Should I keep RAID0

[keving98]

I built a computer for a photographer friend last September that I
have in my shop for a check up while she's on vacation.

These are the specs:
Asus Core 2 Extreme X6800
Asus P5WDG2-WS Professional MB
Asus GeForce EN7600GT Video Card
(2) 1GB Corsair XMS2 DDR2 800 SDRAM
(3) Seagate Barracuda 320 GB 7200 RPM 16MB Cache SATA 3.0Gb/s Hard
Drives
Windows XP Media Edition 2005

I had originally installed (2) of the Seagate drives, one for data and
backups and one for the OS.
I always felt bad about not installing a third hard drive so she would
have (2) in a RAID0 configuration for the OS and (1) hard drive for
backups so I did this while she's on vacation. I used the Marvell RAID
controller. There is an Intel RAID controller as well.

She often works with massive RAW image files in Photoshop CS2 and I
thought the RAID0 would help her.

After getting the drives installed and configured I ran HD Tach to
test the configuration.

According to HD Tach the burst speed went from 243 mb/s in non-raid
mode to 116 mb/s in RAID0 mode. Random access in non-raid went from
14.0ms to 13.8ms in RAID0 mode. Average Read went from 65.4 MB/s in
non-raid mode to 67.5 MB/s in RAID0 mode.

You can view the screen shots here: http://www.cageynet.com/hdtach.htm
(Non-raid is on top)

My question is: Is she better off with RAID0 or non-RAID? It's tough
for me to tell just by going back and forth between configurations (I
still have the original boot drive set aside). I realize HD Tach is
pretty old. Are there better HD benchmarks available somewhere?

She is the type of person who will download and install anything that
presents itself and then blame the computer for the sluggish
behavior. That can't be changed. The best that I can do is to get
this computer running at maximum performance and pester her into using
Drive Image 5 religiously.

Any help would be greatly appreciated.

 

[Bob Willard]

I built a computer for a photographer friend last September that I
have in my shop for a check up while she's on vacation.

These are the specs:
Asus Core 2 Extreme X6800
Asus P5WDG2-WS Professional MB
Asus GeForce EN7600GT Video Card
(2) 1GB Corsair XMS2 DDR2 800 SDRAM
(3) Seagate Barracuda 320 GB 7200 RPM 16MB Cache SATA 3.0Gb/s Hard
Drives
Windows XP Media Edition 2005

I had originally installed (2) of the Seagate drives, one for data and
backups and one for the OS.
I always felt bad about not installing a third hard drive so she would
have (2) in a RAID0 configuration for the OS and (1) hard drive for
backups so I did this while she's on vacation. I used the Marvell RAID
controller. There is an Intel RAID controller as well.

She often works with massive RAW image files in Photoshop CS2 and I
thought the RAID0 would help her.

After getting the drives installed and configured I ran HD Tach to
test the configuration.

According to HD Tach the burst speed went from 243 mb/s in non-raid
mode to 116 mb/s in RAID0 mode. Random access in non-raid went from
14.0ms to 13.8ms in RAID0 mode. Average Read went from 65.4 MB/s in
non-raid mode to 67.5 MB/s in RAID0 mode.

You can view the screen shots here: http://www.cageynet.com/hdtach.htm
(Non-raid is on top)

My question is: Is she better off with RAID0 or non-RAID? It's tough
for me to tell just by going back and forth between configurations (I
still have the original boot drive set aside). I realize HD Tach is
pretty old. Are there better HD benchmarks available somewhere?

She is the type of person who will download and install anything that
presents itself and then blame the computer for the sluggish
behavior. That can't be changed. The best that I can do is to get
this computer running at maximum performance and pester her into using
Drive Image 5 religiously.

Any help would be greatly appreciated.

Your results indicate that the Marvel RAID controller is a loser, not that
RAID is bad. Is there, perhaps, an update available to Marvel's firmware
to fix its serious bandwidth limitations?

I'm not familiar with that Marvel gear. Does it plug into vanilla PCI?
If it uses vanilla (32b 33MHz) PCI for DMA, then it is pretty hopeless,
and I'd not use it with fast SATA HDs. If it plugs into a faster version
of PCI (e.g. 64b 66MHz) and if your MB has such, then it might be useful.

 

[keving98]

Your results indicate that the Marvel RAID controller is a loser, not that
RAID is bad. Is there, perhaps, an update available to Marvel's firmware
to fix its serious bandwidth limitations?

I'm not familiar with that Marvel gear. Does it plug into vanilla PCI?
If it uses vanilla (32b 33MHz) PCI for DMA, then it is pretty hopeless,
and I'd not use it with fast SATA HDs. If it plugs into a faster version
of PCI (e.g. 64b 66MHz) and if your MB has such, then it might be useful.

Thanks for the response.

Actually, the Marvell 61.xx RAID controller is built into the Asus
board, along with a separate Intel ICH7R RAID controller. I'm not
sure why there are two different RAID options but just by chance I
picked the Marvell. I'm going to install the drives on the Intel
ICH7R controller and see what happens.

Asus isn't offering a firmware update for the Marvell controller but I
will take a look at Marvell's Web site and see what they have.

Thanks again for the response.

Kevin G

 

[Arno Wagner]

I built a computer for a photographer friend last September that I
have in my shop for a check up while she's on vacation.

These are the specs:
Asus Core 2 Extreme X6800
Asus P5WDG2-WS Professional MB
Asus GeForce EN7600GT Video Card
(2) 1GB Corsair XMS2 DDR2 800 SDRAM
(3) Seagate Barracuda 320 GB 7200 RPM 16MB Cache SATA 3.0Gb/s Hard
Drives
Windows XP Media Edition 2005

I had originally installed (2) of the Seagate drives, one for data and
backups and one for the OS.
I always felt bad about not installing a third hard drive so she would
have (2) in a RAID0 configuration for the OS and (1) hard drive for
backups so I did this while she's on vacation. I used the Marvell RAID
controller. There is an Intel RAID controller as well.

She often works with massive RAW image files in Photoshop CS2 and I
thought the RAID0 would help her.

After getting the drives installed and configured I ran HD Tach to
test the configuration.

According to HD Tach the burst speed went from 243 mb/s in non-raid
mode to 116 mb/s in RAID0 mode. Random access in non-raid went from
14.0ms to 13.8ms in RAID0 mode. Average Read went from 65.4 MB/s in
non-raid mode to 67.5 MB/s in RAID0 mode.

Assuming you mean MB above, not mb (that would be ''milibits'', i.e.
1/1000 of a bit), what you see is the result of adding a RAID
controller to satisfy marketing. Basically worthless. IN
all likelyhood the RAID controller sits on a pretty slow PCI
bus.

You can view the screen shots here: http://www.cageynet.com/hdtach.htm
(Non-raid is on top)

My question is: Is she better off with RAID0 or non-RAID? It's tough
for me to tell just by going back and forth between configurations (I
still have the original boot drive set aside). I realize HD Tach is
pretty old. Are there better HD benchmarks available somewhere?

Non-RAID, by all means!

She is the type of person who will download and install anything that
presents itself and then blame the computer for the sluggish
behavior. That can't be changed. The best that I can do is to get
this computer running at maximum performance and pester her into using
Drive Image 5 religiously.

Seems to me this lady need the experience of a) loosing a disk
to disk fault and b) loosing a disk to installing something
malicious. Some people only learn when it hurts enough.

Arno

 

[Folkert Rienstra]

Your results indicate that the Marvel RAID controller is a loser, not that
RAID is bad. Is there, perhaps, an update available to Marvel's firmware
to fix its serious bandwidth limitations?

I'm not familiar with that Marvel gear.

Does it plug into vanilla PCI?

What does 243 MB/s suggest to you?

If it uses vanilla (32b 33MHz) PCI for DMA, then it is pretty hopeless,

But still much better than 70MB/s.

and I'd not use it with fast SATA HDs. If it plugs into a faster version
of PCI (e.g. 64b 66MHz) and if your MB has such, then it might be useful.

Apparently not

 

[Folkert Rienstra]

Assuming you mean MB above, not mb (that would be ''milibits'', i.e.
1/1000 of a bit), what you see is the result of adding a RAID
controller to satisfy marketing. Basically worthless.

IN all likelyhood the RAID controller sits on a pretty slow PCI bus.

Another moron that can't interprete 243MB/s, and couldn't bother to look up the specs.

http://www.asus.com/products4.aspx?l1=3&l2=11&l3=248&model=1289&modelmenu=1

 

[Bob Willard]

What does 243 MB/s suggest to you?

What is suggested to me is that the HD was directly connected to a
standard SATA-II port, completely bypassing the RAID thingie. Just my
assumption, since the OP didn't really say.

But still much better than 70MB/s.

Uh, no. ISTR measuring quite a few PCI thingies that didn't do DMA much
better than ~70 MB/s on vanilla PCIs; some were noticeably worse.

 

[Folkert Rienstra]

What is suggested to me is that the HD was directly connected to a
standard SATA-II port, completely bypassing the RAID thingie.

Nope, that was nowhere suggested.

Just my assumption, since the OP didn't really say.

Exactly. As he didn't need to.

Uh, no.
Yup.

ISTR measuring quite a few

With the emphasis on *few*.

PCI thingies that didn't do DMA much

So what.

better than ~70 MB/s on vanilla PCIs;

Those are the 'PCIs' that are labeled as broken.
The ones that are working as supposed to, manage to do well over 100MB/s.

some were noticeably worse.

No, really?
Maybe because you weren't measuring PCI bus bandwidth at all?

And what's with this "PCI DMA".

 

[keving98]

Nope, that was nowhere suggested.


Exactly. As he didn't need to.





With the emphasis on *few*.


So what.


Those are the 'PCIs' that are labeled as broken.
The ones that are working as supposed to, manage to do well over 100MB/s.


No, really?
Maybe because you weren't measuring PCI bus bandwidth at all?

And what's with this "PCI DMA".

I moved the drives over to the Intel ICH7R controller and it made the
difference:

Burst Speed: 270.6 MB/s from 116 MB/s
Random Access: 13.1ms from 13.8ms
Average Read: 101.7 from 67.5 MB/s
CPU Utilization: 2% from 8%

I've also updated the HDTach screenshots: http://www.cageynet.com/hdtach.htm

Thanks for all the responses.

Kevin G

 

[keving98]

Nope, that was nowhere suggested.


Exactly. As he didn't need to.





With the emphasis on *few*.


So what.


Those are the 'PCIs' that are labeled as broken.
The ones that are working as supposed to, manage to do well over 100MB/s.


No, really?
Maybe because you weren't measuring PCI bus bandwidth at all?

And what's with this "PCI DMA".

I moved the drives over to the Intel ICH7R controller and it made the
difference:

Burst Speed: 270.6 MB/s from 116 MB/s
Random Access: 13.1ms from 13.8ms
Average Read: 101.7 from 67.5 MB/s
CPU Utilization: 2% from 8%

I've also updated the HDTach screenshots: http://www.cageynet.com/hdtach.htm

Thanks for all the responses.

Kevin G

 

[keving98]

Nope, that was nowhere suggested.


Exactly. As he didn't need to.





With the emphasis on *few*.


So what.


Those are the 'PCIs' that are labeled as broken.
The ones that are working as supposed to, manage to do well over 100MB/s.


No, really?
Maybe because you weren't measuring PCI bus bandwidth at all?

And what's with this "PCI DMA".

I moved the hard drives over to the Intel ICH7R controller and it made
the difference:

Burst Speed: 270.6 MB/s from 116 MB/s
Random Access: 13.1ms from 13.8ms
Average Read: 101.7 MB/s from 67.5 MB/s
CPU Utilization: 2% from 8%

I also updated the HDTach screenshots: http://www.cageynet.com/hdtach.htm

Tbanks for all your responses.

Kevin G

 

[Folkert Rienstra]

[snip]


I moved the hard drives over to the Intel ICH7R controller and it made
the difference:

But not the difference that's expected.
There still appears to be some bottleneck, that or HD Tach is broken.
The graph reminds me of Diskspeed32, it too shelved the STR of my SCSI
drive up until the very end where the actual STR was lower than that shelf.
HD Tach showed it as expected so the problem was in Diskspeed32's pro-
gramming.

Burst Speed: 270.6 MB/s from 116 MB/s

Good, so that should allow higher STRs as well.

Random Access: 13.1ms from 13.8ms

Average Read: 101.7 MB/s from 67.5 MB/s

Unfortunately that's the maximum read as well. That's not right.
You should see a similarly sloped curve as that from the single drive. Only
the last 100Gig of raid follows that line for the last 50 of the single drive.

The question would now be, does it feel much faster now, since 100MB/s
vs 70MB/s is not all that huge a difference and they are both plateaued
so it should not make much difference where files reside on the drives,
except for the very end.

If not, HD Tach may be correct and there is something not quite well with
that system. Else HD Tach is at fault.