SATA Drives-Not up to specifications

[Gamma1]

Hi:

There has been a lot of marketing hype about SATA drives compared to IDE's. There have many tests
including ones done right here, that show that EIDE-5, ATA-100, 7200 RPm, 8mb cache are better than SATA 150's, in this light I have decided to share a little information that has always been there, except sometimes it get put aside to do other things.

Here I go :

Transfer rates between disk drive and System boards memory by direct access UDMA are leveraged
by several factors, the speed of the drive, the cache size of the drive, the adapters ect, Here the focus is on reality not fantasy and I cover , the information that will allow you to make your own decision.
DISK Drive Maker's claim SATA 150 has a maximum transfer rate of 150 MB/s. This is totally a "marketing hype" and is flase in every sense of the word and technology. I use ATA-5 7200 RPM, 8 mb cache and its really good and getting less expensive all the time, if SATA -150 was faster I would get, the facts are it is not and here is the proof:

Drive Technology Transfer harness number of wires Maximum speed of transfer

SCSI 320 ,, 80 wires , 320 MB/sec
SCSI 160, 68 wires ' 160 MB/sec
SCSI , fast , ect ,, 50 wires ' to 80 MB/sec
ATA -5 ,, 40 wires , to 100 MB/sec
SATA 150 ,, 7 wires , to 150 MB/sec (?)


You can see here that there is a pattern and its quite clear , THE GREATER NUMBER OF TRANSFER WIRES , the "FASTER THE TRANSFER SPEED POSSIBLE, its quite clear that SATA 150 is not possible and that with the reduced number of data transfer wires, the speed of these drives, when driver cheat codes and other cheat codes are eliminated is far less, these disk drive companies like others say " NEW Technology" and expect we the public to believe there statements and allow them to make purchasing decisions for us,,,well I'm not impressed by there lack of engineering talent to make faster desktop models available, everyone already knows, "disk drives are the slowest in data transfer, inside the computer", of course one could argue its PCI or ISA, ect, the point is SATA cannot transfer anywhere near 150 MB/s and is slower than ATA 5 with same speed, size and cache size. What can done to correct this? Thats reengineer and redesign every SATA 150 (7 data wire connector) so that the transfer wires
are increased to at least ATA -7 with 50 wire harness, otherwise they violate, the entire SCSI Industry, and there engineers, who have pioneered higher speed transfer, its quite obvious these companies, with just another suit and tie in there are not being truthful. SCSI has been around since the 1920' and 1930's here SCSI is American standard for all disk drive comparisons.
Thanks

Gamma1

 

[smthmlk]

yes, more wires does often times increase the amount of total bandwidth a cable can carry. however, this is not always the case. A wire, or medium, for which signals transfer through depend on the bandwidth of the medium. If your medium has a high bandwidth range (in Hz, such as 550Mhz for cat6 for example) vs. a much higher bandwidth of a fiber optic cable, more channels of data can travel over it. Also, the shielding of the cable may also help to prevent interference, reducing the number of burst errors on the cable, and therefore reducing the number of retransmitted packets.

i have actually built a computer with the OS on a WDC SATA drive, and the performance difference is noticeable. So, it prompts some questions: Is the bandwidth of the 7 wires in the serial ata cable higher than those 80 wires in an eide cable? Considering how small and the high quage (high guage = small wire, btw) of each wire in an ide ribbon cable, it might be true. However, I've never cut a SATA cable in half to see how thick those 7 wires are. The cable is decently thick, however, and if there are only 7 wires in it, you might expect them to be of a fairly decent guage (lower than those found in ide ribbons, meaning thicker wires, meaning most likely a higher bandwidth for each, and therefore a higher possible data-rate).

Im not saying your information on SATA is false, but I think there's more to it then just 'the number of wires in the cable'. Perhaps someone else who knows more specifics about serial ata cables can give us some solid information on them vs. ide ribbon cables. From my own experiences with SATA, however, I really have to say it seems a lot faster to me but people can 'imagine' things

 

[jagec]

Originally posted by: Gamma1

You can see here that there is a pattern and its quite clear , THE GREATER NUMBER OF TRANSFER WIRES , the "FASTER THE TRANSFER SPEED POSSIBLE, its quite clear that SATA 150 is not possible

So are all P4's slower than all Athlons because they have fewer IPC?

There's more to it than just number of wires.

 

[Swanny]

Unfortunately you're not right.

SATA uses a serial interface.

SCSI and PATA (ATA) use parallel interfaces.


The interfaces are completely differently so you can't compare number of wires.

 

[Lonyo]

Try here

So why would parallel transmission be replaced by a serial method? After all, the most common serial ports on computers (the COM ports) are being replaced because they are too slow. However, their replacements are USB and FireWire, both of which are based on serial technology. Believe it or not, the fastest computer memory, RDRAM, is serial based, as is the networking protocol, Ethernet. Here we are talking about data transmission rates from 100-460 MB/s. In addition to the possible increases in speed, there is a great reduction in the number of wires required for the connection. USB 2.0 uses only four wires but has a maximum data capacity of 460 MB/s, So the design of the new Serial ATA (SATA) hard drives has an entirely new connector to the drive and the motherboard. The SATA connector has only 7 pins compared to the ribbon cable connection of 40 or 80 pins. In addition, the maximum cable length is 1-meter (39.37 inches) compared to the 18 inch maximum for parallel ATA. The cable is keyed to fit the motherboard and drive so there can be no error in the way it is connected. Data transmission requires only a 250-mV connection compared to the 5V used in parallel ATA. The reduction in the number of wires also allows an increase in signal wire diameter, which reduces resistance and impedance. This increases the speed of signal propagation.

The transmission of data through these ribbon cables occurs in parallel mode. This means that transmission can occur over multiple wires at the same time. However, as transmission speed increases the high frequency signals develop problems such as cross-talk and signal reflection, among others. ATA drives are self-terminating. This is to reduce the problem with signal reflection. However, when a single drive is attached to the middle connector on the two-drive cable, reflection problems increase. When the change was made from 40 wire to 80 wire cables, even with the modified connector, problems increased. The smaller wire diameters increased the electrical resistance of the system and limited cable length to 18 inches. Folds in these ribbon cables also increase the probability of electrical problems and signal degradation. And, if the cable is crimped, transmission wires can be easily broken.

 

[SUOrangeman]

Is that rubbish I smell?

If I am not mistaken, we're talking about *THEORETICAL* maximums here. In other words, these various specs posted indicate the fastest transfer rate support by the implementation. How many real products actually brush up against these maximums? In fact, how many hard drives actually get close to the max during a burst? HOW ABOUT NONE.

The only knocks on SATA at this current time is that it is slightly more expensive than its IDE coutnerparts (when there is an equivalent) and the somewhat flimsy connectors. The "missing" command-queueing is not a disadvantage when comparing to IDE.

Only now are we seeing SATA controllers being embedded into chipset (remember, onboard SATA isn't always equivalent to an inherent feature of the motherboard chipset). Once a majority of motherboards offer a chipset-based SATA solution, SATA should beging to distance itself from our old, dear friend.

Anyway, I'll put my first-generation Raptor (riding a 66Mhz bus) up against any single-drive IDE solution you canimagine. StorageReview tells me that the closest IDE competitor is WD's own WD2x00JB.

-SUO

 

[Mark R]

SCSI has traditionally been a parallel interface - multiple signals wires are used to carry the data. Early versions of SCSI were available in standard and 'wide' versions. The wide versions used more wires to carry twice as much data. The different speed grades (fast, ultra, ultra 160, ultra 320) changed the speed at which data was sent down each wire (20 MHz, 40 MHz, 80 MHz and 160 MHz respectively).

50 pin connections were used for 'narrow' SCSI, and 68 pin for 'wide'. Ultra 160 and ultra 320 scsi are only available in 'wide' and both use 68 pins. Hot swappable drives and enclosures use 80 pin connectors which include power and ID controls.

IDE only came in 'narrow' versions, and used 40 wires. When ATA 66 was launched, the high-frequency signals started to suffer from cross-talk - the signal in one wire would leak into its neighbours. The solution was to add 40 more wires. On the 80 wire cable, every other wire is connected to ground and acts to shield the signal wires from each other. No information is carried down these extra wires.

The higher speed ATA variants also put stricter limits on impedence - which meant that wires had to be of higher quality generally. (Note that impedence is not resistance, and the change in wire size didn't alter the impedence significantly).

Serial ATA instead of making the communications channel wider, instead made the channel much narrower but much, much faster (1.5 GHz). A special cable is needed to cope with this speed. In the SATA cable which has 7 conductors - 3 of them are used as shields. The other 2 are arranged in pairs - with one signal being sent on 2 wires to reduce interference. There is one signal connection from PC to drive, and one signal connection from drive to PC. Each signal pair has its own shield. The third shield goes round the whole cable.

The most important specifications of the SATA cable are the quality of the signal pairs (tightness of binding and accuracy of twisting). The size doesn't really matter - although the recommended sizes are 28 and 30 AWG (the same size as the wires on a 40 or 80 wire IDE cable).

 

[Pariah]

Originally posted by: Gamma1
SCSI 320 ,, 80 wires , 320 MB/sec
SCSI 160, 68 wires ' 160 MB/sec
SCSI , fast , ect ,, 50 wires ' to 80 MB/sec
ATA -5 ,, 40 wires , to 100 MB/sec
SATA 150 ,, 7 wires , to 150 MB/sec (?)

Parallel interfaces:

U320 SCSI comes in both 68 and 80 pin configurations. The extra pins in the 80 pin (SCA) format are for power only to be plugged into back plains instead of having a seperate power connector that 68 pin drives have.

50pin SCSI top out at 40MB/s (Ultra2 SCSI), while 40 pin also tops out at 40MB/s (UW SCSI).

ATA-5 is 80 wire (40 pin), not 40 wire.

Serial interfaces:

SATA
SAS
IEEE1394(firewire)

Parallel and serial interfaces are not comparable for what you are trying to prove.

Basically everything you said is wrong, as previous posters have pointed out, so I won't bother repeating what they've said.

 

[Zebo]

Only thing that sucks about SATA is the cost of cables in a large case because those tiny cables they give you with MOBO do NOT do the job. I spent $6 ea for 36" cables for them .


Speed wise just go to storage review and see how WD or hitachi SATA drives dominate thier IDE counter parts.

 

[imported_NoGodForMe]

Excellent explanations.

SCSI costs twice as much as other drives, plus you need an Adaptec controller, but it's well worth it.

My Seagate Cheetah drives have come thru for years, and they are fast.

Review sites usually don't care about hyping scsi drives, because they're usually benchmarking games, and the only thing SCSI would help is load times. But I see the differance doing normal stuff. Plus I'm first into a level when a new map loads. In UT2K4, I spawn, and then see the message that says, "Waiting for other players." When the same map restarts for the next round, it's me and a few other people in the map first, then everyone else a few seconds later.

 

[Hyperlite]

Originally posted by: Swanny
Unfortunately you're not right.

SATA uses a serial interface.

SCSI and PATA (ATA) use parallel interfaces.


The interfaces are completely differently so you can't compare number of wires.

furthermore...
They both excel in certain areas. SATA drives have gotten to the point now where, for example, a 10k raptor SATA drive can beat out a 15k SCSI drive in Read speed and burst speed, where as SCSI appears to excel in write speed and seek time. this information comes straight from a benchmark performed by Maximum PC, June Issue. the 10k SCSI Atlas drive couldn't compare to the 10k SATA Raptor. I quote "The fact is, these high-velocity SCSI drives arn't made for sequential transfers, as their high rotational velocities are ment to provide enormous benefits in multi-user environments with heavy random access workloads."

The fact is, SATA appears to be the SCSI for home/desktop/laptop applications. once again, they both appear to excel in certain areas.
Depending on the type of test OF COURSE A SCSI 320 DRIVE WILL BE FASTER. but a SCSI 320 drive isn't very practical for a home application. the prices of course can't be compared....i just can't see a reason to pit SATA against SCSI...completely different applications IMHO.

 

[AndyHui]

This is an apples to oranges comparison with a low speed parallel interface against a high speed serial interface. Simply counting the number of wires is flawed.

 

[mechBgon]

this information comes straight from a benchmark performed by Maximum PC, June Issue. the 10k SCSI Atlas drive couldn't compare to the 10k SATA Raptor. I quote "The fact is, these high-velocity SCSI drives arn't made for sequential transfers, as their high rotational velocities are ment to provide enormous benefits in multi-user environments with heavy random access workloads."

I wonder if Maximum PC has ever heard of SCSI's "Tagged Command Queueing" My Cheetah 15k.3 is almost three times as fast as a Western Digital Special Edition ATA drive in some real-world (repeat, real-world) tasks that I use it for. Unless the Ratpro is more than three times as fast as an 800JB Special Edition, it had better bring its own Vaseline if it wants to race against a Cheetah 15k.3 in my system. :evil: And in the realm of 15k SCSI drives, the Cheetah 15k.3 isn't the fastest either.

Too bad Maximum PC is content to merely scratch the surface...

 

[Zebo]

Originally posted by: mechBgon

this information comes straight from a benchmark performed by Maximum PC, June Issue. the 10k SCSI Atlas drive couldn't compare to the 10k SATA Raptor. I quote "The fact is, these high-velocity SCSI drives arn't made for sequential transfers, as their high rotational velocities are ment to provide enormous benefits in multi-user environments with heavy random access workloads."

I wonder if Maximum PC has ever heard of SCSI's "Tagged Command Queueing" My Cheetah 15k.3 is almost three times as fast as a Western Digital Special Edition ATA drive in some real-world (repeat, real-world) tasks that I use it for. Unless the Ratpro is more than three times as fast as an 800JB Special Edition, it had better bring its own Vaseline if it wants to race against a Cheetah 15k.3 in my system. :evil: And in the realm of 15k SCSI drives, the Cheetah 15k.3 isn't the fastest either.

Too bad Maximum PC is content to merely scratch the surface...

It's pretty fast even beating your cheetah in some benchmarks

 

[Hyperlite]

Originally posted by: Zebo

Originally posted by: mechBgon

this information comes straight from a benchmark performed by Maximum PC, June Issue. the 10k SCSI Atlas drive couldn't compare to the 10k SATA Raptor. I quote "The fact is, these high-velocity SCSI drives arn't made for sequential transfers, as their high rotational velocities are ment to provide enormous benefits in multi-user environments with heavy random access workloads."

I wonder if Maximum PC has ever heard of SCSI's "Tagged Command Queueing" My Cheetah 15k.3 is almost three times as fast as a Western Digital Special Edition ATA drive in some real-world (repeat, real-world) tasks that I use it for. Unless the Ratpro is more than three times as fast as an 800JB Special Edition, it had better bring its own Vaseline if it wants to race against a Cheetah 15k.3 in my system. :evil: And in the realm of 15k SCSI drives, the Cheetah 15k.3 isn't the fastest either.

Too bad Maximum PC is content to merely scratch the surface...

It's pretty fast even beating your cheetah in some benchmarks

mechbgon, select the High-end test, and put those 2 in a head to head test and the results are obvious, unless there is some thing about your system that makes your cheetah THREE TIMES AS FAST. and look back at what i said earlier, then look at the server mark suite of tests. thats exactly what i said, the SCSI dominates in server applications, and nothing else. furthermore, the Raptor beats the Cheetah in the majority of the other tests.

 

[Pariah]

They both excel in certain areas.

SATA the interface doesn't excel in anything compared to the SCSI interface except cost.

SATA drives have gotten to the point now where, for example, a 10k raptor SATA drive can beat out a 15k SCSI drive in Read speed and burst speed

No on both counts, unless you are talking about older generation 15k drives. All 4 current gen 15k drives have faster max and average STR than the 2nd gen Raptor.

I quote "The fact is, these high-velocity SCSI drives arn't made for sequential transfers, as their high rotational velocities are ment to provide enormous benefits in multi-user environments with heavy random access workloads."

And how exactly do high RPM's hurt sequential transfers? Their conclusion isn't really wrong, but their reasoning for the results are. The 2nd gen Raptor is just as fast as current 15k drives in most home users situations, while SCSI will crush the Raptor in a number of situations most of us will never see, mainly due to firmware optimizations that are tuned for different environments. However, again, it's not SATA interface that has caught up with SCSI, it's the Raptor that has. Drives have a much greater affect on home user performance than the interface. It should also be noted that "current" gen 15k drives are up to 18 months old now. 3 of the 4 SCSI makers are currently sampling new drives that should be available soon, with specs that appear to leave even the 2nd gen Raptor in the dust. (100MB/s peak STR, 5.0ms access time)

 

[nitromullet]

Originally posted by: Pariah
..it's not SATA interface that has caught up with SCSI, it's the Raptor that has...

Among many good points made in your post, I am glad the you mentioned this as well. There is a misconception that SATA is just fast because it is SATA. The Raptor is elite among HDs in that respect, and the only reason for this is because it was intended for the enterprise market. That being said, I'm sure that WD isn't too upset about the fact that it has become the choice drive for enthusiasts.

 

[mechBgon]

Originally posted by: Hyperlite

Originally posted by: Zebo

Originally posted by: mechBgon

this information comes straight from a benchmark performed by Maximum PC, June Issue. the 10k SCSI Atlas drive couldn't compare to the 10k SATA Raptor. I quote "The fact is, these high-velocity SCSI drives arn't made for sequential transfers, as their high rotational velocities are ment to provide enormous benefits in multi-user environments with heavy random access workloads."

I wonder if Maximum PC has ever heard of SCSI's "Tagged Command Queueing" My Cheetah 15k.3 is almost three times as fast as a Western Digital Special Edition ATA drive in some real-world (repeat, real-world) tasks that I use it for. Unless the Ratpro is more than three times as fast as an 800JB Special Edition, it had better bring its own Vaseline if it wants to race against a Cheetah 15k.3 in my system. :evil: And in the realm of 15k SCSI drives, the Cheetah 15k.3 isn't the fastest either.

Too bad Maximum PC is content to merely scratch the surface...

It's pretty fast even beating your cheetah in some benchmarks

mechbgon, select the High-end test, and put those 2 in a head to head test and the results are obvious, unless there is some thing about your system that makes your cheetah THREE TIMES AS FAST. and look back at what i said earlier, then look at the server mark suite of tests. thats exactly what i said, the SCSI dominates in server applications, and nothing else. furthermore, the Raptor beats the Cheetah in the majority of the other tests.

Reality > benchmarks. In a real-world I/O-intensive task, my Cheetah does in 45 seconds what a WD Special Edition does in 120 seconds. You don't like it? Tough. I bought the hardware to perform well under heavy I/O loads, not to play benchmarking games, and it's pretty obvious to me on a daily basis that I made the right choice.

Don't get me wrong, I would pick a 74GB Raptor over any other ATA drive (cost notwithstanding), and for many purposes it would undoubtedly be a wonderful performer while offering lots of capacity for the dollar. I might pick one up sometime as a bulk-storage drive, in fact, and also run my real-world work tasks on it to finally settle the question of how it compares to a SCSI drive in the real-world stuff where drive performance matters. Heh, but by then I might have a Fujitsu MAU-series SCSI drive too... :evil:

 

[glugglug]

With the exception of the Raptors, WD SATA drives are NOT NATIVE SATA. Since there is actually NO BENEFIT to SATA as of yet (other than cable size &amp; production costs), pretty much every manufacturer other than Seagate decided to add a parallel ATA <-> SATA converter to the controller boards on their existing parallel ATA drives - the drive actually still communicates on the parallel interface, which has this chip in between it and the SATA connector.

So any difference you are seeing is clearly from improvements in the physical drive technology (track size, RPMs, etc), NOT the SATA interface.

 

[XplosiV]

And there i was thinking they planed to take SATA beyond 150, upto i think 500 over the next 18 months or so, SATA 150 was sposed to be the earlyest form of SATA for public retail with developments comming to bring i think SATA 300 &amp; 500 in the future. If that happens, what will beat it then ?

 

[Sonix7]

SATA is just appearing in the desktoop market and the first version SATA-150 maybe does not showing all the potential existent on it, but soon, when SATA-300.SATA-600..and so on. this technology will entirely demolish the PATA interface.
It's true that there are a few disk out there with native SATA support. It's also true that E-IDE ATAPI 7 (133MB/sec) it's still very competing and cheaper. I think E-IDE found its end because it uses the PCI bus to transfer and we all know that PCI maximum bandwidth is just 133mb/s, so...in summary there is a plenty future for this new tech.

If anyone wants to become an expert on this art, visit this link http://www.lostcircuits.com/advice/sata150/ ...Huge of Unknown information...very interesting