Procedure
The first set of benchmarks were run without doing any modifications on the Velociraptors with the exception of RAIDing them together. This is done in the bios of the motherboard. The settings that I selected for the RAID were RAID 0 with 128 KB stripe size.
Second set of tests were doing with a mod similar to the one in the Seagate Mod Article we published some time ago. After deleting the RAID partition and resetting the drives, I proceeded to downsize the drives to about 1/2 their original capacity – approximately 300 GB give or take. This was done with a software from Seagate (Seagate Tools) DOS Version (not Windows Version).
Steps to short stroke your drive:
- To do this you first have download the Seagate Tools which can be found here.
- After downloading the .iso burn the iso onto a CD with any program.
- Insert the CD into the your CD/DVD/Blu Ray/..etc drive and restart your computer.
- Assuming your system is configured to look at your drive before loading the Operating System, you should see a window that lists all the drives in your system.
- Select the first Velociraptor drive and go to the Toolbar>Advanced>Set LBA number. The software will tell you the maximum amount of LBAs your drive has, whatever that number is = MAX advertised capacity of the drive in the case of the Velociraptor that number is ??? this equates to about 600.1 GB. Dividing LBA number by 2 will yield our desired hard drive capacity (~300 GB).
- Enter that number into the box and press enter.
- For the second drive repeat steps 5 and 6.
Testing
I used the following benchmarks in for all the tests I ran on the drives:
- HDTune Pro 4.6
- CrystalDiskMark
You guys missed the fact that the Patriot Torx 256GB SSD does not even exist.
The rigth way to compare was to use a 256 MB on the HDD (128 Mb from each HDD).
Anyways, it should be compared against a decent SSD (No Patriot SSD).
That would favors the HDD by short strocking, increaing the 4 Kb random speed, which is the more important statistic.
Clearly this article was written to generate some buzz, comments, links, etc to improve the websites standing in search engines. The more the better. Great SEO imo…
Wow. The amount of effort expending while missing the point here is staggering. You wrote
“While random reads and writes (4 KB) segments are important, most of the information on the hard drive is stores in clustered blocks.”
This is a stretch, any data to back this up? Any real world benchmarks you’d like to share? I think one chart of PCMark or similar would make the reality apparent.
The raptors ultimately are slower, 10x to 20x slower than the raptor raid. I’m talking about the 4k reads/writes. An SSD could boot up an OS or load a game 10x to 20x faster than that raid that you spent almost the same amount of money on.
Do some real world benchmarks. Time how long it takes to boot up windows or linux OS, how long it takes to load a level in F.E.A.R. or favorite game, and how long it takes to install a virtual machine or the OS on the system.
You’ll find the raptors are painfully slow compared to the SSD.
This article is highly misleading! A bunch of synthetic benchmarks focusing on raw throughput favour RAIDed magnetic storage? Who’d have thought?!
How about you show me some real-world benchmarks? Windows startup time, gaming, Photoshop, databases, app installation — anything that actually benefits from solid state storage.
This article completely misses the entire point and usage model of an SSD.
Sure, if you tailor the tests, sticking to sequential only and leverage cost per gb as you metric you can make a set of hard drives look better than an SSD – but what are you going to use this array *for*?
Assume you’re using the pair of raptors or SSD for an OS drive. In this case, You firstly don’t need to be using a costly 256GB SSD, a 120/128GB will more than suffice (use a 2TB WD Green or Black to store all your big files). One of the OCZ Vertex or Vertex 2 drives can be had for around or just over $200; less than the cost of a single raptor. In this usage model, 4K performance (especially writes) rules, and in this case a single SSD will outperform even your RAID of raptors by at least 3-5x.
Assume instead you’re not looking for OS performance, but large file storage; in this case, the extra cost of the raptors over more standard hard drives is unjustified, and you will attain 2/3 or more of the performance of a raptor drive at a fraction of the cost.
Who are you trying to say this particular array is for? For what purpose are you going to use such large sequential reads and writes for, completely ignoring the fact that you could instead RAID two 128GB SSDs together for lower cost than your 256GB drive and then soundly trounce your raptor array in all metrics?
Please do some more research before publishing such a misconceived article.
Only the sequential transfers is faster…where is your killer???
???
Who did this review? This is ridiculous.
Your first benchmark comparing the raptors vs the SSD is misleading in the fact you are using different file lengths. I’m guessing to make your “ssd killers” look reasonable and worth the write up of the review. MISLEADING.
As well as when your raided raptors are @ 6+ms access time, the SSD has another 6ms to catch up and perform multiple extra read and write cycles to destroy the “difference of speed” you are seeing in the benchmark which is already hardly negligible as it is.
SSD is about insanely fast read speeds coupled with hardly any access time reading and writing files.
It is possible to get HDD close to performance of a SINGLE SSD, but it will never outperform the access time, which is one of MANY advantages of the SSD compared to HDD.
Nice try, but this is far from a “How to build an SSD Killer with traditional HDD.”
Give the money and free products involved in this project to me and I’ll provide your site with something a real technically minded person would enjoy and be intrigued to read.
I believe this article is misleading. Raptors in RAID 0 are not as fast as SSDs in anything but sequential transfers. Additionally the Patriot Torqx drive as of 12/29/2010 can be outclassed by drives from every manufacturer including the Patriot at the same price point or less.
I like the Raptors in RAID 10 in file and VM hosts but they do not fit into the personal computer landscape well anymore. When you start talking system responsiveness, application load times, installs, updates, virus scans, high IO DB servers, etc there is a better tool. Instead of Raptors for sequential stuff I’d go with 2TB 7200rpm drives and double the spindle count for the same cost.
While an SSD does cost more for considerably less storage there is absolutely no doubt that it holds the performance crown. Just look at the 4K QD32 reads – the RAID setup manages 5 and the SSD manages a whopping 65, that’s 13x faster. Then look at the seek times, which determine how responsive the system feels – the RAID achieves 1.32ms and the SSD 0.089, that’s 15x more responsive.
The RAID setup may score higher peaks but the SSD will be noticeably faster in real world usage. That’s without even taking into account the noise and heat generated by the Velociraptors, versus the silent and cool running SSD.
This article may be well intentioned but it is seriously off base.
dont forget the power required to run one SSD ~ 2 watts versus the power to run two velociraptors ~ who knows how much.