Serial ATA - Now!
- Thread starter Bozo Galora
- Start date
Originally posted by: ST4RCUTTER
Nice...now show me where I can buy an SATA drive?
Didnt you read? It comes with an adapter.
I read, and like everyone else here wondered "what's the point"
It's like buying a car that can do 1,000 MPH then only driving on 25 MPH roads.....
Sure in 10 years they may upgrade the roads, but by then I'll get another car..........
Adul
Elite Member
The PCI bus maxes out at 133mb/s, and since you rarely transfer a series of huge files, theres lots of room for intermediate and small file transfer speed increase with the controller chip and firmware bios on the add in card.
All drives are the same basics internally, including IDE, SCSI and Serial. Its the interface, bus, and protocol that counts.
Current IDE HDD's max out about 46MB/s at outer edge of disc in transfers, SCSI 60 - 70, and burst to 80-100.
I have seen Winbench scores on "real" serial ATA drives that had the chipset support on mobo that showed 102 MB/s TRANSFERS at start of disc (not burst) which is double current ATA.
My only question would be if one needs the chipset/mobo bios/on mobo connector to get these speeds. In other words, does on mobo serial ata still use the PCI bus or a dedicated one like AGP - if so we will soon need PCI-X.
zephyrprime
Diamond Member
- Feb 18, 2001
- 7,512
- 2
- 81
It's not just about performance. SATA also frees us from the master/slave thing. I've had so many friends that have had problems with that in the past.
And we'll need the increased bandwidth someday. Need 3gio first.
Actually the Western digital Jumbo Buffer hard drive hits 49.3 according to storagereview.com...
"I have seen Winbench scores on "real" serial ATA drives that had the chipset support on mobo that showed 102 MB/s TRANSFERS at start of disc (not burst) which is double current ATA."
Those results are either bogus, or of RAID arrays, not single drives. For those results to be possible using a 7200RPM drive, you would need platter density over 100GB/platter which isn't happening anytime soon.
Those results are either bogus, or of RAID arrays, not single drives. For those results to be possible using a 7200RPM drive, you would need platter density over 100GB/platter which isn't happening anytime soon.
Note the name of the picture, "3ware." It's a RAID setup. If you look at the picture before it, you'll see it's a 4 drive RAID array.
http://watch.impress.co.jp/pc/docs/2002/0514/3ware_06.jpg
http://watch.impress.co.jp/pc/docs/2002/0514/3ware_06.jpg
Serial ATA is all but completely meaningless right now. Why are some of you going "ga ga" over it?
The only thing that interests me besides the novelty factor is the slim, unobstructing cable. Bus bandwidth has always outpaced hard drive throughput, except in high end striped arrays, so the performance issue is nil nil nil.
We have rounded cables today so the slimness of ATA cabling isn't terribly compelling especially, if I understand the updated interace correctly, you'll need twice the number of cables.
Twice the cables.
Twice the controllers (now 1 controller per device, vs 2 with parallel. Compare to 63 on Firewire, etc, this is progress?)
Still got to route those power cables. For me they've always been at least as much of a bother as the ribbons themselves.
Gotta run adapters with current drives, and if you get a SATA drive to use in the future you need adapters to use it in a parallel ATA computer now. Plus the power converters since SATA requires new power connectors, so even after you get a SATA drive and SATA controller you'll need converters. Eventually power supplies will have these, but then will you have more or less flexibility with your power connectors than you have now.
Yeah, that all sounds like a LOT less hassle than setting master/slave jumpers. Pfft.
--Mc
Twice the controllers (now 1 controller per device, vs 2 with parallel. Compare to 63 on Firewire, etc, this is progress?)
Still got to route those power cables. For me they've always been at least as much of a bother as the ribbons themselves.
Gotta run adapters with current drives, and if you get a SATA drive to use in the future you need adapters to use it in a parallel ATA computer now. Plus the power converters since SATA requires new power connectors, so even after you get a SATA drive and SATA controller you'll need converters. Eventually power supplies will have these, but then will you have more or less flexibility with your power connectors than you have now.
Yeah, that all sounds like a LOT less hassle than setting master/slave jumpers. Pfft.
--Mc
"Twice the cables."
How is there twice the cables? One for data and one for power for each drive. Or are you simply meaning since each drive has a seperate data cable? If so that still isn't twice (on 2 drives that would be 4 cables vs today's 3).
"Twice the controllers (now 1 controller per device, vs 2 with parallel. Compare to 63 on Firewire, etc, this is progress?)"
Where did you get these numbers. I think you are mixing up your terminology. With current PATA you have 1 controller which implements 2 channels which can handle 2 drives. With SATA you have 1 controller which implements x channels for x drives (1 channel per drive .... it's point-to-point) depending on what the manufacturer chooses to implement.
You should all really read this paper:
IMPLEMENTING SERIAL ATA TECHNOLOGY
and check out the other information @ The SerialATA Working Group
Thorin
How is there twice the cables? One for data and one for power for each drive. Or are you simply meaning since each drive has a seperate data cable? If so that still isn't twice (on 2 drives that would be 4 cables vs today's 3).
"Twice the controllers (now 1 controller per device, vs 2 with parallel. Compare to 63 on Firewire, etc, this is progress?)"
Where did you get these numbers. I think you are mixing up your terminology. With current PATA you have 1 controller which implements 2 channels which can handle 2 drives. With SATA you have 1 controller which implements x channels for x drives (1 channel per drive .... it's point-to-point) depending on what the manufacturer chooses to implement.
You should all really read this paper:
IMPLEMENTING SERIAL ATA TECHNOLOGY
and check out the other information @ The SerialATA Working Group
Thorin
I will take point to point with the tiny cable any day compared with a daisy chained ribbon cable.
Now that SATA is starting the rollout... we need to beat the drum for SAS (Serial Attached SCSI) as it starts its evolution to bring Serial technology to SCSI class drives. I guess SATA II deserves some airtime too....
Now that SATA is starting the rollout... we need to beat the drum for SAS (Serial Attached SCSI) as it starts its evolution to bring Serial technology to SCSI class drives. I guess SATA II deserves some airtime too....
Sorry thorin, twice as many DATA cables. Yeah, they're skinny, but you need 1 per drive. Not a big deal I guess, just have to make sure to keep more supply in your workbox.
Yes, 1 controller per device. I phrased that extremely poorly. With parallel you get two devices per controller. With serial you get 1. Which means four physical controller chips on your motherboard to support as many ATA devices tomorrow as you can today. When it's all integrated into the southbridge then maybe we'll see some space savings on the motherboard. At the moment though it's not there. Are you confident the chipset makers are going to supply you with 4 S-ATA controllers in your southbridge in the future? I'm not. The board designs sampled so far with S-ATA have had just 2 controllers. Looks like some upcoming boards may have 1 or 2 S-ATA controllers along with the traditional P-ATA controllers still present and enabled.
My main problem with S-ATA is that as long as it's been yapped about this was the best that could be delivered? The speed "issue" is a nonissue with current drives and won't be one for quite some time, S-ATA doesn't need to be rushed out for it. Data security, well, unless you have an IBM Deathstar the improvements in that area seem nice, but not critical. It's point to point, yes, but like said here "Serial ATA is a point-to-point process. This is going even further away from SCSI than where ATA is now. Instead of having sixteen, or seven, or two devices (master and slave) per port, Serial ATA can only have one device per port. A controller card would have between two and six ports with a four pin connection for each. Hubs are not allowed." What's so great about point to point? Now if it was peer to peer like scsi I might give some more credit, but well, it's not. Data cable lengths of a meter, but you still have to chase the device around with an additional power connector, how 1980s is that? Yeah, I recognize USB and Firewire are primarily for external use where power cables are even more of a bother, but to me it makes so much more sense to incorporate both power and data onto the same cable that S-ATA looks like the already five year old kludge that it is.
I'd much rather see internal 1394 (Firewire) developed than S-ATA. SCSI's never been price competitive with ATA, 1394 can be. S-ATA simply carries on the tradition of P-ATA and USB, while saving pennies instead of dollars. blah. How do you get excited about that?
--Mc
Yes, 1 controller per device. I phrased that extremely poorly. With parallel you get two devices per controller. With serial you get 1. Which means four physical controller chips on your motherboard to support as many ATA devices tomorrow as you can today. When it's all integrated into the southbridge then maybe we'll see some space savings on the motherboard. At the moment though it's not there. Are you confident the chipset makers are going to supply you with 4 S-ATA controllers in your southbridge in the future? I'm not. The board designs sampled so far with S-ATA have had just 2 controllers. Looks like some upcoming boards may have 1 or 2 S-ATA controllers along with the traditional P-ATA controllers still present and enabled.
My main problem with S-ATA is that as long as it's been yapped about this was the best that could be delivered? The speed "issue" is a nonissue with current drives and won't be one for quite some time, S-ATA doesn't need to be rushed out for it. Data security, well, unless you have an IBM Deathstar the improvements in that area seem nice, but not critical. It's point to point, yes, but like said here "Serial ATA is a point-to-point process. This is going even further away from SCSI than where ATA is now. Instead of having sixteen, or seven, or two devices (master and slave) per port, Serial ATA can only have one device per port. A controller card would have between two and six ports with a four pin connection for each. Hubs are not allowed." What's so great about point to point? Now if it was peer to peer like scsi I might give some more credit, but well, it's not. Data cable lengths of a meter, but you still have to chase the device around with an additional power connector, how 1980s is that? Yeah, I recognize USB and Firewire are primarily for external use where power cables are even more of a bother, but to me it makes so much more sense to incorporate both power and data onto the same cable that S-ATA looks like the already five year old kludge that it is.
I'd much rather see internal 1394 (Firewire) developed than S-ATA. SCSI's never been price competitive with ATA, 1394 can be. S-ATA simply carries on the tradition of P-ATA and USB, while saving pennies instead of dollars. blah. How do you get excited about that?
--Mc
You don't or shouldn't. But that won't matter. You'll see a ton of people here proclaiming it's the second coming simply because they're conditioned into believing if it's new, it's great.
Serial ATA is just the evolution of the interface. It will allow EIDE to scale. That's all.
McCarthy,
Rest assured that you will see 4 ports per southbridge (and in some cases more). The initial ICH-5 only has 2 ports of SATA but also has standard PATA. Your picture of one chip per port is not likely. Stand alone SATA controllers will likely have between 6 and 8 ports per chip. A single port SATA part would be horrible economically since small silicon geometries would certainly make it a pad limited design. The sweet spot seems to be the 6-8 range. Less is pad limited and more drives to flip chip or thermal issues.
Rest assured that you will see 4 ports per southbridge (and in some cases more). The initial ICH-5 only has 2 ports of SATA but also has standard PATA. Your picture of one chip per port is not likely. Stand alone SATA controllers will likely have between 6 and 8 ports per chip. A single port SATA part would be horrible economically since small silicon geometries would certainly make it a pad limited design. The sweet spot seems to be the 6-8 range. Less is pad limited and more drives to flip chip or thermal issues.
McCarthy "Yes, 1 controller per device. I phrased that extremely poorly. With parallel you get two devices per controller. With serial you get 1. Which means four physical controller chips on your motherboard to support as many ATA devices tomorrow as you can today. "
?No ... I'll quote driectly from the White paper.
"The Serial ATA interface replaces today?s 80 pin ribbon cable with a 4 conductor cable. Rather than sending out data in parallel, the data from the controller is serialized, and sent out as a differential signal pair to the target disk device. The disk also sends data on a differential pair back to the host controller. Simultaneous transmission occurs on both channels, from host to disk and disk to host. Because of this, Serial ATA is a point-to-point link, and only supports a single device per controller interface, in contrast to the primary and secondary support of two drives from a single parallel ATA port. Controllers can address multiple devices, but each device requires a separate, dedicated port."
? As I said one controller (chip) multiple devices, it's up to the manufacturers to decide which particular controller (chip) they implement (and how many ports it can handle).
McCarthy "Are you confident the chipset makers are going to supply you with 4 S-ATA controllers in your southbridge in the future?"
? Absolutely.
McCarthy "The board designs sampled so far with S-ATA have had just 2 controllers. Looks like some upcoming boards may have 1 or 2 S-ATA controllers along with the traditional P-ATA controllers still present and enabled."
?Again there is a very good reason for this ... quoting the white paper again since they said it so perfectly.
"The board designs sampled so far with S-ATA have had just 2 controllers. Looks like some upcoming boards may have 1 or 2 S-ATA controllers along with the traditional P-ATA controllers still present and enabled."
McCarthy "but you still have to chase the device around with an additional power connector, how 1980s is that? "
?I'm sorry but I really don't see your point here. Please show me a USB/Firewire/SCSI HD (internal or external) that doesn't have a seperate power connector? Further to not seeing your point I think you're completely misinformed and I quote the announcement from Molex on their SATA connector: Molex Launches Connector Line for New Serial ATA 1.0 Architecture
"Molex?s Serial ATA connector line features differential pair signaling technology. It supports 1.8?, 2.5? and 3.5? form- factors with both cabled and direct docking systems. The combined signal and power interface, with built- in hot-plug capability, is ideal for blind- mate applications, including notebook media bays and RAID storage units."
?Perhaps your confusion revolves around the fact that you haven't actually seen a SerialATA drive yet, only PATA drives on a SATA controller card...and yes obviously PATA drives still require seperate power connections as they always have.
McCarthy "I'd much rather see internal 1394 (Firewire) developed than S-ATA. SCSI's never been price competitive with ATA, 1394 can be. S-ATA simply carries on the tradition of P-ATA and USB, while saving pennies instead of dollars. blah. How do you get excited about that?"
Well I'm sorry to see that you've already decided that there are all these problems with SATA even though you haven't seen demonstrable performance or taken into account any of the other benefits like ultra low electrical requirements (250mV signaling), hot swapping, etc etc... But when it's been here for a while I'm sure that you will see the benefits. Some reading on the technology and associated announcements may help you get excited about it... But maybe not (to each their own).
Thorin
?No ... I'll quote driectly from the White paper.
"The Serial ATA interface replaces today?s 80 pin ribbon cable with a 4 conductor cable. Rather than sending out data in parallel, the data from the controller is serialized, and sent out as a differential signal pair to the target disk device. The disk also sends data on a differential pair back to the host controller. Simultaneous transmission occurs on both channels, from host to disk and disk to host. Because of this, Serial ATA is a point-to-point link, and only supports a single device per controller interface, in contrast to the primary and secondary support of two drives from a single parallel ATA port. Controllers can address multiple devices, but each device requires a separate, dedicated port."
? As I said one controller (chip) multiple devices, it's up to the manufacturers to decide which particular controller (chip) they implement (and how many ports it can handle).
McCarthy "Are you confident the chipset makers are going to supply you with 4 S-ATA controllers in your southbridge in the future?"
? Absolutely.
McCarthy "The board designs sampled so far with S-ATA have had just 2 controllers. Looks like some upcoming boards may have 1 or 2 S-ATA controllers along with the traditional P-ATA controllers still present and enabled."
?Again there is a very good reason for this ... quoting the white paper again since they said it so perfectly.
"The board designs sampled so far with S-ATA have had just 2 controllers. Looks like some upcoming boards may have 1 or 2 S-ATA controllers along with the traditional P-ATA controllers still present and enabled."
McCarthy "but you still have to chase the device around with an additional power connector, how 1980s is that? "
?I'm sorry but I really don't see your point here. Please show me a USB/Firewire/SCSI HD (internal or external) that doesn't have a seperate power connector? Further to not seeing your point I think you're completely misinformed and I quote the announcement from Molex on their SATA connector: Molex Launches Connector Line for New Serial ATA 1.0 Architecture
"Molex?s Serial ATA connector line features differential pair signaling technology. It supports 1.8?, 2.5? and 3.5? form- factors with both cabled and direct docking systems. The combined signal and power interface, with built- in hot-plug capability, is ideal for blind- mate applications, including notebook media bays and RAID storage units."
?Perhaps your confusion revolves around the fact that you haven't actually seen a SerialATA drive yet, only PATA drives on a SATA controller card...and yes obviously PATA drives still require seperate power connections as they always have.
McCarthy "I'd much rather see internal 1394 (Firewire) developed than S-ATA. SCSI's never been price competitive with ATA, 1394 can be. S-ATA simply carries on the tradition of P-ATA and USB, while saving pennies instead of dollars. blah. How do you get excited about that?"
Well I'm sorry to see that you've already decided that there are all these problems with SATA even though you haven't seen demonstrable performance or taken into account any of the other benefits like ultra low electrical requirements (250mV signaling), hot swapping, etc etc... But when it's been here for a while I'm sure that you will see the benefits. Some reading on the technology and associated announcements may help you get excited about it...
But maybe not (to each their own).Thorin
That's a nice molex connector. But then so is this one. Now which is it, combined or separate data/power cabling?
From this article.
--Mc
From this article.
--Mc
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads
- Started by Tigerick
- Replies: 23K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
-
Facebook
Twitter