Raptor Raid0 74GB slow as can get - HELP [Archive]

g0dM@n

02-05-07, 08:42 AM

Check this out and be amazed:

31.9 burst
18.6 avg read
http://i5.photobucket.com/albums/y176/godman114/SlowRaptors020507.jpg

My main rig with these puppies has been SLOW AS HECK starting up for MONTHS now. I hadn't had time to do a format, and finally decided to do it as I watched the bowl last night.

I got everything installed after HOURS... still boggled my mind. I had done a full format on my C: partition at 30gb, and left my other partition. The C: partition was created first a long time ago, so it should be the faster part of the HDDs.

I had set up the raid0 16k stripe long ago... think if I rebuild it that it will fix this problem? I just don't get how it could be THIS SLOW.

I was actually thinking about selling the raptors b/c my maxtor 80gb sata150 raid0 config in one of my other computers BLOWS these away!!

There's gotta be something to fix this. I don't want to have to install everything ALL OVER again, and I just recently bought Norton Ghost 10.0, so hopefully I can ghost my new C: partition to my 320gb storage drive, and rebuilt the raid... will rebuilding the raid fix my problem?

Thanks in advance.

Maviryk

02-05-07, 08:47 AM

g0dM@n

02-05-07, 08:49 AM

I can't see the image (blocked because I'm at work), what're the numbers saying?

Also which raid controller on the DFI are you using on each setup? Intel chipsets outperform all the other chipsets hands down.
nvidia controller and i listed the #s above pic... =/

the spike is the partition (obviously)... i'll relist:

random access 8.4ms
avg read 18.6mb/s
burst 31.9mb/s

look at the red... disgusting...

dominick32

02-05-07, 08:53 AM

With your exact setup this is what I achieved on your nvraid controller and with 16k stripe two 74G raptors. This is the older (8mb) revision so you should be even slightly faster on the sustained reads:

http://www.overclockers.com/articles1297/pic3.jpg

The bottom line indicates a single drive.

As you can see, something is definitely wrong with your setup. Unfortunately it may sound like a borked drive. So you say you selected a 16K stripe? Did you ever fully scan the drives for errors individually before raiding them?

If I remember correctly, there was a member on here named TuskenRaider who might have some data on formating, partioning, and creating a raid array with one faulty drive. He said the performance was affected obvoisuly and tremendously. He returned the drive to WD, got a new drive and everything shot right back up to normal benchmarking and performance again.

I think you are going to need to test everything on an individual level again. I would boot into windows on a standard SATA drive with both raptors dissasembled from the raid array. You can than use HDTACH on each drive individually.

Raptor 1 should be: 65 to 85 mbs depending on 8mb or 16mb revision
Raptor 2 should be: 65 to 85 mbs *** ** ** ** ** **
Respectively^

After you run HDTach on both drives indivually I would scan them thoroughly for errors, run WD's drive recovery and error detection tool.

If you still cannot find anything wrong with the drives after:
1. Testing Individually in HDTACH
2. Deep Scanning and Formatting
3. Running WD's tools

Than there is either something wrong with the NVRaid controller or setup. Although I have never seen this happening before as it is usually a fairly easy setup process.

Hope this helps a tad. I know raid can be frustrating sometimes.

Dom

g0dM@n

02-05-07, 09:03 AM

Dom, definitely worth a try... I don't need to format the drives before this testing do i? Just hook them up one at a time to another PC?

I have three DFI NF4 setups... so I could throw one of these raptors in another setup to test.

By the way, in the meantime I knew this would probably be useful, so I went ahead and ran HDTach 32mb on my Seagate Perp 320gb drive on that same PC... I think this HDD is hooked up to the other sata ports (assuming silicon image?):

random access 16.9 ms
Read 57.0MB/s
Burst 201.1MB/s

http://i5.photobucket.com/albums/y176/godman114/SeagPerp320_020507.jpg

How did you run the tach on both drives on the same graph? I woulda just done that... oh well.

I don't mind @ all sending a HDD back if that will help me out... I just can't believe I was so lazy to let this go on for MONTHS!!! You guys have no idea... it's been at least 6 months like this, but I just didn't want to spend the time on it... on my spare time I either cleaned room, spent time with gf, or played 360, lol.

Dom, let me know if I should just slap a raptor in one of the other rigs for testing, or if it has to be formatted first (since it's part of a raid - assuming the MBR asks for other drive?) I don't know much about this stuff...

dominick32

02-05-07, 09:06 AM

Dom, definitely worth a try... I don't need to format the drives before this testing do i? Just hook them up one at a time to another PC?

...

Before I respond to your entire thread I want you to try one thing:

Go into the bios (no you do not need to physically format or disconnect anything) all you need to do is:
1. Go into bios.
2. Disable the NVRaid controller
3. Select your standard SATA boot drive and boot into windows
The RAID file system on both drives will be perfectly intact and available for testing so you do not need to reformat anything. It will just boot windows with two blank or invisible drives. (but HDTACH will be able to read both platters even if they are unavailable to windows)

GO GET THAT TEST DONE!!
PS- Hopefully you read this: ***You do not need any other rigs for this testing***

g0dM@n

02-05-07, 09:11 AM

Before I respond to your entire thread I want you to try one thing:

Go into the bios (no you do not need to physically format or disconnect anything) all you need to do is:
1. Go into bios.
2. Disable the NVRaid controller
3. Select your standard SATA boot drive and boot into windows
The RAID file system on both drives will be perfectly intact and available for testing so you do not need to reformat anything. It will just boot windows with two blank or invisible drives. (but HDTACH will be able to read both platters even if they are unavailable to windows)

GO GET THAT TEST DONE!!
PS- Hopefully you read this: ***You do not need any other rigs for this testing***
I'm doing all of this remotely. I'm at work... unfortunately I cannot get into BIOS remotely. :(

I'll try to do this during my lunchbreak... around 1pm... I'll run HDTach on each drive individually (leaving them both on the nvidia sata ports, but I will disable nvidia raid)... then I'll take a SS of each, and post them here.

On a side note, does anyone know anything about Norton Ghost 10.0? I hope I can ghost this... spent so much time reinstalling EVERYTHING (i mean everything literally... have lots of progs I use).

Thx.

dominick32

02-05-07, 09:16 AM

On a side note, does anyone know anything about Norton Ghost 10.0? I hope I can ghost this... spent so much time reinstalling EVERYTHING (i mean everything literally... have lots of progs I use).

Thx.

In all honesty, if you do have a borked up drive indeed. By the way Raid 0 works, you will not be able to ghost any image from this setup successfully. You will be in theory copying over a messy dinner plate from your original setup.

Something else to think about: When I switched from single raptor to raid 0, my XP install time was almost cut in half!!! :mad: As well as xp and game loads. So, it should be faster for you on your next install if we can figure out this little issue you are having.

g0dM@n

02-05-07, 09:28 AM

In all honesty, if you do have a borked up drive indeed. By the way Raid 0 works, you will not be able to ghost any image from this setup successfully. You will be in theory copying over a messy dinner plate from your original setup.

Something else to think about: When I switched from single raptor to raid 0, my XP install time was almost cut in half!!! :mad: As well as xp and game loads. So, it should be faster for you on your next install if we can figure out this little issue you are having.
aww man... and it was SUCH A RELIEF to have everything reinstalled again...

#1 reason i reinstalled is this (really weird)... my browsing was SLOW AS HELL. every other computer in the house, the internet was fast... i tried both ETHERNET on this pc and both were slow... tried diff ports on switch, tried diff cable... no go. after my reformat last night, my browsing was fast again.

man... i spent so much time... oh well...

and i know how fast raptors are... it USED to be fast a long time ago, lol... who knows what the heck happened.

i guess a borked drive is the best explanation...

so if and when this is all fixed up (however that is)... i'll need to know how to ghost... i finally have a real ghosting program and it better work like i want! :)

btw, i figured out how to list both HDDs in one benchmark:

http://i5.photobucket.com/albums/y176/godman114/Raptors0vsPerp_020507.jpg

dominick32

02-05-07, 09:39 AM

aww man... and it was SUCH A RELIEF to have everything reinstalled again...

#1 reason i reinstalled is this (really weird)... my browsing was SLOW AS HELL. every other computer in the house, the internet was fast... i tried both ETHERNET on this pc and both were slow... tried diff ports on switch, tried diff cable... no go. after my reformat last night, my browsing was fast again.

man... i spent so much time... oh well...

and i know how fast raptors are... it USED to be fast a long time ago, lol... who knows what the heck happened.

i guess a borked drive is the best explanation...

so if and when this is all fixed up (however that is)... i'll need to know how to ghost... i finally have a real ghosting program and it better work like i want! :)

btw, i figured out how to list both HDDs in one benchmark:

http://i5.photobucket.com/albums/y176/godman114/Raptors0vsPerp_020507.jpg

Something else stupid that we could be overlooking. Maybe its a software issue. Do me a favor and run this synthetic benchmark: ATTO Disk Bench

I cant seem to locate a link for it. Also, run sisoft sandra HD benchmark. Lets verify that these numbers are indeed correct.

Dom

g0dM@n

02-05-07, 09:56 AM

Just downloaded ATTO Benchmark, and running it... I changed total length to 32mb - should I have? I'm assuming the large the length, the more thorough the test (but again I don't know much about this stuff)... I'll post screenshot when it's done.

And I have to download sisoft... gonna do that now.

g0dM@n

02-05-07, 10:02 AM

ATTO Benchmark (changed size to 32mb - hope i should have done that)

This was the 30gb C: (main) partition on the 74gb raid0

http://i5.photobucket.com/albums/y176/godman114/RaptorMainATTO_020507.jpg

dominick32

02-05-07, 10:04 AM

g0dM@n

02-05-07, 10:18 AM

LOL.. Something is definitely wacky man. Sorry for the bit of humor, I have just never seen a raptor raid 0 array perform like this before.
Lets go back to the previous step. We need to test both drives on an individual basis.

Dom
so what did this ATTO show you, though? Now I'm trying to figure it out myself. :D

here's 4mb ATTO:

http://i5.photobucket.com/albums/y176/godman114/RaptorMain4mbATTO_020507.jpg

dominick32

02-05-07, 10:21 AM

It is still showing what looks to be an average sustained read of close to 15 to 20 MB/s

That means that you definitely have two applications: Atto and HDTach proving that there is definitely something wrong with your raid setup.

g0dM@n

02-05-07, 10:44 AM

g0dM@n

02-05-07, 11:03 AM

Here's screenie of Sisoft:
http://i5.photobucket.com/albums/y176/godman114/Sisoft_74RaptorRaid0_Read_020507.jpg

And I've exported results to CSV (comma delimited) if need be. XML kept exporting as TAB, so I just stuck with CSV since I don't know what .tab is.

*edit*
I tried WRITE benchmark on sisoft and it says "Error: Disk must be empty to be write-benchmarked", so duh... I didn't know that, lol.

dominick32

02-05-07, 11:04 AM

Yes, again proving that somethings definitely wrong. We need to test individual drives to continue working at a solution.

ATTO = 19 MB/s
Sisoft = 32 MB/s
HDTach = 18.6 MB/s

Those results seems on par for a mid to late 1990's 5400rpm hdd.

lilneel12

02-05-07, 11:11 AM

g0dM@n

02-05-07, 11:18 AM

i had the same EXACT problem with nvidia raid and 2 74gig raptors

i disabled command queing and disabled reading and write cache options in the sata controller in device manager. This made a huge difference. Make sure you disable the settings for both raptor drives.
gonna give that a whirl right now... so i guess i do this to the NVIDIA RAID CLASS CONTROLLER in DM... where the heck did you find out to do that...? yes, i'm a HDD newb...

i bought these raptors based on reviews and they rocked at first... just not being nice to me lately...

*edit*
Here's the only change I made (for now)... I did "Disable Tagged Queuing", but didn't touch the other as I don't know what that is... I don't see where to disable read and write cache options, lilneel...
http://i5.photobucket.com/albums/y176/godman114/RaptorDM_020507.jpg

Jon

02-05-07, 11:33 AM

Disabling Command Queuing is what I have to do on my Ultra-D. Performance is dismal until I do that. Never had to disable the caching though.

lilneel12

02-05-07, 11:35 AM

click on the properties of scic and raid controllers
then click on nvidia sata controller
there should be 2 of them,
make sure you find both raptors and disable the three things i mentioned b4

lilneel12

02-05-07, 11:36 AM

i did some tests and disabling the cacheing actually made burst and avg transfer speed faster.
but the hdtach graph wasn't as smooth as it was when cache was enabled.

-is the smoother curve any better?

g0dM@n

02-05-07, 11:55 AM

Disabling Command Queuing is what I have to do on my Ultra-D. Performance is dismal until I do that. Never had to disable the caching though.
okay i found command queuing (i did it before but sorry i couldn't post b4 you lilneel)... the screenshot i posted is the wrong thing, but i found it in nforce4 sata controller (had to look around but wasn't too tough)... i am rebooting now and tried disabling command queuing and read/write cache... if this works out, i'll try re-enabling read/write cache (dont even know what that does) and make sure that command queuing is all i needed to change.

anyone care to explain what these do? i'll be VERY happy if i don't have to reinstall everything and can fix with these minor changes.

g0dM@n

02-05-07, 11:59 AM

You guys rock!!!!!!!!!!!!!!!!!!!

I did a quick test cuz I g2g to lunch now:
http://i5.photobucket.com/albums/y176/godman114/RaptorWorking020507.jpg

Now I just need to know if I should keep read/write cache disabled or not.

I'm now so happy... now the question is why does command queuing do this...

tuskenraider

02-05-07, 12:04 PM

It's gonna be the CQ slowing you down, and I'm waiting for the elation you'll have in a moment............;)

Edit...........beat me to it..........glad to hear!

As far as why CQ does it, I don't know, but it's chipset(NF4) related only it seems. The funny thing is a just got my NF4 Ultra-D running over the weekend and CQ had no effect whether it was enabled or disabled for me.

Jon

02-05-07, 12:10 PM

Command queuing is only useful in certain applications (not software). For benchmarks, it's worthless since it's not being utilized for what it was made for - multithreaded access.

For typical use, command queuing only adds to read/write latency causing you and I to experience what we do when it's enabled - a fraction of our drive's capabilities.

As far as why CQ does it, I don't know, but it's chipset(NF4) related only it seems.

That would be interesting and I can't really comment on it. I don't have it available on any other systems and haven't really researched it. Would be nice to know if it's chipset-related though.

tuskenraider

02-05-07, 12:16 PM

Would be nice to know if it's chipset-related though.I can honestly say I've responded to over a dozen threads about this exact problem specifically with an NF4 chipset mobo. Until this week, I hadn't had an NF4 board, but had remembered seeing a post about it some time ago and was able to direct people to the solution. I have never seen a post about this problem with another type of chipset and CQ hasn't affected my benchmarks for whatever reason while testing everything out over the weekend.

g0dM@n

02-05-07, 12:50 PM

Here is a more thorough HD Tach Run:

Command Queuing Disabled
Read Cache Disabled
Write Cache Disabled

http://i5.photobucket.com/albums/y176/godman114/RaptorWorking_NoRead-WriteCache_NoC.jpg

Command Queuing Disabled
Read Cache Enabled
Write Cache Enabled
http://i5.photobucket.com/albums/y176/godman114/RaptorWorking_NoCQ_020507.jpg

It's better, but UGLY!!

Should I unclick "Let BIOS select transfer mode"?

http://i5.photobucket.com/albums/y176/godman114/SATAprop020507.jpg

tuskenraider

02-05-07, 01:00 PM

Should I unclick "Let BIOS select transfer mode"?No. The BIOS detects what the chipset and driver should run at. I'd leave it alone. It's Window's settings that usually screw things up and slow things down, as you've seen.

At this point, your graphs aren't "ugly" , they're about right on for the drive.

g0dM@n

02-05-07, 01:21 PM

No. The BIOS detects what the chipset and driver should run at. I'd leave it alone. It's Window's settings that usually screw things up and slow things down, as you've seen.

At this point, your graphs aren't "ugly" , they're about right on for the drive.
well aren't my curves a little much? I'm gonna try disabling the read/write cache again and see how it does...

So this means that anytime I do a windows install, the drives are running crappy until I get to desktop to disable CQ? pfff....

tuskenraider

02-05-07, 02:12 PM

well aren't my curves a little much? I'm gonna try disabling the read/write cache again and see how it does...That curve is what is expected as your STR rates drop torwards the inner platter of the drive. Those big dips are likely programs or processes accessing the drive during the test. Realistically, it won't have much impact on the benchmark. The best I've seen a pair of 8MB cache Raptors in RAID0 average in HD Tach is about 129MB/s, so you're right there performance wise.So this means that anytime I do a windows install, the drives are running crappy until I get to desktop to disable CQ? pfff....On an NF4 RAID controller at least..........

g0dM@n

02-05-07, 02:28 PM

On an NF4 RAID controller at least..........
Ya... that i figured... thx... oh well.

I guess my windows installation will be the only slow part from now on... i could swear I didn't have this problem before... it's weird that I didn't before and I do now... maybe b/c I'm using newer nf4 drivers... next time I'll HDTach before installing drivers...

this was my best run so far...
http://i5.photobucket.com/albums/y176/godman114/RaptorWorking_NoRead-WriteCache_-1.jpg

I'm keeping read/write disabled as well... since my scores are SLIGHTLY better without read/write cache (though i still don't know what they TRULY do)

dominick32

02-05-07, 03:04 PM

WOW. Awesome news. I am glad to see the solution was something simple that we overlooked initial! Congrats bro.

Dom

thefly

02-05-07, 10:25 PM

g0dM@n

02-05-07, 11:00 PM

I tried disabling these settings as well, but found that disabling the read&write cache caused my atto scores to suffer a lot. This wasn't apparent in hdtach mind you. Anyway, best setting for my board seems to be command queuing disabled... Read&Write enabled.
i guess i'll have to try... AGAIN... cuz i think my read/write is disabled... HDTach did nicely without read/write cache...