RAID 0 worth it for gaming?

For gaming, would two 80g 7200.2rpm HDs in RAID 0 configuration give better performance then 1 160g 7200.2rpm HD? As in any HD based tasks are performed noticebly faster?

I am aware that with RAID 0 I could be risking the loss of both drives if one dies but its not a large concern for me.

Thanks.

 

While yes it will be faster, is the extra risk and extra cost worth a 5-15% jump in HD performance?

 

RAID mostly helps in games when loading maps/zones. Once in game, your experience won't change noticably, if at all..

In the days of Anarchy Online, I used to have a large advantage over other's having my raid 0, because my zone times were nearly instant. Some FPS games also benefit like BF2/2142, because it gives you time to pick your gear etc before the round begins, so you can start taking flags the second the match starts.

 

I understand that RAID 0 is more risky than RAID 1, but is it really and more risky than having a single hard drive that might fail anyway?

 

If a single drive fails, you only lose data on that drive. If a RAID-0 drive fails, you lose data from both drives in the array. So you double the chances of failure, and double the data loss.

 

Are there any figures on what are the chances/how often a drive fails? Has yet to happen to me with any computer I've owned over the last 10 years.

 

Overall its pretty random, but since laptops can see much more wear and tear from bumps or drops failure rates are much higher than in a desktop application.

 

The likely hood of a single drive failing is quite low, but by being dependant on two drives, you effectiveky double the risk. RAID 0 offers very little to a gamer, because it can only help with load times, and even those are not decreased a massive amount. Its better to consider a faster single hard disk, then doubling two up in an array.

 

I am currently using a Sager NP9260 with a RAID 0 array (2x 80GB 7200RPM SATA). I have to say that while I do notice a performance increase over my laptop's 100GB 7200RPM, the difference is not huge. Games load maybe 20% faster and startup is quicker too. However, the difference is not substantial enough to justify doubling the risk of data loss. I would recommend a RAID 1 array - that's the way to go.

 

I guess I am considering it primarily because I play mostly MMORPG's and all of those, as far as I know, involve some sort of zone loading. I've had many times with my old laptop where I would be left behind on the zone line.

 

A RAID 0 array may actually be slower for that purpose because seek times are slower with a RAID 0 array. That is just a guess though. I would go with RAID 1 with a pair of 7200RPM hard drives.

 

Err.. so raid 0 allows you to access the data faster when you find it but the actual finding may take longer then normal?

 

Yep. (Technically, the seek time only goes up a little bit, while the transfer speed more or less doubles, so it might result in a net gain in performance... Depending on how important seek time is vs transfer speed in your particular case... Generally speaking, RAID 0 will be maybe 3% faster than a single drive though.

 

Games use alot of small files normally not a few big files.

Raid 0 was designed to help with the transfer of big files, its a super good thing if your a video editor, 3d vector worker, ect that work with huge files.

But for a game its not so great, you have to set the strip size of the raid array, and at the normal optimal setting it wont be able to strip game files down between the two drives, make it smaller and you just fragmented the files into such small pieces you may actully lose performance or gain nearly nothing out of it except more heat, more power use, and the prementioned 2x chance of drive falure.

Raid 0 is one of the biggest myths in the computer world, and unfortunatly those who have it always get effected with a placebo and go scream to the world how fast it is when they run a synthetic benchmark and see 1.8x there normal max transfer rate...

 

RAID 0 as ViciousXUSMC said is designed to boost throughout for bulk read/write. The benefit are mostly seen in heavily multitasked servers that is IO bound. If the request cannot be "mutiplexed", it is useless.

We see similar things in dual core/hyperthreading. If you run a single threaded application, dual core/hyperthreading is useless.

RAID 0 however can be very useful to make a DIY flash SSD(in fact there was just a link about a new adaptor shown in computex). have a software RAID 0 over 2+ flash drive can be a very good way to compensate for the lower throughput of flash drive comparing with HDD, yet have the much faster seek advantage for burst, small read.

Not sure if Windows support software RAID on removable device though.

 

Hmm sounds neat chimpanzee, mind posting the link?

Im very closley watching the market of SSD's because I want one. Cheapest I have seen is about 500$ for the 32gb model. ~10x the life of a hdd, 2x the speed, and 1/2 the power use. SSD is the future of notebook storage I think, and will probably be adopted by desktop performance enthusiast for faster OS startups and game loading.

 

follow this:

http://forum.notebookreview.com/showthread.php?t=130571&highlight=apacer

My take about the product:

It is a very good idea but not yet feasible. Seems to be a IDE/SATA device so can only be used to replace HDD in notebook but with the current size of CF card, it doesn't even meet the basic requirement of Vista.

I believe in a years or two time when we see 16G CF at reasonable price, it would be an interesting product. By then, I would probably replace my HDD with it and use an external USB enclosure for less accessed large files like videos etc.

I shall soon play with the idea using standard USB thumb drives running linux and see what kind of result I can get.

 

I was speaking on MMO's and zone loading from years of gaming experience using RAID 0 and non RAID 0 machines. Games like everquest and anarchy online that had larger zone files loaded much faster using raid, the newer games like WoW though, use tiny files so they can load seemlessly in the background and your zone's are mostly transparent (other then in instanced zones).

Some games still see benefit from a raid though, Guild Wars, as well as many FPS's with large texture files (some as large as 512mb) are improved by the raid greatly.

 

What FPS games? I have played almost all of them and never had any loading issues. Always super fast, longest wait I have is BF2142/BF2 and im stil one of the first 3 in every game.

I think having more ram is more imporant than anything when it comes to textures.

 

The data still has to get from your HDD to your RAM... that's where RAID speed comes in handy... The point of lot's of ram is that you don't have to keep taking info from the HDD, when you have both you are even better..

Wait for these new games coming out for DX10... Then run them on Vista (some of which are only coming out for Vista) and I think the RAID will start to be appealing again..

The hardware cycles against the software, always has, probably always will. Basically right now we are building hardware that isn't taxed by the software because we are on the peak for a massive change. Once we drive into DX10 fully you will see many games coming out that will again tax the system's hardware for a year or two, then a new peak where the hardware is beyond the software again.

 

What is the point of a RAID 1 for a home user (mainly a gamer in this case)? It's strictly for redundant data, in fact your performance is normally decreased slightly because of the real time drive cloning...

Did I just misunderstand your post?

 

Hmm well if sombody is running DX10 and all that jazz they have new hardware probably, there are drivers out there now that are so fast on there own, it would challenge older raid 0 setups and avoid any of the problems.

Again its only a few games in the first place that will probably benifit from Raid 0 much if any at all. I say get a SSD and be done with it

 

That's where RAID speed comes in handy IF (and this is a big if, because it's hardly ever true) the files are arranged so that RAID 0 can actually make use of its higher transfer speed.

Again? RAID 0 has *never* been particularly appealing. It has *never* shown serious performance gains for regular desktop use (or gaming). And DX10 (which happens on the GPU, not the harddrive) certainly won't change that.
The *only* factor that determines whether RAID 0 is faster than a single drive, is the layout of the files to be loaded. And that is completely independent of OS, and of DX version.

Er, what gives you that idea? Different parts of the hardware may be taxed at different times, yes, but the hardware as a whole is always taxed. If it wasn't, it'd run the game faster, and then it would be taxed.

And the hardware certainly isn't "beyond the software".

RAID 1 *might* slightly improve your seek time (because it only may only need to find the data on one of the disks, so the first one to return a result is used), but generally, yes, it is purely for redundancy.
But isn't that beneficial to a home user? Most home users have plenty of data they don't want to lose. If you browse this forum, you'll see more than a few people crying because they lost their anime collection/documents/photo albums or whatever else.

But true, if you simply don't care about your data, then RAID 1 is not for you. In that case, I'd go with a single drive.

 

I understand DX10 is GPU powered, but it's also a huge library set used by software developers to enable the use of that hardware, you misread what I stated.. what I mean is that whenever new stuff is released it always takes the software developers time to write GOOD code, usually towards the END of the product life cycle. It's not just the code writers faults, it's also poor area's in the DX(insert version here) libraries themselves. This has been this way since DX was first released in Win95...

The thing I said about the cycle of hardware... Look at the current games on the market.. An NV7600 is more then enough to run them, 7950GTX's are the shiznit in notebooks, and 8800's are on the desktop side. These GFX cards are WAYYYYY overkill for the software available today(and they will be perfect for new games like Crysis). Now go back a bit to when Oblivion or FEAR was released. There were basically no GFX cards that could keep up, infact the entire system had a hard time.. Hense the flip flop back and forth. You can find lot's of information on this stuff if you go back through time from present articles, to the "dawn of PC gaming"(DX).

Speaking of Oblivion, that is a great game for testing how and when RAID0 will benefit..

 

Hmm, are you saying that because RAID 0 is so "new", developers need time to learn to take advantage of it? If so, I'm going to have to disagree (again).
First, RAID 0 is not new. Second, RAID 0 does not pose any new challenges to developers. It's still the *exact* same file I/O code they have to use, and it still works in the same way and they have to do the same things to get good performance. And third, even if it was something developers needed time to "learn", they still wouldn't do it because most people don't have RAID 0.

Not at max settings. A reasonable mid-range card has pretty much *always* been able to run every game currently out, if you're willing to sacrifice graphics level. (And for a fairly obvious reason. Games are made to be sold. It's a business. And it is not good business sense to release a game that can *only* run on high-end cards)
So I'm sorry, can't agree with that one either...
Game developers pay a lot of attention to what GPU's are coming up, so that they can make games that run well on midrange cards *when the game comes out*, while still being able to make the most of a high-end card.
True, sometimes they aim a bit too high or too low, and sometimes, we get a few months without any games that really improve graphics. But that's not a cycle, that's just random variations. And trust me, game developers have absolutely no problems bringing even the fastest, most extreme PC to its knees. They still have to optimize like hell and cut corners and cheat and sacrifice awesome features to get acceptable framerates. And there are still dozens of nice graphical features standing by just waiting for the day when GPU's get fast enough. So no, todays GPU's are definitely not overpowered.

No, it's a great game for testing whether RAID 0 will benefit Oblivion. (But of course, only if you benchmark both with and without RAID 0, where all other hardware and software running is identical. Running just RAID and saying "Oh, this feels much faster" doesn't really tell you much)
Every game has its own file structure and file loading code, and decides for itself when to load resources. So you can't really draw any general conclusions from looking at RAID performance in individual apps. Best you can conclude is that "in the game we tested, performance is so and so", or "on average, in the 10 games we tested, you gain/lose x% performance". But it always applies only to the games you tested.

 

you mentioned oblivion, my load times are nearly non existant(in oblivion) with raid 0, but it does vary between users. i don't know why my mate, with almost identical setup has twice as long load times????? if your worried about data loss go for a raid 5 set up.

 

How did you decide a paragraph speaking of DX technology/libraries and software development is stating that RAID 0 is new?

I'm done discussing this matter with you because this seems to be going nowhere.. You are bringing up things I'm clearly not stating, and you are pulling extremist remarks about examples I'm using. They are examples for a reason, which I actually have used HD performance testing software for a company I previously worked with for testing the benefits of gaming and RAID 0. You must use popular games as an example of the benefit of a technology when testing in order to connect the data to the consumer, so I don't see how or why you wish to discredit my use of RAID 0 for Oblivion, which has been measured as performing better both synthetically, and "user experience" from using RAID 0 on systems from the generation of available hardware when the game was released. You can find this sort of information yourself using google. Be careful which reviews you check though.. because platter size and number of r/w heads plays a HUGE bit into the ability for RAID 0 to be beneficial, and raid 0 setups using 1.5Tb are going to be slow no matter what you do. The controller card itself is also a contributor to the speed. In most home user systems they have RAID 0 through a software controller that either get's piped through the northbridge along with the CPU, memory, and the rest of the I/O which will limit the ability to benefit (Also taking CPU cycles to manage the array, as well as system memory to help buffer the syncronization), or through the southbridge which get's limited access rights compared to the northbridge. But this isn't the topic.. because a dedicated RAID 0 PCI card really should be used for testing if you want to see the benefits, since the PCI bus in modern systems is directly linked to the memory and CPU.

I'm not sure where you were going with my comment about hardware technology lifecycles either, because the average guy playing games ISN'T playing on the latest hardware, and ISN'T running the game maxed out. Most people I find still prefer to play games @ 1024x768, which is why all the benchmarking software still defaults to that resolution.

 

^^damn, i find myself agreeing with that

 

I have never had to wait for any loading in oblvion with a single drive. Id assume I would if I had less system ram or it didnt load things ahead of time.

 

Err, I tried to guess what you meant.
First I assumed you meant that DX10 somehow affected, or depended on, RAID performance. You said that wasn't the case. Then I guessed maybe you meant that both were similar, as "new technology that developers have to get used to".
I'm sorry, but I can't read minds. Getting offended because people don't understand what you mean isn't very constructive...
Why don't you explain what you meant then?

And that means your benchmarks are perfect? Most companies are extremely good at making biased benchmarks. Some do it intentionally (to make their, or their client's products look good), and some do it unintentionally. And yes, sometimes, a company makes valid benchmarks.
But as I said before, testing RAID performance in individual games can *only* ever tell you about RAID performance in those particular games. There is nothing in common between games in general when it comes to disk I/O performance. And any benchmark that tries to establish such a "general rule" is automatically invalid.
You might as well try to compare the average number of legs on living creatures, and try to deduce something relevant from that. There's just no relationship between your samples.

I don't, and I didn't. All I said was that your benchmarks in Oblivion only tell you how RAID 0 performs in Oblivion. They're very accurate for this. But it tells you nothing about how it performs in Doom 3, Crysis, Dawn of War or WoW. With GPU's or CPU's, you can benchmark, say, 5 games, and then claim to have found the "overall" performance of the GPU/CPU in question. That is because GPU's and CPU's are used in pretty much the same way by all games, which makes comparisons valid. Every game renders a sequence of frames as fast as possible, they do it all the time, and they do *roughly* the same things during each frame. And they all rely on DirectX or OpenGL to do all the GPU stuff.
RAID (or disk I/O in general) just doesn't work like that. There is no common ground there. Each game uses its own I/O code written separately and serving different purposes, and driven by different game logic. In general, if one game runs faster with GPU X than GPU Y, then it is safe to conclude that X is faster than Y. (Or at least if you test with 2 or 3 games)
With RAID, not so. One game might show 20% better performance. Another might lose 10%.
You could test RAID 0 with 200 popular games, and it *still* wouldn't tell you anything useful about the 201th.

Simply that they aren't regular enough to qualify as cycles. There'd have to be some kind of pattern to it, some kind of reason for it to cycle. There isn't. It is basically random when new games come out that push the GPU, and they try to aim for supporting the latest hardware at the time when they're released. Sometimes the GPU they rely on gets delayed, so there's nothing that can run them at max settings at release, and sometimes the game gets delayed (or the developers misjudged the performance of unreleased hardware) and end up with a game that can be easily run on most systems. But there's nothing cyclical to it. Purely random variations.

 

You tried to guess what I meant? Why did you need to? The entire paragraph was about software libraries..

The company I worked for didn't produce anything other then benchmarks, we were a 3rd party, sort of like an audit service that was hired by companies that DID make technology. They paid us because we had all the systems and the software, we just needing them to tell us what they wanted to see(edit: by this I mean what sort of tests they wanted), or offer us their custom item's so we could included them for results. I promise you a good number of clients did not like the results they got from us, but it's ok because it made them rethink what they were doing.

I still don't get the point in mentioning what I said about Oblivion.. I listed the first game that came to mind that has easy to see results for the benefit. (Albeit not so much with modern dual core's and whatnot, which is why I clarified with stating the results have to measured on the hardware of the time of the games release).

And as for there not being a lifecycle to the hardware you see on the shelf.. I work for the largest division at Intel and I'd challenge you tell anyone of the hundreds of product managers that their life cycle doesn't exist or that it isn't valid. My group does nothing but design and maintain the tools used by these managers to stick to their life cycle model.

Btw.. english isn't my first language, so don't take offense to any of it if it comes across wrong. I edit almost every post I make about 6 times trying to correct stuff.