Serial ATA Beginning to Appear

[Technoid]

No but the limit on ata100 is roughly 80mb/s…. (that slipped from my mind in may last “transmission” )
And considering that, it’s not strange why we are moving from ATA100 to SATA!

[Technoid]

Originally posted by Wombat
What's wrong with ATA133? The address extension was necessary.

It did prolong ATA!

They should have gotten SATA out erlier...

[Marshmallowman]

gurm , get 2 drives on the same channel, run two benchmarks at the same time one for each drive. You will see the perfomance drop.

35MB/s may the max throughput, but two drives fighting over the same bus(plus other overhead) you will not get maxium perfomance from the drives..

whats a faster setup in a 2 drive RAID (AT100)

2 drives on the same channel
or 2 drives on separate channels?

[CHHAS]

Of course performance will drop while accessing two drives on the same IDE chain, only one of them can be active at the time.
Separate your drives if you need to access them hard simultaniously.

[Marshmallowman]

Yes that is what I am saying, two drives well push the current ATA spec.

Thas probably also why they are going to go with a single channel per drive on serial ATA...

at 150MBps it will quite a few years before that gets saturated..and they already have 600MBps specc'ed out for the future.

[Gurm]

Marshmallowman,

You're confusing maximum throughput with the nature of IDE.

Only one of the drives can, technically, access at a time. No matter how much bandwidth is on the bus, you're NEVER going to get (Max*2) out of the drives... because you can't have concurrent access.

'Course, you can't have concurrent access in SATA either, as far as I know. So ... that argument is moot.

- Gurm

[Marshmallowman]

well yes, the "nature of IDE" IS limiting the usable bandwidth which is what I was saying.
We are playing around with semantics here, ATA100/133 spec support 2 drives, in this configuration you cannot get maxium performance out of the drives, therfore it is being limited by the interface....in real terms

Anyways, serial ATA is cool.

[The PIT]

Seagate have announced a new barracudda drive (The link I had no longer works I'm afraid) that supports serail ata. They claim programs will load 44% faster. That last bit I find doubtful as every new thing is XXX% faster.

[Gurm]

Umm... I'm not playing semantics.

If you think that serial ATA (it's called SERIAL for a reason) will allow concurrent accesses and therefore speed up current-gen drives, you have another think coming.

- Gurm

[The PIT]

The new drive is a serial ata drive with 8mb cache. I made no claims for faster speed I just said that I don't believe it. Thanks for correcting my spelling mistake by the way.

[Nuno]

Actually there´s no point on ATA133 or even ATA100. It´s only marketing.

Besides no drive can sustain that amount of data transfer (as Gurm pointed) what´s the point on having a 133 Mb/sec IDE bus if the data will have to go through the PCI bus that was a theoretical max bandwidth of... 133 Mb/sec? Add it some overhead, add the bandwidth required by your PCI devices (NIC, modem, soundcard and so on), add the handicap of running a VIA chipset (that can´t sustain PCI transfers higher than 70-80 Mb/sec), and you see what you have left for your IDE transfers.

That reminds me of the good old discussions about people getting very annoyed because not getting AGP4x while using SDRAM PC100...

What the PC arquitecture needs soon enough is a new standard for the PCI bus, 133 Mb/sec were really overkill on 1993, but now it´s getting a bit short.

I don´t really know much about serial ATA, but if it´s connected to the good old PCI bus, what´s the point?

[Technoid]

PCI has'nt been used as "mainboard freeway" for atleast 2-3 chipset generations....

There is already several new "PCI" buss standards

[Gurm]

PIT:

I wasn't actually picking on your spelling, but pointing out that SATA is SERIAL, not PARALLEL, and therefore no faster with current-gen drives.

Technoid:

You can claim that all you want, but let's be honest - PCI throughput is limited to well under 100MB/sec. on virtually all chipsets. Less on VIA.

- Gurm

[Wombat]

Over 100MB/sec on Intel (and especially SiS) boards. Doc posted up the numbers a while back.

[The PIT]

The newer via chipsets have seperated the ide channel from the pci bus (According to George Beese whos helped out a lot via users) . However there is a bug on the chipset that keeps interupting the controller and therefore limiting it's transfer. Even so you only going to get average transfer rates of between 20 -35mb's and burst of 70 - 90 depending if you patch or not.
Other controllers that do go through the pci bus need a patch which basically resets the latancy timer. The controller then can stay on the bus longer before being kicked off. For some reason my promise has now got set back to orginal settings but I haven't noticed any differance.
George B has posted a lot on this over on Via forums about 6 months ago.
Interstingly if have a drive udma5/6 on a pci controller and one on the newer via ide controllers the transfer across the bus is more like the older udma33.

Gurm I never said it was I was just quoting part of Seagate blurb I saw posted. I don't believe will be quicker. One thing it will do is make wiring a lot easier to tidy up.