IDE vs. SATA - Multitasking

[Trackr]

Hey, guys.

I'm building several computers and I happen to realize I still have two old 320GB IDE drives from 2007 on hand.

They both work perfectly fine with and without the IDE-to-SATA adapter.

But for some reason, I have this belief in the back of my head that IDE drives are slower than SATA drives, kind of like 5400RPM drives are slower than 7200RPM drives, and would thus bottleneck, say, a Q6600 in multitasking.

What do you guys think? Am I just wrong? Does the IDE-to-SATA adapter make a difference?

Thanks!

 

[mwroobel]

Hey, guys.

I'm building several computers and I happen to realize I still have two old 320GB IDE drives from 2007 on hand.

They both work perfectly fine with and without the IDE-to-SATA adapter.

But for some reason, I have this belief in the back of my head that IDE drives are slower than SATA drives, kind of like 5400RPM drives are slower than 7200RPM drives, and would thus bottleneck, say, a Q6600 in multitasking.

What do you guys think? Am I just wrong? Does the IDE-to-SATA adapter make a difference?

Thanks!

Well, if you live and die by benchmarks, sure the SATA will be faster. The ISE/SATA adapter will cost you a few points on the interface overhead. As to a bottleneck with a Q6600, it depends on what you are doing. For large, sequential transfers sure it can make a difference, but it depends on what you are comparing it to. Would I use it as my primary OS drive, no. Could you find use for it for an extra 320GB of space, sure.

 

[msitpro]

In my experience, SATA drives that properly support NCQ offer far superior (very noticeable) multitasking perf. than IDE drives.

Early SATA drives didn't support NCQ very well either, especially Western Digital. (probably due to their lack of experience in enterprise SCSI drives at the time which already used Command Queueing)

Obviously AHCI mode must also be enabled. IDE mode (as the IDE interface) doesn't support any command queueing.

 

[Trackr]

Well, if you live and die by benchmarks, sure the SATA will be faster. The ISE/SATA adapter will cost you a few points on the interface overhead. As to a bottleneck with a Q6600, it depends on what you are doing. For large, sequential transfers sure it can make a difference, but it depends on what you are comparing it to. Would I use it as my primary OS drive, no. Could you find use for it for an extra 320GB of space, sure.

So, basically exactly what I thought - it's not fast enough to be an OS drive.

But the question is - why? What precisely is slower on the IDE drives? I mean, it's the same drive, just a different interface.

In my experience, SATA drives that properly support NCQ offer far superior (very noticeable) multitasking perf. than IDE drives.

Early SATA drives didn't support NCQ very well either, especially Western Digital. (probably due to their lack of experience in enterprise SCSI drives at the time which already used Command Queueing)

Obviously AHCI mode must also be enabled. IDE mode (as the IDE interface) doesn't support any command queueing.

I don't know if my drives support NCQ.

Actually, I have my SATA drives running under IDE mode. ACHI mode caused issues with Windows starting up, if I remember correctly.

 

[Sebastianem]

ACHI mode must be enabled in BIOS prior to Windows installation. If not, when you change BIOS to ACHI after the fact, Windows will not start.

IDE hard drives (assuming ATA-7 and you utilize an 80 conductor cable rather than 40 conductor) will be limited to 133 MB/sec on the PCI bus and that bandwidth is shared amongst all the devices on the bus. Insofar as I know, IDE drives never came close to saturating the bus unless some sort of RAID solution was used.

 

[Trackr]

ACHI mode must be enabled in BIOS prior to Windows installation. If not, when you change BIOS to ACHI after the fact, Windows will not start.

IDE hard drives (assuming ATA-7 and you utilize an 80 conductor cable rather than 40 conductor) will be limited to 133 MB/sec on the PCI bus and that bandwidth is shared amongst all the devices on the bus. Insofar as I know, IDE drives never came close to saturating the bus unless some sort of RAID solution was used.

So then WHY are IDE drives slower than SATA drives?

 

[msitpro]

So then WHY are IDE drives slower than SATA drives?

Assuming platter amount, density, cache size, etc. are all equal the benefits of SATA for a hard drive are as follows:

1. If sequential transfer rate off the drive's platters exceed the 133MBps theoretical IDE limit then the advantage is obvious

2. SATA has Native Command Queueing, IDE does not

3. Burst speeds from cache can be processed much faster over SATA; as point 1.


Cant think of anything else....


Obviously newer drives come in SATA ONLY, so you cant compare as platter density is vastly different.

 

[Trackr]

Assuming platter amount, density, cache size, etc. are all equal the benefits of SATA for a hard drive are as follows:

1. If sequential transfer rate off the drive's platters exceed the 133MBps theoretical IDE limit then the advantage is obvious

2. SATA has Native Command Queueing, IDE does not

3. Burst speeds from cache can be processed much faster over SATA; as point 1.


Cant think of anything else....


Obviously newer drives come in SATA ONLY, so you cant compare as platter density is vastly different.

Well, according to that, multi-tasking should not be slower on IDE.

So, is it? And if so, why?

 

[dustNbone]

Read about NCQ and you may understand

http://en.wikipedia.org/wiki/Native_Command_Queueing

 

[Benny Blanco]

I think I remember when SATA drives first came out, seeing some benchmarks that showed no advantage to the SATA drives because at the time, drives (like someone else mentioned) were not even saturating the PCI bus limit of 133MB/s. On top of that, SATA 1.0 (150MB/s) was hardly an improvement over IDE (in throughput, that is).

Eventually SATA drives started pushing the limits of SATA 1.0 in burst speeds, but I think even today, platter drives still don't max out even the SATA 1.0 spec on sustained reads. High performance SSD's at this point would obviously be severely bottlenecked by SATA 1.0, but earlier SSD's and even some budget SSD's today don't break 150MB/s.

In fact, I also remember early SATA drives not even being true SATA drives. They were simply integrating IDE-SATA adapters onto the drive's mainboard!!

[stuff about IDE traffic being on the PCI bus was not correct, unless an add-on PCI card or additional IDE controller on the motherboard. Removed, thanks to correction by Aesma and Shikami!]

But IDE (Parallel ATA) and SATA (Serial ATA), if all else is equal such as # of platters, areal density, cache, and whatever else (msitpro already mentioned), then the advantages are what he mentioned, and then it really comes down to the particular drive itself and whether or not it can run with the big dogs

If you're not aware of the basic differences between PATA and SATA and why PATA needs 40/80 wires and SATA needs 4 (or something like that), google it (I'm tired of typing LOL)

Hopefully my memory isn't failing me here. If I'm incorrect on anything I've said, hopefully someone will correct me.

 

[thefreeaccount]

Back when disks came in both IDE and SATA flavors, there was usually a performance cost for enabling NCQ; performance did not "break even" until you hit huge (10+) queue depth. You won't see that kind of queue depth in normal desktop operation. Ex:
http://www.storagereview.com/ST3500641AS.sr

All else being equal, IDE = SATA no NCQ > SATA w/NCQ

 

[Trackr]

I think I remember when SATA drives first came out, seeing some benchmarks that showed no advantage to the SATA drives because at the time, drives (like someone else mentioned) were not even saturating the PCI bus limit of 133MB/s. On top of that, SATA 1.0 (150MB/s) was hardly an improvement over IDE (in throughput, that is).

Eventually SATA drives started pushing the limits of SATA 1.0 in burst speeds, but I think even today, platter drives still don't max out even the SATA 1.0 spec on sustained reads. High performance SSD's at this point would obviously be severely bottlenecked by SATA 1.0, but earlier SSD's and even some budget SSD's today don't break 150MB/s.

In fact, I also remember early SATA drives not even being true SATA drives. They were simply integrating IDE-SATA adapters onto the drive's mainboard!!

Fortunately, SATA's advantages are not limited to the drive alone. IDE drives used the PCI bus and SATA has it's own bus to the southbridge. We're essentially talking about more lanes of traffic which will allow more traffic and less bottlenecking. Check the second photo on this page on WikiPedia

In theory, using an IDE drive with a IDE-SATA adapter could offer some advantage if the SATA controller is on the PCI-express bus (this might not be the correct terminology). If the SATA controller is an add-on PCI card, or possibly even on the motherboard yet wired to run through the PCI bus, you would lose this advantage. And how much of an advantage is it, you ask?? I have no idea, LOL.

But IDE (Parallel ATA) and SATA (Serial ATA), if all else is equal such as # of platters, areal density, cache, and whatever else (msitpro already mentioned), then the advantages are what he mentioned, and then it really comes down to the particular drive itself and whether or not it can run with the big dogs

If you're not aware of the basic differences between PATA and SATA and why PATA needs 40/80 wires and SATA needs 4 (or something like that), google it (I'm tired of typing LOL)

Hopefully my memory isn't failing me here. If I'm incorrect on anything I've said, hopefully someone will correct me.

Ah, so you're saying that it's because SATA gets its own bandwidth while IDE has to tote itself on the back of the PCI bus..

I suppose that could account for my perception of why IDE drives are inferior.

Back when disks came in both IDE and SATA flavors, there was usually a performance cost for enabling NCQ; performance did not "break even" until you hit huge (10+) queue depth. You won't see that kind of queue depth in normal desktop operation. Ex:
http://www.storagereview.com/ST3500641AS.sr

All else being equal, IDE = SATA no NCQ > SATA w/NCQ

I have a new SATA HDD running at the moment with NCQ (I think - all I know is it's under AHCI), and I'm not really seeing a difference from a SATA drive under IDE and an IDE drive on a 40-pin cable..

 

[mwroobel]

If you're not aware of the basic differences between PATA and SATA and why PATA needs 40/80 wires and SATA needs 4 (or something like that), google it (I'm tired of typing LOL)

I actually have some milspec 40 pin/120 wire ATA cables that had 2 grounds for every pin laying around somewhere.

 

[ckryan]

In my experience, SATA drives that properly support NCQ offer far superior (very noticeable) multitasking perf. than IDE drives.

Early SATA drives didn't support NCQ very well either, especially Western Digital. (probably due to their lack of experience in enterprise SCSI drives at the time which already used Command Queueing)

Obviously AHCI mode must also be enabled. IDE mode (as the IDE interface) doesn't support any command queueing.

I had a WD Raptor 36GB in 2003 that didn't properly support NCQ.

All I can say is, with the price of HDDs going up and the price of SSDs going down, the best option is to get a SSD. Use the IDEs for storage.

But that costs money.

 

[Aesma]

In its hay day PATA/IDE was integrated into the southbridge just like SATA is now, so it was not using PCI bandwidth, unless it was on an additional controller.

Now we've come around and on modern motherboards still having IDE, it's on an additional chip connected by PCIe, the chipset doesn't support it natively anymore.

 

[Benny Blanco]

I actually have some milspec 40 pin/120 wire ATA cables that had 2 grounds for every pin laying around somewhere.

NICE. I never even heard of such a thing. Definitely [H]

In its hay day PATA/IDE was integrated into the southbridge just like SATA is now, so it was not using PCI bandwidth, unless it was on an additional controller.

Now we've come around and on modern motherboards still having IDE, it's on an additional chip connected by PCIe, the chipset doesn't support it natively anymore.

Wow, I had no idea PATA ever had it like that. Not bad there PATA! Sorry you lost your mojo when SATA came around.

I suppose that could account for my perception of why IDE drives are inferior.

I have a new SATA HDD running at the moment with NCQ (I think - all I know is it's under AHCI), and I'm not really seeing a difference from a SATA drive under IDE and an IDE drive on a 40-pin cable..

And there you have it. IDE drives can easily still hold their own It ultimately comes down to the question of what are you using it for. Basic PC stuff? PATA (/IDE) is just fine and dandy. If you think NCQ is some critical performance enhancing feature, think again. It certainly is for servers where data is accessed differently than desktop/workstation.

I'd definitely use an 80pin cable though. 40pin cables are so 1990's Seriously, look up IDE cables to see why 80pins are preferred. Some mobo's autodetect the lack of an 80pin cable and give a warning.

Back when disks came in both IDE and SATA flavors, there was usually a performance cost for enabling NCQ; performance did not "break even" until you hit huge (10+) queue depth. You won't see that kind of queue depth in normal desktop operation. Ex:
http://www.storagereview.com/ST3500641AS.sr

All else being equal, IDE = SATA no NCQ > SATA w/NCQ

The math equation is a little confusing but I think you're saying, performancewise,

IDE is equal to SATA (without NCQ)
IDE is superior to SATA (WITH NCQ)

I'm assuming you are talking about drives only around the PATA/SATA transition? (2003-ish?) Or even today?

 

[danswartz]

"IDE is superior to SATA (WITH NCQ)"

Did you mean inferior?

 

[Deleted member 214115]

But the question is - why? What precisely is slower on the IDE drives? I mean, it's the same drive, just a different interface.

There are a few differences with native and compatible mode controllers (http://msdn.microsoft.com/en-us/windows/hardware/gg463530). One of the major differences with SATA when comparing it to IDE is the very fact that SATA is first party DMA. Meaning that IDE still had DMA logics that were not capable of handling the transaction completely independently. The rest are listed in the link, and add other abilities that are supported with SATA such as NCQ for a better understanding of SATA's robustness.

IDE hard drives (assuming ATA-7 and you utilize an 80 conductor cable rather than 40 conductor) will be limited to 133 MB/sec on the PCI bus and that bandwidth is shared amongst all the devices on the bus. Insofar as I know, IDE drives never came close to saturating the bus unless some sort of RAID solution was used.

I think I remember when SATA drives first came out, seeing some benchmarks that showed no advantage to the SATA drives because at the time, drives (like someone else mentioned) were not even saturating the PCI bus limit of 133MB/s. On top of that, SATA 1.0 (150MB/s) was hardly an improvement over IDE (in throughput, that is).

What many do not know here-apparently one did-is that the hard drive controller has been separated from the PCI bus long ago. Actually, many of the I/O logics are segregated from the main host bus to prevent arbitration contention of a single bus. You can note in many of the chipset logic diagrams that the hard drive controller has a line that is independent of the bus that is used in the host system. If it was part of the bus, such as a controller that is integrated on the motherboard, you will see it attached as so.

Hard drives don't have the ability to saturate data transfers due to mechanical limitations, and most likely never will since they will be giving way to SSD's which can. However, the improvements do help with transactions that take place, but are in a sense minuscule and do not seem to be an improvement when measured via benchmarks. But are an improvement and it increases the robustness of the controller.

Now we've come around and on modern motherboards still having IDE, it's on an additional chip connected by PCIe, the chipset doesn't support it natively anymore.

I would not say support natively due to the disconcerting issue of a compatible mode and native mode controller. I would say the IDE logic is not integrated or is not supported anymore.

 

[Benny Blanco]

Thanks Shikami for that great explanation!

"IDE is superior to SATA (WITH NCQ)"

Did you mean inferior?

No, I meant superior, lol... Thefreeaccount is talking about implementations of NCQ/TCQ in SATA drives (early ones I believe, but waiting for his response). All us enthusiasts were hoping for increased performance from TCQ/NCQ, especially with the 10,000RPM raptor, but it ended up being no increase, and in fact a slight decrease. I don't recall when exactly it was resolved but for some time, people were disabling TCQ/NCQ. As Thefreeaccount said, it NCQ doesn't really start to shine until 10+ que depth, and that's where it starts to shine (in servers). I don't know that but I'm taking his word for it

 

[danswartz]

I must have misunderstood. I thought I had read that was back on older drives?

 

[Old Hippie]

All us enthusiasts were hoping for increased performance from TCQ/NCQ, especially with the 10,000RPM raptor, but it ended up being no increase, and in fact a slight decrease.

I was extremely disappointed when my Raptors couldn't take advantage of the AHCI features.

The AMD/Nvidia controller actually reduced performance but I don't recall any difference with the Intel controller/drivers.

 

[Benny Blanco]

I was extremely disappointed when my Raptors couldn't take advantage of the AHCI features.

The AMD/Nvidia controller actually reduced performance but I don't recall any difference with the Intel controller/drivers.

Sometimes being an early adopter sucks! LOL...

Regarding the Nvidia controller, the installation of the chipset drivers offerred you the option to install Nvidia's IDE driver, or keep the standard IDE controller drivers included in XP. The correct choice was to choose the latter, LOL...

Ahhh, nForce2.... the good olde days

 

[Old Hippie]

Ahhh, nForce2.... the good olde days

I had the series 4 and I don't remember anything good about that Nvidia chipset.

 

[Trackr]

So.. I guess in the end.. what you guys are saying is,

That IDE is equal to SATA.

 

[drescherjm]

That IDE is equal to SATA.

Maybe in your case.

Does the IDE-to-SATA adapter make a difference?

No. An adapter will not make your old drives faster or less of a bottleneck.

 

[darkeen]

Interface wise the big difference (as others have pointed out) is the command queuing. In essence it lowers the average latency of read/write operations under heavy random read/write usage. In your case your IDE drives will run as fast as an IDE drive can and with the command set an IDE drive was designed for. The adapter might introduce some latency, so if possible go native.

 

[thefreeaccount]

No, I meant superior, lol... Thefreeaccount is talking about implementations of NCQ/TCQ in SATA drives (early ones I believe, but waiting for his response).

Yes - sorry, that's exactly what I meant. In the review I linked (http://www.storagereview.com/ST3500641AS.sr), you can see that performance decreases ~3% across the board with NCQ enabled:

 

[Red Falcon]

HDDs from 2007? Neither would achieve the speeds that SATA-I or PATA-133/100 would allow, so it wouldn't really matter anyways.

Even if those exact same drives were on SATA-I, they would run the exact same as PATA.

 

[westrock2000]

The biggest thing I remember being a headache with IDE's was that only 1 device could talk at a time. Its not as big a deal now, but back then when all our games ran on CD-ROM's the harddrive would access, then the CD-ROM would access, then HDD, then the CD. You could just sit there and watch the lights all take turns. Add in another HDD, like in the case of a system HDD, a games HDD and an optical and it just got worse.

One of my first experiences with SCSI was a system with a OS HDD, a media HDD and a optical drive. I installed a game from CD to OS and formatted the secondary drive all at the same time and all of them had their lights going. It was marvelous.

That to me was the real reason to go with the expensive SCSI back then.

But as mentioned NCQ fixed this is modern drives and now no one in the enthusiast world gives a shit about SCSI.

Also, the real world issue with IDE is that its simply been dead for too long. I do not believe a 1 Terabyte drive was ever made in IDE. Thats when SATA started to come about. So all those drives, 320GB, 500GB, 720GB whatever had much lower density then a modern drive. Thus despite their size are just simply not going to be able to keep up with a modern drive. Even my "green" 1.5TB drive can sustain almost 100MB/s.

 

[hummilaygay]

SATA VS IDE Cables
The SATA cable or Serial ATA cables are only 7 pins wide compared to the IDE Ribbon Cable which is 40 pins. The SATA cables are also longer then the ribbon cables (some up to 1m) which make them more useful in large cases. The PATA or IDE cables are only 45cm long and can connect two hard drives to the motherboard while the SATA cable connects only one hard drive to the motherboard.
Speed
The next step in IDE vs SATA is to work out the data transfer speeds. IDE (PATA) transfers data from the speeds of 5MB/sec up to 133MB/Sec (ATA100/133). SATA however data transfer speeds range from 150MB/sec to SATA II which doubles it to 300MB/sec.
Jumpers
IDE vs SATA in the aspect of jumpers again goes the way of SATA. With the IDE cable you need to always mess around with the jumpers since there may be two hard drives or two CD drives or one of each on the one cable. With SATA there is only the one cable so there are no jumpers to mess around with. If you are new to computers then a SATA hard drive will be easier to install.
Summary
In summary of SATA vs IDE, the SATA cable is smaller, faster and therefore better. It can reach further then a IDE cable and will also reduce the obstruction to air cooling which IDE cables can contribute to. The cables also transfer data much faster (more crucial if you have two or more hard drives) and are easier to install since there are no jumpers. The only downfall of the SATA is that they easily fall out of their positions and that you can only have one drive per cable.

View more at: http://www.techyv.com/questions/which-best-harddisk-sata-or-ide

 

[drescherjm]

SATA however data transfer speeds range from 150MB/sec to SATA II which doubles it to 300MB/sec.

It took until 2011 for that to matter much. I mean in 2011 the first SATA drives (with 750 GB or larger platters) that could max out SATA1 speed were created.

The next step in IDE vs SATA is to work out the data transfer speeds. IDE (PATA) transfers data from the speeds of 5MB/sec up to 133MB/Sec (ATA100/133).

There are revisions for IDE that are faster than 133 MB/s (used for flash based devices not PCs).

One thing you have to remember with this discussion is that manufacturers have stopped producing ide hard drives except for maybe a few drive models from years ago. I know of no 2011 models with an ide interface so we can debate the advantages of sata versus ide but it in the long run its only talk of old hardware.

 

[Deleted member 214115]

The biggest thing I remember being a headache with IDE's was that only 1 device could talk at a time. Its not as big a deal now, but back then when all our games ran on CD-ROM's the harddrive would access, then the CD-ROM would access, then HDD, then the CD. You could just sit there and watch the lights all take turns. Add in another HDD, like in the case of a system HDD, a games HDD and an optical and it just got worse.

One of my first experiences with SCSI was a system with a OS HDD, a media HDD and a optical drive. I installed a game from CD to OS and formatted the secondary drive all at the same time and all of them had their lights going. It was marvelous.

If it was PIO then the device would not be released from the bus till its transaction was complete, and most CD-ROM drives were PIO not DMA back during the Triton chipset days. So the act of segregating the CD-ROM to one channel was common. However, unless it was NT you technically could not send a read and write to two separate IRQ's concurrently. Buffering prevents this from seemingly being worse than it can be. To note, this is why in the MS References Manuals it was common to place the pagefile or another pagefile on another channel, which is basically another IRQ to be read and written to in tandem with the system partition. When the bus mastering drivers were released and CD-ROM's were finally DMA it was still a common practice to have them on different channels, but was unnecessary due to DMA. There were even benchmarks that were done concerning this argument in-the-day and it proved that there was no loss as long as bus mastering drivers were used.