Intel Tigerton - first look at current and upcoming enterprise server propositions

[HEXUS]

Core Microarchitecture has been around over a year but it's only now that Intel is introducing a four-socket solution. In our first look at current and upcoming server propositions for the enterprise market, we check out the fastest Tigerton Xeon CPU, the X7350, and the one with the lowest power draw, the X7345, to try to see if these strong CPU cores can overcome the disadvantage of an ageing system-interconnect design

Read more.

[DR]

... now that is a beast of a Server


HEXUS has recently changed a LOT of it's old infrastructure to Core based systems - it destroyed the old Opteron/Xeon/P4 stuff we were using...

I'm sure our server admins would die for a box like this....

... still will be good to see it against Barcelona...

[Funkstar]

But the real question is: how well does it Fold?

[DR]

I wish I could answer that...

[Duke3d87]

I heard that XeonMP would get the Penryn core eventually.

And where did they get the info about a Nehalem based XeonMP next year?

Intel said that sometime next year, the XeonMP gets the Penryn core and probably a year after Peryn for XeonMP, we'd get Nehalem for XeonMP.

[DR]

... we have some MP details coming - we will be updatin very soon

[James Thorburn]

Duke3d87, yes it appears that is correct, it does appear that next year the 45nm Penryn core will arrive in the Dunnington processor.

Unfortunately this pushes out CSI equipped parts into 2009, so while there is going to be some upgrade potential for Caneland systems the positive of this is off-set by the fact Intel aren't going to be moving their multi-processor systems - where they need it most - to a point-to-point interconnect for even longer than we had hoped.

[Duke3d87]

Originally Posted by James Thorburn

Duke3d87, yes it appears that is correct, it does appear that next year the 45nm Penryn core will arrive in the Dunnington processor.

Unfortunately this pushes out CSI equipped parts into 2009, so while there is going to be some upgrade potential for Caneland systems the positive of this is off-set by the fact Intel aren't going to be moving their multi-processor systems - where they need it most - to a point-to-point interconnect for even longer than we had hoped.

That's the way that it's been for a while now. Intel claims that the need for the additional year is because of validation. The other reason is probably to hack their chipset for more busses like this years launch. However, being that Nehalem wouldn't or shouldn't require any fancy chipsets as it has an IMC and CSI the validation theory seems more correct. That's what an Intel employee said. They've had tigerton for a very long time.

As for upgrades, there are company that are pushing out low voltage FB-DIMMs an whatnot. Also, the XeonMP line is a pretty small sector of the market anyway. Bit-tech said that Intel might have delayed X38 for a 45nm shrink. And while that's some extreme speculation, maybe Intel would do that for the newest chipset too?

[James Thorburn]

Originally Posted by Duke3d87

Bit-tech said that Intel might have delayed X38 for a 45nm shrink. And while that's some extreme speculation, maybe Intel would do that for the newest chipset too?

A 45nm chipset? Current Bearlake chipsets are 90nm, shrinking two nodes is far from trivial and why would you use 45nm capacity on a chipset?

[Duke3d87]

Originally Posted by James Thorburn

A 45nm chipset? Current Bearlake chipsets are 90nm, shrinking two nodes is far from trivial and why would you use 45nm capacity on a chipset?

I know it sound strange but i'm just posting one of bit-tech's theories. I can't post the link cause i think i only have 3 posts.

[DR]

... X38 has not been delayed and is still going to be launched I'm going to do some digging - I'd be amazed to see a 45nm chipset...

[Duke3d87]

me too considering most of the time, CPUs are given the latest process.