Intel may add its own memory controller

This would indeed be quite interesting news, but isn't the BTX standard designed in such a way that the distance between memory and CPU is large enough to require a separate memory controller (northbridge)?

I think it would be a great idea; I don't know, however, how much this implementation will indeed help an architecture like Netburst or even Dothan. I'd venture a guess that it will matter less on netburst than on A64; however, Dothan should benefit more from this than netburst cores...

The true benefit is kind of unpredictable at this point. What Intel<i> does</i> need, and there's no question about that, is a point-to-point system bus. Like PCIe is point-to-point. And like Hypertransport is point-to-point.

I think Intel would only catch up with AMD if they managed to introduce (still in 2005) a tweaked Dothan core (assuming it's tweakable for performance, eh), in dual and single core versions, with onboard memory controllers, point-to-point busses and fully functional, speedy 64-bit extensions. Even so, they'd still have to put out some extra goodies in order to really strike AMD's performance leadership; they're technically quite behind. A nice 65nm transition may help (if they don't make it prescott-like). Dual core, I suspect, won't save them, however.

wasn't there a project by intel regarding onboard memory controller a few years back? they experimented with Timinan? i forgot the code word.

------
Prescott 3.2E 1MB L2 HT
1GB PC 3200 Dual channel(PAT)
Asus P4P800 Bios 1016
PNY Geforce 6800 GT 256MB DDR3
60,823 Aquamarks

> but isn't the BTX standard designed in such a way that the
>distance between memory and CPU is large enough to require
>a separate memory controller (northbridge)?

Hmmm.. it makes it a bit harder perhaps, but definately not impossible. One simple but effective solution is to simply put the cpu where the northbridge is supposed to go, its still positioned in the "thermal module" and should allow enough space for heatsink/airflow. BTX isnt ideal for ODMC, but its not a complete showstopper either.

>I think Intel would only catch up with AMD if they managed
>to introduce (still in 2005) a tweaked Dothan core
>(assuming it's tweakable for performance, eh), in dual and
>single core versions, with onboard memory controllers,
>point-to-point busses and fully functional, speedy 64-bit
>extensions.

Is that all ? LOL.

> Even so, they'd still have to put out some extra goodies
>in order to really strike AMD's performance leadership;

Apparently, according to you its not  
I think that would be more than enough to keep up. Not that I expect it really, I don't think we will be seeing a hypertransport/ODMC mesh like AMD's that quick from intel. I also believe they may not go this route at all, and instead do something fancy with FB-Dimm. Integrating the memory controller onto Dothan AND adding a new bus seems far too much of a redesign to be implemented so quickly (not too mention EM64T and SSE3 as well). Especially when you believe (like I do) Dothan wasn't meant for this role until relatively recently. Its not like they canned Tejas and server variants 4 years ago.

Still, looking further ahead (2007) I'm very curious what intel will do... I wouldn't rule out P2P and/or ODMC for that timeframe at all. Could be a nice fit for the promised shared Xeon/Itanium platform perhaps.

= The views stated herein are my personal views, and not necessarily the views of my wife. =

You're right. I'm guessing Intel will not be able to release a truly innovative product even in the whole 2005, let alone 2004. I'd be surprised if they came up with something really interesting.

I'm thinking that Intel will only be on its two feet by 2006, or maybe, if they hasten up deployment of new technologies, by late 2005.