether the application they are using it on benefits. It improves multiple user access throughput, it improves sustained transfer throughput but it doesn't really improve most desktop uses much if at all. You either need a lot of parallel reads or a lot of sustained reads to make it worth while, and I don't personally have anything else that can do 500MB/s like my SSD can, if I copy files over the network I am entirely dominated by the cable, even the hard drives in the NAS can do 200MB/s+ reads and writes but the interface + software overhead has me at about 100MB/s. So its a really specific class of problems that benefit from RAID and if you aren't doing one of them its kind of a waste of time.

 

[smakme7757]

I see, with that said, since I rarely transfer files at all, RAID is not ideal for my situation.

I am breaking the RAID array and reinstalling Windows 7 and enabling RAPID mode with the new Magician 4.4

As a system drives it's not much use. A good use case for RAID and SSDs is for Virtual Machine storage where the increase in IOPS is really useful. But if this is just for a gaming machine, then there really isn't much point with RAID 0 1TB SSDs.

A good argument could be that loading times in games would decrease as the increase in throughput would help there. But that's about the only advantage i can think of.

 

[G73S]

Even with increases in 4k throughput (which is where the OS tends to be locking up) you often find that you are reaching a CPU limitation.

One of the tests I did a few years ago was with Java compilation with a big project. I tested it with an old and slow 80GB hard drive, tested it with a more modern 1TB drive and then an Intel G1 80GB SSD and finally a Vertex 2 120GB. While there was a definite improvement going from the hard drives to the Intel SSD going from that very early SSD to a much faster one made no difference in performance. But when I varied the CPU clock speed I saw linear improvements. But the overclock didn't help as much on the hard drives, it was far from linear although there were still improvements. Basically the SSD had improved 4k performance so much that it was no longer the problem and compilation was dominated entirely by clock speed and CPU performance (single threaded unfortunately).

So it might very well be the case that the snappiness improvement people see from SSDs is relieving the bottleneck from the hard drive but then more SSD performance doesn't directly translate into actual application performance. You can see that in the non artificial benchmarks most sites do. If you look at the pc application test suite the differences are minimal between the different SSDs, so while file copy improvements OS startup and application startup doesn't, and its because the SSD isn't the problem, its the CPU.

I see people RAIDing SSDs and I often wonder whether the application they are using it on benefits. It improves multiple user access throughput, it improves sustained transfer throughput but it doesn't really improve most desktop uses much if at all. You either need a lot of parallel reads or a lot of sustained reads to make it worth while, and I don't personally have anything else that can do 500MB/s like my SSD can, if I copy files over the network I am entirely dominated by the cable, even the hard drives in the NAS can do 200MB/s+ reads and writes but the interface + software overhead has me at about 100MB/s. So its a really specific class of problems that benefit from RAID and if you aren't doing one of them its kind of a waste of time.

Thank you so much for this detailed explanation. Makes perfect sense.

The only benefit I see with RAID is the ability to club my 3 1TB SSDs into one large or 2 large partitions as to not need to split the data across multiple partitions.

I am gonna reinstall Windows 7 Pro in a bit I am not happy with my Windows 8.1 Pro's performance, every single benchmark is lower than Windows 7, everytime I restart, due to the fake fast startup which everyone brags about, yes you get to the desktop faster, BUT services and drivers then start to slowly load. That makes my WIFI take from 30 seconds to 1 min just to connect but in Windows 7 it connects right after the reboot.

 

[BrightCandle]

I am gonna reinstall Windows 7 Pro in a bit I am not happy with my Windows 8.1 Pro's performance, every single benchmark is lower than Windows 7, everytime I restart, due to the fake fast startup which everyone brags about, yes you get to the desktop faster, BUT services and drivers then start to slowly load. That makes my WIFI take from 30 seconds to 1 min just to connect but in Windows 7 it connects right after the reboot.

You can adjust the delayed service starts if you like in the services section of the control panel. There is an option for start (delayed) and just start and that may very well improve your network interface if its currently set to delayed. Worth a try before you reinstall. Benchmarks aside I find Windows 8.1 is faster overall in actual feel, the multimonitor is better and some games really seem to benefit from it (BF4). The delayed startup is so you get to a usable desktop faster from turn on, and it does for me boot a lot faster and suffers from none of the awful early boot performance problems we saw in Vista where it gets bogged down with all the immediate service loads. Even Windows 7 delays service loads to compensate for the problem Vista introduced.

Really the last time you got good performance after boot was XP, ever since then more and more of the parts of the OS are being delayed to allow fast boots and a usable desktop when you get there. But it also means for benchmarking you do need to wait a little while for the OS to be done in the background, doesn't mean the OS is slower just means for normal day to day use its better.