PDA

View Full Version : 8KNXP ITE suffers BAD performance...


modem
July 24th 03, 02:23 AM
Earlier today I decided to do a comparison of the ITE Gigaraid IDE
chipset vs a standard Promise UltraTX2 IDE PCI adapter. I did this
after I read a few comments over the past few months that suggest the
ITE Gigaraid chipset is very subpar.

All of my tests were ran on a clean fresh Windows 2000 installed
system on my new GA-8KNXP system with 1GB of Corsair XMS memory. I
used HD Tach for the benchmarking and without further waiting... here
is the results:

(I appologize for no screenshots, I didn't have my network up to copy
them over)


ITE GigaRaid IT8212 (Drivers version 1.0.0)
--------------------------------------------------------------------------------------------------------
Drive Access Time Burst Speed CPU Utilization
--------------------------------------------------------------------------------------------------------
M 30GB 19.0ms 30.9mbps 2.7%
M 40GB 12.2ms 30.9mbps 4.4%
M 80GB 13.2ms 30.8mbps 0.9%
M 80GB 11.7ms 31.0mbps 2.1%

( M = Maxtor )
( = UDMA 5 )
( = UDMA 6 )



Promise UltraTX2 PCI Adapter card (Driver version 2.00.0.42)
--------------------------------------------------------------------------------------------------------
Drive Access Time Burst Speed CPU Utilization
--------------------------------------------------------------------------------------------------------
M 30GB 19.9ms 86.9mbps 1.9%
M 40GB 12.3ms 116.5mbps 2.2%
M 80GB 13.6ms 86.1mbps 7.0%
M 80GB 11.5ms 116.5mbps 4.1%


As you all can see there is QUITE a huge discrepency regarding the
performance of the ITE chip vs a standard Promise IDE adapter. I
can't be sure if this is a driver issue, a chip BIOS issue or what.
If someone knows where to get later drivers I'll give them a test.

Brad

Tim
July 25th 03, 05:48 AM
That need to do some fixes pronto. I have never seen a bios init take so
long (ITE in non RAID config).

- Tim


"modem" > wrote in message
...
> Earlier today I decided to do a comparison of the ITE Gigaraid IDE
> chipset vs a standard Promise UltraTX2 IDE PCI adapter. I did this
> after I read a few comments over the past few months that suggest the
> ITE Gigaraid chipset is very subpar.
>
> All of my tests were ran on a clean fresh Windows 2000 installed
> system on my new GA-8KNXP system with 1GB of Corsair XMS memory. I
> used HD Tach for the benchmarking and without further waiting... here
> is the results:
>
> (I appologize for no screenshots, I didn't have my network up to copy
> them over)
>
>
> ITE GigaRaid IT8212 (Drivers version 1.0.0)
> --------------------------------------------------------------------------
------------------------------
> Drive Access Time Burst Speed CPU Utilization
> --------------------------------------------------------------------------
------------------------------
> M 30GB 19.0ms 30.9mbps 2.7%
> M 40GB 12.2ms 30.9mbps 4.4%
> M 80GB 13.2ms 30.8mbps 0.9%
> M 80GB 11.7ms 31.0mbps 2.1%
>
> ( M = Maxtor )
> ( = UDMA 5 )
> ( = UDMA 6 )
>
>
>
> Promise UltraTX2 PCI Adapter card (Driver version 2.00.0.42)
> --------------------------------------------------------------------------
------------------------------
> Drive Access Time Burst Speed CPU Utilization
> --------------------------------------------------------------------------
------------------------------
> M 30GB 19.9ms 86.9mbps 1.9%
> M 40GB 12.3ms 116.5mbps 2.2%
> M 80GB 13.6ms 86.1mbps 7.0%
> M 80GB 11.5ms 116.5mbps 4.1%
>
>
> As you all can see there is QUITE a huge discrepency regarding the
> performance of the ITE chip vs a standard Promise IDE adapter. I
> can't be sure if this is a driver issue, a chip BIOS issue or what.
> If someone knows where to get later drivers I'll give them a test.
>
> Brad

Tim
July 26th 03, 01:20 AM
It seems, for some really odd reason that you have to have the ITE
configured as RAID before the Windows installer will run. Same for other IDE
RAID.

Seems picky to me.

You can't configure a non-raid config on any of these controllers when they
are set to RAID in bios. IE I thought JBOD was where each drive was an
ordinary individual drive, instead it seems to be drive spanning (IE a
variant of RAID 0 - Striping), but alas no, these controllers insist you
have 2 or more drives before JBOD is permitted.

JBOD is as dangerous as RAID 0 - twice the probability of a drive failure
(with 2 discs) and so loss of data.
Strange, everyone goes for performance, I go for data safety - RAID 1 - if
any RAID.

The device scan with ITE is about as slow - if not slower than SCSI. But at
least with SCSI in the same time it will pick up 15 devices, not four. Way
way too slow. They need to fix this.

- Tim


"Bob Davis" > wrote in message
. ..
> Well, when first installed this mobo I tried loading the ITE drivers, but
> XP-Pro wouldn't take them despite numerous tries. At this point, after
> reading this newsgroup for six weeks, I'm glad I failed--as I've heard
> nothing good about this controller.
>
> "Tim" > wrote in message
> ...
> > That need to do some fixes pronto. I have never seen a bios init take so
> > long (ITE in non RAID config).
> >
> > - Tim
> >
> >
> > "modem" > wrote in message
> > ...
> > > Earlier today I decided to do a comparison of the ITE Gigaraid IDE
> > > chipset vs a standard Promise UltraTX2 IDE PCI adapter. I did this
> > > after I read a few comments over the past few months that suggest the
> > > ITE Gigaraid chipset is very subpar.
> > >
> > > All of my tests were ran on a clean fresh Windows 2000 installed
> > > system on my new GA-8KNXP system with 1GB of Corsair XMS memory. I
> > > used HD Tach for the benchmarking and without further waiting... here
> > > is the results:
> > >
> > > (I appologize for no screenshots, I didn't have my network up to copy
> > > them over)
> > >
> > >
> > > ITE GigaRaid IT8212 (Drivers version 1.0.0)
> >
>
> --------------------------------------------------------------------------
> > ------------------------------
> > > Drive Access Time Burst Speed CPU Utilization
> >
>
> --------------------------------------------------------------------------
> > ------------------------------
> > > M 30GB 19.0ms 30.9mbps 2.7%
> > > M 40GB 12.2ms 30.9mbps 4.4%
> > > M 80GB 13.2ms 30.8mbps 0.9%
> > > M 80GB 11.7ms 31.0mbps 2.1%
> > >
> > > ( M = Maxtor )
> > > ( = UDMA 5 )
> > > ( = UDMA 6 )
> > >
> > >
> > >
> > > Promise UltraTX2 PCI Adapter card (Driver version 2.00.0.42)
> >
>
> --------------------------------------------------------------------------
> > ------------------------------
> > > Drive Access Time Burst Speed CPU Utilization
> >
>
> --------------------------------------------------------------------------
> > ------------------------------
> > > M 30GB 19.9ms 86.9mbps 1.9%
> > > M 40GB 12.3ms 116.5mbps 2.2%
> > > M 80GB 13.6ms 86.1mbps 7.0%
> > > M 80GB 11.5ms 116.5mbps 4.1%
> > >
> > >
> > > As you all can see there is QUITE a huge discrepency regarding the
> > > performance of the ITE chip vs a standard Promise IDE adapter. I
> > > can't be sure if this is a driver issue, a chip BIOS issue or what.
> > > If someone knows where to get later drivers I'll give them a test.
> > >
> > > Brad
> >
> >
>
>

Bob Davis
July 26th 03, 01:55 AM
"Tim" > wrote in message
...

> It seems, for some really odd reason that you have to have the ITE
> configured as RAID before the Windows installer will run. Same for other
IDE
> RAID.
>
> You can't configure a non-raid config on any of these controllers when
they
> are set to RAID in bios. IE I thought JBOD was where each drive was an
> ordinary individual drive, instead it seems to be drive spanning (IE a
> variant of RAID 0 - Striping), but alas no, these controllers insist you
> have 2 or more drives before JBOD is permitted.


Well, I had everything set in the bios for RAID, had the two drives ready on
the two ITE controllers, and had the array created. I don't know what else
I could've done to tell the system, Windows, and the world I was trying to
install WinXP on the two RAID drives, which were the only two in the
computer. They were set up as an array prior to the OS installation, and
ITE was seeing them as such.


> JBOD is as dangerous as RAID 0 - twice the probability of a drive failure
> (with 2 discs) and so loss of data. Strange, everyone goes for
performance, I go for data safety -
> RAID 1 - if any RAID.


I never worried much about the chances of an array failure in my old system,
which ran RAID0 (Highpoint on an Abit mobo). Every Saturday I cloned the C:
drive in Norton Ghost, left the cloned drive in the system (Win98SE), and
backed up important files (email, WAB, business databases, etc.) every
evening. If the RAID crapped out I could be up and running in 10 or 15
minutes without even any major hardware swapping. When my IBM 75GXP died
about 18 months ago I lost nothing and was back to normal in minutes.

You can't safely leave a cloned drive with an active partition in a system
running WinXP and presumably W2K, but Win9x has no problem at all with such
an arrangement. Now after I do my weekly clone I remove the drive before
rebooting into WinXP, put the clone on the shelf (with three others that are
rotated), and perform my daily backups to a dedicated backup drive that is
permanently installed in the system as D:. It doesn't contain all files
like before, but just those dynamic files I would have backed up before onto
the cloned drive. I'm not running RAID now, but if C: dies I would use the
latest clone, then move the dynamic files from D: back to C:, essentially
accomplishing the same thing as before, just with a few more steps to the
procedure.


> The device scan with ITE is about as slow - if not slower than SCSI. But
at
> least with SCSI in the same time it will pick up 15 devices, not four. Way
> way too slow. They need to fix this.


I seem to recall the ITE scan to be slow, but so was the Highpoint on my old
system, and also for the SCSI adapter I still have installed. The SCSI
(Adaptec AHA-2930) takes at least 15 seconds to scan when it isn't optioned
to boot from CD, in which case it takes seven seconds longer while it checks
the CD drive on SCSI-ID 0 every time for a bootable disk. I normally have
that option disabled for obvious reasons, as I rarely need to boot from a
CD.

Timothy Drouillard
July 26th 03, 02:05 AM
Dumb question
Did you remember to load the RAID drivers at the start of installing Windows
when it asks you if you want to add support for any other controllers
(something like that).

"Bob Davis" > wrote in message
.. .
>
> "Tim" > wrote in message
> ...
>
> > It seems, for some really odd reason that you have to have the ITE
> > configured as RAID before the Windows installer will run. Same for other
> IDE
> > RAID.
> >
> > You can't configure a non-raid config on any of these controllers when
> they
> > are set to RAID in bios. IE I thought JBOD was where each drive was an
> > ordinary individual drive, instead it seems to be drive spanning (IE a
> > variant of RAID 0 - Striping), but alas no, these controllers insist you
> > have 2 or more drives before JBOD is permitted.
>
>
> Well, I had everything set in the bios for RAID, had the two drives ready
on
> the two ITE controllers, and had the array created. I don't know what
else
> I could've done to tell the system, Windows, and the world I was trying to
> install WinXP on the two RAID drives, which were the only two in the
> computer. They were set up as an array prior to the OS installation, and
> ITE was seeing them as such.
>
>
> > JBOD is as dangerous as RAID 0 - twice the probability of a drive
failure
> > (with 2 discs) and so loss of data. Strange, everyone goes for
> performance, I go for data safety -
> > RAID 1 - if any RAID.
>
>
> I never worried much about the chances of an array failure in my old
system,
> which ran RAID0 (Highpoint on an Abit mobo). Every Saturday I cloned the
C:
> drive in Norton Ghost, left the cloned drive in the system (Win98SE), and
> backed up important files (email, WAB, business databases, etc.) every
> evening. If the RAID crapped out I could be up and running in 10 or 15
> minutes without even any major hardware swapping. When my IBM 75GXP died
> about 18 months ago I lost nothing and was back to normal in minutes.
>
> You can't safely leave a cloned drive with an active partition in a system
> running WinXP and presumably W2K, but Win9x has no problem at all with
such
> an arrangement. Now after I do my weekly clone I remove the drive before
> rebooting into WinXP, put the clone on the shelf (with three others that
are
> rotated), and perform my daily backups to a dedicated backup drive that is
> permanently installed in the system as D:. It doesn't contain all files
> like before, but just those dynamic files I would have backed up before
onto
> the cloned drive. I'm not running RAID now, but if C: dies I would use
the
> latest clone, then move the dynamic files from D: back to C:, essentially
> accomplishing the same thing as before, just with a few more steps to the
> procedure.
>
>
> > The device scan with ITE is about as slow - if not slower than SCSI. But
> at
> > least with SCSI in the same time it will pick up 15 devices, not four.
Way
> > way too slow. They need to fix this.
>
>
> I seem to recall the ITE scan to be slow, but so was the Highpoint on my
old
> system, and also for the SCSI adapter I still have installed. The SCSI
> (Adaptec AHA-2930) takes at least 15 seconds to scan when it isn't
optioned
> to boot from CD, in which case it takes seven seconds longer while it
checks
> the CD drive on SCSI-ID 0 every time for a bootable disk. I normally have
> that option disabled for obvious reasons, as I rarely need to boot from a
> CD.
>
>
>

July 26th 03, 08:42 PM
Exactly! My friend editted the actual file, and changed the path - he
didn't browse to it, because as you stated, that doesn't work. If I
can get a copy of the file from him, I'll try to post it.

>That was the case here, acting like it couldn't find that file. I
>downloaded the drivers from the Gigabyte site and put them on a floppy, but
>it still couldn't find it no matter where I put it, and I tried every folder
>(including root) of the floppy disk. I finally gave up after wasting
>several hours futzing with it.
>
> wrote in message
...
>
>> There is a problem with the .oem setup file on the drivers disk /
>> cdrom. I was working on the file with a friend of mine, and he
>> figured out the paths are wrong. If you fix the file, it will let you
>> install the drivers using the F6 and S key procedure as listed in the
>> RAID manual that came with the motherboard.
>

July 27th 03, 06:42 PM
I heard back from my friend - states he found an easier way:

After you create the driver disk, make another directory on the floppy
and name it: ITE_RAID Then copy the 3 driver files from the XP dir to
the ITE_RAID directory, and to the root directory of the floppy disk.

Even if you don't need this info anymore, I'm sure lots of other
people out there would find it handy.

Nobody

>Thanks, but ITE RAID reportedly runs like a dog anyway, so I'm not even
>going to bother. My single Maxtor runs fine for now.
>
> wrote in message
...
>> Exactly! My friend editted the actual file, and changed the path - he
>> didn't browse to it, because as you stated, that doesn't work. If I
>> can get a copy of the file from him, I'll try to post it.
>>
>> >That was the case here, acting like it couldn't find that file. I
>> >downloaded the drivers from the Gigabyte site and put them on a floppy,
>but
>> >it still couldn't find it no matter where I put it, and I tried every
>folder
>> >(including root) of the floppy disk. I finally gave up after wasting
>> >several hours futzing with it.
>> >
>> > wrote in message
>> ...
>> >
>> >> There is a problem with the .oem setup file on the drivers disk /
>> >> cdrom. I was working on the file with a friend of mine, and he
>> >> figured out the paths are wrong. If you fix the file, it will let you
>> >> install the drivers using the F6 and S key procedure as listed in the
>> >> RAID manual that came with the motherboard.
>> >
>>
>

Bob Davis
July 27th 03, 09:34 PM
> wrote in message
...

> I heard back from my friend - states he found an easier way:
>
> After you create the driver disk, make another directory on the floppy
> and name it: ITE_RAID Then copy the 3 driver files from the XP dir to
> the ITE_RAID directory, and to the root directory of the floppy disk.
>
> Even if you don't need this info anymore, I'm sure lots of other
> people out there would find it handy.


Well, I may not need it now, but I've saved for later use in case GB gets
their act together and revives ITE RAID. As it is now I wouldn't use it
even if I could get it to work. Thanks for the input.