Nvidia RAID 0

[Dawgdoc]

Helping a friend out with setting up his system.

This is a 4 X 250gb Seagate Barracuda 7200.10 with Nvidia RAID 0 on an EVGA 780i FTW mobo.

I helped him set up raid over a VOIP (Ventrilo) and he installed windows himself.

He says he has checked the jumpers to make sure he is getting his full 3.0Gb/s xfer rate.

What do you think of this HD benchmark?

CPU usage looks WAY to high, and IMHO I think the speed is crap.

What do you think? Suggestions?

Also FYI, disk write-back caching HAS been enabled, and the OS is Vista Ultimate x64.

http://img166.imageshack.us/my.php?image=hdtuneos8.jpg

Thanks!

[bigdogchris]

I just got off from using Raid 0 on a 750iFTW board. I used 2 WD6400AAKS drives. I'll never do it again. I'm back on 1 drive now.

His benchmarks look 100% accurate. Are you on a 64k stripe? (hopefully) Software assisted raid has a 12-22% cpu utilization on synthetic sequential read benchmarks. As for the throughput, that looks as expected for those drives.

He should of went raid 5 though if he had 4 drives. Only would of had about 750GB of space but if 1 drive fails, he can still boot. Performance probably would be a little better also.

[Syntax Error]

4 Disk RAID0 isn't going to have 4x the performance, there are diminishing returns.

The chipset and motherboard for the RAID controller also makes an impact here, I've tested NVIDIA stripe with SSDs and I got the "shark tooth" benchmarks like your friend's been getting as well. NVIDIA stripe isn't the most robust of RAID out there, and that's to be expected from FakeRAID.

[bigdogchris]

Don't forget the instability. I'd be afraid of using 4 drives on that shit.

[nitrobass24]

Looks about right...
1. 250gb SATAs arent that fast to start with
2. CPU usage is at 20%....was something else going on in the background?
3. The jaggedness seems to coincide with all NVidia Stripe sets.
4. NV Stripe just sucks

[AreEss]

1) nV RAID is almost purely software.
2) nV RAID is entirely pure crap.
3) nV RAID is going to eat your data.

Oh, and 20% CPU is low for that trash.

[Dawgdoc]

Thanks for the input thus far guys. It is appreciated.

To dispel a few misceptions that are here though....

1. RAID 5 is never faster than RAID 0. Not possible.

2. The size of the drive has very very little (if anything) to do with the speed. Its the RPMs and the cache so saying a 250gb drive isnt "that fast" is an incomplete/inaccurate statement.

3. Nvidia RAID is going to eat your data??? Please explain that one......

I do appreciate all of the information. I am not really an Nvidia RAID person and have limited experience with it. I have reasonably extensive experience with Intel RAID, but my friend has an Nvidia based board so that really doesnt do him any good here :)

Looks like you guys here on [H] are big proponents of Nvidia RAID? NOT :)

Any additional input appreciated. Ive had several other ppl on other forums say that this even looks slow for Nvidia RAID so if anyone has suggestions Id love to hear them.

Thanks!

[Sipster]

This is what I get with my Nvidia raid with 2 80GB 7200.11's.......

min: 71.4
max: 125.2
avg: 104.5
access time: 15.0
burst rate: 165.1
cpu usage: 20.0%

I have 2 new 7200.11 320GB drives coming today......I think I'm done with Raid and I'll just have one for OS and games and the other for backup.

[dandragonrage]

2. The size of the drive has very very little (if anything) to do with the speed. Its the RPMs and the cache so saying a 250gb drive isnt "that fast" is an incomplete/inaccurate statement.


The most important factor is data density, and most 250GB drives are lacking in that category.

[xX_Jack_Carver_Xx]

My NVidia Raid0 has been running flawlessly since it started on my NForce2 board... and Abit NF7-S, moved to a DFI Infinity NF2, then on to a DFI Infinity NF4, and now onto the Foxconn NF590 I'm using now. So 2x36G Raptor raid0 has been running without incident since 2002.

Gee I gues that Nvidia raid really sucks. Cough. :eek::rolleyes::p

[dandragonrage]

My NVidia Raid0 has been running flawlessly since it started on my NForce2 board... and Abit NF7-S, moved to a DFI Infinity NF2, then on to a DFI Infinity NF4, and now onto the Foxconn NF590 I'm using now. So 2x36G Raptor raid0 has been running without incident since 2002.

Gee I gues that Nvidia raid really sucks. Cough. :eek::rolleyes::p

Hey guys, my Yugo has been a great car and has been running without incident for years now. How dare you think it sucks?

[Dan_D]

I just got off from using Raid 0 on a 750iFTW board. I used 2 WD6400AAKS drives. I'll never do it again. I'm back on 1 drive now.

His benchmarks look 100% accurate. Are you on a 64k stripe? (hopefully) Software assisted raid has a 12-22% cpu utilization on synthetic sequential read benchmarks. As for the throughput, that looks as expected for those drives.

He should of went raid 5 though if he had 4 drives. Only would of had about 750GB of space but if 1 drive fails, he can still boot. Performance probably would be a little better also.

I'd advise against NVRAID 5. Its' pretty terrible in terms of performance. All onboard RAID 5 is pretty bad but NVRAID 5 is horrid. RAID 5 is also going to always be slower than RAID 0 or 0+1. RAID 5 requires parity calculations which is why its' so slow.

[bigdogchris]

I'd advise against NVRAID 5. Its' pretty terrible in terms of performance. All onboard RAID 5 is pretty bad but NVRAID 5 is horrid. RAID 5 is also going to always be slower than RAID 0 or 0+1. RAID 5 requires parity calculations which is why its' so slow.All of the benchmarks I've seen shows raid 5 as a very good choice for a good combination of performance/reliability. But that's server hardware vs software assist /shrug

[dandragonrage]

All of the benchmarks I've seen shows raid 5 as a very good choice for a good combination of performance/reliability. But that's server hardware vs software assist /shrug

That is more of a difference than you might think. Even the server hardware varies wildly in performance.

[Dan_D]

All of the benchmarks I've seen shows raid 5 as a very good choice for a good combination of performance/reliability. But that's server hardware vs software assist /shrug

I was specifically referring to onboard RAID solutions like the ICH9R, ICH10R, and NVRAID. They are basically crap for RAID 5 performance. However I've seen RAID 5 performance from LSI and Areca cards that's onpar if not superior (in most respects) to onboard RAID 0 performance out of NVRAID and even Intel's ICH9R/ICH10R RAID solutions.

[AVT]

FWIW, what I get on ICH9R with dual 40gb's is:

Min: 59.0
Max: 116.2
Average: 97.3
Access Time: 14ms
Burst Rate: 132.5mb/s
CPU usage: 4.6%

[Elf_Boy]

re: NV raid 0

I have an m2n32 sli delux, nvidia 590 chipset.

I have been running raid zero since xp. I picked up the board when it first hit the market.

The only issues I ever have are with the nvidia drivers and thier sucky support.

The hardware has run flawlessly. Two drives raid 0. I dont see any real hit to my CPU either.

What benchmark software are you using? I'll see what results I get.

[B239]

Reason for Editing

[Dan_D]

I get 390 mb/s on avrage with the HD Tuner (Raid 5)
http://www.youtube.com/watch?v=hZXz57HOB-Q

On NVRAID?

[B239]

Reason for Editing

[Dan_D]

Areca ARC-1220 PCI-Express x8 SATA II Controller Card RAID level 0, 1, 1E, 3, 5, 6 (if RAID 6 engine supported) and JBOD

x8 Western Digital Caviar SE16 WD5000AAKS 500GB 7200 RPM 16MB Cache SATA 3.0Gb/s Hard Drive

ASUS RAMPAGE FORMULA LGA 775 Intel X48 ATX Intel Motherboard - Retail
Onboard EPU provides industry-leading energy efficiencyle

Intel Core 2 Quad Q9450 Yorkfield 2.66GHz 12MB L2 Cache LGA 775 95W Quad-Core Processor - Retail

LIAN LI PC-A77B Black Aluminum ATX Full Tower Computer Case - Retail

BFG Tech BFGE981024GX2OCXE GeForce 9800 GX2 1GB 512-bit GDDR3 PCI Express 2.0 x16 HDCP Ready SLI Supported Video Card -
Retail
OCX overclocked to 680MHz / 2100MHz &
includes HDMI cab

I was going to say that's not possible on NVRAID.