Do you have a question? Post it now! No Registration Necessary. Now with pictures!
- Posted on
May 31, 2010, 11:29 pm
rate this thread
Re: SATA6 card that works with Gigabyte's EP5-45-UD3p?
The problem with both SATA III and USB3 chips, is the choice of
interfaces they decided on. (And no, PCI would be a particularly
bad choice for this, as desktop PCI is 133MB/sec max :-) )
The PCI Express they're using, is capable of more than that,
but there are still issues they haven't addressed very well.
This is a SATA III card, carelessly plugged into a PCI Express x1
revision 1 (250MB/sec lanes) motherboard. The bottleneck is
the bus, and the SATA ports won't work any better than the
SATA II ports on the Southbridge. This is a waste of $35 for
the controller card, if you do it this way.
----------- Drive #1
PCI Express x1 Rev.1 slot --------------- SATA III
----------- Drive #2
If the motherboard has a Revision 2 slot, the picture looks like this.
A Rev.2 slot is backward compatible with Rev.1, so if a Rev.1 type
card is plugged in, the speed automatically drops to 250MB/sec.
The SATA III card would be Rev.2, so it should be trying to run
at the 500MB/sec lane rate. I don't have a copy of the PCI Express
spec, but one of the issues with Rev.2, is the PCI Express clock
must be extremely low jitter for Rev.2. And I think that issue is what
makes it a problem to make Rev.2 more prevalent in motherboard
designs (for *all* the slots, not just video slots). Many motherboards
now, support a mix of slot revisions, with most of the x1 slots being
Revision 1. So you can't find that many actual x1 Rev.2 slots.
There is still a potential bottleneck here, but is isn't as bad.
This is still worth doing.
----------- Drive #1
PCI Express x1 Rev.2 slot --------------- SATA III
----------- Drive #2
On a P45 based motherboard, the Northbridge has a PCI Express x16 interface
which can be split into two pieces. You can connect that interface to
two video card slots, each running at x8. But the important aspect
of this, is the interface supports Revision 2, and the two halves
of the interface can run at different speeds. (One slot can run a
Rev.1 card, while the second video card slot runs a Rev.2 card.
Intel made them effectively independent of one another.)
What the Intel datasheet doesn't make clear, is whether any external
bifurcation chips, makes a difference to the operating mode. Bifurcation
is in the data path, but since I don't know what chip numbers are used,
I don't know if the data is reclocked or not. (In this picture, the
four tiny chips to the left of the blue video card slot, are the
bifurcation chips used for lane switching between slots.)
The GA-EP45-UD3P motherboard comes in three different revisions.
I grabbed one motherboard manual, but they don't go into that level
of detail, to verify in fact, that the x8/x8 mode supports
Rev.2. The Intel P45 datasheet looks supportive of running
Rev.2, so you could do this with your two video slots.
This is how I'd do it, if buying a SATA III controller card
for that motherboard.
Video_Slot_#1 ---- PCI Express Video Card (running x8 Rev.2, 4GB/sec)
Video_Slot_#2 ---- SATA III Controller Card (running x1 Rev.2,
The PCI Express x1 slots on the GA-EP45-UD3P, come from
an ICH10R. And the ICH10R supports Revision 1 interfaces.
So you wouldn't particularly want to use the x1 slots on
the motherboard, for a SATA III controller. The 250MB/sec limit
of the ICH10R provided lanes, would make the SATA III card
pointless. You'd be just as well off in that case, using
the existing SATA II ports on the Southbridge.
So the video card slot is the only reasonable alternative
slot, to be plugging in the SATA III card.
If the SATA III chip had PCI Express x4 as an interface,
the chip would see a reduced potential sales market.
But at least early adopters would get to see every ounce
of transfer performance from at least one SATA connector
on the card. If they'd used an x4 interface, the minimum
bandwidth would have been 1GB/sec, and 2GB/sec with
Rev.2 lanes. Which would have given users a better
taste of performance.
The highest transfer rate to date on a SATA III interface,
is 355MB/sec, but I suspect when Southbridges come out
with native SATA III on them, we'll see a slight improvement
on that. The current SATA III card is still fast, but some
day, it might be eclipsed (slightly) by other forms of
SATA IIi solution.
HighPoint Rocket 620 $35
Picture of card - PCI Express Revision 2 x1 slot, two SATA III ports.
Using one of those cards for regular (rotating) hard drives,
is a waste of time and money. Only the burst transfer
speed would change, which is of no consequence. Users
want blazing fast sustained (large file) transfer rates,
and hard drives are still severely limited by head to
media rate (about 125MB/sec).
Currently, the only storage device worthy of one of those
cards, is this. 355MB/sec read, 215MB/sec write, $679.99.
This is a good test for the $35 controller card.
Other manufacturers will come out with SATA III SSDs, and
there is a good chance the write rate will be higher on the
next ones that come out.
One question, would be to what extent the 355MB/sec read rate
of the Crucial C300, is affected by the 500MB/sec Rev.2
limitation of those (early) controller cards. It would be
interesting, some day, to see if that drive does any better,
when a real SATA III Southbridge comes out.
It is just as easy, to plug a couple regular SATA II SSD
drives into the Southbridge and do RAID 0. Two of these
in RAID 0, would beat the performance level of a single
Crucial C300. RAID 0 is a less reliable storage mode,
in that either drive failing, leads to array failure.
And with SSDs anyway, you'd want to have backups to
fall back on, in case of trouble. Like a firmware issue...
(Fun with SSDs...)
- » Newest MemTest86, MemTest86+ incompatible with Sandy, Ivy chipsets?
- — Next thread in » Computer Hardware