Seagate's NCQ

[derekblankmccoy]

Anybody know anything about this? Click

SEAGATE SHOWS SATA WITH NCQ CAN BEAT 10K SATA; SEAGATE NCQ SHIPS NEXT MONTH

SCOTTS VALLEY, Calif.?04 May 2004? Seagate (NYSE:STX) today at WinHEC demonstrated in head-to-head tests using Intel's IOMeter benchmark that its Barracuda 7,200-rpm Serial ATA (SATA) hard drive with Native Command Queuing (NCQ) technology equals or exceeds the performance of the latest 10,000-rpm SATA hard drives on the market. Seagate also announced it is shipping its Barracuda 7200.7 SATA hard drives with NCQ technology to direct OEM customers next month. The demonstration can be seen this week at Booth 411 at WinHEC in Seattle.

 

[bluewall21]

About NCQ or seagate?

 

[derekblankmccoy]

Seagate, of course, never heard of such a company.

 

[jrphoenix]

This is awesome. I have been waiting for the new NCQ drives to start showing up.... this will be a big performance booster

 

[mechBgon]

So they're claiming they can outperform the WD Raptor (in IOMeter at least)... interesting! By what kind of margin, I wonder. I remember them claiming the Barracuda ATA IV would be the fastest 7200rpm ATA drive on the market when it came out, too, which it clearly was not, nowhere close.

Now the $64000 question: does this capability work in all situations on all SATA controllers, or does it require awareness/cooperation on the part of the controller?

 

[Pariah]

This means practically nothing to any of us. IOMeter is a glorified access time benchmark that tests the drives ability to handle an extreme amount of very random accesses that in no way mirrors anything any of us will ever be doing with our computer. It's not surprising that Seagate chose the Raptor, because it doesn't support ncq. Any remotely recent ncq supporting drive regardless of speed will beat any non-ncq capable drive in IOMeter.

If you peruse the performance database over at SR you will see how little affect it has on non-server benchmarks. Look at the beating the original Raptor takes in the server benchmarks vs similar generation 10k SCSI drives which are ncq enabled, and then look how it does against them in the workstation benchmarks, slightly beating them. Now compare the Raptor to 7200RPM ATA drives, where it basically beats them proportionally to their rotational increase in server benchmarks, while in workstation benchmarks which don't benefit as much from great random access, the margin of victory actually decreases.

If you're expecting anything above a minimal performance gain from ncq in anything but server benchmarks, you're going to end up disappointed.

NCQ requires a supporting controller and drive.

 

[jrphoenix]

Originally posted by: Pariah
This means practically nothing to any of us. IOMeter is a glorified access time benchmark that tests the drives ability to handle an extreme amount of very random accesses that in no way mirrors anything any of us will ever be doing with our computer. It's not surprising that Seagate chose the Raptor, because it doesn't support ncq. Any remotely recent ncq supporting drive regardless of speed will beat any non-ncq capable drive in IOMeter.

If you peruse the performance database over at SR you will see how little affect it has on non-server benchmarks. Look at the beating the original Raptor takes in the server benchmarks vs similar generation 10k SCSI drives which are ncq enabled, and then look how it does against them in the workstation benchmarks, slightly beating them. Now compare the Raptor to 7200RPM ATA drives, where it basically beats them proportionally to their rotational increase in server benchmarks, while in workstation benchmarks which don't benefit as much from great random access, the margin of victory actually decreases.

If you're expecting anything above a minimal performance gain from ncq in anything but server benchmarks, you're going to end up disappointed.

NCQ requires a supporting controller and drive.

Hello. I hope this isn't a lame question. Would this mean that I would have to purchase a controller (like something made by Promise) to use NCQ? Are these controllers built into any upcoming motherboard chipsets (Nvidia, VIA, SIS)?

Thank you!

 

[Pariah]

I'm not aware of any current SATA controllers that support NCQ. It isn't in the original SATA spec. Seagate says the following in their press release:

""The Native Command Queuing technology is one of the most anticipated new performance capabilities delivered by the Serial ATA II Working Group in addressing the needs of modern multi-threaded environments," said Knut Grimsrud, Serial ATA steering committee chairman and Intel senior principal engineer."

The release, however, makes not mention that this new drive is SATA II, and there definitely aren't any SATA II controllers available, so I have no idea what Seagate is doing with this drive. Maybe they'll bundle a compatible controller like older ATA drives that required a new controller for 48bit LBA support.

 

[bluewall21]

Originally posted by: derekblankmccoy
Seagate, of course, never heard of such a company.

bleah

NCQ is cool, HDDs are finally getting more efficent. Think about the elevator analogy. Then you'll realize why NCQ is such a smart thing.