Bechmark HD Tests – Dominick V. Strippoli
To hear some of the results from other review sites about how one Raptor is better than 2 and a RAID configuration makes no noticeable difference, I decided to do my own testing to prove otherwise.
I own the best video card on the market right now, basically the fastest retail dual core AMD processor on the market besides the FX-60 (x2 4800+ @ 2.6 Ghz), 2 gigs of RAM with excellent overclockability – basically a top of the line configuration, if you want to call it that.
The only thing bottlenecking my computer was the hard drive. Heck, the only bottleneck to most non-RAID systems today is the hard drive. I am the individual that wants the maximum possible performance out of my PC even if it means spending the extra amount of money on 2 Raptors for a RAID array, or buying the latest and greatest video card every 3 months.
I just couldn’t believe the “one Raptor is better than 2” argument to be true. So what did I do? I went out and I ordered my first Raptor. I was a little disappointed with my single Raptor results. In actuality, my old SATAII 250GB drive performed the same way in real world usage.
Here is a benchy of my first single Raptor yesterday:
Being crazy about PC performance, I was disappointed with my single Raptor. So what did I do? I ordered another Raptor and I just finished setting up a RAID 0 configuration with 16k striping. I want to put the Anandtech.com: “RAID is not meant for the desktop” statement to rest or at least challenge it.
I am very anal retentive when it comes to my PC not performing the way I like, as you can see with my impulsive second Raptor purchase. I also understand that the review site in question did not factor in a lot of variables during that testing, including the various different serial ATA controllers that all perform at completely different levels, some differences are very substantial.
My own example is that my Asus A8N-SLI motherboard has 2 controllers with 4 SATA ports on each controller for a total of 8 SATA ports. The first is the nVidia RAID controller with a maximum of 300 MBps, and the Silicon controller with a maximum of 150 MBps. We are basically looking at burst rates or theoretical maximum transfer rates when comparing maximums on the controllers, but it was proven that various controllers definitely affect sustained transfer rates, and obviously burst rates. So by me using the silicon controller, effectively speaking I could have downgraded the performance and not reached the full potential of a Raptor striping array.
The first thing that I noticed after the RAID-0 setup was the Windows XP installation speed. The computer performs tasks so rapidly that I find myself turning my back for a few seconds and realizing that a task is completed. The Windows XP installation, not counting formatting, took approximately 10 to 15 minutes. LOL! It was ridiculous. On my WD2500, this would take around 30 minutes.
The next thing I noticed on my PC after I was completely done with the XP Professional install and completely configuring all of the hardware drivers and standard software was the shutdown and reboot.
This next feat is incredible: After the Asus BIOS flash screen pops up and the RAID0 array is loaded, obviously everyone knows that the Black Windows XP loading screen comes up. With my new Raptor RAID setup (All Kidding Aside), the Windows XP boot screen flicks on and off for a total of 1 second. It takes about 2 seconds to load Windows XP Professional to the login prompt. Everything in the OS environment feels fluid and instantaneous. All of you doubters about RAID on the desktop really have to try it for yourself.
I was skeptical at first, but the performance difference realized on screen was like going from a Pentium-III 800 Mhz to a Pentium-4 3.2 Ghz processor. It was just incredible. Remember, this is coming from a guy who thought his single Raptor drive purchase was worthless (and I still do). Fear and Quake 4 shaved off seconds on load times and the PC in general just feels snappy. It’s the way I always wanted my computer to feel but never realized how I would go about attaining it. Well, I finally realized it.
This may sound crazy to some of you, but a simple addition of 2 Raptor 74 Gig hard drives in RAID 0 has put my PC up to a new level of performance. I hope you can appreciate the pure power of these drives combined.
On to the benchies:
My first HDTach 3.0 benchmark with the 2 X Raptor 74G RAID 0 array:
An average read of 129 MBps is just incredible. I doubled the performance in sustained transfer rates over a single Raptor. I think the most important thing to note is that from 0 to 27 Gigabytes, the Raptors maintain a sustained rate of over 140 MBps. Considering the fact that I do not use more than 27 Gigabytes on the OS 40 GB partition, my average throughput should be around 140 steady.
Another thing that I noticed was that random access time did not even flinch. The other review site in question states that you will lose a load off of random access time when moving to a RAID 0 setup. My times stayed identical from one drive to the next. Again, you see the burst speed of 217 MBps. If I was using the Silicon controller, I would have been held back to only 150 MBps technically bottlenecking and limiting my RAID setup.
Here is a direct comparison of a single Raptor vs my 2 Raptors in a RAID configuration. As you can see, I have doubled the performance spec and capability of a single Raptor and my random access time has remained the same:
Finally, my atto results:
In the end after everything, I definitely recommend RAID to anyone with a single Raptor. If Anandtech.com said that RAID is not for the desktop, I say this:
RAID is not for the office professional or “stay at home mom” desktop. RAID is for the desktop user that wants the maximum possible performance squeezed out of his machine – the overclocker, the hardware fanatic, the geek.
Theoretically, even if my 2 Raptors performed exactly the same as a single Raptor and displayed a 2% to 5% improvement over a non-RAID system, that 2% to 5% means a heck of a lot to a person like me who wants maximum performance out of my PC. If that was the scenario and it was a small gain, was it worth the money? To some of you, it isn’t. To me, it is.
My simple conclusion would be:
1 X WD2500 = Baseline SATA hard drive. Good performance, 7200 rpm
1 X Raptor = No noticeable performance increase, or very little.
2 X Raptor (RAID 0) = Incredible gains realized, noted in every aspect of PC usage.
DISCLAIMER: Please note – this is based on an opinionated review merging synthetic benchmarks with “real world” analysis based on my personal opinion. Your results may differ depending on your mobo/card controller, Raptor 74G model numbers, and/or your technical computer configuration. I made my best efforts to try and not be biased whatsoever towards the RAID configuration. My best advice is this: If you are skeptical on the so called “Real World Performance”, my only recommendation is to try it out for yourself.
Previously in Part 1 of this article, I explained in a more general and non-technical aspect that I had gained a very sizable performance increase when switching from one single Raptor drive to two drives in RAID 0. This increase was notable in almost every aspect of PC usage. I ran some synthetic benchmarking tests including HDTach 3.0 and IOMeter and discovered that there was a double in synthetic performance increase when switching to the RAID array and my random access time had remained the same.
I appreciate all of the e-mail feedback that I have received on my Part 1 article and this Part 2 is based on a lot of the information and testing that many of you requested that I do. One of the main issues was the synthetic testing that was done in Part 1 of the article. We all know that synthetic benchmarks for HDD’s mean little in comparison to how the actual drives perform in real world usage in a single user environment. Most synthetic benchmark apps are tailored for server level I/O’s and although in my Part 1 you saw a double in average sustained read, in actuality that may make the RAID array look significantly better in a single user environment then it really was.
Based on that, I decided to rip apart the array and start testing from scratch using an old fashioned stop watch timing technique. Again, this comparison is between a single WD74 Raptor vs. dual WD74 Raptors in RAID 0 with 16k striping.
Preliminary System Specifications:
- CPU: AMD Athlon X2 4800+ (2.4 GHz)
- Motherboard: Asus A8N-Sli Premium
- Video Card: ATI X1900 XT
- Memory: Corsair XMS 2 Gigabytes PC3200 (DDR400)
- Power Supply: Silverstone Zeus ST56ZF
The actual timer used was a Casio hand held stop watch. The testing procedure was done exactly the same for both setups. I had all testing tools and applications on a WD2500 250 Gig Serial ATA II hard drive left installed in the system.
First, the Windows XP Professional with SP2 operating system was installed on the drive/s. Immediately into the operating system, the Nforce4 motherboard drivers were installed. Direct X 9.0c was installed, Athlon Dual Core Processor Drivers and XP Dual Core Hotfix were all installed, sound card drivers and finally the catalyst video card drivers. The rest of my testing applications were then installed and all of my read/write rar, zip, avi, and jpg files were copied over from the WD2500 to the current setup.
Immediately following final copies and installs, the setup was defragged twice and the system was restarted and prepared for initial testing. Again, this was repeated exactly the same on the single drive and on the dual RAID drives.
*Please Note* This article was completed using a hand held stop watch timer and although my results are very close to being dead on, you always have to factor in a slim margin of error. Especially when it is physically impossible for the human brain to make completely accurate timings. My results are as close as humanly possible.
Windows XP bootup was where the RAID array really displayed a dramatic increase in loading performance, as I stated in Part 1 of the article. These times were measured over a span of 10 shutdowns and startups between application benchmarking on both of the setups. The average over a span of 10 bootups and shutdowns was recorded into the data field and chart that you see above you.
For the application loading tests, I restarted the computer before each application load. Photoshop obviously displayed the most dramatic increase in loading time over the rest of the applications, but the RAID array proved to definitely increase the snappy feeling of application and file loads. The stop watch was clicked on as soon as the program was double clicked from the desktop and clicked off as soon as the application ceased loading. Each application was timed, and the PC was restarted twice to get two separate readings. The reading you see on the chart is the average reading of the two on both setups.
Gaming load testing was generally completed the same way as previous application load testing except for the fact that only one timed reading was taken from each game. The RAID array, to my suprise, shaved off more than just a few seconds on these games. Ten Seconds off a game load is rather significant, in my eyes. Again, I was not expecting these kind of gains from the RAID array. The games were timed from the first level loading screen and the computer was re-booted between each game. One timed reading was taken from each game on each setup.
Falcon 4.0 Allied Force was chosen for its heavy loading impact on any gaming rig. I took one reading from loading the game from the Windows XP desktop and I took my second reading loading the Flight Training Takeoff level. The same procedure was used on the RAID array. Again, you are looking at one timed reading from this game on each setup.
For the file copy read/write test, four individual AVI video files were present on the setups after our initial copy from the WD2500. The files lay dormant on the drive until the testing. Each file was simply copied/duplicated directly from its folder on the same drive to the Windows XP desktop. One reading was taken and the computer was re-booted in preparation for the following file size. The process was repeated until all 4 avi files were timed. As you can see (especially with a larger file), the RAID array performs significantly better.
The File Copy Intensive Test was just as the title explains. A folder containing 4.31 GB of mixed graphic images, video files and executables of all different file sizes was simply copied/duplicated from its dormant folder on the Raptor/Raptor RAID drives to the Windows XP desktop. One time measurement was taken. As you can see, with a file intensive task the RAID array obviously performs much better than the single drive.
The Winrar extraction test was taken using a rar file that had a compressed amount of jpg picture files. The file had 271 MB of compressed images. The test was timing how long it took for the computer to extract these jpgs from the rar file. Three timing measurements were taken and the computer was re-booted between each extraction. You are viewing the average extraction time between the three runs.
The installation to drive (from drive) test was timing how long it took to install 3D Mark 2003 from an executable on the desktop of the single drive and RAID setup. The RAID array improved executable installation time dramatically. Only one timed reading was necessary on this test. The timing was started as soon as the OK button was depressed in the setup program and was finished after all of the progress bars had come to 100% and finished.
PCMark 2005 is a program that measures the complete performance of every aspect of your computer. As per the software developer:
“PCMark ®05 is everything you need to reliably and easily measure the performance of your PC and determine its strengths and weaknesses.” (Futuremark.com)
Measuring the performance of my PC with the single Raptor installed netted a score of 5876 marks. By simply installing the Raptor RAID 0 array, I increased my PC Mark score to 6235. If you are familiar with Futuremark and this benchmarking tool, you will know that a gain of almost 400 PC Marks is substantial from simply adding a RAID array.
The total time to complete the PCMark05 benchmark was taken from start to finish and only one reading was necessary. The RAID array knocked off a full minute vs. the single Raptor.
Sisoft Sandra is another synthetic benchmarking application for measuring hard drive performance. The comparison is between the Drive Index Rating, as per this software. Almost a 50% improvement was netted from the RAID array over the single drive. The benchmark was completed one time on each setup and the results are displayed in the chart.
One measurement was taken for the above chart. This test measured how long it took the computer to complete the Sisoft Sandra Filesystem “Drive Index” test. The test was simply timed from start to finish.
HDTach3.0 is another synthetic file system benchmarking application. You are viewing the results of the average sustained read betweem the two Raptor setups. Its no secret that the RAID array doubles the sustained transfer rate in these synthetic benchies.
Here you see the unchanged Random Access time between the two setups.
The following testing was done for a placebo effect. I wanted to see if a RAID array had any positive or negative effect at all on software during highly dependent RAM/CPU priority situations.
The Gaming Frames Per Second measurement displays both the single drive and the RAID array performing exactly the same in general gaming. For testing, each game was set at 1024 x 768 with 4XAF. Softshadows was disabled in FEAR. The built in FPS measurement tool was used for recording FEAR benchmarks and the HOC Quake 4 tools were used to record FPS for Quake 4.
This is a placebo test measuring how long it takes a CDROM drive to install Pinnacle Studio 9 to the test rig. As everyone would be expecting, there is very little to no noticeable difference in install speed between the RAID array and the single drive. The CDROM drive is the bottleneck or “placebo” in this testing, so to speak.
As per the manufacturer:
“By combining high quality 3D tests, CPU tests, feature tests, image quality tools, and much more, 3DMark05 is a premium benchmark for evaluating the latest generation of gaming hardware (Futuremark.com).
3DMark05 is one of the synthetic benchies that is known to be heavily dependent on the video card, rather than the CPU and RAM. For obvious reasons, this software was also chosen to be part of the placebo effect for this article. As you can see, the scores remained exactly the same from a single drive compared to a RAID array.
Our last placebo test is heavily dependent on the processor and RAM. I simply compressed 512 MB of jpg picture files into a RAR file using Winrar by Rarsoft, Inc. Only one measurement was necessary. Timing was simple and measured from start to finish of compression. The last placebo test definitely holds true, as the hard disk should have little to no effect on compression time.
As you can see, using my individual PC components including processor, motherboard, and on board RAID controller:
Two Raptors in RAID was definitely worth the money for me personally.
Timed data concluded only what I already believed to be true about how much better the RAID array was performing on my computer vs. the single Raptor. When I was reading my feedback about Part 1 of this article, a reader sent me an e-mail a week or two ago and pointed out that in a RAID 0 configuration especially, any minor difference in motherboard, processor, or RAID controller, including the way you actually setup the striping on your array, is going to give you completely different results from mine, or anyone else for that matter.
That is a great point, and although it is a simple a statement as it sounds, it is vastly important to the end results in this article and should definitely play a part in your own consideration of building a RAID setup for yourself.
This Part 2 article gave me a completely different outlook on RAID and whether or not it really should be for the desktop user. Factoring in failure possibility of the array and monetary capability of the majority of readers, I really don’t recommend a RAID array for the average desktop user.
But there is a catch to that last statement:
As stated previously in Part 1, I recommend “RAID for the desktop user that wants the maximum possible performance squeezed out of his machine – the overclocker, the hardware fanatic, the geek.”
That is exactly where I stand on this subject. As far as a single Raptor vs. a Raptor RAID setup. There is now currently no doubt in my mind that two 74 G Raptors in RAID 0 perform much better in a single user environment than the individual 74 G Raptor. Be prepared in the coming months for a Part 3 – I plan on introducing the new 150 Gig Raptor X into the mix.
Dominick V. Strippoli (aka “Dominick32”)
Check out my site on muscle cars SVT Snake.com