What's with all the RAID-0?

Everything I've read has always concluded that RAID-0 is mostly pointless for gaming yet I see a lot of people on these forums with such setups.. Let me paste a quote from a knowledgable poster on storagereview regarding this exact topic:


"Raid does not make games run faster. Consumer raid gear is not tested or matched to particular drives or drive firmware revisions. Consumer RAID zero often results in system performance that is slower than that of a single drive. Most users don't know how to tune RAID 0 for their usage profile. Most users don't even understand how to profile their storage usage or what a profile is.

RAID 0 is not RAID, and "more than" doubles the risk of data loss. With a single drive, you have the MTBF of the drive to worry about. With Consumer RAID, you have to worry about the drives, and whether "Bob" from PromiseTech tested the latest controller firmware with your drive firmware (guess what... He didn't), in addition to testing with whatever stripe/cluster size you "guessed" at when you created the array.

When is RAID 0 appropriate? Easy. It's appropriate when the required sequential read/write transfer rate of the application profile exceeds that of a single drive. Examples would be multitrack audio recording, raw digital video capture and real-time editing, and in-line data transformation. Pay special attention to "read/write", as a "read only/mostly" profile would demand RAID5.

We're all glad that you're happy with your rig. Please stop suggesting it as a solution to others, as you will be ultimately responsible for their added expenses, their data loss, and their time. Any performance increase that you have seen can just as easily be gained by adjusting your disk subsystem configuration to match the read/write profile of the application."


So my question is, are the people who are setting up these RAID setups just uninformed? Or are all you RAID users also using your rigs as a heavy workstation?

I don't mean to offend anyone, I'm just curious..

Thanks.

Flamewar in 3..2..1

From all I have read, people have conducted experiments with RAID-0 and some came up showing a benefit from R-0 others showed the opposite. Whether or not R-0 improves gaming is a topic that is often debated religiously on this forum. There are some that suggest that Experiment X shows clearly that R-0 is the best thing since sliced bread, other say that Experiment Y indicates that R-0 may be detrimental to performance.

In the absence of undisputed evidence that R-0 does not improve gaming, I think that people ought to make up their own mind. I notice that the way this topic is debated reminds me a lot of religious debates. Whether or not R-0 is good for gaming seems to be strongly correlated to which experiment people believe.

I think that the best course of action when people ask for advice concerning R-0 is to mention the reduced reliability and that there is evidence supporting multiple hypothesis w.r.t. gaming and R-0.

So my question is, are the people who are setting up these RAID setups just uninformed? Or are all you RAID users also using your rigs as a heavy workstation?

Click to expand...

People get the desire to try it themselves as computer enthusiasts and that's kinda the people you find here, no? And a simple stopwatch can tell you whether or not it's beneficial for you in your own system. It's when people start citing articles to prove their point of view either for or against RAID0, or put their own value/cost sytem on it, which really has no relevance to performance in your own system, that things get ugly. At this point, I just say I've had positive results with RAID0 in 4 setups and hopefully the person who considers it will too.

Well first, most enthusiasts at home don't earn money with their home rigs (it's not what brings home the paycheck), so the bad parts of RAID (double chances of failure) are overlooked... also why they overclock to points of instability and strap 6 pounds heatsinks to their shit.

Next, learning new shit. I had no idea how to setup RAID, or SATA for that matter... and I wanted to know, dammit. Tooling around with stuff is the best way for ME to learn. Bought two cheap HD's and did the damn thing. Learned a lot.

Next, the performance is debatable. You will read that it does things better, some things worse. You can change your stripe size and adjust to your needs. Large files, large stripe size, etc.

For me, it was a combination of the above. I don't care what e-people say, I wanted to KNOW. I wanted to know that RAID 0 is a huge waste of money, or was the best thing since sliced bread. It was a good learning experience, and now that I have it, I'm not getting rid of it. It is faster than my old single hard drive setup. I don't care about the extra power consumption or chance for losing my data. I have another hard drive to throw my "needed no matter what" data on to. Which is 13 gigs of music. Big whoop. I'm not setting up a nice backup solution for 13 gigs of music.

Sure RAID 0 isn't for everyone, but the complete animosity towards those that use it and constant thread shitting is for the birds. Let people do what they want, deliver the facts to them, but don't piss and moan when they do it anyway. Those people are like parents... jesus. Some people on here REALLY care a WHOLE LOT, that OTHER people are putting THEIR data at risk....

1) Fun
2) faster than my old setup
3) feel more geek-like
4) why in the hell not if you don't care about data loss
5) why use modded BIOS's, 6 pound heatsinks, FUCKING WATER TO COOL YOUR SHIT, overvolted GPU, chipsets, and CPU's, and hacked video card drivers and BIOS's but care about about "doubling the chance for data loss".

[H]yphenated [H]ardness.... you guys should be ashamed.

drizzt81 said:

In the absence of undisputed evidence that R-0 does not improve gaming, I think that people ought to make up their own mind. I notice that the way this topic is debated reminds me a lot of religious debates. Whether or not R-0 is good for gaming seems to be strongly correlated to which experiment people believe

Click to expand...

So I'm the first person in BF2 and BF2142 because of something other than raid?

Sorry, but raid makes a HUGE difference for me in loading games, namely BF2 (Running 3x80gb SATAII in R-0). Huge being a few (5 to possibly 10 ) extra seconds. Enough that I can be halfway, if not all of the way to a flag and capturing it before even half of the other team joins the game.

There is a big diffrence between the preformance of software based raid 0 configs and hardware based raid configs. My hardware based solution is MUCH faster than a single drive... I've proved that to myself by doing benchmarks, etc... I get about 1.8 times the speed of a single drive on large file transfers, either reading or writing.

I've also tested software solutions under various OS's - I was unable to get more than 1.3 times the speed in a purely software based solution - but it's still an increase. Software based solutions do take up quite a bit more CPU recources though while transfering files.

Now gaming - faster hard drive speed probably won't do anything for framerate or make you a better gamer. If you're actually playing through a game though (offline game), maploads will be MUCH faster + it will be more of an enjoyable expiriance. Also, when switching maps on on-line games, I'm generally the first or second one to make it to the new map... I think the above 1.8 times the hard drive speed accounts for that.

DriveEuro said:

So I'm the first person in BF2 and BF2142 because of something other than raid?

Click to expand...

The $2k-ish machine in your sig has nothing to do with it, of course. $2k of practically brand-new machine is hardly average for BF2(142), I'd wager. And your internet connection? What does BF2 do while it's loading? Is it disk-bound or not?

mdameron said:

Sure RAID 0 isn't for everyone, but the complete animosity towards those that use it and constant thread shitting is for the birds. Let people do what they want, deliver the facts to them, but don't piss and moan when they do it anyway. Those people are like parents... jesus. Some people on here REALLY care a WHOLE LOT, that OTHER people are putting THEIR data at risk....

Click to expand...

... because OTHER people come here and COMPLAIN when their DATA is LOST and they want it BACK. When someone suggests raid 0 and everyone jumps on him and says "what about data loss!?!" that is delivering the facts. Denying that raid 0 increases probability of data loss is simply not realistic.

teststrips said:

There is a big diffrence between the preformance of software based raid 0 configs and hardware based raid configs. My hardware based solution is MUCH faster than a single drive... I've proved that to myself by doing benchmarks, etc... I get about 1.8 times the speed of a single drive on large file transfers, either reading or writing.

Click to expand...

If you're using your mobo's solution, is that hardware or software? Nforce 4 chipset... board is in sig.

unhappy_mage said:

... because OTHER people come here and COMPLAIN when their DATA is LOST and they want it BACK. When someone suggests raid 0 and everyone jumps on him and says "what about data loss!?!" that is delivering the facts. Denying that raid 0 increases probability of data loss is simply not realistic.

Click to expand...

Again, very good points. I just got bit in the ass by not having a back up (at work though).

RAID 0 is fun... for home use with non-essential data. I wouldn't use it at work without a beast of a backup plan though. I don't need fast HD's at work though.

teststrips said:

There is a big diffrence between the preformance of software based raid 0 configs and hardware based raid configs. My hardware based solution is MUCH faster than a single drive... I've proved that to myself by doing benchmarks, etc... I get about 1.8 times the speed of a single drive on large file transfers, either reading or writing.

Click to expand...

There's no doubt that raid-0 is far faster than a single drive in sequential read-writes of large files and such.. That is already well known.

The point of my post was asking why people who use their rigs for mostly pure gaming are getting raid-0.. Because you are putting your data at more risk doing so and getting what seems to be no extra performance from it whatsoever..

Sure you may transfer large files once in awhile, but is the extra speed you gained from that worth the increase in chance of data loss and the like?

roper512 said:

The point of my post was asking why people who use their rigs for mostly pure gaming are getting raid-0.. Because you are putting your data at more risk doing so and getting what seems to be no extra performance from it whatsoever..

Click to expand...

me said:

5) why use modded BIOS's, 6 pound heatsinks, FUCKING WATER TO COOL YOUR SHIT, overvolted GPU, chipsets, and CPU's, and hacked video card drivers and BIOS's but care about about "doubling the chance for data loss".

Click to expand...

There are performance benefits to RAID 0 in games. Maybe not solitaire or CS 1.6... but DoD:S and HL2 I notice a great difference from my old single drive setup.

mdameron said:

If you're using your mobo's solution, is that hardware or software? Nforce 4 chipset... board is in sig.

Click to expand...

It's a little of both.....
In this case your CPU acts as the actual RAID logic engine with the software that tells it how to do that being loaded with the boards BIOS and the chip that the drives plug into is handling the communications only.

A TRUE software RAID like in linux implements everything at the kernal level instead of at the BIOS level like your board does - so it uses a little more CPU to overcome the things that hardware is doing for you.

A TRUE hardware RAID will have it's own processor for a logic engine and completely offloads any of this work from the CPU, thus improving your OS's performance because it isn't having to share CPU cycles like the previous 2 cases.

unhappy_mage said:

The $2k-ish machine in your sig has nothing to do with it, of course. $2k of practically brand-new machine is hardly average for BF2(142), I'd wager. And your internet connection? What does BF2 do while it's loading? Is it disk-bound or not?

Click to expand...

Actually u_m RAID 0 does help more than the $2k of other components he's got or his internet connection.

I've verified this playing BF2 and CS:S on my own computers. And the reason is that the maps for FPS games are all very large single files that contain a bazillion different textures and all of the sound and 3d models for the map. RAID 0 is the proven winner at moving large files.
And they are local files not requiring download, CS:S only transfers like 8-12Kbps for each player on a map - BF2 is higher on the transfers though, but not a huge amount.

I've been consistently in the first 2-5 players on a map of CS:S with a 50-60 ping and playing on a socket 478 PresHott 3.0Ghz with only 1 gig of RAM and an ATi 9600 AGP video card. But had 2 of the old 36GB Raptors in RAID 0.
Playing against guys with 12 pings and AMD Dual core CPU's, 2 Gig of RAM, PCIe 16X video cards and only one nice large hard drive.

Knew one guy that played with 4 of the old 36GB Raptors in a RAID 0 configuration. And his machine loaded the map files FAST!

roper512 said:

There's no doubt that raid-0 is far faster than a single drive in sequential read-writes of large files and such.. That is already well known.

The point of my post was asking why people who use their rigs for mostly pure gaming are getting raid-0.. Because you are putting your data at more risk doing so and getting what seems to be no extra performance from it whatsoever..

Sure you may transfer large files once in awhile, but is the extra speed you gained from that worth the increase in chance of data loss and the like?

Click to expand...

All maps for FPS games like BF2, BF2142, CS:S, HL2, DOD:S are single LARGE files - the speed increase applies directly to opening them. And for some of these games you'll only be on a single map for only 5 minutes.
Even some games like Oblivion are built on the same format of the area maps being single large files which have to be loaded into memory and then operated on.

How much risk am I putting data that belongs to an online FPS that I can't hit a save point in? Or a game that I downloaded from Steam?
Or a game that all I have to do is re-install?

Sure, I might lose my save games - but the games themselves are all on the install disks.
Windows is on it's install disk also, but I would lose some time installing it and configuring it. And the information isn't sensitive like tax and bank account data would be.

Plus, if I have a single large drive as my system's data drive, I can use backup software to image the RAID 0 array to it, or run a scheduled Synctoy run to a backup folder for my savegames and cfg files. Or to a large USB drive. Rather frequently, even do it in the background.

Another thing you've got to remember about those of us that get into gaming seriously, we are always tweaking our computers. Buying the fastest video cards, trying to OC our systems, playing with our network configurations, trying out different memory sticks, and other kinds of stuff that requires us to basically re-install all of our games and OS's all the time anyway.
We're trying to do ANYTHING that will give us just about any tiny edge in our chosen games. Regardless of the possible data loss.
It's only data that's on a backup somewhere and the time to reinstall/reconfigure it. Thats actually part of the fun.

The government isn't going to stall if I have to rebuild my harddrive one more time.

And most of us have more than 1 rig also. I, personally have 4 running from the same KVM switch.

This is just from my own personal experience:

The only thing RAID 0 ever affected for me was load times. I saw somewhat reduced load times in games that had large map files and such. That's the biggest benefit of RAID 0: sequential reads are substantially faster, for obvious reasons. A well-defragged array on a good controller can drastically reduce the time it takes to load a new game.

What RAID 0 doesn't help (that I've ever seen) is paging. In fact, I've read several times that if you're using 0 for an OS drive that you should put your paging file on a separate drive; what it needs is low seek times and taking it off your heavily-used OS drive would fix that.

If I had money to burn, I'd have three arrays:

one for the OS and some file storage
one for apps (mounted to the Program Files folder)
one for the paging file and additional storage

Ok, here we go again. I love my RAID-0, and seeing this arguement, I figured I would try to generate a "real-world test" of a game to see if my RAID-0 was in fact speeding things up.

First of all, as we say in the Sports Car world, my "butt dyno" says my Windows XP SP2 system, OVERALL, feels much faster running C: on a RAID-0 setup.

So given that, here we go.

I decided to use Halo, because I didn't already have it installed and some of the .map files are fairly large, 70+ MB.

1. Install HALO to my Files drive, a 320GB WD Caviar WDC WD320 SATA 8MB Cache

2. Apply the HALO 1.06 patch.
3. Add "-timedemo" to the desktop shortcut, and launch the game.

4. Using Add/Remove Programs, completely uninstall HALO and then go and delete the G:\Halo file from the drive.
5. Install HALO to my Windows XP Boot drive (C, a pair of WD WDC1200JB PATA 8MB Cache drives on an Iwill SIDE RAID100 Controller Card (HighPoint based) in RAID-0 config, default stripe size of 64k

6. Apply the HALO 1.06 patch.
7. Add "-timedemo" to the desktop shortcut, and launch the game.


As you can see, simply installing the game to and running it from a RAID-0 drive increased the timedemo benchmark average by 6.5 FPS. While not mind-blowing, it's an 8 or percent increase over the single drive. So that's with an older game ... I can't imagine what it would do to modern game loading times. Will it make the FPS increase while you're actually PLAYING the game? Probably not... but the loading times are night and day. Hell, the game even installed faster.

It's also worth noting that I've done nothing to the Windows XP paging file setup. It's managing it's own paging file on the RAID-0 drive, which is where I installed Windows XP in the first place, and is the C: drive.

I hope this helps.

sn3ak3rp1mp <---- zips up flame suit

I have a few things you didn't cover about your benchmark results.

• How full was your 320GB disk? Is it fragmented?
• How consistent were these results across multiple runs?
• Was your system networked? Was stuff running in the background?
• How does real gameplay compare to timedemo results?
• Since Halo does level loading stuff in the middle of levels, what was disk usage like during the timedemo?

Thanks for doing the tests, but I think they're not very relevant to recent games - Halo is a console port (which most PC games aren't), isn't very heavy on the video card, and generally runs slower than other DX9 games IIRC.

Constructive criticism only Every benchmarking setup I've seen has had some of these apply to it to some degree.

sneakerpimp -

I'm not really familiar with Halo benchmarking, but couldn't that increase in FPS also be from having the benchmark cached? I know that many people run a benchmark more than once to get their true scores.. If not I'd be interested to see what score your single drive got if you ran it one more time, it would probably be much closer if not identical to your raid setup.

I don't see how a hard drive raid array can increase framerates by that much unless there's some large files it needs to load during the benchmark and those files are not cached.\

But hey what do I know

rodsfree said:

All maps for FPS games like BF2, BF2142, CS:S, HL2, DOD:S are single LARGE files - the speed increase applies directly to opening them.

Click to expand...

The benefits don't have anything to do with large files; they're come into play for large, seqential reads.

Looking at BF2, I found that the levels aren't single large files. When loading a level, there is indeed a pretty big map file being loaded -- but then there's dozens of much smaller texture and mapping files. The BF2 installs these all as big GZIPped files, and then unpacks them as they're loaded. It turns out that some perf gain is realized by minimizing the I/O with the compression, even though unpacking the file causes CPU load.

But the bottom line is that a large file has nothing to do with the issue; a large file might be read in small requests (which result in little RAID0 benefit), or might not be read sequentially. I observed BF2 using both non-sequential reads and quite short reads, and both of these facts lead me to doubt RAID-0 could help BF2.

Raid 0 for life for me. I don't care about those who don't notice the benefits from it, maybe you are doing something wrong, maybe you should make sure both of your hard drives are the exact model and have same firmware. I just installed 2x320gig Seagate Barracuda 7200.10 drives to replace my old 74gig raptors and these cuda's are faster in my opinion. I'm getting 116mb/sec read compared to 94mb/sec with my old raptors. Games like BF2 load extremelly fast, I load in a server before anyone, and im pretty much alone in the game for about 10sec. Things like antivirus programs, adware tools, scanning is cut by almost half completing the job. Like i mentioned in a diff thread, you have to properly setup your raid0 in order to benefit from it, and YES, if there is something wrong, ex; raid drivers, chipsets, software, Raid will hinder your performance but only because of human error.

mikeblas said:

The benefits don't have anything to do with large files; they're come into play for large, seqential reads.

Looking at BF2, I found that the levels aren't single large files. When loading a level, there is indeed a pretty big map file being loaded -- but then there's dozens of much smaller texture and mapping files. The BF2 installs these all as big GZIPped files, and then unpacks them as they're loaded. It turns out that some perf gain is realized by minimizing the I/O with the compression, even though unpacking the file causes CPU load.

But the bottom line is that a large file has nothing to do with the issue; a large file might be read in small requests (which result in little RAID0 benefit), or might not be read sequentially. I observed BF2 using both non-sequential reads and quite short reads, and both of these facts lead me to doubt RAID-0 could help BF2.

Click to expand...

Not doubting your research - but take your conclusions over to the CAL forums and they will run you out of town.

The preferred setup for FPS games in CAL is a 2-4 drive RAID0 setup.

I think that there must be a reason why.

What's "CAL"?

mikeblas said:

What's "CAL"?

Click to expand...

Cyber Athletes League - The guys that are turning FPS games into a pro sport with cash prizes and sponsorships and all that stuff. Fata1ty is a prime example.

meh..data loss. I've been running my rigs in raid 0 for years, never had a drive fail on me dammit. So periodically i just take a bunch of magnets out and run them up and down my harddrives so I don't get spoiled ...

it's becoming more and more tempting to try out raid-0 for the first time I want to try 2 raptor 150's in raid-0.. But I want to do it right the first time.. Can anyone point me to any good raid-0 guides on how to set it up properly and configure it for my usage?

Thanks!

rodsfree said:

Not doubting your research - but take your conclusions over to the CAL forums and they will run you out of town.

Click to expand...

How would they explain the observations?

rodsfree said:

The preferred setup for FPS games in CAL is a 2-4 drive RAID0 setup.

I think that there must be a reason why.

Click to expand...

What is the reason? Mass hysteria is a reason, but not good one, for example.

Interesting example sn3ak3rp1mp, but I don't think it was truly "apples to apples"

mikeblas said:

What is the reason? Mass hysteria is a reason, but not good one, for example.

Click to expand...

e-penis. Seriously. If someone can match you on processor speed, RAM quantity, and GPU make/model, you can at least have more drives than him.

If they really thought drive speed made that big a difference, they'd have 15k SCSI drives.

edit: and use multiple arrays to divide disk access into OS, app, and paging file.

O actually blame the game coders for some of the benefits of R-0
It would be OK if it adjusts the FPS only, but it seems a pseudo cheat to make it so that you can "buy" an extra 5-10 seconds on games like BF2 just by having a R-0 setup and being allowed to join before everyone else.

not exactly fair play...
fair would be to spawn the first couple of guys together. (at the very least)

but faster loading time are definetly a benefit

PopeKevinI said:

e-penis. Seriously. If someone can match you on processor speed, RAM quantity, and GPU make/model, you can at least have more drives than him.

If they really thought drive speed made that big a difference, they'd have 15k SCSI drives.

edit: and use multiple arrays to divide disk access into OS, app, and paging file.

Click to expand...

Some guys do have the 15k SCSI's.
But is is cheaper to get the Raptor's or a lot cheaper to get any of the new 16MB cache drives, like the new Seagates or Western Digitals. So more people do that. Building a system for gaming is still a balancing act to try to build the best rig that you can while still staying within the budget that you have.

And granted - the drive system is generally the last place a gamer spends his money, so he is trying to get the most bang for his buck at that point.

rodsfree said:

Some guys do have the 15k SCSI's.
But is is cheaper to get the Raptor's or a lot cheaper to get any of the new 16MB cache drives, like the new Seagates or Western Digitals. So more people do that. Building a system for gaming is still a balancing act to try to build the best rig that you can while still staying within the budget that you have.

And granted - the drive system is generally the last place a gamer spends his money, so he is trying to get the most bang for his buck at that point.

Click to expand...

A few years back a company I worked for did some extensive tests with RAID 0 and 5. We found that most application and game access was only marginally improved by RAID; the better bang for the buck would be a single faster drive, since seek time is far more crucial than linear read speed, which only comes into play on large, unfragmented files.

A competitive gamer could easily justify $600 on a disk system. Many of them spend more than that on GPUs in a year. That would buy a controller and three 36 GB Fujitsu 15k drives...one drive for the OS, one for games, and one for paging. By separating the three seek-intensive tasks, you'd have a much faster system than running all three off the same spindle or array.

Note: Yes 36 GB is a bit small when it comes to games; you might only fit six or eight modern games on that drive. I'm talking about a system specifically built for competition, not for running 30 different games. And of course larger drives are available, up to 150 GB, for more money.

mikeblas said:

How would they explain the observations?

What is the reason? Mass hysteria is a reason, but not good one, for example.

Click to expand...

Mike - Gamers generally don't test their rigs to death, especially when they could be fragging instead of testing.
They generally try to duplicate the rigs of the guys that are pwning them - and through "natural selection" - for want of a better term - the RAID 0 systems have came out on top.

So if you want to call it "mass hysteria" that's ok with me.

This is like the discussion about the aerodynamics of a bumblebee. Based on all of the math and science that we know a bumblebee shouldn't be able to fly - but try telling the bumblebee that.

So, based on all of the hard facts that we know, a single drive should out perform a dual drive RAID 0 setup at everything, except large file manipulation. And currently, everybody's definition of large files is in the +100MB range, if not even bigger.

But, the empirical evidence of a large body of gamers says that this is not true.

Why???

I don't know. I'm not a hard drive engineer, but if I were to hazard a guess, I'd have to say that our definition of a large file is the cause. Depending on the RAID controller I can adjust my stripe size down to 32KB or in some cases even smaller. 128KB is a common stripe size and so is 64KB. I, personally, haven't seen any controllers that go bigger than 128KB, they are probably out there - I just haven't ran across them.
Knowing this, I also know that if I have a file that is slightly larger than (1 X my stripe size) then the controller is going to break it into 2 chunks and put one chunk on each of my 2 drives. And when I try to read that file, 2 different heads are going to be accessing it instead of just 1 - which increases my read speed - not quite twice as fast - the two chunks still have to be reassembled at the controller, and the 2 chunks might not be symmetrical in size, but still faster than a single head can do it.

And, again this is just a guess, I don't have any data to back it up and I'm not going to disassemble my current gaming rig to test it just for shits and giggles. But is seems to be logical to me.

So, were any of the files in your BF2 map folder bigger than 128KB?

PopeKevinI said:

A few years back a company I worked for did some extensive tests with RAID 0 and 5. We found that most application and game access was only marginally improved by RAID; the better bang for the buck would be a single faster drive, since seek time is far more crucial than linear read speed, which only comes into play on large, unfragmented files.

A competitive gamer could easily justify $600 on a disk system. Many of them spend more than that on GPUs in a year. That would buy a controller and three 36 GB Fujitsu 15k drives...one drive for the OS, one for games, and one for paging. By separating the three seek-intensive tasks, you'd have a much faster system than running all three off the same spindle or array.

Note: Yes 36 GB is a bit small when it comes to games; you might only fit six or eight modern games on that drive. I'm talking about a system specifically built for competition, not for running 30 different games. And of course larger drives are available, up to 150 GB, for more money.

Click to expand...

Good to know.....
But, the majority of the gamers that I know, myself included, and you apparently based on your sig ( nod to RAID 0 cudas) aren't in the top 2% of competitive gamers and also I don't have $600.00 to spend on disks. So, for <$200.00 I can get 2 120-250GB 7200rpm drives that have 16 MB cache that I can RAID 0 and get the similar, if not better, performance to a single 150GB Raptor, have more total drive capacity, and be cheaper.

Now, I do know guys that have iRAMS for page file disks and multiple Raptors for their OS disks and program disks. But they also have 2 of the new Nvidia 8800's in SLI and a lot more money tied up in their systems than I do.

I haven't ran across many gamers that run SCSI though, not even the guys that sink a mint into their rigs.
Not sure why.

rodsfree said:

So if you want to call it "mass hysteria" that's ok with me.

Click to expand...

Someone observing a bumblebee can say it's flying or not. But what we're talking about is how fast it might be flying, and that's impossible compare without a quantitative measurement. A concensus is not a measurement, and therefore isn't empirical evidence.

Your bumblebee is a perfect example of this. We didn't know how insects flew until we more carefully observed their wing movement. Once it was measured, then simulated, scientists realized that insects really weren't breaking the laws of physics. (See http://www.news.cornell.edu/releases/March00/APS_Wang.hrs.html for a neat animation of the pressure zones created by a dragonfly wing; they twist when flipping back.)

rodsfree said:

So, were any of the files in your BF2 map folder bigger than 128KB?

Click to expand...

Yes. But as I've explained before, file size is irrelevant; what matters is the physical I/O request size, and that the access is actually sequential and not random. You can look at the numbers I've posted and see that the requests BF2 makes of the filesystem are rarely sequential and almost always smaller than a cluster.

You'd really have to stopwatch time installation on the same drive in RAID and not in RAID. Then you'd have to time level loading times for the same level in a given game in order to do a real world test of RAID 0 vs. non-RAID.

I do RAID testing on motherboard controllers all the time using I/O Meter. It's in that part of the mobo reviews most people skip. I use the same drives for single drive benchmarks, and then for RAID-0 testing. I've found some difference between RAID and non-RAID, and quite a bit of difference in some instances between the chipset RAID controller and whatever onboard controllers were added off the chipset.

Currently I use two Samsung Spinpoint 40GB drives, which are actually very quick on their own. They give my 74GB Raptors a pretty good run for their money. They support NCQ and SATA 3G. We run with a stripe size of 16k The differences are always there in the tests, but I can't say I've really noticed the difference. On my home rig, I've run with a single Raptor, and dual Raptors on RAID 0. The performance is basically the same from what I can tell, except level load times in a lot of games are MUCH shorter on the RAID0 setup, which is why I continue to use it.

rodsfree said:

Good to know.....
But, the majority of the gamers that I know, myself included, and you apparently based on your sig ( nod to RAID 0 cudas) aren't in the top 2% of competitive gamers and also I don't have $600.00 to spend on disks. So, for <$200.00 I can get 2 120-250GB 7200rpm drives that have 16 MB cache that I can RAID 0 and get the similar, if not better, performance to a single 150GB Raptor, have more total drive capacity, and be cheaper.

Now, I do know guys that have iRAMS for page file disks and multiple Raptors for their OS disks and program disks. But they also have 2 of the new Nvidia 8800's in SLI and a lot more money tied up in their systems than I do.

I haven't ran across many gamers that run SCSI though, not even the guys that sink a mint into their rigs.
Not sure why.

Click to expand...

The reason is due to lack of performance improvement by using SCSI drives vs. SATA drives. Not to mention there is added complexity of working with a SCSI setup. It isn't hard if you know what you are doing, but most people have never worked with it. Additionally, there is the HUGE cost per megabyte increase over even Raptor drives. Not to mention there is the cost of the controller as well. Good SCSI cards aren't cheap. Plus you can't deny there is also a lack of free PCI slots on most motherboards. By the time you grab a pair of 8800GTX's, and a Creative X-Fi, you are about out of expansion slots on most motherboards. You may have one or two left over PCIe x1 slots if you are lucky. In my case, I have a single PCIe x16 slot running at x8 speeds left over. My SCSI controller happens to be PCI-X (which I don't have any slots for, and I am also out of regular PCI slots), or I'd be running SCSI drives right now.

I've stated this in other threads before. My 10,000rpm Ultra 320 Fujitsu SCSI drive actually out performs my Raptors pretty easily in all the tests I've done. RAID 0 or not. Granted these are IO Meter tests and aren't real world. The controller I have is a true hardware solution, and even though it's old (LSI MegaRAID U320-2 64MB Ram/Battery Backup) it performs very well. Plus I have 4 additional 10,000 RPM SCSI drives I could throw on it, but those are older U160 drives. On a seperate channel for data storage in a RAID5 setup, they'd be pretty sweet.

rodsfree said:

Good to know.....
But, the majority of the gamers that I know, myself included, and you apparently based on your sig ( nod to RAID 0 cudas) aren't in the top 2% of competitive gamers and also I don't have $600.00 to spend on disks. So, for <$200.00 I can get 2 120-250GB 7200rpm drives that have 16 MB cache that I can RAID 0 and get the similar, if not better, performance to a single 150GB Raptor, have more total drive capacity, and be cheaper.

Now, I do know guys that have iRAMS for page file disks and multiple Raptors for their OS disks and program disks. But they also have 2 of the new Nvidia 8800's in SLI and a lot more money tied up in their systems than I do.

I haven't ran across many gamers that run SCSI though, not even the guys that sink a mint into their rigs.
Not sure why.

Click to expand...

My RAID 0 is still there because I don't have the capicity to back up 240 GB and break the array

The beauty of building a good disk system is that it lasts years as a high-end component, as opposed to six or twelve months as is the case with most things. My system is aging a bit right now so I wouldn't spend the money on new disks at the moment, but after the major upgrade I'm planning for summer I probably won't be replacing much for a year or more...which means I might actually look into the 15k SCSI drives for my own system.

I've never compared putting two slower drives up against a somewhat faster drive to see where the advantages were. Our tests were all with 15k drives, because we were looking for the fastest configurations we could find.

Dan_D said:

I've stated this in other threads before. My 10,000rpm Ultra 320 Fujitsu SCSI drive actually out performs my Raptors pretty easily in all the tests I've done. RAID 0 or not. Granted these are IO Meter tests and aren't real world. The controller I have is a true hardware solution, and even though it's old (LSI MegaRAID U320-2 64MB Ram/Battery Backup) it performs very well. Plus I have 4 additional 10,000 RPM SCSI drives I could throw on it, but those are older U160 drives. On a seperate channel for data storage in a RAID5 setup, they'd be pretty sweet.

Click to expand...

Well, what you've got is hands-down the best drive and controller in their class. We tested all the major SCSI drives and found Fujitsu to be just slightly faster (but above margin of error) than the rest. We tested all the major SCSI controllers and found that LSI enjoyed a nearly 10% speed advantage over the rest. We also tested Gb NICs and found that Intel's were the fastest, just FYI

A 10k SCSI will always outperform a 10k SATA because the interface is faster. I don't know the tech specs well enough to discuss exactly why, but SATA drives with near-identical specs are slower than their SCSI counterparts.

Our drive tests consisted of large linear file transfers, fragmented file copies, and random seek tests. The LSI/Fujitsu combo was unbeatable.

edit: LSI also seems to have the lowest overhead on its RAID 5...the slowdown between RAID 0 and 5 was nowhere near as significant with LSI than with any other card.

uberwurst said:

Interesting example sn3ak3rp1mp, but I don't think it was truly "apples to apples"

Click to expand...

Thanks ... I was just trying to throw something together quickly, for my own satisfaction. I did reboot in-between installations, so I doubt caching was an issue.

I wouldn't say it's dead on apples to apples, but it's not apples to oranges either. The only variation between the two tests was where the game was executed from. I stopped all background services before executing the timedemo.

I don't have a stopwatch

The 320 should not have been fragmented, as that is my data storage volume, and is used mainly for large files like DIVX movies, etc. It had about 50GB free at the time I ran this test.

The point was this (and since I'm not a benchmarking lab, take this as my own personal conclusion): I feel I have proven to myself that my RAID-0 can read game files (i.e. maps, images, etc.) faster than I could on a single drive... be that either my 320 SATA or just an individual 120 PATA. Any reference material which explains RAID-0 will back that conclusion up.

Since I do not have any other hardware to test on, YMMV. But I invite others to run the same type of comparison on their own hardware, with other games.

Cheers!
sn3ak3rp1mp

I use HDTach to test performance of my RAID-0 and I get 161.5 MB/sec burst and 104.4 MB/sec average read. Now when I just run my WD 160GB SATA drives as normal, I get 92 MB/sec burst and 54 MB/sec average read. So in my case RAID-0 does make a differance. I also notice a huge difference loading program with one versus the other.

spartus4 said:

I also notice a huge difference loading program with one versus the other.

Click to expand...

What were the measurements in program loading?