New Nehalem socket is... large.

[Borisz]

Originally Posted by cottonvibes

i hope the "integrated memory controller" is more beneficial than harmful.
since AMD chips already have one, i theorize thats 1 reason they can't OC as good.
but i have no evidence to support this.
just that the more the CPU has to do, the less you can overclock it because it would be more prone to errors.

The only reason AMD can keep up is because they have IMCs. Once the Nehalem hits, they'll be raped, unless they have an epic ace up their sleeve.

[skid]

Originally Posted by Borisz

The only reason AMD can keep up is because they have IMCs. Once the Nehalem hits, they'll be raped, unless they have an epic ace up their sleeve.

hmm.. i dunno. you might be right... but i thought amd were ahead for a long time because they had better internals (microcode etc), not only because of the imc. we'll see when nehalem is released.

[John Aiton]

Originally Posted by cottonvibes

R.A.P. totally copied you john :O

totally

actually, now I think back it was a signature.
Probably the jc pics from a nehalem 8 core + 8 virtual cores

hyper transpo....sorry, hyper threading of course

[Borisz]

Originally Posted by skid

hmm.. i dunno. you might be right... but i thought amd were ahead for a long time because they had better internals (microcode etc), not only because of the imc. we'll see when nehalem is released.

They were ahead because Intel was making **** chips (P4) and took too long to wise up.

[cottonvibes]

Originally Posted by Borisz

The only reason AMD can keep up is because they have IMCs. Once the Nehalem hits, they'll be raped, unless they have an epic ace up their sleeve.

what does having IMCs have to do with AMD keeping up with intel?
the IMC dosn't matter if the chips perform worse.

[Linktothepast]

Nehalem's biggest benefit would be multi- threading applications. Estimates put that to 2X the performance of a similarly clocked penryn, basically due to the included hyperthreading per core . That means that a Dual core Nehalem will have about the performance of a Quad core Penryn of the same clock speed with multithreading applications . That is quite a feat, if things go as Intel claim. But we should probably hold a small basket as always and see what happens .

[T D]

Hyper threading? That's partly what made the P4s run as hot as hell with only a small performance increase. HT isn't any kind of replacement for 2 more cores.

My hopes aren't high for nehalem's HT. But the P4s with HT ran hot, and a 45nm won't run as hot as them so HT will be better than the P4s but not enough to match a quad core penryn. I think. All speculation

[John Aiton]

Originally Posted by T D

Hyper threading? That's partly what made the P4s run as hot as hell with only a small performance increase. HT isn't any kind of replacement for 2 more cores.

My hopes aren't high for nehalem's HT. But the P4s with HT ran hot, and a 45nm won't run as hot as them so HT will be better than the P4s but not enough to match a quad core penryn. I think. All speculation

actually the HT is way better implimented this time and the thermal ratings are around the same as a go stepping q6600

all speculation?
why?

do you know about the leaked benchmarks and samples?
Plus intel just demo'd a 3.2Ghz Nehalem system

[cottonvibes]

theres leaked benchmarks out?

[T D]

I meant what I said was speculation.

Still think that HT on DC Nehalem won't be as good as quad penryn. It will be better than P4s HT implementation, that's for sure.

[John Aiton]

Originally Posted by T D

I meant what I said was speculation.

Still think that HT on DC Nehalem won't be as good as quad penryn. It will be better than P4s HT implementation, that's for sure.

emmm, there's no nehalem dual core

they come in 4 core and 8 core models and with HT
so really 8 or 16 virtual cores

[T D]

Ahh.. I'm really not being clear today. I mean havendale. Still sort of counts as nehalem.
I guess Linktothepast made the same mistake I did.

[John Aiton]

it's still to early to tell things for sure, but I guess nehalem is not much faster clock for clock than penryn

Based on a leaked superpi test, NOT the stupid fake one that's everywhere where a nehalem scores sub 8 seconds (sooo fake)

but with twice as many cores

[cottonvibes]

i don't think you can trust any leaked benchmarks.
its hard enough trusting legit benchmarks.

[Fadingz]

I can 'leak out' anything I want, and, believe me, someone will trust me over the net and spread the words for me

[John Aiton]

Originally Posted by cottonvibes

i don't think you can trust any leaked benchmarks.
its hard enough trusting legit benchmarks.

I'm not talking about your usual leaked benchmarks,
practically confirmed by a guy who goes by the handle jc



he's proved he has nehalem chips and the superpi times were not too favourable

[cottonvibes]

i always wonder how the hell people get ahold of these things.

but if you think about it,
not only would he have to sneak off with the CPU.
but he'd also need a mobo that supports it.

[runawayprisoner]

Originally Posted by cottonvibes

i always wonder how the hell people get ahold of these things.

Hint: be a seller.

but if you think about it,
not only would he have to sneak off with the CPU.
but he'd also need a mobo that supports it.

And Intel would gladly supply one... to "certain people."
Sadly, I'm not one of those... but I hope I can be one next year.

As for the Nehalem pic, no comment. (I didn't think anyone would be brave enough to show them)

[skoreanime]

One of my friends was mailed an engineering sample Pentium 4 chip once (once they were starting to roll out) . Stuff like that can happen

[Borisz]

Originally Posted by cottonvibes

what does having IMCs have to do with AMD keeping up with intel?
the IMC dosn't matter if the chips perform worse.

IMC is the one thing AMD has over Intel right now. Intel is still pounding them in the back, even though the core 2 is bottlenecked by the old FSB standard. Remove that bottleneck and gangraping will commence. AMD will be the butch side, unless they can make the Phenom suck less phenomenally.