Intel SSD (X25-E & X-25M) RAID 0 Benchmarks

[RealTelstar]

per request, ATTO results (X25-E, Raid 0)

1. Are you using the latest version (2.34)?
2. You should put Total Length to 32mb and I/O to 'neither'. The latter is essential.

Please redo the test with these settings.

[SkOrPn]

I don't think you'll find ANY PCIe cards that are NOT videocards able to use PCIE 2.0 bandwidth. This was specifically introduced for video. Given the glacially slow rate of uptake of PCIe by hardware controller manufacturers changing from PCI, and the probability that they are going into boards that can only support PCIe 1.0 anyway (and the chip issue mentioned below), I bet there's little chance of a PCIE 2.0 controller card soon.

Halve those numbers and you might be closer. And even then the throughput is limited by the IOPs chip on the card for wider bandwidth slots, not the slot width, so IIRC PCIe 1.0 x4 is about the best bandwidth you'll get whatever card you have and even if it's designed for/plugged into x8.

Although I do agree with you and I was not aware that these cards are not already PCI-e 2.0 I think with that said Im sure all raid chip manufacturers are hard at work making sure they are up to speed soon. SSD's are obviously going to saturate the SATA II bus by the end of next year and if you have say 32 SSD's on a controller your going to want every MB/s and IOP you can get for your money. Im sure Adaptec would agree with me... I bet we will be seeing x16 pci-e 2.0 cards by 2010 easy...

[alfaunits]

2GB/s IOPs, however cruious in the text, aren't really useful in bulk of the cases... except file copy essentially. With that much disk I/O, the processor and memory will become the big bottleneck
It's getting very fast random read/write performance that most want now and will want in the near future.

videy, any luck with IOMeter?

[m^2]

2GB/s IOPs, however cruious in the text, aren't really useful in bulk of the cases... except file copy essentially. With that much disk I/O, the processor and memory will become the big bottleneck

Usually chipset becomes the limit first.

[imperium]

You said the controllers maximum throughput is 666MB/s, but the burst rate is around 3000MB/s. Do controllers have a burst rate to?

Could someone explain?

[FuriousSalesman]

This thread convinced me to go out and get a pair of X25-M SSD's. I was going to wait until the X25-E price drop, but I don't think that 64 gb from two SSD's would be enough. The M's are going to replace my slow My Seagate 7200.11. I am hoping for some similar benchmarks!

[alfaunits]

I wish videy would post the IOmeter results

[IanB]

You said the controllers maximum throughput is 666MB/s, but the burst rate is around 3000MB/s. Do controllers have a burst rate to?

Could someone explain?

If the data requested is in the controller cache, then you'll get burst speeds reading it across the bus into main memory. Otherwise, the speed of the onboard processor will determine the limit on how much data can be read sequentially off-disk. If it's having to do parity calculations, there is a limit on how many can be done per second. Bigger cards use beefier (or dual) processors, clocked higher, not just more cache.

[imperium]

So what about having a ARC-1231ML card with 2gb cache with these two Intel ssds (X25-E)? Is that the most uber setup in this case or can you get a better controller card?

[SkOrPn]

If the data requested is in the controller cache, then you'll get burst speeds reading it across the bus into main memory. Otherwise, the speed of the onboard processor will determine the limit on how much data can be read sequentially off-disk. If it's having to do parity calculations, there is a limit on how many can be done per second. Bigger cards use beefier (or dual) processors, clocked higher, not just more cache.

I was going to answer this question earlier, but then it occurred to me that the ICH9R does not have any dedicated controller cache, this burst looks to be normal write caching and/or Advanced Performance settings as provided in the Intel Matrix Storage Manager software or Windows Device Manager. If you uncheck these options you will see the true burst of the ICH9R and storage drives, I think... If I have them checked I get similar results on my Raptors...

[alfaunits]

You are correct. On ICH, the "burst" is essentially a memory copy

[IanB]

I would guess that the ICH driver reserves some system memory for cache.

[SkOrPn]

I would guess that the ICH driver reserves some system memory for cache.

Yeah I was thinking along the same lines...

[FuriousSalesman]

Hey guys, I just received my pair of X-25M SSD's from newegg. I ran HD Tune read test and got max 539 mb/s, min 397.6 mb/s, and 514.2 mb/s average. 1996.1 Mb/s burst rate.

This was on a ICH10R on a Gigabyte X58-UD5. I could not seem to use the write test. When I tried to use it, HD Tune would tell me that "Writing is disabled. To enable writing please remove all partitions. Please check the manual for more information" Does any one know what the issue is?

Edit. used 64k blocks and windows vista ultimate.

[Chosen.]

To enable writing please remove all partitions. Please check the manual for more information" Does any one know what the issue is?

Yes, if you read the error message carefully, you will see that the answer is included in the error message: you need to remove any partitions from the RAID0 array.
If you already installed your OS on it, you cannot run the write tests.
If your OS is still on another HDD, then go to Disk Management, delete the partition(s)/volume(s) from the RAID0 array and run the HDTune write benchmark.

[Levish]

Yes, if you read the error message carefully, you will see that the answer is included in the error message: you need to remove any partitions from the RAID0 array.
If you already installed your OS on it, you cannot run the write tests.
If your OS is still on another HDD, then go to Disk Management, delete the partition(s)/volume(s) from the RAID0 array and run the HDTune write benchmark.

bear in mind you'll loose whats on the drives, so make backups if you need

[FuriousSalesman]

bear in mind you'll loose whats on the drives, so make backups if you need

I installed a fresh copy of vista, I haven't accumulated anything important.

Yes, if you read the error message carefully, you will see that the answer is included in the error message: you need to remove any partitions from the RAID0 array.
If you already installed your OS on it, you cannot run the write tests.
If your OS is still on another HDD, then go to Disk Management, delete the partition(s)/volume(s) from the RAID0 array and run the HDTune write benchmark.

I did read the error message, but I didn't understand the part about the partitions. This is also the first time I have ever done RAID. I made the array and just installed on it. I was not aware that there were any partitions.

[Chosen.]

I installed a fresh copy of vista, I haven't accumulated anything important.
I did read the error message, but I didn't understand the part about the partitions. This is also the first time I have ever done RAID. I made the array and just installed on it. I was not aware that there were any partitions.

Well, if you really want to test the writes with HDTune, the only solution would be to install your OS on another disck, boot from that disk and remove the partitions on the RAID0 array. As previously stated, all data on the array will be gone after that and then after your tests you will have to re-install the OS.

[FuriousSalesman]

Well, if you really want to test the writes with HDTune, the only solution would be to install your OS on another disck, boot from that disk and remove the partitions on the RAID0 array. As previously stated, all data on the array will be gone after that and then after your tests you will have to re-install the OS.

So, if I understand this right, I cannot do a write test if there is an OS on the array?


However, I think the write performance for the X25M is predictable. I don't think I have seen it higher than 75MB/s for 1 and 150MB/s for 2. I probably don't need to run a write test.

[SkOrPn]

Have any of you seen OCZ's latest announcement? The new High-Performance Vertex series SSD drives with completely new controller and up to 64mb of cache. Nearly as fast as the Intel drives but at 1/2 the cost or less.

http://www.ocztechnology.com/aboutocz/press/2008/320
http://www.ocztechnology.com/product...ata_ii_2_5-ssd

Sure hope to see some in the wild soon...

[alfaunits]

There are already two topics on the OCZ drives, keep this X25-M/E related.

[SkOrPn]

There are already two topics on the OCZ drives, keep this X25-M/E related.

Ok, thanks for the info. I thought this was just a SSD topic, not specific to Intel. But thanks for pointing this out to me.

[n0shutz]

The only ATTO bench is in raid so here it is in non raid =)
Sorry i could not resist!!!

I tried to sqeeze as much info in the screen shot but i know some details are still left out like... no page file. Well if you guys need other x25-e benches in non raid just reply or let me know. Ok thats all.

[Boogerlad]

btw, does onboard intel raid have long initialization times?

[Brahmzy]

btw, does onboard intel raid have long initialization times?

No, it's extremely fast. A non-issue.