RAID 0 or faster processor for increased desktop performance

D

Dark Prodigy

Jawbreaker
Joined
Mar 10, 2006
Messages
2,803
Doing a small workstaion type build and can't make the decision whether to go with RAID 0 or faster processor for increased desktop performance.

Possible setup 1: pentium D 820 and 2 Samsung Spinpoints in RAID 0

Possible setup 2: 3800+X2 and 1 Samsung Spinpoint

Computer will be mainly for office type work, no gaming. Microsoft Office, home networking, web surfing etc...

Help me out... thanks.
 
My thoughts are:

either dual core will give you a smoother multi-tasking enviroment and both have plently of horsepower for your listed uses, good call :D

The raid 0 will give you slightly faster load times on your applications but keep in mind you just doubled the chance of losing all your data if a drive crashed. Unless you had a application that constantly was going to the HD I would recomend against it.

You didnt mention how much memory you were going to install. For your purposes I would go with 2GB of memory then see what you have left for the other options and then I would tend to go with faster processor and one drive.
 
I'd actually say that memory is your most important asset. RAID0 only improves upon sequential transfers, and most disk latency in your case is going to be seek times, which RAID actually hinders if anything (especially software RAID, which can really kill latency.) You might notice, at best, a 20% speed increase for loading larger applications. Actually running an office app will show little if any improvement, and overall efficiency may decrease.

For the most part, RAID0 is only truly beneficial for manipulating very large files (think video editing.) It can provide a performance increase for gaming PCs, but the return on investment is relatively modest.

You also lose some reliability with RAID0, don't forget. If either drive fails, you lose everything. Statistically, RAID0 is the most failure-prone of any disk configuration.

CPU speed increases also show diminishing returns for basic productivity software once you get past the budget CPUs. Even on a cheap modern system, the only thing that'll come close to taxing the CPU is the occasional Java or Flash component on a webpage. For the average user, CPU is the one place where there's almost always power to spare.

Memory, however, is something you almost can't have too much of. More memory means both better disk caching and less chance that you'll have to read anything from swap. The former helps considerably with disk I/O, and the latter can make the difference between a system that screams and a system that crawls. Windows XP really likes to have a full gig available, and if you're a hardcore power user, you can leverage more (though for most people, 1G is more than enough.)

IMO, if in doubt, more RAM is usually the best place to sink your money into.
 
Since this is somewhat of a budget build, I'm going with 1GB Corsair ValueRam (mem prices these days are ridiculous). I understand the risks of RAID 0. I have RAID 0 on 2 of my other machines for the better part of a year with no problems.

Or maybe I should simply go with the E6300 Conroe processor. I understand the importance of memory, but adding another 1GB of memory will be around 40% more expensive than choosing either processors over the 820 D and 1 HDD.
 
hmm I love corsair value ram for business builds. one bad stick out of hundreds.

I love my conroe too but thats almost overkill for office and not overclocking it is a sin (just to clairify, I would not overclock a machine used in an office environment, well not much anyway ;)

LOL ok - final advice, use the cheaper processor one drive one GB and take the extra money and take girlfriend/wife/dog/whatever and go have a good time.
 
I simply want my programs to open and close quickly. I hate it when there is that delay between pressing a button on the start menu til the app opens up, making it feel 'sluggish'. And what directly contributes to that... RAM? Faster processor? HDD performance?
 
hard drive rpm and ram are your main culprits. and fragmentation. processor speed has no real noticable effect on the type of work you want to do.
 
Whatsisname said:
RAID 0 is completely useless.
Click to expand...
Literally absolutely wrong. It is not 'completely' useless at all.


IMO, RAID-0 is worth it and having gone raid 5 years ago, no way in hell would I go back. Why? One example; it's the reason in BF2 I've already taken the first flag and am moving onto the next before all the non raid muppets even get into the game. BF2 sequentially reads all the key game data files after load to verify you're not cheating. This is a sheer read stress and with RAID-0 you gain a massive load time advantage and one which, as I've noted, gives you a massive game advantage. Can utterly change how a map plays out.

It also helps all the little things, windows loads a bit faster, everything is a bit snappier, pagefile access is improved, it all adds up.


The reason most people don't bother with raid is not because of the supposed lack of added performance, but because they lack the effort to get it setup. Given how cheap hard drives are I don't understand why people are so hesistant to get 2 smaller drives and RAID them up.

As for reliability - lol. People are always quick to mention the literal fact that 2 drives double the chances of losing your data 'cos if ones goes, it's over. But that's never happened to me once. Not even when I had 4 drives in Raid-0 on an overclocked PCI bus at 41mhz. 1 cutting edge high rpm drive is more likely to fail than 2 average speed decent drives. Probably cheaper to get the 2 lower speed drives as well.
 
spine said:
Literally absolutely wrong. It is not 'completely' useless at all.


IMO, RAID-0 is worth it and having gone raid 5 years ago, no way in hell would I go back. Why? One example; it's the reason in BF2 I've already taken the first flag and am moving onto the next before all the non raid muppets even get into the game. BF2 sequentially reads all the key game data files after load to verify you're not cheating. This is a sheer read stress and with RAID-0 you gain a massive load time advantage and one which, as I've noted, gives you a massive game advantage. Can utterly change how a map plays out.

It also helps all the little things, windows loads a bit faster, everything is a bit snappier, pagefile access is improved, it all adds up.


The reason most people don't bother with raid is not because of the supposed lack of added performance, but because they lack the effort to get it setup. Given how cheap hard drives are I don't understand why people are so hesistant to get 2 smaller drives and RAID them up.

As for reliability - lol. People are always quick to mention the literal fact that 2 drives double the chances of losing your data 'cos if ones goes, it's over. But that's never happened to me once. Not even when I had 4 drives in Raid-0 on an overclocked PCI bus at 41mhz. 1 cutting edge high rpm drive is more likely to fail than 2 average speed decent drives. Probably cheaper to get the 2 lower speed drives as well.
Click to expand...

Last time I played BF2 at a lan event, I was into the game quite a bit sooner than everyone else using my RAID-0 Raptors. Though that is not conclusive, because my system is generally quite a bit more powerful than all the PC's of my friends.
 
Adidas4275 said:
raid5 w/ 3 drives
= speed, redundancy
Click to expand...
the truth of that statement depends heavily on the way that RAID-5 is implemented: software vs. hardware. Also, according to most literature, it seems that RAID-5 write speeds are not not excellent.

spine said:
As for reliability - lol. People are always quick to mention the literal fact that 2 drives double the chances of losing your data 'cos if ones goes, it's over. But that's never happened to me once. Not even when I had 4 drives in Raid-0 on an overclocked PCI bus at 41mhz. 1 cutting edge high rpm drive is more likely to fail than 2 average speed decent drives.
Click to expand...
Probability and statistics are difficult subjects. It appears that many people like make the mistake to generalize from their, commonly rather limited, sample size. This often leads to statements like "I have had n thingymabobs and none broke." Often times n is a number that is less than 10, while the population of thingymabobs is in the double digit millions. I am curious how likely this sample is a representative one.
 
drizzt81Probability and statistics are difficult subjects. It appears that many people like make the mistake to generalize from their said:
I totally understand that. But I was making a point that the failure rate of drives is way over exagerated and so RAID-0 isn't as massively dodgy and unstable as people like to make out. I have to say I've never had a drive die on me.

Actually, recently, one of the drives in my current SATA RAID-0 array overheated* and now has some physical bad sectors on it. I painstakingly worked out where they were and partitioned around that area. Fully formatting the area didn't work as I tried that first. It wrote off a few bad sectors but some remained. Anyway, point is, even with that physical failure, I kept all my data no problem. I've since added cooling to the drives and haven't had a problem since.

If a drive properly dies, it dies. Raid or no raid, you're fucked either way.


*Off the point; the drives overheated because I'd wrapped them in foam to keep the noise down. Without direct contact to the metal case housing to use as a heatsink they got pretty damn hot.
Click to expand...
 
spine said:
But I was making a point that the failure rate of drives is way over exagerated and so RAID-0 isn't as massively dodgy and unstable as people like to make out. I have to say I've never had a drive die on me.
...
If a drive properly dies, it dies. Raid or no raid, you're fucked either way.

*Off the point; the drives overheated because I'd wrapped them in foam to keep the noise down. Without direct contact to the metal case housing to use as a heatsink they got pretty damn hot.
Click to expand...
While your experience has been great with RAID0, it is, however, anecdotal. While working at a data center (while a student), I replaced/rebuilt countless drives/arrays. While the failure rates of drives while in raid arrays may seem like an exageration to you, I can assure you, it is not. They're not exagerating failure rates, they're telling you the probability of data loss -- its as simple as that. The above statement (that RAID0 doubles your chances of data loss) is true because of simple math. Nothing is certain, everything is probable.
 
Yes, of course it's going to be double the risk, no denying that. I'm just saying that even then the actual risk isn't that great. Buy a decent reasonable speed drive, raid it up for 2 years and move on and you'll probably have no problems.

I'm just making the point that people who stay away from raid sheerly because of this doubled chances are being overly cautious.

Well whatever, I'm fine with the 'risks', if others are not it's their call I spose.
 
spine said:
I totally understand that. But I was making a point that the failure rate of drives is way over exagerated and so RAID-0 isn't as massively dodgy and unstable as people like to make out. I have to say I've never had a drive die on me.
Click to expand...
and I have had ~4+ die on me. Yet you won't see me running around telling everyone in the world to RAID-1 their drives, since harddisks are bound to fail, do you? Both of our views are most likely very limited and due to a large number of variables that can influence the time of failure, it is most likely a random process. As such individual experiences with small sample sizes are mostly irrelevant when it comes to the expected value. The probability of a hard drive dieing may be rather small, but regardless of its magnitude, if losing one drive will cause the data on both to disappear, the probability of this is still approximately twice that of losing only data on a single drive.

spine said:
If a drive properly dies, it dies. Raid or no raid, you're fucked either way.
Click to expand...
I never suggested otherwise. However, the magnitude of data loss and its probability is different:
If you are losing 2x the data with 2x the probability, given a constant cost of data, the expected loss from a RAID array is 4x that of a single drive. The expected loss from two independent disks is 2x that of a single drive, given that all drives are of the same size. Of course, you can store more on a raid array or the two independent disks....

Regardless of this, I shall take my advice that I gave to Djnes recently and stop bothering and leave this thread. In the end, I have nothing to gain/ lose from proving my point.

Have a nice day.
 
820/RAID0 = meh

3800+ X2 = meh

C2D E6300 = similar price to the X2 with better performance.
 
Buying a lower end Netburst 820 processor or "low performance" X2 3800+ seems foolish and would negate any benefit that Raid0 would give you. I would go with an E6300 or even E6400, crank the clocks up a few hundred MHz and call it a day.

I do agree that Hard drives are the MAIN bottleneck these days when it comes to desktop performance. It is extremely noticeable when the hard drive is seeking constantly and the LED on the case is solidly lit.... I think this is what causes a lot of the lag when opening programs/extracting multiple large files/multitasking.

I don't think RAID0 is really the big fix for this problem, I think we'll just have to hope that hard drives advance to be faster and although perpendicular recording is definately a bit of a performance boost, we need to see the whole mechanical storage medium abolished but that will take a long, long time.

I think your best bet is to wait until the Hybrid drives come out later this year (or maybe they will be delayed who knows) and that way if you get Vista, you will be at an advantage where the solid state cache will be used heavily, decreasing hard drive spin ups and seek times.
 
filthysanchez said:
gigabyte I-RAM with the most important programs on there?
that'd be pretty quick.
Click to expand...

Not in the budget.


So.. here is what I'm thinking now. P 820 D, 1 GB Corsair ValuSelectRam, 2 Samsung Spinpoint P series HDD's RAID 0'ed, Win XP Pro, XFX 6500 PCI-e (for larger DVI monitor), 350w Cooler Master PSU. Dual core, inexpensive, workstation type specs.

I do agree that Hard drives are the MAIN bottleneck these days when it comes to desktop performance. It is extremely noticeable when the hard drive is seeking constantly and the LED on the case is solidly lit.... I think this is what causes a lot of the lag when opening programs/extracting multiple large files/multitasking
Click to expand...

I agree also. Raid 0 gives 'snappier' desktop performance. I've RAID 0'ed 4 PC's and they none of them seems sluggish in the least bit. And plus for this particular workstation I'll probably employ an external HDD for daily backups at a later date, so potential data loss isn't that much of a concern.
 
InternalStorm said:
I do agree that Hard drives are the MAIN bottleneck these days when it comes to desktop performance. It is extremely noticeable when the hard drive is seeking constantly and the LED on the case is solidly lit.... I think this is what causes a lot of the lag when opening programs/extracting multiple large files/multitasking.
Click to expand...
if this is something you do on a frequent basis, may I suggest investing in two dedicated harddrives? one for the source files and one for the destination (i.e. compressed) files. Another option would be to increase your system's RAM and configuring the compression utility to be aggressive with RAM usage, i.e. only write to disk after the in-memory compression is done.

InternalStorm said:
I think we'll just have to hope that hard drives advance to be faster and although perpendicular recording is definately a bit of a performance boost, we need to see the whole mechanical storage medium abolished but that will take a long, long time.
Click to expand...
Would you mind indulging me on what is special about perpendicular recording that allows it to boost performance more than the usual data density increase?
 
You must log in or register to reply here.
Back
Top