Win2000 and ICH8-8R-9-9R-10-10R

[jaclaz]

Err... Humm... Cough, cough.

Well, after installing Xp in dual-boot after W2k, I just noticed inadvertently that Ntldr and Ntdetect had changed their version, but as Xp needs its own files, I couldn't revert to W2k's ones, you see. And then, when measuring boot times, 5.1 file versions were in place for W2k as well, which makes the comparison unfair, as I would revert to 5.0 if using just W2k, you know.

2k3 files seem very close to Xp on my mobos.

Make yourself a few floppy images with the different versions of NTLDR and NTDETECT.COM and BOOT.INI and boot them through grub4dos.

This way you can compare any version (of course you can use only backwards compatibility).

Quick sum up:

http://msfn.org/board/index.php?act=ST&amp...04&t=127900

AND useful should you have a 2K with a registry bigger than around 16 Mb.

jaclaz

[blackwingcat]

I fixed Intel Matrix Storage Manager 7.6 for Windows 2000.

I missed in inf file 'Intel® ICH9M-E/M SATA AHCI Controller'.

[Molecule]

@ pointertovoid ... a very helpful (practical-realistic) topic

Terabyte's Image for DOS ("IFD") will restore a system image unattended (by batch command)

by default IFD will recognize SATA drives as AHCI (or RAID), but for greater speed TB allow a user to bypass AHCI and use BIOS direct

image and restore speeds are competive ... so I figured Terabyte tech support would have struggled with this

so I asked if it made any difference if the source and target drives were on the same or different controller chips ... (e.g. I could hang both source and target satas on ICH10, or have one on ICH10 (SATA only) and the other on JM363 (SATA+IDE) ...

Well, SATA does not have contention down from the controller, but can still

have it through the PCI bus. However, IFD isn't using overlapped IO or

accessing multiple drives at exactly the same time so it shouldn't matter.

Whichever BIOS interface (if not using BIOS (direct) ) is faster would matter

more.

then I asked if there was a speed difference between configuring SATA as IDE (bios handles drives) versus configuring SATAs with AHCI (drivers handle drives) ...

If your BIOS is fast (using UDMA) then using them will be the same (via the

BIOS option). If your BIOS is slow, moving it to AHCI mode, IFD can access

it directly (via BIOS (direct)) and that uses UDMA speeds by default. AHCI

being a standard, a single AHCI driver can be used on multiple systems.

basically, it looks like if you want a faster harddrive, first get a faster mobo with faster bios (ide-udma)

If you have a fast bios, and don't need a RAID array, it looks like adding AHCI driver is not really necessary

[jaclaz]

If you have a fast bios, and don't need a RAID array, it looks like adding AHCI driver is not really necessary

JFYI:

http://www.msfn.org/board/install-usb-wins...4-page-590.html

jaclaz

[Molecule]

thanks jaclaz ... how you remember stuff from page 32 of a 34 page thread AMAZES me! I'm on a dialup, so a thread like that can take hours just to dl ... meanwhile my addled and festering old brain turns into a gigantic fart and disappears like a wizard out a hole missing in the top of my head or something ...

I heard that raid arrays had to be setup during clean install, but I did not know that IDE and AHCI could be swapped post install. Makes sense though (I guess). My ICH10 bios has an option to setup SATAs as AHCI sole, e.g. the options are [iDE, AHCI, or AHCI+RAID]. I'm using w2k, and a prior thread somewhere (arghh) suggested that since there were significant internal differences between ICH7 and 10 chips, the safer bet was to push the risk towards the OS, thus use the ICH10 driver. That's illogical to me. My bet is that ICH7 hooks are as a subset of ICH10 hooks, so the ICH10 chip can handle anything that a ICH7 driver can throw at it. And the ICH10 driver might be able to throw stuff back at the w2k OS that it is not ready for. The driver engineers at Intel are probably the only ones who know for sure. In between I end up with cognigive dissonance, which is resolved by sticking to IDE.

There's a lot of sniffy advice out there ... then cognitive dissonance starts ringing in my brain like merry Christmas Bells ...

Edited November 29, 2009 by Molecule

[jaclaz]

Well, NCQ has not much sense when imaging, but it does seems like having a relevance in day-to-day usage.

jaclaz

[pointertovoid]

Meanwhile I gratefully downloaded a newer version of BlackWingCat's improved driver.

I got it there:

http://blog.livedoor.jp/blackwingcat/archives/813816.html (explanations in Japanese, didn't help me)

http://files.me.com/theblackcat/o5v5q6 (download file iata76_cd2kb.zip)

which bears the same version number 7.6.0.1011 (from Intel)

still has IaStor.sys last modified on 20090416

but has IaStor.inf and IaAhci.inf last modified on 20100515 (hope they are the most recent by now).

And as fantastic good news, I see in the material list of the Inf files:

Ich6 to Ich10 (with all sub-details m, d, r...) AND Pchm and Pch,

which must be the kindof southbridges H55 H57 P55 Q57 or "series 5" intended for the Core i5 on socket 1156.

Does that work? Can we, using this new version, run W2k with Ahci on a Core i5 Clarkdale at full throttle?

Thanks!

[blackwingcat]

Hi pointertovoid.

I try to run this driver on QM57 and PM55 Chipset.

It works fine.

Sincerely.

Does that work? Can we, using this new version, run W2k with Ahci on a Core i5 Clarkdale at full throttle?

[pointertovoid]

Thank you so much!

For the answer, and primarily for the driver. It serves day for day. Really nice from you!

[RJARRRPCGP]

Meanwhile I've received my mobo with chipset Ich10r and know answers to (2) and (6), here they are:

(2) Without any driver, Ich10r runs at full throttle in IDE mode, thus enabling W2k. With 500GB Seagate 7200.12 drives, Atto sees the expected contiguous read speed of 134MB/s and huge throughputs on small files, absolutely identical to what it sees on Xp with Ahci mode chosen in the Bios, and with improvement at Q=10 and Q=4 versus "no overlapped I/O", so the Ncq works even with Ide settings - which isn't that surprising, as the equivalent of Ncq was introduced with Ata/5 (Udma/100). HdTach also sees buffer read speeds of 260MB/s so transmission is at Sata300 speed.

This seems to be true with my Acer Aspire M5630. IDE mode did NOT cap it to UDMA 133. Hooray. (ICH 7?) (2007)

With my Asus P5QL Pro, IDE mode seemed to cap it to UDMA 133! (ICH 10) (2008)

Yours is looking good!

Edited November 28, 2010 by RJARRRPCGP

[pointertovoid]

Here's the kind of performance obtained with Blackwingcat's driver - as good as the XP driver, or even better!

An Intel SLC SSD: 30GB X25-E, on an ICH10R host, measured here by AttoDisk with Q=4.

And now, a Raid-0 of two X25-E on an ICH10R, again with Blackwingcat's driver.

this was with a stripe size of 4kB and an Ntfs volume aligned on megabytes, created by GPartEd.

Measurements by IOMeter are equally impressive, especially if its accesses are aligned on multiples of 4kiB for the Raid - not really cheated, since this is the cluster's size. On a single X25-E disk, alignment of volume or accesses brings nothing - as this disk is already a Raid over 10 channels, not 8 nor 16.

In both cases, an F6 diskette made with this driver was able to install W2k on the disk or group - but only if the first volume was created by the W2k installation disk, not by GPartEd.

Well done, Blackwingcat!

[pointertovoid]

I bought four 600GB Velociraptors - the fastest Sata mechanical disk as these used parts were, errr, less expensive than usual

Notice the arm's speed, much improved at the 450GB and 600GB series. And, with the driver ported by BlackWingCat to run the ich10r, they ruffled my hair

Don't trust Atto's maximum throughput too much, since the VRaptors have 64MB buffer each and this version of Atto tests over a 256MB file... WinBench99 must be more accurate, as usual. Anyway, aligning Ntfs to 1MiB boundaries did improve Atto's results BUT W2k installs only on the volume it created, at sector 63 I suppose.

Also interesting: the group exceeds 2TiB which W2k can't handle directly, but with Intel's Matrix Raid Bios (recent enough!), I could define two pseudo-disks smaller than 2TiB each, and only then it seems to work. That is, 2k wouldn't install at all in >2TiB, but by slicing the group, it did; I was lecherous enough to create the second pseudo-disk at around 2TiB distance from the Win volume and overlapping it, and a slow format of the higher volume didn't destroy the lower volume. So at least for Raid groups between 2TiB and 4TiB, this would be a solution.

They didn't start W2k (F6 install easy) appreciably faster than a single VRaptor does, and certainly slower than my X25-E Ssd , but they allow to browse folders of uncompressed pictures very quickly, and are excellent at launching games - where Ssd don't have the capacity.

Well done, BlackWingCat!

[tomasz86]

Very nice access time Better than mine (6.2 ms) on a SAS 15 000 rpm drive.