More dualcore goodness from Intel

[Scali]

Just noticed this: http://www.theinquirer.net/?article=23643
Sadly it's from the Inquirer, but the screenshot looks real enough anyway.

So it looks like the dualcores will be able to reach at least 3.73 GHz in the next generation... They can already do that now, so looks like 65 nm is going smoothly.
They also say that the Blackford chipset offers both CPUs on the board a dedicated 1066 MHz FSB, so that would make it far more competitive with the Opteron than the current Xeons.

 

[USMC2Hard4U]

Whats supprising is... the INQ has been right on the money the past few months.... so this could be good news.

 

[xonik]

So it looks like the dualcores will be able to reach at least 3.73 GHz in the next generation... They can already do that now, so looks like 65 nm is going smoothly.

I'm sorry if I missed the reference, but where did it say that the cores were built with the 65 nanometer process?

so that would make it far more competitive with the Opteron than the current Xeons.

Theoretically, but I'd like to wait for the benchmarks to see. The 1066 MHz system bus didn't do a whole lot of good for the latest NetBurst iteration, which leads me to believe that everything is not how it seems.

 

[Scali]

I'm sorry if I missed the reference, but where did it say that the cores were built with the 65 nanometer process?

The article says they are Dempsey cores, which are 65 nm.

Theoretically, but I'd like to wait for the benchmarks to see. The 1066 MHz system bus didn't do a whole lot of good for the latest NetBurst iteration, which leads me to believe that everything is not how it seems.

It's more about the fact that each CPU has a dedicated FSB than about the fact that the FSB is bumped up to 1066 MHz. Currently a two-CPU system runs with a single shared 800 MHz bus, so in the worst case each CPU only has 400 MHz FSB effectively.
So better too much FSB than too little... also note that these are dualcores, so the cores have to share the FSB, so 1066 MHz may have more of an effect here than on a single core.
In short, you can't compare this situation at all with the 1066 MHz FSB that was on the P4EE.

 

[xonik]

In short, you can't compare this situation at all with the 1066 MHz FSB that was on the P4EE.

Don't be so dramatic! My comment was merely to point to a loosely related situation and show that the real-life performance didn't measure up to the theoretical performance at all. Certainly this could be the case with the independent FSBs. Of course the new Blackford chipset will make the Xeon more competitive with the Opteron--the question is whether or not your claim of "far more competitive" will hold true. I have no wish to see the Xeon fail, mind you, but I have my doubts with regards to the NetBurst architecture as a whole, currently.

But we're all gentlemen here, I think that our theoretical discussion here is stimulating in its own right.

 

[markku]

well, the brief anandtech review with N0 step (2M l2 cache) xeons showed that they were
quite competitive with opterons already, winning 2 tests out of 3.
though amd is on their own league now with dc opterons.

was dempsey supposed to have some kind of arbitration bus (a la amd's srq) ?
if doesn't seem to be similar to next dualcore p4 (presler) as it has 2x1M cache,
vs preslers 2x2M (rumours again)..

and i don't see anywhere an indication that xeons will get 1066 fsb next year..

 

[Donnie27]

and i don't see anywhere an indication that xeons will get 1066 fsb next year..

The one at Tyan was running without an apparent hitch for two days, and had two Dempseys in it, dual core with 1MB cache per core. The chips were running at an indicated 3.73GHz on a 1066FSB. CPU-Z was not recognising it because it was coming up as a 90nm Nocona core which it is not, but that is no big shock.

Donnie27

 

[Scali]

the question is whether or not your claim of "far more competitive" will hold true.

I think you're being overly sceptic. The worst case will go from 400 MHz FSB to 1066 FSB per CPU. Since one of the weak points of the Xeon is that it gets bandwidth-starved, this will surely be a considerable improvement. It's no HyperTransport yet, but now the bandwidth will be roughly comparable between Opteron and Xeon.