Do I need a fast system drive if i also have a RAID 0 array?

It certainly helps general response time to use a fast system drive in addition to a data drive in RAID. The thing is, file transfer speed isn't really all that important in many system applications, there simply isn't enough data in any given file for the drive to get fully up to speed. What does make a big difference is seek time, and that's why raptors shine in system performance. So if you can afford it, go for it.

OHH, Seek time - I hadn't been looking at that factor. Thanks.

The SATA drive i am looking to replace my IDE system drive with ( Western Digital Caviar SE16 WD2500KS 250GB) has a seek time of 13.5 ms. This versus my IDE drive seek time of 14.1ms.

I dont think thats much of a difference so its not worth me swapping the drives and swapping the OS from EIDE to SATA.

A fast system drive is always faster, but is it worth it? In your case, I suspect not. Once you have loaded your application, it will not be accessed much. With four work drives, I would like to see you conduct a test. 1) Run your application putting the inputs on one set of drives, and your outputs on a second set of drives. 2) Run your application on the raid-0 system as planned. If your application is primarily sequential in nature, as I suspect, I'm betting that you will do better without raid-0. With raid, the disk arm will be constantly be moving from source to target. With separate drives, it can stay on the cylinder longer, and there may be some sequential cache read-ahead. If you can try this test, please post your results.

MAny SATA drives do 3gb/s while many ides only do 1gb/s...thats why...

Ahem.... OS? Is it just 'magically' going to load?

I think you've gotten you read/write specs confused for the two types of drives listed.

I'm not a graphics guru, (there are a few on here) but given what you've said you'd gain just a wee bit by upgrading your system drive to SATA. What you'd gain most is future-proofing your system as most (if not all) new MB's will only include one PATA port and the rest will be SATA.

However, your existing drive will not hurt you.

geofelt, im doing video editing, and ive never heard of anyone putting there application on the RAID array - if thats what you are trying to suggest.

People that build high performance systems for editing, always have the system drive with OS and apps, and then a RAID 0 for media. -video audio images.

fishquail, I am not suggesting raid of any kind. I don't think it is worth it, in general. Many here swear by it though, and I thought that if you had the means to test out several configurations, then we all could learn something from the real world, instead of synthetic benchmarks. The loading of your application/program from the system disk should be a minute part of what you do, compared to the processing of your data. Spending much time or dollars optimizing that is porbably not worthwhile. Testing out different configurations of hard disks for your data, on the other hand, might be very useful. I suspect that using four independent non-raided drives might better than any raid configuration. But I could be wrong .

I spent AAAAAAAAAAAGES looking thru benches to find out which drive is actually fastest and best to put in a RAID0 array, ages.


I came up with this:

Seagate ST3250820AS 250GB 7200RPM SATA300 8MB Cache


If you check benches like Tom's. Its actually extremly fast for a 7200RPM drive. In a RAID0 array you get 500GB total space aswell. Im putting 2 in mine in RAID0.

I have yet to see a realistic benchmark of raid. Firstly, they are synthetic workloads which may, or may not reflect what you and I might want to do. Next, they compare systems of different sizes. We know that the fastest part of a hard drive is on the outer cylinders. If you compare two drives in raid-0 to one, the two drives will have more data on the fast parts. Even then, the synthetic transfer rates look impressive for raw synthetic data transfer, but the application workloads show hardly any difference. For a system like the op's, I think he will be vastly faster with inputs on two drives, and outputs on two drives, all w/o raid. I would love to see a real world test.

"MAny SATA drives do 3gb/s while many ides only do 1gb/s...thats why..."

Did everyone miss this comment? or is everyone just ignoring it?
I'm not sure how to approach this one, but here goes.

SATA theoretically can handle ~3 Gb/s (haven't checked it in a while, not sure on this number) while PATA theoretically handles ~133 Mb/s. The problem is that there are no hard drives made in these configurations that can saturate the bus. I'm not sure, but I think even the latest SCSI drives wouldn't saturate either one.

SATA has become the new standard as it has other advantages over PATA but the theoretical max transfer speed was not one.

I hope this made sense,

HTH

Well, video editing benchmarks do show improvement with RAID0, as they should. Particularly uncompressed video, so that there's no processor load. This is the situation where RAID 0 shines.

OP: A couple of comments:
1) You seem to be spending quite a bit on the system, so it might actually make sense to get a dedicated RAID card. It will probably outperform nVRAID and allow for easier migration
2) In terms of the main system drive. The apps you mention to benefit from faster drives - their startup does take a bit. The main question is if it's worth spending additional $$ to get slightly better performance, and the answer to that question really depends on what kind of drive you have (the EIDE one). UATA 133 is not a limit, I don't think, yet, but if the drive is older then the new ones are probably faster. Raptor could be overkill; it's cost / capacity or even price / performance ratio is pretty bad by today's standards. There are probably better options.

My $0.02