it makes you wonder what the hel* was the point of the latter test? The former shows that with NCQ, yes theres virtuallly NO difference in the rar/zip times. However the former test and subsequent discussion pointed that NCQ had huge impact on the number of mails imported? So from this we can conclude that the NCQ/NO NCQ tests in the latter Seagate article should have been performed on a machine where CPU isn't the limiting factor (dual core or HT). But instead they were performed on AMD crappy POS which doesn't benefit that much on NCQ as its not designed for multitasking! And no notes regarding how much mail were imported during the zip crunching.
So essentially the test by Purav Sanghani is critically flawed when it comes to NCQ performance in the non-AMD single core setups.
Strange thing is that I never heard of 7200.7 having NCQ, I have this drive myself. However I also have Maxtor Diamondmax 10 and boy does this beat the 7200.7. I hope you'd test some different NCQ hdd's in these multitasking scenarios.
Haha, I freaking knew it. I remember Anand talking about this very same thing quite some months ago. That dual core would aid multitasking but that I/O bandwidth was still a limiting factor in how much dual core could actually make a difference. So when So reading the hints you dropped in the blog before this one got me thinking, could some new feature in recent HDDs be aiding performance in combination with dual core? NCQ baby.
Ha! When I read your earlier blog post and you mentioned that some other factor was influencing your multitasking scores, NCQ was the first thing that came to my mind. I was unsure, however, because most of the benchmarks for command queuing show little to no improvement. Nice to see the benefits, finally.
Using WinTrace32 and RankDisk (both part of the discontinued Intel IPEAK Storage Performance Toolkit), I have measured huge differences (up to 30 percent) in I/O performance in some multi-tasking scenarios. Your stopwatch timings confirm that these differences in I/O performance potentially have a significant impact on the user experience.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
7 Comments
Back to Article
sixpak - Thursday, April 21, 2005 - link
If you carefully look at results of ncq/no ncq athttp://www.anandtech.com/storage/showdoc.aspx?i=23...
and compare to
http://www.anandtech.com/cpuchipsets/showdoc.aspx?...
it makes you wonder what the hel* was the point of the latter test? The former shows that with NCQ, yes theres virtuallly NO difference in the rar/zip times. However the former test and subsequent discussion pointed that NCQ had huge impact on the number of mails imported? So from this we can conclude that the NCQ/NO NCQ tests in the latter Seagate article should have been performed on a machine where CPU isn't the limiting factor (dual core or HT). But instead they were performed on AMD crappy POS which doesn't benefit that much on NCQ as its not designed for multitasking! And no notes regarding how much mail were imported during the zip crunching.
So essentially the test by Purav Sanghani is critically flawed when it comes to NCQ performance in the non-AMD single core setups.
Or how else you explain that
http://www.anandtech.com/storage/showdoc.aspx?i=23...
states the Maxtor Diamondmax looks to have multitasked differences like this:
NCQ 68.837 NO NCQ 68.604
"While the Maxtor DiamondMax 10 performed the best out of all the drives, its NCQ performance was slightly lower than with the feature disabled"
That statement totally contradicts YOUR tests which say:
"Seagate Barracuda 7200.7 NCQ - 25.2 minutes
Seagate Barracuda 7200.7 no NCQ - 33.6 minutes"
I think this matter needs a much more thorough investigation with TOP test bed, not any crappy AMD single core.
Also the test should include the new version of WD SATA Caviar w/NCQ that probably bests Diamondmax 10. WD3200JD
Turnip - Wednesday, April 6, 2005 - link
And I'm pretty chuffed I mentioned this in comments on the original "It's been a long weekend" entry! :) Whoohoo! :)As to the particularly variable results... I guess that with a single threaded, single core processor there is always a small variance between tests.
As more and more factors (dual core, dual threading) are introduced into a test, I guess it's reasonable that the variance increases.
When you get a lot of things going on, Windows can be an unpredictable beast at the best of times.
Still. It's all good fun. Well. That's what we all tell ourselves, anyway ;)
Keep up the good work!
sixpak - Wednesday, April 6, 2005 - link
Strange thing is that I never heard of 7200.7 having NCQ, I have this drive myself. However I also have Maxtor Diamondmax 10 and boy does this beat the 7200.7. I hope you'd test some different NCQ hdd's in these multitasking scenarios.Bill - Wednesday, April 6, 2005 - link
Haha, I freaking knew it. I remember Anand talking about this very same thing quite some months ago. That dual core would aid multitasking but that I/O bandwidth was still a limiting factor in how much dual core could actually make a difference. So when So reading the hints you dropped in the blog before this one got me thinking, could some new feature in recent HDDs be aiding performance in combination with dual core? NCQ baby.GhandiInstinct - Wednesday, April 6, 2005 - link
Off I go to return my 74g Raptor...jbond04 - Tuesday, April 5, 2005 - link
Ha! When I read your earlier blog post and you mentioned that some other factor was influencing your multitasking scores, NCQ was the first thing that came to my mind. I was unsure, however, because most of the benchmarks for command queuing show little to no improvement. Nice to see the benefits, finally.Femme Taken - Tuesday, April 5, 2005 - link
Wow, nice work.Using WinTrace32 and RankDisk (both part of the discontinued Intel IPEAK Storage Performance Toolkit), I have measured huge differences (up to 30 percent) in I/O performance in some multi-tasking scenarios. Your stopwatch timings confirm that these differences in I/O performance potentially have a significant impact on the user experience.