Hitachi T7K250 3Gbps results and an explaination

[Scottz]

According to several reviews, 3Gbps transfer speeds with current SATA-II drives may not buy you much. But what the heck. If your hardware can handle it, why not use it?

I have x2 Hitachi T7K250 250Gig drives that I'm going to use for Video Editing so I thought I'd try it out.

I found this review in regards to using the T7K250 and 3Gbps transfer speed.
http://www.hardcoreware.net/reviews/review-283-1.htm

The T7K250 doesn't negotiate the speed with the controller and is shipped with the default 1.5Gpbs speed enabled. This is to be more compatible with controllers that don't support 3.0Gbps.

It mentions the Hitachi Feature Tool that you'll need to set the speed (On Hitachi drives of course). Use the link in the review to download it.

Side Note 1: I need to mention to make sure to really power off your system (pull the power cord or use your power supply switch) after changing the setting or else it will revert to 1.5Gbps. Boot the Feature Tool again and check to make sure the 3Gbps option stuck.

Side Note 2: As it says, DO NOT change this if your controller doesn't support 3Gbps transfer speeds.

As the review shows, in a single drive setup the burst rate is improved, but that's it.
Here's my HDTach test for a single drive:

1.5Gbps (Red) and 3.0Gbps (Blue) Long Test:


Ok, so not that much of a big deal.
Average Read/Write are identical, but burst speed has improved.

But what about RAID-0?

I was trying out 128K stripe due to my files being 100MB - 7Gig.
In 1.5 mode, I was seeing alot of jagged edges in theory due to the onboard drive cache being more optimized for smaller blocks (turns out 32K is the sweet spot).

1.5 mode - Single Drive (Red) - 128K Stripe RAID-0 (Blue)


Compared to:

3.0 mode - Single Drive (Red) - 128K Stripe RAID-0 (Blue)


So I got about another 10MB per second on the Read/Write, a better burst rate (not sure how real-world that factors in), and an improvment on how it handles large block sizes.

This pretty much confirms the review.
What I didn't see in the review was how 3.0Gbps helped out with larger stripe sizes, so that was a bonus.

The additional help in RAID-0 may not be stellar, but it is an improvment.
I definatly wouldn't run out to buy drives based on this.

But if your disks and controller support it, you might want to consider using it. Just don't expect a huge jump in performance (especially on a single drive).


My Setup:
Supermicro PDSGE - Intel 955X chipset
Intel ICH7R SATA RAID controller
Pentium-D 840 (3.2Ghz)
2Gig DDR2 667
Nvidia 6600 PCIe
Matrox RTX100
RME 9632 Audio

WD360GD 36Gig Raptor (boot drive)
x2 Hitachi T7K250 250Gig (Raid-0)

-Scott

 

[Scottz]

Well, I've just done some more looking around on the SATA-II spec and found that port multiplying is in there. I even found some SATA-II hubs already on the market.

Now the 3Gbps makes since and why many don't see any advantages for it.

If you've ever run a SCSI RAID, you'll understand immediatly what it means.

While drives for SCSI-Ultra320 never hit the 320MBps (Byte, note the big B) on their own, it was allowing more drives to talk on the same cable without interfering with each other.

If you get x4 drives on the same cable, you need a faster transfer speed so each drive doesn't have to wait to send/receive it's information.

The transfer speed between the drive and controller needs to be faster than what the drive itself can do on it's own.

i.e. The drive may work great at a sustained rate of 50MB per second, but if you have x4 drives on the same cable running at 50MB and the transfer speed was 50MB per second, then only one drive could talk to the controller at a time leaving the other 3 drives waiting. You would need at least 200MB+ per second transfer speed on the same cable so all drives could send/receive data at the same time.

Increase the transfer speed to 320MB per second and you have some breathing room plus room for a couple of more drives on the same cable.

What's that got to do with SATA-II 3Gbps transfer speed?

SATA-II 3Gbps is also known as SATA-II 300 (for 300MB per second).
(SATA takes 10 bits per byte)

With the support for port multipliers (Sata-II hubs if you will), you can add more drives to a single controller port.

Your motherboard may have 4 SATA ports on it, limiting you to only 4 devices. Throw a port multiplier on one of the ports and you can add more drives to a single controller port.

More drives talking on the same port requires faster transfer speeds to allow all of them to talk at the same time.

So with the Hitachi drive here, theoretically (depending on who's math you use) I could get x4 drives on a single port with room to spare.

x4 Hitachi T7K250 running at 55MB per second = 220MB per second.

Anyway, this is why we aren't seeing much benifit to the 3Gbps speed today. Not just that the drives haven't caught up to it, but rather their starting to add more options for more devices on the same port/channel/cable like their older SCSI cousins.

It explains the better burst speed and not overall read/write performance when 3Gbps is enabled.
At least to me anyway 8)

 

[guy12357]

hey scottz,

im wondering if you tried games on with the raid 0 setup. Im thinking about whether to have 2 hitachi sata II hard drives in raid 0 or 1 western digital raptor 74gig hard drive. Im wondering if the raid 0 sata II hard drives are fast enough for gaming.

 

[DD_Jay_AZ]

SATA-II 3Gbps is also known as SATA-II 300 (for 300MB per second).
Yea, I know the math is wrong but unfortunatly that's how the industry is refering to both.

Actually, it's not wrong.

It takes 10 bits to transmit 8 standard bits, the extra 2 bits are for synch and error control if I recall my SATA specs.

For Serial ATA, 3Gbp/s = 300MBp/s

The reason for the 3Gbp/s is for pure marketing value of bigger numbers sounding better.

Enjoy!

 

[Scottz]

It takes 10 bits to transmit 8 standard bits

AHHH!!! Ok... I get it.
Was wondering about that.

So it's my math that's wrong

I'll take that bit out.

 

[DD_Jay_AZ]

You made me dig this up since I couldn't remember.

http://www.serialata.org/docs/serialata%20-%20a%20comparison%20with%20ultra%20ata%20technology.pdf

Why 8 = 10 for SATA

Directly from the horse's mouth no less!

 

[DD_Jay_AZ]

btw, your spot on about the bandwidth per channel

Look into a Silicon Image 3132 controller and 3726 Multplier.

I've envisioned eventually getting 4 to 6 SATA-II drives, 2 of each going through a 3726 and the two 3726's going into each port of the 3132, which is attached directly to the PCI Express bus. Of course this would be in a RAID-0.

My 2x Seagate RAID-0 pulls about 75MBp/s now, I figure with the setup here, I could expect about 150MBp/s throughput at least.

BTW, the Sil3726 is a 1-to-5 multiplier, I could theoretically hook up 10 drives to my single Sil3132 controller.. seek times would be disgusting, but OMG THINK OF THE THROUGHPUT!!!!

Just can't find ANY Sil3726 based products available in the U.S. yet, bummer!

 

[Scottz]

Thanks... I get it now.

( 3000 x .8 ) / 8 = 300

80% of the 3Gpbs is our data.

Which if 10bit encoded, is really 240MBps.
That's marketing for ya.

Takes me back to the old baud vs. bps days.

 

[Scottz]

Look into a Silicon Image 3132 controller and 3726 Multplier.

You know, I just got the SIIG PCI Express RAID which uses the 3132.

I was surprised to see my CPU usage jump to 60% when copying files in RAID-0.

I also couldn't get the burst rates that I was seeing in HDTach with the ICH7R.

There was no way to confirm the 3Gbps was being used, the SIIG configuration doesn't show it. It's a new card and their app to configure it is a bit raw.

I was about to take it back, but thought I'd ask you how yours is working out?
Maybe I got something wrong or should use a different driver?

 

[Scottz]

hey scottz,

im wondering if you tried games on with the raid 0 setup. Im thinking about whether to have 2 hitachi sata II hard drives in raid 0 or 1 western digital raptor 74gig hard drive. Im wondering if the raid 0 sata II hard drives are fast enough for gaming.

I have the 36Gig Raptor which is a little slower than the 74Gig.
The RAID 0 still does much better than the 74Gig buy itself.

Take a look at this web review (page 3).

http://www.cluboverclocker.com/reviews/hard_drives/hitachi/T7K250/index.htm

And that's without the 3Gbps enabled.

 

[DD_Jay_AZ]

Wow! 60%?

I just did a simple test with a 590MB file, copied from desktop to root C:, then copied from desktop to a secondary 160GB PATA drive.

Either copy operation never broke 15% CPU useage.

I'm also using an AMD/nVidia based system that has the Sil3132 integrated onto the motherboard PCB. Might effect something.

Driver Version: 1.3.0.9
Driver Date: 10/18/2005
Driver Provider: Silicon Image
BIOS Version: No idea, it's the release BIOS for the Abit AN8 32X, don't want to reboot to find out.

I wouldn't use the SiiG driver unless you had to, I've never had good experiences with their drivers. Only reason I've ever used them is because they made some hard to find style of component for a laptop that only their driver would work with.

It also wouldn't hurt to update your card's BIOS, SiiG is notorious for not updating production cards firmware.

With the Silicon Image driver, there is a "Device Info" tab that you can verify what generation of SATA your drives are running in.

Linkage:
http://www.siliconimage.com/support/supportsearchresults.aspx?pid=32&cid=3&ctid=2&osid=4&

Try this driver and let me know if you have better results.

Also, the ICH7R uses "Matrix" RAID, Intel's little whacky RAID protocol that actually works quite well. However, it only truly shines if you install the Windows based applications for it, like the Application Accelerator.

 

[Scottz]

Yup, the SIIG driver and BIOS are way out of date.

Driver: 1.2.3.1
BIOS: 7.2.13

The current Silicon Image is:

Driver: 1.3.0.9
BIOS: 7.2.30

BUT... both the Silicon Image and SIIG driver won't allow updates to the BIOS through the "Silicon Image ATA Controller Properties". I'm getting a "CHIP: Unknown Size: Unknown" in the Flash BIOS tab.

Even SIIG's driver download claims to include a BIOS, but doesn't.

ARG! There using a a flash chip that either isn't supported by Silicon Image for the 3132 or there's something else with the PCIe version.

I'll keep looking into it.

 

[DD_Jay_AZ]

Just use the Silicon Image BIOS & Utility to do it, should work fine. No guarantees though. Like I said.. not a big fan of SiiG devices, had plenty of problems with them.

Linkage:
http://www.siliconimage.com/support/supportsearchresults.aspx?pid=32&cid=15&ctid=2&osid=0&

As soon as your BIOS and driver are up-to-date, try HDTach again, I'm betting your CPU usage will drop dramatically and your overall performance will increase.

You'll just want to use the "UPDFLASH" utility to update the firmware, shouldn't hurt the SiiG BIOS that way. Of course, have the compatible driver ready.

Let me know how it works out.

 

[Scottz]

Thanks again,

I had found those tonight and the dos based one doesn't work either.
It can't find the card even with the -ID3132 being set. If I recall, PCIe cards have their own issues being seen from DOS (different routine than scanning the normal PCI bus). Why PCInfo isn't being developed in DOS anymore.

Confirmed that the SIIG's are using a non-compliant Silicon Image flash which leaves it up to SIIG to come up with a util to do this (according to Silicon Image).

There's another thread going on over at Storage Review where it's been confirmed. I got a ticket open at SIIG, but as another user has reported, they don't have an answer yet.

If I don't get an answer quick, it's going back to the store and hopefully I can find another PCIe to compare against.

Thanks again for your help.

 

[Scottz]

Well SIIG is one of those companies that doesn't respond to anything, so the card is going back.

Should have known, but to others, stay far away from that card.

One last thing I noticed and a bit curious about is in the Hitachi Feature Tool that Ultra DMA 6 133MB is selected.

Being that SATA is a serial extension of the ATA spec, I'm speculating that you will never see a single drive really go above 133MB burst (or average more than 66MB per sec) due in part by the legacy ATA spec.

That sound right?