Sign in with
Sign up | Sign in
Your question

SATA 3.0 Useless?

Last response: in Storage
Share
July 2, 2009 12:38:40 PM

Hi,




So I just realized that my Seagate Barracuda 7200.10 had a jumper setting so that if I remove the jumper I can use it at SATA 3.0 standards . But I saw no difference in the benchmark from HDTune.


When I was using SATA 1.5 the read was 75 mb/s and after I changed to SATA 3.0 the read speed is the same 75 mb/s . And I neither felt or saw even the slightest difference and yes my MOBO supports SATA 3.0!

So, what really does this SATA 3.0 do over the SATA 1.5?

More about : sata useless

July 2, 2009 1:54:59 PM

It is SATA 3Gbit/s and 1.5Gbit/s so it will allow information to move between the hard drive and the motherboard quicker. But as im guessing the max read on your hard drive is 75mb/s then the speed it connects with will not matter as it is below the 1.5Gbit/s.
a b G Storage
July 2, 2009 11:41:55 PM

depends on what drive you hook up to it

Doesn't matter if your racing in the salt flats if your driving a yugo
Related resources
July 3, 2009 5:09:47 AM

Quote:
depends on what drive you hook up to it


I don't get what your saying. What do you mean by that? And what kind of hard drives will use 3gbit? Does it depend on the RPM of the spindle. And what's the use of my Seagate barracuda offering me a 3gbit jumper setting when it cannot even perform at that speed?
a b G Storage
July 3, 2009 5:41:24 AM

I meant that different drives read at different speeds. Traditional drives peak at about 100; ssds peak around 220.

Current top end ssd drives already exceed standard sata bandwidth.

The rpm of the spindle has more to do access times than throughput.

As far as the jumper it is most likely there for compatibility issues. Also in addition to the increased bandwidth I would assume there are more features in the sata standard.
July 3, 2009 6:04:01 AM

What it means is the speed to and from the CACHE on your drive. It IS indeed twice as fast, but then once the 8 0r 16 or 32 meg buffer is emptied (.1 seconds roughly for a 32 meg buffer, or .05 for a 16 meg buffer) the information can only come off the hardrive as fast as it can be read from the platters.

The buffer or CACHE can only write to the hard drive as fast as the mechanical portion can write.

Look at it like this, get in your car for a 2 mile trip, drive a quarter mile at 180 mph and after that the other 1.75 miles at 60 mph. guess what your overall time is almost the same as if you just drove 60 the whole time.
July 3, 2009 6:05:28 AM

klsdivan said:
It is SATA 3Gbit/s and 1.5Gbit/s so it will allow information to move between the hard drive and the motherboard quicker. But as im guessing the max read on your hard drive is 75mb/s then the speed it connects with will not matter as it is below the 1.5Gbit/s.


yes, it's the speed between the motherboard and the cache on the drive.
July 3, 2009 1:05:26 PM

It is interface speed. Not drive speed. It means the the 3.0 interface has twice the bandwidth of the 1.5 interface. Solid State drive, which are in fact hard drive with memory module instead of mechanic, platter and head, will saturate easily the 1.5 bandwidth, and will saturate the 3.0 bandwidth. But in no way standard mechanic drive will saturate the interface, even the 1.5.

The exception being readind from cache. Cache is memory on the drive, not limited in speed by the mechanical component. But cache is so small that is don't affect much the day to day operation.

What the newer sata interface brought is more feature, like hotswap, ncq,... But as ong as drive won't become really faster, the speed will not really be affected.

running on a highway don't make you running faster than on the sidewalk..
a b G Storage
July 3, 2009 6:11:40 PM

pat said:


What the newer sata interface brought is more feature, like hotswap, ncq,..


wasn't that sata 2
July 3, 2009 7:34:50 PM

sata 2 doesn't exist officially, the real name is sata 3.0 Gb/s. The sata3 (third generation) will be SATA6.0 Gb/s
a c 127 G Storage
July 4, 2009 1:24:09 PM

No power on the SATA connectors yet, a missed chance! Especially with SSDs requiring 0.01 - 0.5W power, this would have been a great opportunity IMO.
July 5, 2009 7:53:25 PM

^would have been nice
December 20, 2009 5:29:17 PM

AS for SATA III It is a good idea with the solid State Drives getting faster. Also one drive company and I do not know what one as I have been so dam busy
Stated that they were going to come out with a RAID Hard Drive. What is a RAID Hard Drive? Well on the larger drives such as a 1 TB or even a 500Gig Drive there are lots of platers. Each Plater has two sides and each plater has a read write and erase head that is not to much unlike a Tape drive head other than it is far more precision.
Now. Normaly a drive will have about 4 to 8 platers. With each one being 2 sided
then you have 8 to 16 surfaces to read and write to.
Also that means 8 to 16 heads to do the reading and writing. As of right now Drives only read and write from one head at a time. This is realy stupid but it is the way things are. Why? Well at first it would cost to add the controler chip
on the drive to make all of the heads read and write at the same time. The data flow on a RAID system splits the data up and writes it to several differant drives at a time drasticly increasing the speed. The same thing could be done inside a single drive. Have all of them read and write data that has been split up on the CPU of the Drive so for example (And this is a very simple example) if you wanted to write 0 1 2 3 4 5 6 7 8 9 A B C D E F (That is HEX for 0 to 16)
On a normal drive one head would be choseen to do the writing to just one side of just one plater using just one head. Now with a RAID hard drive that has 8 platers with two sides each and 16 heads then each head would get one of the 0-F charachters and all of them would write at the same time. So the speed increase would be almost 16 times faster for both reading and writing just as it is with a RAID controler and 16 Differant drives. Price has been the main issue but if these were mass produced the price would drop very fast.
With just one drive with this technology a very small CPU or in this case it is caled a controler chip would not have to be very powerfull. In fact the hardware raid cards use RISC chips mostly and they are only runing at 400 MHz. 3Ware and LSI both make very good HARDWARE based RAID cards. These cards are realy cool and I use them even if only using one drive as they do not use the CPU of the host computer to RUN the RAID. Regardless it would be cool if all Rotory RAM storage devices (Hard Drives) used this technology. If they did then we would need a SATA V or SATA VI or so.
Anyway also the cards I mentioned are SCSI to the computer and Sata to the drives. SATA and ATA Drives both use the main computers CPU to run most of the hard drive. SCSI does not do that. A SCSI Drive and SCSI controler have their own CPU or as it is called on a periferial a Controler chip.
Anyway THINK back to ATA 66 and then ATA 100 and ATA 130.
there was no differace in speed as all drives in those days were too dam slow to use all of the bandwidth of the ATA 66. So the 100 and the 130 is and was just marketing. Unless the drive is one fast mama it will not mater if it is SATA II or SATA II. If you want more speed always use a Hardware controler card for the Drves. I do that with everything I put in my computers. Sound Cards and even some network cards have their own CPU so they do not rob the host computer of CPU cycles that can be so valuable in the middle of a good shootem up game. Also not only the CPU cycles but the added interupts also slow a system down. So I use a Sound blaster XFI as it has its own CPU and I use a 3 Ware or LSI SCSI / SATA Controler as it has its own CPU and also 3 differant Network cards have their own conroler chip and offload the work from the main CPU. With these type of cards you can make a fast system realy fast and a slower system very fast. IT is also a good investment. I have had one of my 3ware cards for over 6 years and it is stil faster than 4 drives can handle in Stipe mode. And I will not go into the differant type or RAID versions there are as this post is already too dam long.
Even if you are only running One drive these cards are a big help.
Jim Davison
December 23, 2009 8:16:37 PM

My 7k160 reads contiguous data at 80MB/s so Ata/100 is useful.

I know of no single host, whether Raid or not, that would rely on the Cpu, except sometimes for Raid-5. But then, a few Xor on a Core2 or Core-i5 are negligible.

Look: 5 disks in Raid-5, each writing 140MB/s, need 420M XOR per second. Each SSE core at 3GHz achieves 16 XOR per tick, or 48G XOR per second. A software Raid-5 driver would take 10% of one single core.

Some Raid-5 controllers have a Cpu, all others have a specialized chip which isn't a Cpu. It does all the job and interrupts the Cpu only when the request is complete, which can be many sectors long - interruptiong the Cpu for each word was called Pio. This has nothing to do with carrying its own Cpu; any Dma, Mdma or Udma host works like this, in conjunction with the "bus master" capability if on Pci.

Hosts integrated in the chipset have all the advantage plus the one of not depending on one more bus, like Pci or Pci-E. They shall be preferred, unless their speed is Ata/33 of course.

Sata300 brings Ahci with Ncq and hot plug. Ncq makes a very noticeable difference. And my 7200.12 read contiguous data at 134MB/s, so Sata150 isn't enough. Much less so for my X25-E.

-----

Hard disk drive integrating Raid through independent arms: yes, but the right method is described there, because it slashes rotational latency.
http://saposjoint.net/Forum/viewtopic.php?f=66&t=1793
http://www.physforum.com/index.php?act=ST&f=5&t=23343&s...
December 23, 2009 8:22:15 PM

What you didn't understand with using all heads simultaneously is the precision issue with hard disk drives. When one head is well placed over its track, the other heads on the same arm aren't, and can't be, because of mechanical deformations.

The whole system works thanks to permanent arm position adjustment, and this adjustment can be done for one head at a time only. This is why two arms are necessary to read from two heads at a time.

Small adjustment as well: in 2009, individual platters have 500GB capacity.
a c 415 G Storage
December 23, 2009 11:19:51 PM

Pointer is right - you can't use the heads in parallel because there's no way to guarantee that when one head is correctly positioned over it's track that the other 7 (or 5 or however many) heads are exactly positioned over their respective tracks. Thermal expansion of the platters moves the tracks relative to the heads, and not all of the platters are heated equally. The drives work based on a constant feedback servo which keeps the head centered over the particular track that's being accessed.

I've worked with mainframe-class drives with large (15-inch) platters that had two independent access mechanisms to improve performance, and also with drives that had two heads per platter on a common access arm to reduce the amount of full-seek travel distance. But in no case did any of these drives ever activate more than one head per access mechanism.
December 30, 2009 7:09:29 AM

OK
So SCSI did not offload most of the work and the Controler Chip on the SCSI card (Real SCSI not Emulated by some driver emulater). Why did IDE cost a lot less than SCSI OR MFM etc..
Talk to ANY Hard Drive Manufacture and you will get the same answer.
IDE costs a lot Less becuse there is no Controler Chip and the Drivers for IDE as well as SATA use the Host CPU to do all of the work. Not to mention the buss is moved from the card to the Motherboard. This is a FACT it is not my or someone
just guessing. If you have two drives and move data from Drive A to Drive B
with SCSI the systems sends instructions to the controler card to handle this job and give progress updates. But the data from Drive A to B does not go accross the Motherboard regarless of what type of slot you are using. With IDE on the of the reason they cost less especialy when they first came on the Market was they used a controler card but the card did not have a controler chip (Its own CPU)
And the same with the Drive itself. So the Host computer and CPU have to do most of the work. IF you move information from Drive A to Drive B in the same Card using IDE then the data goes from Drive A to the Card and then to the Motherboard and through the slot and accorss the bus to the CPU and RAM just like most programs and then it has to go back across the SAME bus to the same
Card through the same Slot and then on to Drive B

Now there is always somone that wants to Argue. That is fine. I just do not want to be involed with a bunsh of idiots that come up with a bunch of BS.
I used to work at Seagate and I happen to know what drives do and how they work. MFM, RLL, SCSI and IDE and there are other formats that are like MFM and RLL that are not used anymore. For example ESDI.

Now as for the Heads that you are talking about having to be in line.
You talk as if there were Tracks drawn out on the platers at the factory.
Well you are only parly correct. However each head is as you say not perfectly eligned with each other. The platter is nothing more than media and there are no tracks until the drive is started and formated. So any head deviation from one head to the other is not an issue. The reason is they are all connected to the same swing arm. The Tracks are drawn when the drive is made and if one haed is a little bit in a differant place on the other side of the plater it will always be in a differant place on the other side of the plater and or on a differant plater. So unless on head were to come loose then all of then are on track if one is on track then all of them are even if the tracks are not exactly in line with the other tracks. They do not have to be. The Tracks are drawn or created after the drive is assembled and formated on a low levele format. (Not something you have to deal with on IDE.
So using each head as if it were a differant drive and adding the Controler chip to make the RAID Calculations can be done and in fact has been done. There just is not any hard drive company that does this as the cost is too high and the posibility of drive failure is higher if the data is across all of the heads at the same time. But it can and has been done.
So I am not going to post on this site anymore as I thought this was a site with some folks that know what they are doing. I thought we could have some stimulating conversation. What we get is persons that talk about things as if they are fact when it is just an idea.
I get so tired of that.
And yes you are correct. There are acctualy several Drives that were faster than ASA 66. But in general the point is the Drive is the bottleneck and the faster conection to the drive is useless for the most part.
this is in general. Even the drives that are faster cannot go faster than ATA 100
and no drive was ever faster than ATA 130.
Jim
December 30, 2009 1:12:26 PM

You believe what you want.
a c 415 G Storage
December 30, 2009 4:00:10 PM

jimmyjoejetter said:
The Tracks are drawn when the drive is made and if one haed is a little bit in a differant place on the other side of the plater it will always be in a differant place on the other side of the plater and or on a differant plater. So unless on head were to come loose then all of then are on track if one is on track then all of them are even if the tracks are not exactly in line with the other tracks.
Read my lips: differential thermal expansion.

The platters expand as they heat. The platters may not all heat at the same rate. As a result the tracks may not line up.

This is exactly why drives do thermal recalibration to optimize seek times.
a b G Storage
December 30, 2009 4:24:03 PM

jimmyjoejetter said:


Now as for the Heads that you are talking about having to be in line.
You talk as if there were Tracks drawn out on the platers at the factory.
Well you are only parly correct. However each head is as you say not perfectly eligned with each other. The platter is nothing more than media and there are no tracks until the drive is started and formated. So any head deviation from one head to the other is not an issue. The reason is they are all connected to the same swing arm. The Tracks are drawn when the drive is made and if one haed is a little bit in a differant place on the other side of the plater it will always be in a differant place on the other side of the plater and or on a differant plater. So unless on head were to come loose then all of then are on track if one is on track then all of them are even if the tracks are not exactly in line with the other tracks. They do not have to be. The Tracks are drawn or created after the drive is assembled and formated on a low levele format. (Not something you have to deal with on IDE.
So using each head as if it were a differant drive and adding the Controler chip to make the RAID Calculations can be done and in fact has been done. There just is not any hard drive company that does this as the cost is too high and the posibility of drive failure is higher if the data is across all of the heads at the same time. But it can and has been done.
So I am not going to post on this site anymore as I thought this was a site with some folks that know what they are doing. I thought we could have some stimulating conversation. What we get is persons that talk about things as if they are fact when it is just an idea.

You're acting as if none of the heads or disks ever move even a single track's distance in relation to each other.

They can. Thermal expansion, vibration, and a wide variety of factors ensure that you can only effectively use one head on each servo mechanism at a time.
March 10, 2012 10:41:28 PM

Pointertovoid said:
What you didn't understand with using all heads simultaneously is the precision issue with hard disk drives. When one head is well placed over its track, the other heads on the same arm aren't, and can't be, because of mechanical deformations.

The whole system works thanks to permanent arm position adjustment, and this adjustment can be done for one head at a time only. This is why two arms are necessary to read from two heads at a time.

Small adjustment as well: in 2009, individual platters have 500GB capacity.

===
I do understand what you are stating. The Drives would not only have to have a differant progrming type.
But Alining the heads even when there is only one ARM is not that hard to do. Of cource there may be defects. However a head is aligned when the factory does the low level format. From then on the Track is set. It does not move even with head expansion the whole asembley can compansate for this as the
Drive is built in almost all cases to have about the Same temp on all platers and heads.
Thus the only things that are an issue is the RAID programing would have to be differant. It would not be as fast as 16 Drives but it could be close. The Drive is programed to Save Data on the Drives differantly.
Never will one head write and the others not. The have to split the data up even and add a charachter in areas that will not revice data if there is only one word thus all heads do not have real data to write. Thus the heads that do not have anything to write and this will not be often will write a chearchter that stands for ( Ignore this Data ). So the other heads that need to write real data are all writing that data
on the same Cylander and sector. Else the whole thing would not work.
Look
It has already been done. IT works. Not sure why it has not been released but it has been done. I think it was Pinacle who did it. Not sure but I think they got in over their head (No Pun Ment) on the Glass Drive RAID. That was also very cool. The point is I am keeping in mind about the heads. The programing has to be differant and the Tracks are not an issue anymore. That was the case in the old MFM and Some of the older SCSI Drives that had " plators and up to 16 inches then there are problems with "Where is the TRACK from Temperature changes. Also one of the reasons a computer room needed to be one temp. On those old Systems the plators stayed outside of the case and also stayed at room temp and room temp did not vary too much.

Anyway I do not want to argue. I shared some data and someone wants to add something that is cool.
But to say I do not understand something that is childs play to me when you do not even know who you are talking to is inconsiderate at best. You could just ask. Hey. Would the Tracks and one Arm be a problem. I did not explain all of this as my post was already short story length.

Anyway you seem smart. Use that Smartness to THINK how it can be done. It is not that it can not be done. It can. But then think about the drive companys. They want to sell more drives. If they did the RAID drive and if the total storage was enouhg on one drive they would loose $. If they charge 12 to 16 times the price for the speed no one would buy it anyway. Thus it does not go to Market. A great deal of the best technology has never made it to market. Look at RAM Bus for example. Even now nothing can touch what it could do 12 years ago. In Computer years that is over 100 years ago.
I just made up the Term Computer years kind of like Dog years. they would be much shorter than yours and my year.
jimmyjoejetter
March 10, 2012 10:48:30 PM

sminlal said:
Pointer is right - you can't use the heads in parallel because there's no way to guarantee that when one head is correctly positioned over it's track that the other 7 (or 5 or however many) heads are exactly positioned over their respective tracks. Thermal expansion of the platters moves the tracks relative to the heads, and not all of the platters are heated equally. The drives work based on a constant feedback servo which keeps the head centered over the particular track that's being accessed.

I've worked with mainframe-class drives with large (15-inch) platters that had two independent access mechanisms to improve performance, and also with drives that had two heads per platter on a common access arm to reduce the amount of full-seek travel distance. But in no case did any of these drives ever activate more than one head per access mechanism.

-----
No
You are wrong and right. But the drive has to be a bit differant. And I have already stated that this has been done damit.
It just never made it to Market. You guys are trying to apply current Drive Technology to something that is differant. Do not do that to youself.
And if everyone thinks the way you are then we would still be riding horses rather than Flying etc..
Open your Mind. there is a way around all problems and hurdles.
Maybe we can think of them. Maybe we can not. But thee is a way around them and mostly we do in time solve them. This one is not hard to do.
But do not use old tech for holding back net tech. If you can not move the rock the build the road around it. But do not stop the road from being built.
And then someday technology will be there to move the rock in a cost efective and time efective manor and you would have your cake and eat it too. I alsways hated that term but it fits here.
Jimmyjoejetter
March 10, 2012 10:58:43 PM

Pointertovoid said:
You believe what you want.


Look
There is always some moron that will challange anything. Think or dont think is up to you.
It just gets old where there is always a NAY sayer. There are ways around these and as I stqated to the other fellow that wants to argue
rather than help with the technology then close your mind up and do not solve the problems. If I had put everything in the post it would have taken
several pages. Do you REaly think a hard drive Engineer that has made drives for years and yes has dealt with Heat and expansion and Controlers and
Software RAID as compaird to hardware raid (And I have a little problem with those two terms. I know what they indicate but they are not very discriptive unless you do not count the Cores of the CPU that you talk of as hardware). Anyway enough. The Company and I think it was Pinacle is not going to make the drive now. So you seem to want to argue. If that is the case then tell yourself you are the winner.
I did not know there were so many folks that have to argue about everything. Why not just ask how the PIO was handled?
Good god
Jimmyjoejetter
March 11, 2012 12:01:46 AM

sminlal said:
Read my lips: differential thermal expansion.

The platters expand as they heat. The platters may not all heat at the same rate. As a result the tracks may not line up.

This is exactly why drives do thermal recalibration to optimize seek times.

===
No kidding. You must be a genius.
Look. Instead of asking how it is done you just keep beating that same old drum. I have dealt with guys like you but I should never have made the post.
I thought Toms hardware readers were a little more advance in thought. What you say I have alredy addressed in Part. I would have been happy to explain it to you in detail how it can work. But instead of asking how would you get around this issue you tell me It can not be done. Well I sure as hell
am glad you are not in charge of desinging new technology. Your paterns of thought are too ridged yet limited. I am tired of listening to drival.

You know
In the old Analog Modem Days I was one of 7 companys that could program a DSP on a Modem. Almost all Modem factorys could not do this and had to get the DSP already programed from someone else.
When we were reaching 19,200 bps There were some idiots that came up with this very strange Math that they Said Proved that the Limit of Any Modem
was 28,800bps. So when we reached beyond that speed in the lab using both our in house test lines and real phone lines with Hardware all over the world to conect with after updating the firmware. I posted it on a website. It was not this one as Tom's hardware was not up yet. Or at least I was not aware of it if it was.
However all I got was these guys telling me that what we just said we did and I certainly know it worked as I had played a major role in programing it. But we would be told things like it can not be done becuase of the limited Bandwidth on a analog line. And then I would get one of the variations
of the BS Math Proff of the limitation. Thus my post by a few was discarded as BS. Well. We went ahead and programed the modem. I could go into detail as to how it worked but you are the expert. I only came up with the programing and made it work. Mr Beck as I always called him of Rockwell was also working on his version as was neil at Bell Labs ( Latter to become Lucent).
It is almost embarasing what I see here. But then again it is kind of good. Why. Well. AS I think of it and it has been awhile I made a great deal of $ off of those modem DSP Chips. I sold them to many modem Companys. You will have heard of most of them. You on the other hand will only think of WHY it can not be done. In stead of asking me how did we deal with that issue or thining about how to deal with it yourself.
This is a good thing for me and a few others. I ran into a person in Hollywood with the same type of thoughts. And I was hurt at first then angrey and now glad. Why. Because this Joker will loose a lot of buisness from me and the companys that will be using my technology.
Keep it up. Keep that Mind closed. close it up. Do not ask. Just tell how things are. You know just enough to think you know everything I guess. Not sure how it works for guys like you.
I just know one thing. I will not post on these boards again.
And one more thing. The Drive was made. I just did not make it to market. And not for the reasons you stated. Those linitations are clearly something
that had been thought of and solved. Hell. I can think of several ways to deal with them as well. And with one arm. And I bet you could. But for what ever reason you do not. Or can not.
Jimmyjoejetter


March 11, 2012 12:14:23 AM

cjl said:
You're acting as if none of the heads or disks ever move even a single track's distance in relation to each other.

They can. Thermal expansion, vibration, and a wide variety of factors ensure that you can only effectively use one head on each servo mechanism at a time.


Not acting. It and I are for real. And unfortunatly so are you. You are not stupid but you sure as hell put a lot of limits on your thouht paterns.
jimmyjoejetter
!