Barracuda 5 SATA review

[LiamC]

http://www.lostcircuits.com/advice/sata150/

Compares a Maxtor D740X to a 30GB engineering sample SATA 'Cuda V

But I have a few concerns about the review.

The 'cuda could have an 8MB buffer. The Seagate PR says 8MB, but the review states 2MB (it is after all an ES). Is there any software out there that will let you measure the buffer?

The software used to test is old. Winbench 98.

The author does use a Maxtor, and in another review on the same site (with a 60GXP) the IBM appears faster because of Tagged Command Queuing and controller issues.

I have emailed the author and hope he will answer.

I was certain that SATA wouldn't do jack performance wise, but if my concerns are unfounded, I will have to revise my opinion.

Why can't someboy send me a drive to review...

 

[timwhit]

When we received the drive from Seagate, we had to promise we would not run any benchmarks or at least not publish them. Well, we lied ..... Let the games begin!

That is by far the best line in the article.

Pretty impressive performance, but I'll reserve judgement until SR has a chance to review the drive w/SATA.

 

[Prof.Wizard]

Although I'm impressed with Barracuda's V performance in these legacy tests in respect to current generation drives, I refrain from talking cause the SR treating is still pending.

By a first look the drive looks pretty impressive (, dear Mercutio)... :wink:

 

[time]

Based on the dramatically skewed results, I reckon it's gotta be 8MB cache.

I'm not surprised that SATA performance is acceptable, but I don't think you can draw any other conclusions from those tests.

However, I can't resist a premature gloat about the access time. Here, only two people believed the Barracuda V would meet its claimed seek time of 9mS (myself and Prof, I imagine). I see that HDTach measured 13.2mS average access, less 4.2mS latency gives ... 9.0mS. :mrgrn:

 

[Prof.Wizard]

I see that HDTach measured 13.2mS average access, less 4.2mS latency gives ... 9.0mS. :mrgrn:

That's right, time.
Well, I wasn't expecting EVER to agree with you after the vacuum tube mobo thread, but there you go...

 

[Tea]

But it isn't going to have a proper warranty - so who cares? I certainly won't be buying any.

 

[Prof.Wizard]

But it isn't going to have a proper warranty - so who cares? I certainly won't be buying any.

Well Tea, if you find ONE producer that gives a good drive with a 3-year warranty then drop me a call to arrange it for you...

Maxtor, Seagate, and WD together control more than 80% of the market. All three have passed to 1-year warranties regarding the IDE products, right?

What about IBM?

 

[NRG = mc²]

Yes but maxtor and WD still have 3 year warranties on their top of range products.

 

[Tea]

And Samsung have great drives and three year warranties.

 

[Prof.Wizard]

OK buy your Samsung and/or your top-notch (if you're so handy with lots of money, Will) WD/Maxtor drives and let me stick with the 1-year warranty Cudas...

 

[NRG = mc²]

Well I was almost always buying top of the range IDE drives anyway... no 5400rpm for ages now.

 

[Prof.Wizard]

I thought the 1-year warranty was extended to anything IDE, not only 5400RPM drives...

 

[NRG = mc²]

Well WD says the JB series will only have 3 years warranty and thats fine with me as theyre only a few bucks more than the normal versions, and I was intending to get JBs anyway.

Maxtor says the top of the line ones will have a 3 year warranty, and unless theyre intending to release several 7200rpm drives then thats what I was going to buy anyway.

 

[Handruin]

But it isn't going to have a proper warranty - so who cares? I certainly won't be buying any.

I would buy the drive with a 1 year warranty if they reduce the price based on warranty level. I'd spend $40 on the drive for a 1 year warranty.

 

[Mercutio]

But they won't, Doug. They'll release a $85 drive instead of a $95 drive and say they're "passing the savings along to the customer."

Uh-huh.

For the rest, I'm guessing that everyone and his brother releases SATA drives with 8MB cache instead of 2MB, all the while claiming that any resulting performance improvement is the result of SATA, rather than the cache.

From time to time I still read hobbyist magazines that claim Ultra133 is "faster" somehow than Ultra100. This will be more of the same.

 

[LiamC]

Just received a reply from MS @ LostCircuits - the drive does have an 8MB buffer - so that's where the perf comes from...

 

[Splash]

...Compares a Maxtor D740X to a 30GB engineering sample SATA 'Cuda V

But I have a few concerns about the review.

The 'cuda could have an 8MB buffer. The Seagate PR says 8MB, but the review states 2MB (it is after all an ES). Is there any software out there that will let you measure the buffer?...

All SATA Barracuda V drives will have an 8MB buffer/cache. The PCB for the SATA drives has a 8MB module mounted on it along with the SATA controller. Being a 30 GB version, this is an EARLY engineering sample drive that will NEVER be available. SATA Barracuda V drives will only be available in 80 GB and 120 GB capacities and the "conventional" ATA-100 Barracuda V drives will be available in 40 GB, 60 GB, 80 GB, and 120 GB capacities.

...I was certain that SATA wouldn't do jack performance wise, but if my concerns are unfounded, I will have to revise my opinion.

Well, like many, you will find that your expectations of this newfangled skinny serial bus to be unfounded. Serial ATA is MORE than just a tweak at the physical layer for ATA, it also has enhancements to the communication protocol that make SATA significantly more efficient -- including a 50% bus speed increase. SATA will bring expensive SCSI-like performance to inexpensive ATA storage devices. The next version of SATA will double bus speed and have more protocol enhancements. The upcoming Serial Attached SCSI will allegedly have "downward compatibility" with inexpensive SATA drives as well. In short, parallel ATA technologies will be resigned to oblivion a lot faster than most might expect.

Why can't someboy send me a drive to review...

If you want a Barracuda V, the "conventional" parallel version is supposed to start shipping around the 7th of October (originally, it was supposed to start shipping "mid-September..." er.. yesterday). The Serial ATA drives are supposed to start shipping about six weeks later (mid/late November).

 

[Splash]

...For the rest, I'm guessing that everyone and his brother releases SATA drives with 8MB cache instead of 2MB, all the while claiming that any resulting performance improvement is the result of SATA, rather than the cache.

No, what will happen with Maxtor and WesternDigi will be that they will release existing and/or new SATA hard drives with the addition of a performance robbing SATA <---> Parallel bridge circuit. Later generations of drives from these companies will have proper SATA controller implementations, but in the short term, they will likely just cede the SATA marketplace to Seagate for the next 8 or so months. However, you WILL see proper SATA II drives from everyone about 14 ~ 18 months from now, as SATA will be very hot stuff in the mass storage industry.

...From time to time I still read hobbyist magazines that claim Ultra133 is "faster" somehow than Ultra100. This will be more of the same.

It is faster, even though the read/write media rate of the hard drive is less than the channel rate. Going from a 100 MB/s to a 133 MB/s channel rate will give you a slight increase in performance because buffer/cache performance will increase.

But, then again, you are up against the communication speed limits of the parallel ATA bus at 133 MB/s. This is a major problem that SATA solves -- signal integrity. In addition to this problem being solved, you can factor in a channel speed boost to 150 MB/s and new protocol efficiencies. The result is that SATA is a clear winner over parallel ATA.

 

[LiamC]

Splash, I would like to be surprised on the upside by SATA but...

Is there a whitepaper floating around that details the efficiency gains in the SATA command protocol?

Seagate are obviously taking the correct move, but do you have any figures on likely performance boost due to the difference in command protocols?

Does the fact that all southbridge from VIA and SiS connect their peripherals onto the PCI bus have an impact? I would imagine that it does. No good having a 150MB/s connection from the drive into a PCI bus at 133MB/s I would think - though happy to educated to the contrary. Not sure how Intel handles it's garden variety PC SB/hubs...

 

[.Nut]

Splash, I would like to be surprised on the upside by SATA but...

Is there a whitepaper floating around that details the efficiency gains in the SATA command protocol?

Seagate are obviously taking the correct move, but do you have any figures on likely performance boost due to the difference in command protocols?

I know of no whitepaper as such on performance increases due to SATA protocol advancements, but it is frequently mentioned -- though less than the channel speed increase with SATA. Instead, I can only offer the fact that any storage engineer knows very well that with tagged command queuing a substantial performance increase can be had with modern multi-tasking operating systems over the sequential I/O method that current ATA uses. Command queuing was introduced with SCSI a long time ago, so whitepapers on the subject have long since aged to a mellow hue of yellow.

SATA II will advance ATA further again into SCSI territory with protocol additions such as advanced scatter-gather, advanced enclosure protocols (temperature, etc.), and even dovetailing into the greater picture of SerialAttached-SCSI and iSCSI (IP storage) as "economy" storage devices. Yes, we will have a notching up of the now-luke-warm iSCSI versus FibreChannel scuffle, likely becoming a battle royale that will involve the dynamic duo of SCSI+ATA (under the iSCSI umbrella) against the bastard child of of what was supposed to have long ago been Serial SCSI -- Fibre Channel. Even further down the road, you'll see SATA with fiber-optic interfaces, along with several more generations that double channel speed each time.

Does the fact that all southbridge from VIA and SiS connect their peripherals onto the PCI bus have an impact? I would imagine that it does. No good having a 150MB/s connection from the drive into a PCI bus at 133MB/s I would think - though happy to educated to the contrary. Not sure how Intel handles it's garden variety PC SB/hubs...

Being that 32-bit / 33MHz PCI can barely sustain a little over 100 MB/s, it had better change. But, even this is beside the point as plugging a decent SATA controller into an "existing" Pentium or Athlon PC with 32-bit / 33MHz will still give you a performance boost just as using any modern SCSI hard drive and host Bus adaptor will. The end result is that your (Serial) ATA hard drive will actually be utilised in a more efficient manner by the multi-tasking operating system! If Via decides to tie their SATA controller to the PCI bus, PCI-X (100 MHz or 133 MHz) will certainly handle SATA effectively.

And as a side note: PCI-X also represents a "clean-up" of the PCI standard over previous generations of PCI just as SATA represents a "clean-up" of ATA. PCI-X is a bit more efficient per clock cycle than the venerable 32-bit PCI. Even with the PCI bus, serial-mania looms with the eventual coming of PCI Express.

 

[LiamC]

.Nut

Doesn't the IBM Deathstar ranges support TCQ? I am under he impression that it was introduced with teh 60GXP. And if it is true, then what is to stop current PATA drives from other manufacturers benefitting? And as TCQ is only optional in SATA (mandatory in SATA-II), isn't it a little premature to start crowing the "benefits" of SATA? Or for that matter, giving the kudos to SATA when it is something else entirely, not actually related to the SATA protocol?