WD740 Raptor and TCQ

[computer]

Guys, does anyone know how to enable/disable TCQ on the WD740? Does it depend upon the controller used and it's then enabled/disabled in the controller's BIOS? I'm finding very conflicting info on this. I came cross an article: http://www.storagereview.com/articles/200507/20050705WD2500KS_8.html and note the comparisons between TCQ on/off in the chart for the WD740. This is why I want to know how it's turned off or on, and how do I find out its status on mine? It obviously CAN be toggled on/off. I have mine on an Asus P4C800-E Deluxe mobo on the Promise controller. I don't see anything in the BIOS or in XP regarding TCQ whether it's on that controller or the ICH5R controller.
Thanks.

 

[computer]

^

 

[John]

Right click My Computer > properties > hardware tab > device manager > look under each of the IDE/ATAPI Controllers and once you find the WD under the primary or secondary channel tabs you'll see "enable command queuing" so uncheck it > ok > reboot when prompted.

 

[Markbnj]

I just tried my Raptor 74g both ways. Using the speed test in the serial ATA driver properties tab for primary channel 0, without command queuing enabled I got 105 burst, 71 sustained. Enabled command queuing and rebooted. Tested again. First shot got 109 burst, 70.3 sustained. Second shot got 112 burst, 69.4 sustained.

So it seems not to make a big difference on my system, if anything burst speeds went up, sustained went down a very little.

Are there any other ramifications of having this enabled or disabled? Any reason to prefer one or the other?

Edit: a decent article that explains it and presents some tests:

http://techreport.com/etc/2005q3/ncq-multitasking/index.x?pg=1

 

[Some1ne]

So it seems not to make a big difference on my system, if anything burst speeds went up, sustained went down a very little.

I think both your small increase and small decrease would fall squarely within your standard deviation if you ran enough tests (in short, tagged command queuing was not responsible for it at all). The test you ran just are not of the type that would expose the benefits of TCQ. Both the burst and sustained speed tests are going to be sequential reads, and in the case of a sequential read, the requests are already optimally ordered. In order to see the benefits of TCQ, you need to either do a random read test, or com up with a way to run two sequential tests simulaneously that access data from different parts of the drive. The benefit of TCQ is in multitasking performance (and also in single applications that have random disk access characteristics, if there even are any). If you have multiple apps all requesting data from different parts of the disk at the same time, TCQ will improve your performance, at least in theory, although like I said, your tests would not expose this fact, because they only consisted of a single application which generated a sequential read pattern.

 

[Markbnj]

Yeah, I think that is exactly right. I may look into doing some additional tests using another tool, but for the moment I will leave it enabled. I expect the real-world access patterns to be nothing at all like those tests, and probably more amenable to being improved by command queuing.

 

[computer]

Originally posted by: John
Right click My Computer > properties > hardware tab > device manager > look under each of the IDE/ATAPI Controllers and once you find the WD under the primary or secondary channel tabs you'll see "enable command queuing" so uncheck it > ok > reboot when prompted.

Thanks John. Obviously I missed that after all these years because I was looking for the words "command". Mine says "Disable tagged queuing" and it's listed under "Disk drives" and not the controller! I was also never asked to reboot, but I did anyway. That's odd.

As for my results, I spent 2 hours + running benchmarks with it disabled (Winbench99, PCmark04, Sandra, AIDA32's HD tests, HDTach), on the WD740 and Maxtor D'max 80gb/8mb/ATA133/7200rpm, then ran only a couple with it enabled, and it's VERY CONFUSING. Some results much better, some much worse! This is on Winbench99, I haven't compared the others yet with it enabled. I can't use my old benchmark results since they were done on a blank unpartitioned WD740 and it of course now has my OS on it and it's partitioned. Obviously, hopefully that makes a HUGE difference because my Winbench99 scores now are HIDEOUS, BOTH WAYS compared to what the were!

My guess is it's not going to make any difference on a typical Desktop PC in normal apps, and it may in file server or web-server apps, or heavy loads.

 

[Pariah]

The controller has to support TCQ, which very few do. You can't just toggle it on/off in Windows device manager. If you're not seeing any performance difference, then you aren't en/disabling TCQ. You do not want to enable TCQ if you're running a desktop as it will drop performance by about 20% which is pretty significant.

 

[computer]

Originally posted by: Pariah
The controller has to support TCQ, which very few do. You can't just toggle it on/off in Windows device manager.

How else do you turn it on/off?

If you're not seeing any performance difference, then you aren't en/disabling TCQ. You do not want to enable TCQ if you're running a desktop as it will drop performance by about 20% which is pretty significant.

It apparently depends upon the drive maker. At that link in my first post, you can see how the WD740 and Maxtor is slower with it ON, but the Seagate is faster with it on.

 

[Pariah]

How else do you turn it on/off?

The BIOS/configuration utility of the controller.

It apparently depends upon the drive maker. At that link in my first post, you can see how the WD740 and Maxtor is slower with it ON, but the Seagate is faster with it on.

I was referring to the Raptor exclusively, which is the only drive relevant to your situation. I don't think any other ATA drive has supported any form of TCQ for years.