Slow SATA II Performance With ASUS P5WD2 and Deskstar drives

[kafkawasnotdreaming]

I have been tweaking my new ASUS P5WD2 Premium for months, and have only one very surprising bottle-neck: my state of the art SATA II drives.

I have four of these:
Two Hitachi Deskstar 80 gb SATA II, mirrored in RAID 1 as boot drive C
Two Hitachi Deskstar 500 gb SATA II on the same controller, but configured as stand-alone units
To facilitate RAID 1, I switched on the system bios's Configure IDE as RAID option. This does not compel all connected drives to RAID; it just makes it possible to enter the RAID utility to configure the drives as you wish.

My Sisoft Sandra scores are a hoot: 50 mb

HD Tach is more merciful, but not in line with published expectations and marketing brou-ha-ha: 60 mb

I have had the drives in and out of RAID, and turned the bios on and off from RAID, with no significant difference in the outcome.

Since I have tested both RAID and non-RAID drives, the RAID striping is not an issue. The same applies to the mirroring argument, which appears irrelevant. It certainly doesn't speed anything up.

Can you believe I replaced a Western Digital 10,000 rpm SATA I Raptor drive with this junk? Now I peer at the Raptor like a long-lost girl friend. Okay, okay. It was like living near the airport. Without the ear muffs. But it was fast, dog-gone-it.

What factors go into turning such a promising technology into cotton candy? Have I been a bad-boy over-clocker, and throttled my system? At least I didn't hook the main board up to the space shuttle's Nitro tank and a utility pole, like that guy in Japan. Besides, pi is unresolvable; who cares how many digits you spit out?

But I digress.

I am using the motherboard's original SATA cables. I heard there were new ones for external SATA, but I don't think they work on the internal fittings.

I do pack my hard drives in insulated Smart Drive cases, but those things have copper heat pipes, and they report temperatures at around 40 C.

I have a humongous power supply, so no contention there.

I loaded the Hitachi Deskstar Feature Tool and set the drives to 3.0 gb plus Spread Spectrum. I turned off unnecessary items such as acoustic management, etc. The settings are loosey-goosey. I have had them on and off a dozen times, and they never seem to say -- especially Spread Spectrum. I have used the drives both as 1.5 gb and 3.0 gb, and nothing really changes.

What system bios settings affect the SATA read/write speed? What are the known issues with bottle-necks and throttles for SATA on these modern PC boards? Any thoughts?[/img]

 

[Bozo]

I must be missing something,
Hitachi gives ~61MB/sec as max substained data rate. Your HD Tach scores are 60MB/sec. Where's the problem??

Bozo :mrgrn:

 

[Handruin]

I'm also missing something. Which drive is producing 60MB/s in HD Tach? If it's the mirrored 80GB pair, then I'd say it's about right. Only some RAID controllers (3Ware for example) offer a performance increase in a mirrored set by load balancing I/O the the different mirrors. SATA 1.5Gb/s or SATA 3.0 GB/s isn't going to make a substaintial difference in performance. Neither drive is capable of saturating the bandwidth. At best you'll gain native command queueing on the 500GB Hitachi's when you use a SATA II 3.0Gb/s depending on the controller.

Can you believe I replaced a Western Digital 10,000 rpm SATA I Raptor drive with this junk? Now I peer at the Raptor like a long-lost girl friend. Okay, okay. It was like living near the airport. Without the ear muffs. But it was fast, dog-gone-it.

Depending on who you ask here, that may be considered a smart move. I've never owned a raptor, but a few people here have had mixed results with reliability.

 

[kafkawasnotdreaming]

The problem is that the HD Tach score is way, way off the Sisoft Sandra score, and the system is quite sluggish. It has physically slowed down since going to SATA. Is Sandra a bad benchmark for SATA?

The speed meaurements are consistent over all four SATA drives. The morroring doesn't dramatically affect this.

And a note on the Raptor: This is my second drive. The original went back after 24 months. I do think they can overheat, or become unstable, over heavy use.

I want to compare my experience with others on this forum, and pick up motherboard bios hints that will get this system back to a gallop. I have successfully overclocked just about everything else.

 

[Bozo]

The average seek time for the Hitachi is 8.8ms. The average seek time for the Raptor is 5.2ms Sustained transfer rate for the Raptor is 72MB/sec.

No amount of tweaking or overclocking is going to make a slow drive go fast. RAID 1 will not increase your performance either.

You can put 2 of the Hitachis in RAID 0 for a performance boost , but the failure of one drive or a file system error will destroy all your data. BUT, RAID 0 will not decrease your seek time. It might even make it slower.

Generally, seek time is what makes a system "feel" fast or slow.

Don't know about Sisoft Sandra vs HDTach

Bozo

 

[kafkawasnotdreaming]

Another power user achieved a Sistoft Sandra hard drive index of 85 with the same brand and drive (albeit mine is 500 gb):

http://www.bytesector.com/data/bs-article.asp?ID=538&page=2

This is double my performance! We have the same motherboard, processor, and Corsair low-latency memory.

The only difference I see is that they are using Windows XP 64-bit Edition. I fond that to be incompatible with 90% of hardware, including major-brand printers such as HP.

 

[Mercutio]

The conclusion you might draw is that SiSoft Sandra isn't a reliable benchmark and/or is measuring something other than disk performance.

 

[Handruin]

That article is testing the Hitachi T7K250 which isn't the same as your 7K500. The T7K250 is only offered in 250GB and 160GB size drives. I own two of the T7K250's @ 250GB. I'll (reluctantly) post scores from sandra 2005 so you can see a comparison. BTW, that site does an awful job of weighting the scale for a benchmark. That graph makes it look like the SATA II interface is more than twice as fast, which it isn't.

 

[Bozo]

Also, that site is using XP64.
XP64 uses the 2003 Server hard drive settings for improved performance.
You are trying to compare apples to bananas.

Bozo :mrgrn:

 

[sechs]

The conclusion you might draw is that SiSoft Sandra isn't a reliable benchmark and/or is measuring something other than disk performance.

Wasn't there some review that found that the Sandra benchmark wasn't determinate?