Thanks for playing, PCIe

*celebrates*


now that you mention it though, pci-e 1x and 4x devices are exceedingly rare. :(
 
Not too surprised- how many of us knew that PCIe was a thrown-together standard with lots of obvious limitations?

...the most obvious one being:
"Let's make a high-speed serial interconnect by running a whole bunch of smaller serial lanes in parallel."

Edit:
...but then I've been complaining about PCIe since March...
 
i've never really cared for pci-e myself, but i didn't know of anything much better. HTX seems like it has the possibility to be better in every way, and reduce board and chipset complexity at the same time. definitly a win in my book
 
My only concern is, how? Will it require a dual-HT capable chip, like today's Opteron 200-series?
 
We'll see how it turns out in reality.

Might end up only surfacing on high end dual and quad CPU stuff, sort of like PCI-X.
 
The AGP crowd is just gunna jump all over this, you can practically hear the "I told you so's"...man, that sounds weird coming from someone who still has an AGP card :p

Anyway, its a little early to be calling this a PCIe killer.
 
Unknown-One said:
The AGP crowd is just gunna jump all over this, you can practically hear the "I told you so's"...man, that sounds weird coming from someone who still has an AGP card :p

Anyway, its a little early to be calling this a PCIe killer.


Well, it's superior to PCIe, that's for sure. And it's a technology thats been around for much longer than PCIe. I have faith in it.
 
Great, another standards war :mad: Like the "AVP" tagline, "whoever wins, we lose." I've got nothing against the HT or PCI-SIG corsortiums, but this isn't helping the proliferation of next-generation peripherals beyond PCI.
 
osalcido said:
asrock will put a pci-e & htx mobo out :D

And ECS will make a PCIe -> HTX adaptor add in card that also expands to provide ISA and AGP slots.
Asian Dub Foundation said:
imo PCIE is jsut a different name for APG16x

It's a little more than that. PCIe is cool in that it's fairly versatile and modular, plus cheap to impliment. It's just not as cheap as HTX and it's significantly slower.
 
robberbaron said:
It's a little more than that. PCIe is cool in that it's fairly versatile and modular, plus cheap to impliment. It's just not as cheap as HTX and it's significantly slower.
IMO, the biggest difference between AGP and PCIe is that where AGP can only write to the card, PCIe actually has the bandwidth to listen for feedback. That opened the door for physics processors and versatile GPUs.
 
mavalpha said:
IMO, the biggest difference between AGP and PCIe is that where AGP can only write to the card, PCIe actually has the bandwidth to listen for feedback. That opened the door for physics processors and versatile GPUs.

just to clarify, isnt the upcoming physics processor card a PCI component?
 
Asian Dub Foundation said:
just to clarify, isnt the upcoming physics processor card a PCI component?

The first one will be, I believe, for compatibility.
 
Yes, but I was referring to the fact that Ageia was preceded by the discovery that graphics cards/chips could be programmed for other things. First MIT developed its own language for communications with the GPU, then a small third-party developer turned the FX59x0 into a sound-processing station.
 
mavalpha said:
...the most obvious one being:
"Let's make a high-speed serial interconnect by running a whole bunch of smaller serial lanes in parallel."
Pretty clear you've never spent time trying to get a 64 bit 133 MHz DDR PCI-X bus to run properly. High speed serial is the way of the future....count on it. Seems they also claim that not using clock and data recovery is an advantage; the slot has a differential clock signal as one of the slot signal groups. Back to matching all the trace lengths to a gnat's ass hair, length wise....and you'd better hope the board fab guys keep the impedances in spec....:rolleyes: Having fought those battles, it sounds like a non-trivial step backwards.

PCI-e is a point to point, differential signalled high speed serial bus....as it appears this one is (I'll read the tech specs at work over coffee today.....:D ) They've done away with the root complex between the end points, but gone back to distributed clocks which has its own rather serious disadvantages.

All in all....it's interesting, but not a fundamental change like the move from large parallel single ended signalling (PCI, PCI-X) to multi-lane differential signalled serial buses (PCI-e).

As they say, we shall see......B.B.S.
 
Back
Top