Do you have a question? Post it now! No Registration Necessary. Now with pictures!
- Posted on
- New computer question
- Gary Walker
May 7, 2014, 4:42 am
rate this thread
I'm searching for a new desktop computer. And, I
have the following question.
If USB 3.0 is downward compatible, do manufac-
turers insist on installing USB 2.0 interfaces
1. They want to confuse their customer.
2. They do not trust the compatibility.
3. To unload their obsolete USB 2.0 inventories.
My favorite is #3, but I need understanding.
Re: New computer question
||| | |
SATA USB3 USB2
The current designs, only provide just enough bandwidth
for the stuff running off the Southbridge. If you beef
up everything, you might have to make the DMI fatter.
As it is, the bandwidth down there is over-subscribed.
If you were careful to craft a good test case, you
could probably have one of those bottom interfaces,
upset the performance of its neighbor.
Not all applications need USB3 speed. My USB keyboard
doesn't need it. My USB mouse doesn't need it. My USB
Bluetooth dongle doesn't use it. It's possible to
still get a benefit from the USB2 ports. They're
cheap, and they don't use a lot of power.
So let's try the experiment. I've just put 12 interfaces
running at 5Gbit/sec on the bottom chip. How big does the
DMI have to be ? What impact does that have on
my SATA drives (if any) ?
If I need to run extra DMI lines, maybe I have to wait
for the next CPU socket change, to add them.
Yet another aspect, is power. The USB ports of today
are powered by +5VSB. The power supply only has around
3 amps on that, to power things (it's really not enough).
Now, imagine I had a chipset with twelve USB3 ports,
at 900mA each. That's close to 11 amps from a 3 amp supply rail.
That's if I do a worst case scenario, and connect up a bunch
of power sucking stuff. Sure it can be fixed, but such a
supply doesn't exist today. In fact, there is plenty of
room to be powering ports from two different rails, with
a different colored connector for the rails suited to
charging mobile devices.
USB2 USB2 USB3 USB3 USB3
+5VSB +5VSB +5VSB +5V ... +5V
| | | | |
Keyboard Mouse iPad (Nine power hungry
(Wake Up (Wake Up charger peripherals with no
Capable) Capable) 900mA wakeup capability, no
charging while asleep)
In fact, the scheme as it exists today, is far from
perfect. I'm surprised there aren't more people dropping
by to complain about some of the edge cases of usage
of what is there now (it would be rather easy to
overload the +5VSB on the supply, as it currently
Think of this as lazy or uninspired evolution.
Yes, eventually USB2 will disappear. But will anyone
pay for it ? If my desktop is $50 more expensive as
a result, what do you think would happen ?
I could do this today for you. Take an uninspired
chipset, add three quad port USB3 chips. Using the
auxiliary power connector on each USB3 card, I can
power the ports with +5V (no overload).
CPU ------------ PCI Express ---- 3 quad USB3 chips
| DMI Switch ||||||||||||
| (use a video Massive USB3 ports
Southbridge slot for its
||| | | bandwidth)
SATA USB3 USB2
(An absurdly deluxe version. You'd like this one :-)
There is no sharing between ports on one of these, which
is why it is $100. Is uses a x4 connector, and a x4 switch
to distribute bandwidth. The switch chip is under the heatsink.
And it supports UASP.)
You've already got some spare PCI Express slots.
And why you'd want one of those cards, to for the
stuff in this article. First report I've seen of
decent USB3 bandwidth (secret sauce is a USB3 UASP
hard drive dock). Kudos to the reviewer for setting this up!
Most reviews in the past, weighed in at ~200MB/sec ballpark.
With that concept, you can build "soft-RAID" in a box,
without paying RAID card prices.
This would be an example of an uninspired card, but it's
only $18. Some bandwidth sharing goes on inside one of these.
This is how they do some of the ones on motherboards we've
got now. Now my "soft-RAID" idea doesn't work quite as well.
Re: New computer question
Thank Paul. You make some very good points.
Sadly, I'm only understanding of just a few
of them. But that's my fault. Perhaps the
USB2 "pushing" is not as big of a conspiracy
as I originally suggested.