Bottlenecked.....What if I?

[Krypto]

I just ran my second instance of PCMark Vantage, and it's transparent as to how bottlenecked I am with the HDD package Im running.

So I want your folks input and opinion.

I currently have a RAID 0 with 2 Seagate 750's (ST3750330AS)

I have 2 Raptor 150's sitting in the office, a WD1500ADFD and a WD1500AHFD. The exact same drive except that one has the acrylic cover for platter viewing.

Here is a Comparison (http://www.newegg.com/Product/Productcompare.aspx?Submit=ENE&N=50001306%2040000014%201035507821&bop=And&CompareItemList=N82E16822136011%2CN82E16822136012% 2CN82E16822148298) of all 3 drive types. I'm just caught up on spec grinding. :bang head

The HDD build I'm running now utilizes 7200 RPM SATA 3G drives that have 32m cache. If I striped the Raptors, I could see how 10,000 RPM would be a performance increase but wouldnt the SATA 1.5 and 16m less cache hinder the performance? or at least offset one another?

[CGR]

I just ran my second instance of PCMark Vantage, and it's transparent as to how bottlenecked I am with the HDD package Im running.

So I want your folks input and opinion.

I currently have a RAID 0 with 2 Seagate 750's (ST3750330AS)

I have 2 Raptor 150's sitting in the office, a WD1500ADFD and a WD1500AHFD. The exact same drive except that one has the acrylic cover for platter viewing.

Here is a Comparison (http://www.newegg.com/Product/Productcompare.aspx?Submit=ENE&N=50001306%2040000014%201035507821&bop=And&CompareItemList=N82E16822136011%2CN82E16822136012% 2CN82E16822148298) of all 3 drive types. I'm just caught up on spec grinding. :bang head

The HDD build I'm running now utilizes 7200 RPM SATA 3G drives that have 32m cache. If I striped the Raptors, I could see how 10,000 RPM would be a performance increase but wouldnt the SATA 1.5 and 16m less cache hinder the performance? or at least offset one another?

No drives even use more than 1.5gb anyway, so that isnt the issue.

Cache is over rated. Some test you MAY see a difference, but you woudlnt see any decrease in real world apps.

[Krypto]

No drives even use more than 1.5gb anyway, so that isnt the issue.

Cache is over rated. Some test you MAY see a difference, but you woudlnt see any decrease in real world apps.

So your saying RAID 0 with Raptors eh?

I'm really looking for increased performance. Maybe I should go for 3 raptors?

[CGR]

So your saying RAID 0 with Raptors eh?

I'm really looking for increased performance. Maybe I should go for 3 raptors?

Yea, thats what I would do in the same situation. I dont think you will see much, if any, performance increase with three drives.

[aja]

As you add drives you increase access times due to the added controller latencies.

So it depends on what type of workload the drives will experience.

Webservers need low access times as they continuously seek and access small files, whereas a HTPC could do with better transfer rates as mostly large files are used.

Games tend to lose performance with RAID 0 in some cases due to the increased access times.

One way to add drives and thereby increase the transfer rates, but not increase the access times is to use a MATRIX RAID setup, or to use faster drives (in terms of access latencies)

So it is up to you, how much money you have and how much time you have to setup the RAID.

I would use those two raptors in RAID 0. Their low access times means that even if they are placed in RAID 0, the added controller latencies will not cause the overall access times to be much higher than a single drive would exhibit.

Adding drives could negatively affect performance if the controller is not up to the task. I saw a review a while ago comparing the effects of adding more drives - there was a linear performance gain for each drive added up to the fourth drive. After that the gains were negligible and 8 drives even showed a performance hit in comparison to 4 drives (obviously the controller latencies kicking in)

However, it is very dependent on the type of load placed on the drives.

[Krypto]

As you add drives you increase access times due to the added controller latencies.

So it depends on what type of workload the drives will experience.

Webservers need low access times as they continuously seek and access small files, whereas a HTPC could do with better transfer rates as mostly large files are used.

Games tend to lose performance with RAID 0 in some cases due to the increased access times.

One way to add drives and thereby increase the transfer rates, but not increase the access times is to use a MATRIX RAID setup, or to use faster drives (in terms of access latencies)

So it is up to you, how much money you have and how much time you have to setup the RAID.

I would use those two raptors in RAID 0. Their low access times means that even if they are placed in RAID 0, the added controller latencies will not cause the overall access times to be much higher than a single drive would exhibit.

Adding drives could negatively affect performance if the controller is not up to the task. I saw a review a while ago comparing the effects of adding more drives - there was a linear performance gain for each drive added up to the fourth drive. After that the gains were negligible and 8 drives even showed a performance hit in comparison to 4 drives (obviously the controller latencies kicking in)

However, it is very dependent on the type of load placed on the drives.

I believe a Matrix RAID is proprietary to Intel. Curious what my board has to combat that?

[Endgame124]

I believe a Matrix RAID is proprietary to Intel. Curious what my board has to combat that?

I'm pretty sure you can just partition your raid 0 array for a similar effect. I believe the benchmarking tools don't respect a regular partition, however, so you will have a hard time benching to prove if I'm right or wrong.

Perhaps someone else has a little more concrete evidence that Partitioning will create a similar effect?

[Krypto]

I'm pretty sure you can just partition your raid 0 array for a similar effect. I believe the benchmarking tools don't respect a regular partition, however, so you will have a hard time benching to prove if I'm right or wrong.

Perhaps someone else has a little more concrete evidence that Partitioning will create a similar effect?


Well that may be why my PCMark scores are so low. I have a partitioned RAID 0.

So what your saying is that benchmarking wont get an accurate reading?

[Krypto]

I just verified that the partition may be the culprit. HD Tune write test would not initiate.

Message:

Writing is disabled.
Please remove all partitions!

[tuskenraider]

Well that may be why my PCMark scores are so low. I have a partitioned RAID 0.

So what your saying is that benchmarking wont get an accurate reading? Typical benchmark programs won't benchmark a partition, just the whole drive space. So if you created a small partition to keep your OS/app files closer together, which shortens the stroking of the drive, you'll in effect have quicker average access times. But when you mention RAID here, the Matrix RAID roaches come crawling out. :rolleyes: Instead of creating partitions, Intel's ICHxR series of RAID controllers can create two arrays on the same set of drives. You can confine the OS/app files to one small RAID0 array like above mentioned partition, but benchmarks programs see arrays as two seperate drives. This allows you to show the improvement you get via a picture of the benchmark and pictures are worth more than words. The effect is similar to why stained glass windows are used in churches...........:-/

[Mr.Guvernment]

the write test only works in the paid version and apparently has some issues with some raid setups as said above

Intel Matrix raid is Intel only, the only thing to compete is an add in card - nvidia raid sucks

[Krypto]

Well I was going to delete the second partition and then extend C.

Then I realized I must have made a goof during the initial install lol.

Can you see it? :beer: You can only extend to the right from what I remember, so that idea just went down the drain.

Not gonna use a partition manager.... might as well do a format reinstall

[aja]

I believe a Matrix RAID is proprietary to Intel. Curious what my board has to combat that?

Oh yeah, forgot about that! sorry :(

[Krypto]

Bah.... Just picked up a Velociraptor on Newegg. Got it shipped overnight so it should be in by Friday. Not gonna mess with the RAID for now, I want to give it some time to see what's to come on the SSD stage.

So if anyone needs HDD's let me know.. I've got 3 Raptors and a couple 750's.

Man... hardrive shopping is a B^&+H. They never perform the way you want 'em to