Parallel vs Serial speed

[markw at mohawksoft.com (markw at mohawksoft.com)]

> markw at mohawksoft.com wrote:
>>> Date: Tue, 06 Feb 2007 21:46:22 -0500
>>> From: David Kramer <david at thekramers.net>
>>> markw at mohawksoft.com wrote:
>>>
>>>> In the above examples, we see that the drives perform similarly, but
>>>> that
>>>> the SATA interface gives an advantage in transfer speeds, but not the
>>>> 100mb/s vs 300mb/s advertised difference.
>>> Here's something I don't undertand.  My gut tells me that a parallel
>>> interface, which sends all the bits at once, should be several times
>>> faster than a serial interface, which must send one bit at a time.  So
>>> how is SATA faster than IDE?
>>
>> Well, obviously, parallel is "faster" than serial, with the caveat that
>> serial can be faster if it does something that parallel can't.
>>
>> Reciting from memory, I recall that SATA uses low voltage differential
>> to
>> communicate. This is MUCH faster and more immune to noise than the
>> simple
>> TTL logic level IDE communication.
>>
>> Theoretically, an 8, 16, or 32 bit cable would be faster, but also more
>> expensive to produce.
>
> There isn't much point to going to parallel bits for SATA (yet);
> 300MB/second is already faster than the drive mechanisms can go.

Yea, if you look at the data sheets, the 7200RPM drives today can get
about 750 mbits/sec to the platters MAX (That is, after the heads are in
place after a seek, and the platter has rotated to the point to where the
block is under the head.)

Well under the SATA or IDE spec. HOWEVER!! If you go back to the original
post, you see something interesting.  The drives I am testing are Western
Digital WD2500JS and WD2500JB, which are identical except for the
interface, and the SATA drives are more than 10% faster, and since the
same computer was used and exact same drive mechanics are involved, and
writing directly to the device, one has to assume that the different
transfer characteristics lead to that.

While there is no way we are getting the full bandwidth of these devices,
because the SATA drive should be about 54 seconds faster than the IDE
drive, (8G/300MB/s vs 8G/100MB/s) (all we really see is about 18 seconds),
its nothing to sneeze at.

>
> However, the parallel principle comes into play in PCI Express. PCIe x1
> is about the same speed as PCI; it's serial, but runs at a much higher
> bit rate than PCI's 33Mbps. PCIe x16, which is typically used for video
> cards, moves 16 bits in parallel, using 16 data lanes that are the same
> speed as PCIe x1; the result is that it's twice as fast as AGP 8X.
> (There is also an intermediate level, PCIe x4, that is used for some
> high-performance peripherals like disk controllers; it moves 4 bits at a
> time.)
>


-- 
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.