Sign in to follow this  
halcyon

"disable Tagged Queuing" In Xp (sata Disks)?

Recommended Posts

I'm running Sil3112 controller integrated on my motherboard with two Samsung SP1614C SATA disks.

I have scoured far and wide and have not found any indication as to Samsung's supporting TCQ (Tagged Command Queuing).

However, when I view the "Scsi Properties" tab under Windows XP Device manager for these disks there is a tick box "Disable Tagged queuing" available. By default this is un-ticked (i.e. TCQ ON).

Now, I have ticked this box (TCQ=OFF) as I'm pretty sure the disks don't support it.

However, I'm curious: does anybody know if the controller (Sil3112) supports TCQ or not? Or is the tick box in properties just a fluke?

Share this post


Link to post
Share on other sites

Tagged queuing is really only for SCSI even though the term is used losely to descibe other forms of queuing as well. The SiI 3112 does not have support for SATA native command queuing and the Samsung drive is a bridge drive to begin with, which leaves some sort of legacy queuing as the one used by IBM / Hitachi as the only solution the tab could apply for.

One other possibility is that the drivers / BIOS of the controller are already looking ahead to the 3512 controller, which does support NCQ but in your particular case I would argue that the checkbox has no functional meaning.

Share this post


Link to post
Share on other sites

SiI and others use SCSI miniport drivers instead of (MS/Intel) IDE drivers. The checkbox is there because Windows thinks it a SCSI adapter, nothing more.

Even Highpoint, which has ATA queuing, doesn't use the checkbox.

Share this post


Link to post
Share on other sites

if you're using it for BSD then there might be hope

i remember hearing about an IDE controller under BSD that'll support TCQ... or i could just be sleep deprived and dreaming

Share this post


Link to post
Share on other sites

Thanks for the replies. It's good to have confirmation.

For those who missed it ("unregistered"), NCQ <> TCQ. I was talking about TCQ, not NCQ. Different things those.

regards,

Halcyon

Share this post


Link to post
Share on other sites

The guys at XBitLabs found that the TCQ tickbox did pretty much nothing (even for SCSI HAs), but then they only had a beta sample of the 74 gig Raptor.

Share this post


Link to post
Share on other sites

i dont see why it wouldnt support both. I wish there was soembody who could provide a clear description of the differences of the different command queing methods. As well as some performance comparisons.

Share this post


Link to post
Share on other sites
i dont see why it wouldnt support both. I wish there was soembody who could provide a clear description of the differences of the different command queing methods. As well as some performance comparisons.

Seagate published a white paper, which was written by somebody I know extremely well, on the differences at Spring IDF 2003.

Share this post


Link to post
Share on other sites
Halycon, you were talking about the SiI 3112 controller which, if any, would support NCQ but it doesn't

unregistered,

I know for a fact that SIL3112 does NOT support NCQ. Silicon Image was quite firm on this and NCQ is not in the specs for this chipset.

However, they do have a PCI-X bridgeable chipset that does support NCQ (SIL3124).

But all along I was talking about ATA CQ, NOT NCQ.

Get those CQs sorted out :)

regards,

Halcyon

Share this post


Link to post
Share on other sites

p.s. The question of performance benefits is really tricky because it depends among other things on the level of hyperthreading or, more precisely, on the number of logical processors present in the system. In any standard application, I wouldn't expect any performance increase from NCQ, just as SCSI drives cannot play out their better performance and advantage derived from TCQ in a single host - single device environment.

On the other hand, especially in two way HT and even more so in some upcoming (still hypothetical) 4-way HT CPUs, I can really see that queuing could make a big difference.

BTW, the above mentioned paper was originally a comparison between legacy queuing, SCSI tagged command queing and NCQ with all the differences in the queing schemes in the context of the environment the drives were supposed to work in. E.g. the out of order data delivery in NCQ and the ModifiyDataPointers in SCSI (that was never implemented) can only be appreciated in the context of the protocol etc... but I am preaching to the prophet here since I know that you are well aware of most of these issues..

What was and still is missing is a broad comparison of the different performance metrics within the different environments and there are technical reasons as well as some, er.. other reasons. I have tried to work on some of these issues but every answer so far has raised more questions and I am just where I was about 2 years ago when I started working on these issues first. I guess, time will tell, that is, after sifting through gazillion of irrelevant benchmarks all over the web, there will be some consensus somewhere .. sometime. B)

Share this post


Link to post
Share on other sites
One other possibility is that the drivers / BIOS of the controller are already looking ahead to the 3512 controller, which does support NCQ but in your particular case I would argue that the checkbox has no functional meaning./QUOTE]

What is with SiI and slow product releases?

The 3512 is a 3112 replacement that was announced Feb 2003, but still not available. The 3124 is a faster PCI-X chip demoed at IDF Sep 2003, but still not available.

SiI 3512 Press Release

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
Sign in to follow this