Nvidia Raid Controller performance

[dfalir]

Guys hi,

I have some info to share with you regarding the performance of the above two controllers, and ask your opinion about what may be wrong or how can i increase performance. I think you will find the following interesting at least. ok, here we go.

In my workplace, a guy bought 4 Western Digital 320GB Disks, and made a raid 0 array, based on his nvidia nforce controller, with athlon 5600+ Socket AM2. He send me the Benchmarks which are measured by Sandra 2007 Pro Engineer.

Physical Benchmark Raid 0 4x WD 320GB: 281 MB/s
File System Benchmark Raid 0 4x WD 320GB: 209 MB/s

Following (our unofficial debate) i measured my 2 WD 250GB that i also have as Raid 0 in the Silicon Image 3132 Controller and also measured my 4 Raptor disks (ADFD (NOT GD) series (16MB cache vs 8 mb)) that i have as raid 0 in my nforce 4 sli chipset. Now read this.

Physical Benchmark Raid 0 2x WD 250GB (Silicon Image): 317 MB/s:!:
File System Benchmark Raid 0 2x WD 250GB (Silicon Image): N/A

Physical Benchmark Raid 0 4x Raptor 36GB ADFD (Nforce): N/A
File System Benchmark Raid 0 4x Raptor 36GB (Nforce): 239 MB/s :!:

Conclusions, and Questions, for you to think about...

:idea: How on earth, can my Raid 0 2x WD 250GB (Silicon Image), outperform an array of Raid 0 4x WD 320GB in the Physical Benchmark??? Is the Nvidia such a crap controller???? how do you explain the difference 317-281??? How is it possible that 2 hard disks outperform 4 hard disks in raid 0? needless to say that for all benchmarks, sandra same edition was used, and stripe size in all raids is the optimal (proposed) one.

:!: The difference in benchmarking File Systems between the Raid 0 4x WD 320GB and the Raid 0 4x Raptor 36GB is only 30MB/s. Shouldn't it be much much more? I get the feeling, that if i used 2x Raptor 36GB in the Silicon Image Controller, than the benchmark of them would outperform all other benchmarks in my test...

What do you think about the above results? Please comment and shed some light...

 

[MagicWok]

It would be a better comparison to put the same PC and HDD setup on different raid controllers. The RAID controllers aren't compared as you quote different HDD both times. The only true way would be to have the same HDD's on same spec machines - the only difference being which controller you choose. Only then are the Raid controllers being compared.

Right now, what you've mentioned has too many variables. Different spec PCs/HDDs which all contribute and mask the result. Unfortunately, whilst the numbers are diffferent - it doesn't tell us much.

 

[immagikman]

I concur. you have to match apples and apples.

 

[dfalir]

Although you both have a major point in your comments i think you are both wrong.

Let me tell you why this is.

a) Generally, comparing a raid 0 4x WD 320GB versus a raid 0 2x WD 250GB on the same machine on the same controller, favors the 4 disk set up. This is widely known. Larger drives are faster.

so, what you say would be correct if i had the better spec machine using the raid 0 2x WD 250GB versus the raid 0 4x WD 320GB on a weaker spec machine and the raid 0 2x WD 250GB proved to be faster. Then you could say that i am not comparing apples to apples.

However, THIS IS NOT THE CASE. The raid 0 2x WD 250GB that is FASTER than the raid 0 4x WD 320GB (on the physical test) is actually placed on the WEAKER spec machine! The strong spec is a 5600+ amd SAM2, with 2 gigs of much faster RAM, whereas the weaker spec machine is an Athlon 4400+ with also 2 gigs but of much slower (3200) ram. And the raid on this spec (raid 0 2x WD 250GB) benchmarks faster that the raid 0 4x WD 320GB which is placed on the faster machine!!!

I therefore SERIOUSLY doubt that if i had the chance to put the Silicon image raid 0 2x WD 250GB on the faster spec machine would actually prove to be slower than the nvidia raid 0 4x WD 320GB, because now that the silicon raid 0 2x WD 250GB has better benchmarks albeit in a much slower machine.
Although it is not apples to apples, i am sure that you get what i mean..
am i right??

 

[immagikman]

I think you are missing our point.

Not speaking for anyone else, Im saying that before you can draw any comparison between the two different RAID controllers, you have to connect them to the same drives. You can't put different drives on different controllers and get a meaningful comparison.

 

[sirrobin4ever]

I don't know how your getting these results, but there is NO way that your 250gb WD's are getting 317mb/s. Seriously. Thats 158.5mb/s, and there is NO way that they could push that much as an average data transfer rate. The 239mb/s from the raptors is much more possible. This comes to about 60mb/s each, which is much more believable.

Best of Luck

 

[dfalir]

The number 317 MB/S is physical benchmark. Not average transfer rate.
and please do not doubt me, if you have a problem your complaints to sandra 2007! I do not know how to attach an image here, but send me an email and i ll send you the benchmarking printscreen. :[/img]-)

 

[Cybercraig]

How many other functions does the NV4 chip have to perform compared to the SI chip? I've always wondered if the SI chip would be faster just because it has no other functions to perform versus the NVidia. :?:

 

[dfalir]

To tell you the truth mate, what you say is probably right, but if someone sais that he offers raid functions, then he ahs to support, not only the function itself, but also the speed of it. If you can not do this, then put 4 serial ata raid on the silicon image and 2 on the nvidia controller and not the opposite as it is now.
Anyway guys, my benchmarkings would be more accurate testing them with the same hd's on the same spec machine. But somehow i get the feeling that Nvidia Controller is a mediocre controller.