RAID 0

[phantom404]

Well for some gamers it might not be worth it but I love being the first one loaded in BF2

 

[Tostada]

Originally posted by: Matthias99
A fairly definitive SR article on this topic.

Finally, we come to the SR Gaming DriveMark, a weighted average of the drive accesses generated by five popular PC games. The flat slopes indicate that gaming uses benefit least from both RAID and command queuing. In fact, at 677 I/Os per second, a 4 drive Cheetah array operating off of the Acceleraid 170 lags a single Raptor running on the "dumbest" of SATA controllers by a margin of 9%.

It's not that it doesn't make it faster -- it just doesn't make enough of a difference in gaming to justify spending any extra money on it, and RAID0 has the issue of reduced overall reliability (although this is usually overblown IMO).

Actually, that article is one of the worst I've seen on SR, and it's pretty indicative of their RAID0 coverage in general. It has next to nothing in the way of actual numbers comparing RAID0 vs. non-RAID setups. They have obviously done a lot of RAID0 benchmarking to write such an article, and it's very strange that they don't bother including the results in their performance database. They really don't have any head-to-head comparison of RAID0 vs. non-RAID that I've seen.

While I agree with them that RAID0 is dumb, they really don't back it up very well, and that's pretty suspcious coming from a site that is essentially a collection of benchmarks.

 

[Matthias99]

Originally posted by: Tostada

Originally posted by: Matthias99
A fairly definitive SR article on this topic.

Finally, we come to the SR Gaming DriveMark, a weighted average of the drive accesses generated by five popular PC games. The flat slopes indicate that gaming uses benefit least from both RAID and command queuing. In fact, at 677 I/Os per second, a 4 drive Cheetah array operating off of the Acceleraid 170 lags a single Raptor running on the "dumbest" of SATA controllers by a margin of 9%.

It's not that it doesn't make it faster -- it just doesn't make enough of a difference in gaming to justify spending any extra money on it, and RAID0 has the issue of reduced overall reliability (although this is usually overblown IMO).

Actually, that article is one of the worst I've seen on SR, and it's pretty indicative of their RAID0 coverage in general. It has next to nothing in the way of actual numbers comparing RAID0 vs. non-RAID setups. They have obviously done a lot of RAID0 benchmarking to write such an article, and it's very strange that they don't bother including the results in their performance database. They really don't have any head-to-head comparison of RAID0 vs. non-RAID that I've seen.

While I agree with them that RAID0 is dumb, they really don't back it up very well, and that's pretty suspcious coming from a site that is essentially a collection of benchmarks.

Um... this whole page from that article is comparisons between single drive and RAID0 setups on the different controllers. What more do you want?

 

[Tostada]

Originally posted by: Matthias99
Um... this whole page from that article is comparisons between single drive and RAID0 setups on the different controllers. What more do you want?

So, you can't read today?

I've already said twice in this thread that it is strange -- given that they have obviously run the benchmarks -- for them not to include the actual numbers in their performance database.

What do you mean, "What more do I want"? I want performance numbers of RAID vs. non-RAID. That page you linked just has line graphs comparing different controllers.

Best... smiley... ever. I love that guy!

 

[Matthias99]

Originally posted by: Tostada

Originally posted by: Matthias99
Um... this whole page from that article is comparisons between single drive and RAID0 setups on the different controllers. What more do you want?

So, you can't read today?

I've already said twice in this thread that it is strange -- given that they have obviously run the benchmarks -- for them not to include the actual numbers in their performance database.

What do you mean, "What more do I want"? I want performance numbers of RAID vs. non-RAID. That page you linked just has line graphs comparing different controllers.

Read the frickin' article! The graphs are comparing different controllers with 1-4 drives on each one (that is, non-RAID versus 2-4 drive RAID0). Those ARE the performance numbers you're asking for! At the bottom they also give numbers for RAID1 and RAID10.

 

[Tostada]

Originally posted by: Matthias99
Read the frickin' article! The graphs are comparing different controllers with 1-4 drives on each one (that is, non-RAID versus 2-4 drive RAID0). Those ARE the performance numbers you're asking for!

I did read the article. And it's not like it wasn't painful -- that's a lot of words!

My point is those are some pretty weak comparisons, and the graphs themselves are just comparing controllers. Looking at those graphs, it's more of an argument for RAID0, because it's pretty obvious that the lines are going upwards as the number of drives increase.

For example, they say gaming benefits least, but the graphs show quite obviously that it does benefit.

I just think it would've been simple for them to include a few RAID0 arrays in their regular performance database. That's all I'm sayin'.

 

[Bobthelost]

I don't think you do want RAID results in the performance database there, you'd have to include RAID results for a variety of different setups as well as different numbers of drives, all of which would clutter up the database needlessly. If RAID is as useful as it's supposed to be all that is needed is a comparison of the best RAID controllers availible and a comparison to a Raptor 150 so you can see just how much benifit you get from it.

 

[Matthias99]

Originally posted by: Tostada
My point is those are some pretty weak comparisons, and the graphs themselves are just comparing controllers.

I'm not sure what you mean, because the graphs on the 'desktop' page are clearly comparing both between different controllers and with the same controller running a single drive or RAID0.

Yes, the article is about SCSI versus SATA RAID, but they also include some numbers on RAID0 versus single drive performance.

Looking at those graphs, it's more of an argument for RAID0, because it's pretty obvious that the lines are going upwards as the number of drives increase.

For example, they say gaming benefits least, but the graphs show quite obviously that it does benefit.

The benefit is pretty paltry for gaming. A four-drive RAID0 of 10KRPM SCSI Cheetah drives is still slower than a single Raptor. Going from one Raptor to a four-drive RAID0 of Raptors bumps their 'gaming' benchmark from ~740 to ~780 -- a whopping 5% improvement, which is nothing considering how much better it makes the STR. Clearly, drive STR is not the limiting factor in gaming situations.

I just think it would've been simple for them to include a few RAID0 arrays in their regular performance database. That's all I'm sayin'.

I think that's a great idea.

 

[Tostada]

Originally posted by: Bobthelost
I don't think you do want RAID results in the performance database there, you'd have to include RAID results for a variety of different setups as well as different numbers of drives, all of which would clutter up the database needlessly. If RAID is as useful as it's supposed to be all that is needed is a comparison of the best RAID controllers availible and a comparison to a Raptor 150 so you can see just how much benifit you get from it.

But just a couple common configurations, like showing two Raptor 74s in RAID0 or two cheap 250GB drives in a RAID0 running off an NForce RAID0...

It's such a nice feature being able to take any two drives on SR and look up all their benchmarks head-to-head. Throwing in some RAID0 setups would help people put things in perspective.

What you're saying, though, is probably why they don't do it. They couldn't really keep all the other parameters the same. If they used their own RAID controller, that would probably over-inflate the benchmarks (since onboard RAID would probably perform worse), and if they used onboard RAID, that would be an alteration of their testbed.

 

[Tostada]

Originally posted by: Matthias99
The benefit is pretty paltry for gaming. A four-drive RAID0 of 10KRPM SCSI Cheetah drives is still slower than a single Raptor. Going from one Raptor to a four-drive RAID0 of Raptors bumps their 'gaming' benchmark from ~740 to ~780 -- a whopping 5% improvement, which is nothing considering how much better it makes the STR. Clearly, drive STR is not the limiting factor in gaming situations.

Perhaps if we could consolidate this into one pretty picture titled "Your RAID is useless" ...

But then, as the AT article shows, RAID0 really dominates the "File Zip Test" and the "Copy Folder Test"! The 7200RPM Barracuda RAID is fully twice as fast as a Raptor 74 copyin those folders!

Of course, those being the only two tests that the RAID0 really dominates, it's really funny when you realize that performance would probably be better if you split the drives and did the ZIP (or copy) from one drive to the other instead of within the array.