On a single channel of SATA when the 4 HDDs are on a single port multiplier, the SATA2 drives will give you the higher throughput performance. The SATA 1 drives (won't matter 4 or 5) will be bottlenecked by the SATA1 spec bandwidth.
If there are multiple channels being utilized, either the performance will be the same. Any performance increase will be likely dedicated to the efficiency of the RAID controller.
Note that the failure risk of your RAID partition would be quadrupled or quintipled.
It would be a software RAID 0 under Windows XP x64 using 2 Built-In Controllers, 4 on the SATA2 and 1 on the SATA1. I was thinking of getting 3 more and putting 4 on each controller, would that help performance if I had 8 SATA1 in Software RAID 0 over 2 controllers?
SATA 2 spec calls for 300MB/s.
SATA 1 spec calls for 150MB/s.
Since its a software RAID, more than likely its using port multipliers.
Since you have 2 controllers (SATA2 hopefully), you probably have 8 SATA ports.
Which means if you RAID 0 4 HDD's, those 4 HDD's will have to share 300MB/s - considering that the HDD's are SATA2. That gives about 75MB/s bandwidth to each drive, which is about right. I think only raptors will hit 80ishMB/s.
Now HDD's typically can handle about 60-80MB/s of traffic (the highend being raptors).
Lets consider SATA 1. Not sure if you have a SATA 1 drive on the controller, if it will just throttle down the spec to 1 for that drive or for the whole controller.. Again, controller dependent, probably hard to find that info.
4 HDDs = 240MB/s-320MB/s. SATA 1 spec is 150MB/s. See the problem?
In short, SATA 1 cannot support more than 3 HDDs in a RAID 0 configuration. It can support it, but you won't be utilizing it very well.
If you insist on using SATA 1 drives, I'd just have 2 HDDs on 1 controller and the other 2 on the other controller. I would not go 8 HDDs in RAID 0.
If you wanna have more than 2-4 drives in a RAID, I'd say go RAID 5 or something. RAID 0 w/ 8 HDD's is asking for drive failure, IMO. Plus I think your software RAID will suck alot of processor cycles.
SATA 2 spec calls for 300MB/s.
SATA 1 spec calls for 150MB/s.
Since its a software RAID, more than likely its using port multipliers.
Since you have 2 controllers (SATA2 hopefully), you probably have 8 SATA ports.
Which means if you RAID 0 4 HDD's, those 4 HDD's will have to share 300MB/s - considering that the HDD's are SATA2. That gives about 75MB/s bandwidth to each drive, which is about right. I think only raptors will hit 80ishMB/s.
Now HDD's typically can handle about 60-80MB/s of traffic (the highend being raptors).
Lets consider SATA 1. Not sure if you have a SATA 1 drive on the controller, if it will just throttle down the spec to 1 for that drive or for the whole controller.. Again, controller dependent, probably hard to find that info.
4 HDDs = 240MB/s-320MB/s. SATA 1 spec is 150MB/s. See the problem?
In short, SATA 1 cannot support more than 3 HDDs in a RAID 0 configuration. It can support it, but you won't be utilizing it very well.
If you insist on using SATA 1 drives, I'd just have 2 HDDs on 1 controller and the other 2 on the other controller. I would not go 8 HDDs in RAID 0.
If you wanna have more than 2-4 drives in a RAID, I'd say go RAID 5 or something. RAID 0 w/ 8 HDD's is asking for drive failure, IMO. Plus I think your software RAID will suck alot of processor cycles.
That's not true, because I have setup 2 SATA1 Software RAID 0's w/ 4 Drives and I get 200MB/s Read/Write sustained w/ 450MB/s Burst on SATA1. Can somebody else with knowledge please post so I can get a real answer, thanks.
3-4 would be the "sweet spot". After that, the gains will get increasing smaller with each drive added. You also greatly increase the odds of losing your data should ANY of the drives fail in RAID 0. As I do, back up your data often. I actually use a seperate IDE drive to store the important things.
I have 3 of these in RAID 0 and they work very well. Just remember to D/L the Hitachi utility to set the drives to SATA II, mine came set to SATA I.
I thought the EVGA MB had a problem but it wasn't.
SATA 2 spec calls for 300MB/s.
SATA 1 spec calls for 150MB/s.
Since its a software RAID, more than likely its using port multipliers.
Since you have 2 controllers (SATA2 hopefully), you probably have 8 SATA ports.
Which means if you RAID 0 4 HDD's, those 4 HDD's will have to share 300MB/s - considering that the HDD's are SATA2. That gives about 75MB/s bandwidth to each drive, which is about right. I think only raptors will hit 80ishMB/s.
Now HDD's typically can handle about 60-80MB/s of traffic (the highend being raptors).
Lets consider SATA 1. Not sure if you have a SATA 1 drive on the controller, if it will just throttle down the spec to 1 for that drive or for the whole controller.. Again, controller dependent, probably hard to find that info.
4 HDDs = 240MB/s-320MB/s. SATA 1 spec is 150MB/s. See the problem?
In short, SATA 1 cannot support more than 3 HDDs in a RAID 0 configuration. It can support it, but you won't be utilizing it very well.
If you insist on using SATA 1 drives, I'd just have 2 HDDs on 1 controller and the other 2 on the other controller. I would not go 8 HDDs in RAID 0.
If you wanna have more than 2-4 drives in a RAID, I'd say go RAID 5 or something. RAID 0 w/ 8 HDD's is asking for drive failure, IMO. Plus I think your software RAID will suck alot of processor cycles.
That's not true, because I have setup 2 SATA1 Software RAID 0's w/ 4 Drives and I get 200MB/s Read/Write sustained w/ 450MB/s Burst on SATA1. Can somebody else with knowledge please post so I can get a real answer, thanks.
Testy aren't we... He seemed to be giving you his opinion and did not claim to be an expert, and if you think you have so much more intelligent information why not explain it yourself?
It is dependent on the number or RAID channels you have on your board, if you have 4 SATA channels, than you theoretically could get up to 600MB/sec of bandwidth, but if 4 drives are running off of 2 channels then 300MB/sec would be the theoretical max transfer rate.
So what you're saying is each channel (controller) is limited to 150MB/s...but I have built a server w/ 4 Raptor 74GB in RAID 0 on 1 channel (controller) with 200MB/s sustained read/write and over 400MB/s burst.
4x 160GB 7200.9 SATAII vs. 5x 150GB Raptor, both setup in RAID0.
Now I ask you, which one is faster?
I'm not talking about that, I'm talking about using THAT drive. It would be 4 of those in RAID 0 on 1 SATA2 controller, vs. 5 of those, running @ SATA1 speed with 4 in 1 controller and the 5th in a 2nd in a Software RAID 0, I also want to know if you guys think 8 of them (4 in each controller running SATA1 SPEED) in RAID 0 Software would be alot faster than 4 or 5, or if any overhead would be too much. I know for a fact it is not each controller is limited to 150MB/s on SATA1, because I have 2 RAID arrays of 4 drives (1 hardware, 1 software) in RAID 0 that get 200MB/s and each are on SINGLE SATA1 CONTROLLERS.
"That's not true, because I have setup 2 SATA1 Software RAID 0's w/ 4 Drives and I get 200MB/s Read/Write sustained w/ 450MB/s Burst on SATA1. Can somebody else with knowledge please post so I can get a real answer, thanks."
The Hitachi drives you have listed are what I have. With 3 in SATA II, burst a little over 336 MB/s and an average rate of 135.8 MB/s. You stated 4 in SATA I is how you achieved your benchmarks? What may I ask program are you using to test? Or, are you using a total of 8 drives to achieve this?
"That's not true, because I have setup 2 SATA1 Software RAID 0's w/ 4 Drives and I get 200MB/s Read/Write sustained w/ 450MB/s Burst on SATA1. Can somebody else with knowledge please post so I can get a real answer, thanks."
The Hitachi drives you have listed are what I have. With 3 in SATA II, burst a little over 336 MB/s and an average rate of 135.8 MB/s. You stated 4 in SATA I is how you achieved your benchmarks? What may I ask program are you using to test? Or, are you using a total of 8 drives to achieve this?
I didn't mean I used that drive, I used 2 different types:
Setup1:
4 WD 74GB Raptor in RAID 0 = ~200MB/s Read/Write
Setup2:
4 200GB (2 Maxtor/2 WD) SATA in RAID 0 = ~200MB/s Read/Write
A friend of mine has 2 of those drives (the hitachis) and he gets sequential 115MB/s under SATA2, so something is wrong if you're only getting 135MB/s using 3.
Like I said, my ULi M1575 RAID controller averages 132MB/sec using only 2 Raptors
I have 7 Raptors, each get ~70MB/s Read/Write, 2 get 137MB/s Read/Write, 4 get ~200MB/s Read/Write, so I would surmise that more would bring more performance.
"A friend of mine has 2 of those drives (the hitachis) and he gets sequential 115MB/s under SATA2, so something is wrong if you're only getting 135MB/s using 3"
I ask again what are you testing with? According to Hitachi, I'm on par with the drives capabilities. Orginally, they were lower because they were set to SATA I. This is why I called Hitachi. Is your friend using software or hardware? I'm using the Nforce4 software RAID. Just because you're using more drives (as with SLI) doesn't mean you'll double or tripple your data rates.
"A friend of mine has 2 of those drives (the hitachis) and he gets sequential 115MB/s under SATA2, so something is wrong if you're only getting 135MB/s using 3"
I ask again what are you testing with? According to Hitachi, I'm on par with the drives capabilities. Orginally, they were lower because they were set to SATA I. This is why I called Hitachi. Is your friend using software or hardware? I'm using the Nforce4 software RAID. Just because you're using more drives (as with SLI) doesn't mean you'll double or tripple your data rates.
It's a Hardware RAID 0, it was tested using SANDRA 2005 benchmarks.
EDIT: All the SLI systems I've tested, have resulted in at LEAST 80% increase in performance.
We can argue about this all night long. If you want to run that many drives, I'd suggest a high end hardware controller so you can maximize your drives capabilities. You are somewhat limited with a software solution and it starts eating into your CPU as well.
We can argue about this all night long. If you want to run that many drives, I'd suggest a high end hardware controller so you can maximize your drives capabilities. You are somewhat limited with a software solution and it starts eating into your CPU as well.
I'm not trying to f*cking argue, I want to know the answer to my question.
We can argue about this all night long. If you want to run that many drives, I'd suggest a high end hardware controller so you can maximize your drives capabilities. You are somewhat limited with a software solution and it starts eating into your CPU as well.
I've done up to 6 drives in a RAID 5 software, it barely tapped the CPU but than again, it's Dual Opteron 64's and I have a 165 Opty OC'd.
Well, the way I see it, you already know it all. I'm offiicially done with this thread. You know, disclosing what your hardware is up front does help. You are comparing hardware and software RAID, not a fair comparasin.
Good luck in your quest.
