Strip size with raptors and adding more drives

[afireinside]

Running 2x 74gb 8mb cache raptors in RAID 0 with a 64k strip on NVraid. MUCH slower than with a 36gb matrix partition on intel. Is 64k the optimal strip size for gaming or should I go with 16k? Also would I see gains from adding another raptor or should I sell them and go with seagate drives?

 

[tuskenraider]

Running 2x 74gb 8mb cache raptors in RAID 0 with a 64k strip on NVraid. MUCH slower than with a 36gb matrix partition on intel.

How so? Feel? Benchmarks? Maybe timed a couple apps loading?

Is 64k the optimal strip size for gaming or should I go with 16k?

I bet it's more setup/controller dependant. I was at 32K for awhile and went to 16K currently, can't say I've noticed a difference. The smaller stripes usually benchmark better but that isn't very helpful.

Also would I see gains from adding another raptor or should I sell them and go with seagate drives?

In sustained transfers, but not in seeks times, so I wouldn't think you'd benefit from it. Why would you go to slower Seagate drives?

 

[afireinside]

Fells slower installing, I can tell it's slower loading, and HDtach is showing 2ms slower seek and 40mb less sustained read. I've heard the seagates are faster but nothing solid.

 

[tuskenraider]

Well there's clearly an issue with the chipset RAID controller to lose that much performance, which changing drives isn't going to correct. I can't really help with that since I haven't used Nvidia's stuff since NForce 2, but I've read somewhere NVidia's cotrollers can be real driver dependant though and maybe that might be something to look into and/or experiment with. Maybe check out the motherboard forums for some additional info or feedback. Good luck.

 

[afireinside]

Well the main reason it's slower is because matrix raid kicks *** and I was able to make a 36gb partition using the fastest part of the disks. Wondering if 3 or 4 drives on nvraid will make up for the loss.

 

[Maviryk]

Well the main reason it's slower is because matrix raid kicks *** and I was able to make a 36gb partition using the fastest part of the disks. Wondering if 3 or 4 drives on nvraid will make up for the loss.

Oh sure... don't even thank me for my help over aim

JP

NVRaid does have a lot of issues though.

Maybe you can get a PCI or PCI-e addon raid controller?

 

[tuskenraider]

Well the main reason it's slower is because matrix raid kicks *** and I was able to make a 36gb partition using the fastest part of the disks.

If you put 50GB of OS/apps on a matrix array or on a standard array on the Intel chipset, they'll both perform just as fast. Just because you cut out the slower part of the drive with the matrix to get the benchmark to avoid counting the inner unused/slower space, doesn't mean the data isn't at the edge of the disk platter in either scenerio. It will be accessed at that same speeds. If you don't have a benchmark of your Raptors in a typical array(single RAID0 array using all the space on both drives) on the Intel chipset, how can you accurately compare it to the NVidia setup? You can't.

 

[Maviryk]

He told me over AIM the STR of his 36's on MATRIX RAID0 was around 130Mb/s whereas the 74's were 100Mb/s through HDTach. That and the 36's were partitioned on the faster 18Gb sectors of each drive.

Aside from the benchmarks, his apps loaded slower.

 

[tuskenraider]

Well I'm confused, he noted a 36GB Matrix array created out of two 74GB drive's in the first post. Now you're saying he had 36GB drives on the Intel system and now has 74GB drives on the Nvidia? Either way, it's the RAID controller causing his problems and so without a change of motherboard or trying an add-in RAID card, the performance isn't gonna match his old setup. Nothing to do with the drives.

 

[greenmaji]

Your actuall performance should be the same on the NV raid0 if you partition the array when installing the OS.

But don't expect it to benchmark the same, the benchmarks ignore partitions IIRC.

 

[tuskenraider]

Your actuall performance should be the same on the NV raid0 if you partition the array when installing the OS.

Well I wouldn't expect the same performance from two different controllers, yet alone on two different mobo's. I get vastly different results between the Promise and VIA controller on my current mobo. As for the partitioning, it isn't going to speed anything up as all the data is written from outside to inside on the disk and as mentioned, you can't benchmark a partition.

 

[afireinside]

Well I'm confused, he noted a 36GB Matrix array created out of two 74GB drive's in the first post. Now you're saying he had 36GB drives on the Intel system and now has 74GB drives on the Nvidia? Either way, it's the RAID controller causing his problems and so without a change of motherboard or trying an add-in RAID card, the performance isn't gonna match his old setup. Nothing to do with the drives.

I have 2*74. I ran a 36gb raid 0 matrix partition and 50gb raid 0 on intel. On AMD I run a 148gb RAID 0 partition.

 

[tuskenraider]

I see. I noted the impossibility of fairly comparing your old Matrix setup to you current RAID setup via benchmarks, though you could fairly compared the old 74GB RAID setup to the new, if you saved some screenshots or something. But without you having some kind of method of recording a loss or improvement of performance beyond "feel", I don't see how to really help you for a "problem" that may not exist.

 

[greenmaji]

Partitioning before installing forces the data to be written to those partitons, and the first partiton is made in the center of the drive much like matrix raid makes a second logical drive.
and yes the differince will be in the controller and ICHxR has had a bandwith advantage for quite some time, but I wouldn't say thats something you can even make up for so I wouldn't even consider it.