Samsung 840 Evo 1TB RAID vs. RAPID Benchmarks

G73S

Senior member
Mar 14, 2012
635
0
0
So here are benchmarks of my two 1 TB Samsung 840 EVO SSDs, both in RAID 0 and in RAPID. The 4K speeds are horrible in RAID 0, how come???

What do you guys recommend me to choose now out of the below setups?

AS SSD Benchmark with IRST 12.9.0.1001 in RAID 0 (W8.1)

zui4xu.jpg


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

1zzrfiq.jpg


Not very happy to be honest. those 4K speeds are horrible

Previous benchmarks with RAPID, 4K is way better:

AS SSD Benchmark with IRST 12.9.0.1001 (W8.1):

28lqc0p.jpg


AS SSD Benchmark with IRST 12.8.0.1016 (W8.1)

24nibdx.jpg


==========================================

Windows 7 Benchmarks:

AS SSD Benchmark with Intel Chipset Drivers 9.4.0.1027 (W7) [RAPID]

21j3k9e.png



AS SSD Benchmark with IRST 12.0.7.1002 (W7) [RAPID]

el5zyp.jpg
 

G73S

Senior member
Mar 14, 2012
635
0
0
From the above benchmarks, seems like RAID sucks compared to RAPID especially in the 4K writes. How do people say RAID 0 is the best for performance? maybe, but not for Evos I guess

I think I may need to do another format and undo the RAId and just stick to RAPID
 

Hellhammer

AnandTech Emeritus
Apr 25, 2011
701
4
81
AS-SSD uses queue depth of 1 for 4K testing, which does not scale with RAID 0. You need a parallel workload (i.e. either large transfers or high QD) to benefit from RAID 0 because otherwise the IO can't be split between two drives.
 
Feb 25, 2011
16,969
1,600
126
RAPID is using memory caching. So yes, your system's DDR3 RAM is faster for random writes than your SATA controller. To say nothing of the kind of optimizations you can do with a RAM cache (like "all writes are secretly sequential writes.")

It doesn't translate into real-world performance. Really. Run some database benchmarks or something and see.
 

BonzaiDuck

Lifer
Jun 30, 2004
16,054
1,682
126
I wouldn't have a hard and fast rule to level against someone else's practices and preferences on this. I've had many RAID0 implementations over the course of several years. Not one of them has failed. But we pretty well know the consequences if one disk goes south.

Enter the SSD with its greater life expectancy or reliability over HDDs. RAID0 seems more attractive. You may get the TRIM feature now, even for such a RAID configuration of two drives or more, with recent boards, and/or BIOS updates plus latest versions of Intel IRST.

Somehow I lean toward the RAPID option for one drive, knowing the other will perform at maybe half the speed but otherwise at spec. But that is also because I've so far chosen to purchase and use one SSD.

You can get a lot of programs and data on a single 1TB EVO. Whether or not you even need a second EVO is your own preference or perspective. So that logic extends to having a single 2TB RAID0 volume. You would still have the spec performance for the second drive in AHCI mode, and something of an edge for the first drive with RAPID.
 

Fernando 1

Senior member
Jul 29, 2012
351
9
81
Recently I have done a lot of benchmark tests with my Z77 desktop system after a fresh install of Win8.1 x64 und using different Intel RST drivers/OROM combinations.
These were the SSDs I used:
  1. 2x256 GB Samsung 840 PRO SSDs as RAID0
  2. 1x512 GB Samsung 840 PRO SSD in AHCI mode with/without RAPID
Here are the best results I got:

A. RAID0 Configuration (2x256 GB Samsung 840 PRO)
w2473v1ge9v7.png

Notes:
OS: Windows 8.1 x64 (fresh install)
System Drive: 2x256 GB Samsung 840 PRO as RAID0 (Stripe size: 128K)
RAID driver: Intel RST v11.2.0.1006 WHQL
Intel RAID ROM: Intel RST v11.2.0.1527

B. AHCI Configuration (512 GB Samsung 840 PRO) with RAPID mode
rmtegr5vhe9q.png

Notes:
OS: Windows 8.1 x64 (fresh install)
System Drive: 512 GB Samsung 840 PRO
AHCI driver: Intel RST(e) v13.0.2.1000 WHQL

C. AHCI Configuration (512 GB Samsung 840 PRO) without RAPID mode
nmnly8sivahn.png

Notes:
OS: Windows 8.1 x64 (fresh install)
System Drive: 512 GB Samsung 840 PRO
AHCI driver: Intel RST(e) v13.0.2.1000 WHQL

Remark:
All my benchmark test results and an evaluation of them can be found here: http://www.win-raid.com/t362f23-Performance-of-the-Intel-RST-RSTe-Drivers.html
 
Last edited:

BonzaiDuck

Lifer
Jun 30, 2004
16,054
1,682
126
Boiled down to the basics, this isn't too much different than what is shown by the OP. Even so "good work!"

It seems no less reasonable to be skeptical of RAPID than to embrace it, and vice-versa.

Since even 16GB of RAM will be more than what is needed by many enthusiast systems, it would seem that the 1 to 2GB used by RAPID wouldn't pose any conflict or competition for resources. I cannot say what scenario we face with half that RAM. It's probably not a significant loss of available RAM, but someone else can assert conclusions about it.

That leaves the crucial (no pun intended) question. "Is it stable and reliable?"

IF it proves stable and reliable, then a choice to use it only depends on whether users have any sense that there is some slight improvement over the straight specification read/write scores of the SSD without RAPID.

There is also the notion of a "placebo effect," whereby someone who deploys RAPID might tend to "imagine" that there is improvement.

I already said so: I have convinced myself that there is a noticeable though slight improvement in SOME operations and desktop usage. I also agree that I MIGHT be subject to the "placebo effect" in this judgment.

But -- if RAPID proves to be STABLE and RELIABLE, it might not much matter. Your choice to use it is either harmless or marginally beneficial; your choice not to use it would be no less in either respect.

Barring the cost of a standalone 500GB SSD versus the cost of a 60GB SSD and a 500GB HDD, I suggest we can be more assured of a different judgment: the large SSD alone or with RAPID is better than ISRT with the configuration of SSD and HDD. At this point, I also think it may be more stable and reliable.
 

G73S

Senior member
Mar 14, 2012
635
0
0
Boiled down to the basics, this isn't too much different than what is shown by the OP. Even so "good work!"

It seems no less reasonable to be skeptical of RAPID than to embrace it, and vice-versa.

Since even 16GB of RAM will be more than what is needed by many enthusiast systems, it would seem that the 1 to 2GB used by RAPID wouldn't pose any conflict or competition for resources. I cannot say what scenario we face with half that RAM. It's probably not a significant loss of available RAM, but someone else can assert conclusions about it.

That leaves the crucial (no pun intended) question. "Is it stable and reliable?"

IF it proves stable and reliable, then a choice to use it only depends on whether users have any sense that there is some slight improvement over the straight specification read/write scores of the SSD without RAPID.

There is also the notion of a "placebo effect," whereby someone who deploys RAPID might tend to "imagine" that there is improvement.

I already said so: I have convinced myself that there is a noticeable though slight improvement in SOME operations and desktop usage. I also agree that I MIGHT be subject to the "placebo effect" in this judgment.

But -- if RAPID proves to be STABLE and RELIABLE, it might not much matter. Your choice to use it is either harmless or marginally beneficial; your choice not to use it would be no less in either respect.

Barring the cost of a standalone 500GB SSD versus the cost of a 60GB SSD and a 500GB HDD, I suggest we can be more assured of a different judgment: the large SSD alone or with RAPID is better than ISRT with the configuration of SSD and HDD. At this point, I also think it may be more stable and reliable.

I wished it was a placebo effect.

But in reality, live example

Backing up a system image using Macrium Reflect Pro (source C: / Desitination: d :)

Using RAID = 170 MB/S
Using RAPDI = 450 MB/S

so in real world usage, RAID 0 was much worse for me

I am formatting now and going back to my shiny RAPID FTW
 

npaladin-2000

Senior member
May 11, 2012
450
3
76
The way RAPID works, unless you overwhelm the 1 GB RAM cache, it's absolutely going to be faster. As was mentioned, there are some things that WILL overwhelm it, particularly SQL DBs, but those will also be able to take advantage of RAID parallelism anyway.

RAPID is an excellent performance booster for most situations though. Though if you have enough RAM a RAM disk can be made bigger than 1 GB. That's a lot more complicated to implement though.