How to install Win7 to an SSD Raid array?

[Mergatroid]

Sorry man, I just tested it for a decent improvement.

My SSD was spanking the hard drive (250ish MB/s is about correct for this generation SATA II SSD) vs one hard drive. I said my buddy's RAID 0 (4 drives) was getting the same performance. Now my SSD RAID 0 is not only spanking my old single SSD, but it's SUPER spanking a single hard drive and out performing my buddy's mechanical RAID. I'm afraid I'm not the one with issues my friend. I have been studying everything I can find regarding SSD RAIDs, bought another 60GB SATA II SSD and set the drives to RAID 0 in hardware on my ASUS P8Z68-V Pro mobo. Worked first time, without issue. Installed Windows, have been reinstalling all my apps. Working great. Boot time is now just over 10 seconds from just over 20 seconds on the single drive (after POST). My average transfer benchmark with HD Tune (read) has gone from 246 MB/s on a fresh Windows 7 install on a single Patriot Inferno 60GB SSD to 461MB/s using the Inferno and a Corsair 60GB in a hardware RAID0 striped volume also on a fresh install of Windows 7.

If you go to Tom's Hardware you will find lots of people using SSD RAIDs, more on other sites as well. You can stand by what you say, I'll stand by what I have tested. Of course I can't say how much of a performance boost you would get by RAID0ing two SATA III SSDs. Because of their high transfer speeds a RAID may run into SATA limitations. I can't say for sure because I haven't tested that. Here's a screen shot of my current HD Tune benchmark.

 

[whs]

Nice nums but are they really worth the complications - and cost. My 90GB OCZ Vertex2 boots in 16 seconds - in a laptop. What else do I need.

 

[Mergatroid]

It's a keeper

The cost was less than adding a newer faster drive, at the expense of having to learn more about RAID and how to set it up. I had some experience already, so I wasn't too bad off. There were no complications. I plugged the drive in, set the BIOS, created the volumes, booted windows installer and from that point on everything was a normal Windows install. Really no more difficult than replacing your boot drive. Kudos to my new ASUS main board for working right the first time. Of course there was additional backup work as a lot of things needed to be backed up manually since you can't restore your old SSD image to the new RAID volume. Anyone who cannot seem to get the RAID working using the BIOS and hardware RAID on their main board should try updating their BIOS. This ASUS board was a pleasure to set up. Easy as pie.

I would agree that a person who's not an enthusiast may be better off not bothering with RAID because of the added complexity, but it does offer a performance improvement for SATA II SSDs in RAID 0. RAID1 on a couple of SSD storage drives could provide added redundancy just as it does with mechanical drives. I could see a nice little NAS holding four 2.5" SSDs in a mirroring RAID setup. If I could afford to go out and pick up a couple of SATA III SSDs I would definitely try RAID0 on them just to bench them and try a few things. If I could see an obvious improvement I would keep them in RAID.

 

[DeaconFrost]

If you go to Tom's Hardware you will find lots of people using SSD RAIDs

That site has been laughable for years now, at least on the real, true enthusiasts boards, like [H]ardForums and Anandtech. Once "Toms" was exposed as giving positive reviews and swaying forum members opinions in exchange for cash, the integrity of the site left in a hurry. It is hardly an enthusiasts site anymore. You also are putting all of your faith in synthetic benchmarks, and as the [H]ardForums debunked 5-6 years ago....doesn't show accurate and real performance for RAID arrays. It was a simple fad that was popular when SSDs came out, but it has since died off in the enthusiast community. Then it was determined to be all hype, just like RAID0 on spinners. That is why you'll see the high end manufacturers of boutique and custom PCs typically shy away from RAID, unless it is specifically requested. For a while, the only real purpose was to make up for small drive capacity....but now that SSD prices have been creeping down, that isn't the case. My new Agility 3 240 GB was cheaper than two 120 GB models of the same drive...and doesn't give me all the instability risks of a RAID0 array, while losing nothing in performance.

 

[whs]

Absolutely right Deacon

 

[Mergatroid]

I'll stick with what I see with my own eyes, thanks anyways. I see the improvement in boot times, and overall performance. Sure, the benchmark won't hold true all the time, and it IS an average read speed with peaks and valleys but the fact is that performance has improved. Argue all you want but the numbers don't lie.

Lets see your measurements showing how it doesn't work. I've shown one of mine. I supply proof and you supply...opinion.

 

[DeaconFrost]

I supply proof and you supply...opinion.

You're missing the point. The proof has been out there for years supporting my "opinion". I'm telling you what is the "accepted" answer. You are showing "proof" using synthetic benchmarks, which isn't actual proof to begin with. If you need me to explain that part, then that tells me you missed the great RAID debates from years back.

The bottom line is, what you choose to do with your system is completely up to you. If you feel you are getting some kind of benefit, go right ahead. However, you need to be very careful over what you actually post for others to read. The proof you seek is out there, and online. Anandtech has the definitive write-up that's considered the final word in the enthusiast community on RAID in a desktop, specifically striping. SSDs have eliminated the need to even consider RAID.

 

[whs]

Deacon, I agree with you in general. I would never stripe SSDs for an OS installation because it makes no sense for the particular R/W pattern of the OS. And since the access time for a Raid is no different than for a single SSD, there is very little, if anything, to be gained.

This may change the day we see large SSDs at reasonable prices that we can use to store large amounts of data. If you move GBs of data around, striping may very well speed things up.

And there is another exception (at least it seems to be). That is the OCZ Revo drive which is a 4 way Raid implemented in the hardware controller. I do not have one of those devices but I remember when Shawn installed one on his system earlier this year, he reported a significant performance improvement over his single SSD (which I assume was a Sata II).

Since then we have seen Sata III SSDs appear and that narrows the gap again.

 

[Mergatroid]

I supply proof and you supply...opinion.

You're missing the point. The proof has been out there for years supporting my "opinion". I'm telling you what is the "accepted" answer. You are showing "proof" using synthetic benchmarks, which isn't actual proof to begin with. If you need me to explain that part, then that tells me you missed the great RAID debates from years back.
However, you need to be very careful over what you actually post for others to read. The proof you seek is out there, and online. Anandtech has the definitive write-up that's considered the final word in the enthusiast community on RAID in a desktop, specifically striping.

Right, in other words you offer nothing to back up what you are saying.

Here is further proof. A fellow using a hardware RAID on 2 x OCZ Agility 3 in RAID 0. He was benching 1000 MB/s.

ASUSTeK Computer Inc.-Forum- AHCI/RAID issues

Page 9.

The more I look into this, the more people I find achieving awesome transfer rates. The more I come back here, the more I see nothing but the evidence I provide. It's perfectly fine to claim there will be no real world benefit, but that doesn't explain why my boot time has almost halved (fresh win7 install vs fresh win7 install), or why the software on that raid is responding faster than it was. You say that benchmarks are no good for performance testing, but that's their entire point, to test performance. Sure, I won't see a 2 x benefit all the time, but so far it's looking like I have achieved an overall increase in performance as measured not only by the benchmark, but also software loading times (such as the aforementioned boot time). You keep SAYING there's no performance increase, yet I keep SEEING it.

As I said, I will have to go with what I see with my own eyes. I have done the testing, I have seen the performance increase, I have made my decision. You keep telling people what some other people did some time in the past, and I will show them my benchmarks and preformance time improvements.

Here's a quote from the article I linked to:

"Hello, I had (have) the gigabyte ga-990fxa-ud3. Coupled with 2 OZC Agility 3 in raid 0. It gave me just over 1000Mb/s. Bought the sabertooth 990FX, did a SE on my disks with a clean install, raid 0. Only get just above 800Mb/s and in write?! Something is wrong as, Sepher said. It's the same raid controller so obviously the problem lies within the BIOS and/or the driver. When i do something on my computer it feels a bit laggish. It did not do this on my gigabyte board!"

Obviously I'm not the only one enjoying great success with RAID0 on SSDs. Looks to me like I'm not the one missing anything.

The bottom line is that performance has improved. I really don't understand how you can argue with that. The clock doesn't lie.

 

[DeaconFrost]

I don't have to provide proof...I'm not the one going against the grain. You are choosing to only read forum posts and articles that fall into the myth. I gave you the site to find the final word, and you ignored it. If you said rain drops were made of metal, I wouldn't need to provide you with proof that they aren't.

What you are believing is the results of synthetic benchmarks that do not accurately measure and test SSDs. I've stated that before, but you've ignored it because it's factual and goes against your "beliefs". It's been common knowledge for at least 5 years that stripped arrays can often hurt regular performance due to higher seek times. That hurts regular Windows performance. Striped arrays were only ever useful in very rare situations, such as a scratch drive for Photoshop or for some unique video editing environments. That's why striped arrays with spinners died out.

Once SSDs came out, you saw people creating arrays in order to extend the drive capacity. Two 60 GB drives were often cheaper than a single 120 GB, and could be striped together for the capacity. In real world usage, the performance didn't change. Now that prices are coming down, and smaller drivers don't have the same price per GB advantage anymore, the reasons for RAIDing them together dissolved.

All of what I said above is fact, whether you choose to believe it or not. I'm not going to continue this "debate", because I still have scars from the "Great RAID Debate" from 5 years ago. You still want to put your faith into synthetic benchies, and that is blindly leading you down the same path people walked 5 years ago. But, like I said previously...what you do and believe on your own system is your own business. When we post here, our job is to give accurate, best accepted advice to people seeking it. When you start posting about how great striped arrays are and how synthetic benchies are making you feel warm and fuzzy about your array...all you do is tell us that your thinking is outdated by those 5 years. I don't say this to sound rude....I say it to give you the chance to do your own reading and researching...from authoritative sources, so you know why striped arrays have fallen out of favor within the enthusiast community.

 

[Mergatroid]

"I don't have to provide proof...I'm not the one going against the grain."

Uh, I provided a benchmark of my results, and boot times. If you want to rebut that you'll have to provide something other than you have.

" I've stated that before, but you've ignored it because it's factual and goes against your "beliefs""

Not true at all. I am the one who stated I am using benchmark software. I also said TWICE that I realize I may not get these same results with everything. It appears you are the one who ignored that even though I stated it twice. Also, if the benchmark is completely invalid then it wouldn't achieve basically the same read results as the drive manufacturers say the drives can achieve when tested individually. How did they achieve those results? They used benchmark software.

" In real world usage, the performance didn't change."

Funny, the measurement of times for loading various software, including boot times, shows you are wrong. It's funny how you say completely silly things like your "fuzzy" comment regarding HD Benchmarks and completely ignore the FACT that software load times and boot times have decreased. Please stop harping on about the benchmark software as I have stated information regarding load times and boot times several times. On top of this, I fail to see why any benchmark software exists at all if its as useless as you claim it is. I know benchmarks are, in effect, an average. Some times you will not be able to achieve those results, and sometimes you will. Overall it's still a performance improvement.

"Now that prices are coming down, and smaller drivers don't have the same price per GB advantage anymore, the reasons for RAIDing them together dissolved"

It was less expensive for me to purchase a 60Gb SATA II SSD and add it together with my current 60GB SATA II SSD and get almost the same performance as a SATA III SSD than it was to purchase a 120GB SATA III SSD.

The simple fact that you keep tiptoeing around is that my performance has improved. You don't want to admit that my boot time has almost halved, something you say can't happen. You don't want to admit that the load times for the software on my RAID have improved as compared to what it was on a single SSD. You keep ranting about the benchmark software, but that is only one of the tests I have performed. As I stated in my last comment, the clock doesn't lie. Performance has improved OUTSIDE THE BENCHMARK SOFTWARE.

I'm sorry, I don't have any further time trying to prove anything to you. I have provided my results. If you don't want to believe them, that's your prerogative. Anyone reading this thread and trying to decide if they should try a RAID or not, all I can say is I have tried it and it works. Other people have tried it, and it worked for them. I don't claim this will work for every drive, and I don't claim everyone will get it to work as well as it's working for me. I only claim that I tried it on two 60GB SATA II SSDs and I'm getting excellent results.

You? You haven't tried it at all, and claim to be the authority on if it will work or not.

Anyone reading this, if you have a nice new main board and it has BIOS level hardware RAID capability, give it a try. What do you have to lose? Theoretically you have backed up your system already. Just be aware you're likely not going to be able to restore an image of your old drive onto a RAID volume. If you are really interested in this, back up everything you need on your C drive. Configure your RAID using the motherboard software. Install Windows to your boot volume. I would recommend not having any other drives connected (outside your RAID drives) while you're doing this to avoid the brilliant Windows 7 software from installing the master boot record on the wrong drive. Install Windows normally. Once you're finished, install the newest Intel RST driver/software. It will show you your RAID. The next version is rumored to support Trim. Currently you will have to rely on the "garbage collection" on the SSDs you use. I would recommend leaving them less than 75% full for decent wear leveling.

Once you have done this, and you have installed all your main board drivers, and any other system drivers you require, you can make an image backup of your boot RAID volume. You are now protected enough to start doing some experiments of your own. First, I would recommend measuring your boot time. It will be amazingly fast. When I went from a single SSD to a SSD RAID0 I almost halved my boot time. If you were using other software such as games on an SSD previously, install one or two depending on how much space you have, on your new RAID0 and measure their start up times. You will be pleasantly surprised.

One last thing I have to say on the subject is this: Don't believe everything you read on the 'net. If someone is offering you advice that seems a little "off", then there's nothing better than trying it yourself and measuring your results. Use various methods of measurement to convince yourself you are getting accurate results. And finally, don't argue with a door knob. No matter how much evidence you give, it will never admit you're right.

If anyone is interested in discussing this further you can reach me as Mergatroid on overclock.net. Send me a private message.

cya

 

[Corazon]

Um, Mergatroid...you sure do talk a lot. You like throwing big numbers around too.

What I haven't seen you mention even once is access times. Not maximum transfer rates in MB/sec. I don't think you can show me a harddrive RAID that comes anywhere close to the typical 0.1ms of even a single SSD.

Can you?

See, the point is simply this - being able to read or write 500 or 1000 MB/sec is no good if you're not actually doing something with that speed. You won't be sitting there all day always copying gigantic files back and forth - only occasionally, and IMHO that doesn't really make all the expense and setup worth it.

Otherwise, you're simply boasting with something that is of limited practical use, which is definitely not the case for SSD access times. That's where SSD is king, no if, no but.

 

[DeaconFrost]

That right there, is why striped arrays died out and lost favor in the enthusiast community. They would only show real performance gains in very unique situations. Most of the people who still use them, don't use them for their system volumes...but as scratch disks.

So, that all being said, Mergatroid, the truth is out there, as Fox Mulder used to say. Whether you choose to update your thinking and learn about the "whole" issue around arrays is up to you. Just because you choose not to believe what's been well-known as fact...that's up to you, but it doesn't make the community wrong.

 

[Mergatroid]

Nice words.

I'll stick with my halved boot time and faster MEASURED loading time for my games.

 

[DeaconFrost]

Nice words.

I'll stick with my halved boot time and faster MEASURED loading time for my games.

It's been said many times, but I'll repeat it once again. What you choose to do with your own system is up to you...but when you post on a forum board...one that people come to for help and advice, you need to stick to tried-and-true facts and good computing habits. If you want to bite off a piece of the placebo sandwich, be my guest. Just don't fluff up the hype for others.

 

[Mergatroid]

Placebos don't work. You have never once provided an explanation for the decreased boot times, nor the faster loading of games on the raid volumes.

The fact is, anyone trying this with slower SATA II SSDs will see the performance improvement for themselves, and I don't think people should dispute it if they haven't tried it.

Try it, make the measurements and then come back and tell me I'm wrong. After all, it only takes a couple of hours to set it up and test it.

 

[DeaconFrost]

Try it, make the measurements and then come back and tell me I'm wrong. After all, it only takes a couple of hours to set it up and test it.

It also requires identical drives, which I no longer have due to the uselessness of SSD RAID. Again, you miss the point. I'm the one making the claim on the side of what's accepted and true, and can attest to that with the positive rep I've gotten out of this thread with people agreeing with me....along with several other performance enthusiast forum boards that have all come to the same conclusion. I don't have to prove anything, because SSD RAID has already been debunked by the industry. It's lone and sole purpose was to increase volume size. Thanks to falling costs of SSDs, along with the flooding in Thailand, it is usually cheaper to get one larger drive than two smaller drives, and you don't tie up two SATA ports, nor do you have the inherent risks of data loss with striping. It's a no brainer.

You are the one going against the accepted info, aka making the wild claim. It's already been proven, with spinners and SSDs that striping increases seek times, which a) are a big part of Windows performance, and b) cancelling out the very reason people switch to SSDs. Increased boot times can be caused by many factors, such as BIOS rev, BIOS config, OS config, etc. Anyone who wants to hang their system's performance on it's boot time is severely misguided anyway. That's like measuring a car's performance by how quickly the engine turns over once the key is turned or the ignition button is pressed. That's why I've tried to explain to you that real world performance is the key...and that's where striping dissolves into pure hype and nothing more. I've said this all before, and you choose to ignore it because it doesn't mesh with your believes. Whether or not you choose to accept it makes no difference to me, but once again, you need to remember that others read this board and expect accurate, correct info.

 

[Mergatroid]

"What can be asserted without evidence can also be dismissed without evidence."

Christopher Hitchens

 

[Logun]

Well I have to thank you both for taking the time to post this up, as I am currently having a helluva time with my RAID0. So now that I am faced with SE my SSDs and starting over I need to reevaluate whether I want to raid them or not.

What is the basic measure I should be looking at? Is it only access time? I agree with the thought process around - read and write speeds are only applicable in specific situations, so I'm looking to test for a more applicable measure.

I wanted to use a raid because of both (perceived) speed and because of convenience of only having to deal with one drive

 

[DeaconFrost]

I wanted to use a raid because of both (perceived) speed and because of convenience of only having to deal with one drive

You would be putting your data at twice the risk of loss, just for a "percieved" speed gain. If you truly only want to deal with one drive, sell both SSDs and get a single, bigger drive. If you check some of the drive's benchmarks, speeds actually increase as the size of the drive increases. This way, you'll only be burning one SATA port, you'll have a simpler setup, and you can still pop in a much larger HDD for data storage.