IDF 2008 Day 1: Intel Nehalem Working at 3.2GHz Pictured

[malware]

I promissed more details on Intel Nehalem yesterday, and now it is time to keep my word. During the first day of Spring IDF 2008, the guys over at HEXUS.net have pictured the first working sample of a quad-core Intel Nehalem processor operating at 3.2GHz (revision A1). The 1366-pin, 731M-transistor 45nm native quad-core model, utilizes 256KB of L2 cache for each core, as well as 8MB of L3 cache. The CPU also integrates triple-channel DDR3-1333MHz memory controller and SSE4 instructions. Like the new 533MHz Silverthorne-based Atom processors, Nehalem will also incorporate Simultaneous Multithreading (SMT) which is also known as Hyper-Threading (HT). Each physical core in a single Nehalem processor is paired up with its own virtual core. As a result, the quad-core processor will be detected to have eight cores (on the picture). Predictions say that this new architecture will offer around 30% better performance, on a clock-for-clock basis, when compared to Core 2, in a heavily-multithreaded environment - HPC and low-end servers, mainly. Current Intel roadmaps list the Nehalem launch date for Q4 2008, with a simulteanous rollout across servers and desktops.



View at TechPowerUp Main Site

 

[rampage]

Edit .. looks good

 

[Duxx]

/facepalm

I want.

 

[Solaris17]

that is like holyshit impressive

 

[Morgoth]

I Want It :d

 

[springs113]

I Want It :d

we all want, but that don't mean we can all have

 

[Morgoth]

well i\m gona buy it when it gets released

 

[farlex85]

Hell yea, its gonna be sweet. Probably expensive as hell though, especially when you factor in the cost of ddr3 (although hopefully that will be down by then). I'm still shooting for early 2010 most likely, unless......

 

[springs113]

Hell yea, its gonna be sweet. Probably expensive as hell though, especially when you factor in the cost of ddr3 (although hopefully that will be down by then). I'm still shooting for early 2010 most likely, unless......

not according to reports, d-ram prices are slated to go up...

 

[Morgoth]

that means ddr3 goes down i hope..

 

[TheGuruStud]

Wouldn't HT split the cache even more? Sounds dumb to me and is just another marketing ploy.

 

[acperience7]

I guess this is what Intel meant by 8 core processors?

Wouldn't HT split the cache even more? Sounds dumb to me and is just another marketing ploy.

I don't think it would since it just creates a virtual core.

 

[springs113]

This is the end for AMD. They will not be abble to keep up. Just getting an OK phenom took them for ever, getting new architectures out at the same rate as Intel is just not something I can see happenning. I will have no choice to change for team blue in my next upgrade.

actually amd is in the lead when it comes to this kind of setup...remember they did start it...now all it takes is perfection of this design...which i do believe they will do...and once they do that there is no stopping them...then a die shrink is eminent...

if you look at it the 9850 and the 9950 does give the q6600 problems and if you really think about this is its main competitor... not the 9300/9450 that some websites are comparing/contrasting them to...i for one think 65nm vs 65nm is a fair comparison...the only time there should be a comparison between 65nm vs 45nm is when we are comparin the same company to see how much better the old vs new tech is....
dont get me wrong 65nm amds vs 65nm is ok but its not an apples to apples contest.

 

[PVTCaboose1337]

This seems quite impressive, but that virtual core thing scares me... the P4 I had had that...

 

[Disparia]

This seems quite impressive, but that virtual core thing scares me... the P4 I had had that...

Scary?

Worked well in my experience.

 

[MiST91]

i need one of those, i wonder how hot it runs and if its any good at overclocking

that'll keep those pesky AMD fanboys quiet for a while (joking, btw)

 

[WarEagleAU]

Wow, nice. See, AMD does have good stuff in the works. Good enough for Intel to take and make their own (maybe tweak it a little). Hopefully they got the L3 right

 

[TheGuruStud]

Wow, nice. See, AMD does have good stuff in the works. Good enough for Intel to take and make their own (maybe tweak it a little). Hopefully they got the L3 right

No kidding. Whoever let that L3 be crippled should be fired. Not to mention that the mem controller doesn't run at core frequency haha. No wonder the phenom fails, I guess they didn't learn with the off die cache in the 90s that was god awful slow.

 

[magibeg]

I'm not sure why we went so off topic about the whole intel vs amd thing but that is one hell of a chip that intel appears to be cooking up. I figure if the chip is as good as its supposed to be intel will silently leak more benchmarks onto the internet just like they did with the core 2's.

 

[Thermopylae_480]

These chips are best utilized in server environments. Gamers aren't going to find much use for this yet. How many of you utilize four cores? This won't provide the same performance increase for gamers and the average user as the Core 2 did. I imagine any performance increase for the average user and gamer will be negligible and will be overshadowed by the cheaper quad cores and dual cores.

I feel people who rush out and buy this chip for games are going to be sorely disappointed.

This chip is designed for serious multitasking, the average users and gamers need quad cores with better individual core performance before they need 4 physical cores and 4 logical cores.

Stay on topic. This thread isn't for flame wars.

 

[cray86]

I enjoy video editing... so I'm a wee bit excited!

In Sony Vegas (my editor of choice) you can define maximum number of rendering threads... making any core / HT / whatever combination adaptable to the program (sense with Video editing, rendering is not real-time dependent like games)

 

[Thermopylae_480]

Sounds like a decent processor for you then.

 

[Wile E]

Man, this thing would be an encoding BEAST.

But the main question on my mind is, how does it OC?

 

[btarunr]

Wouldn't HT split the cache even more? Sounds dumb to me and is just another marketing ploy.

and how is that?

It's not about cache being split, just like in the Conroe where there's a common cache for two cores (think two threads basically) and each thread has its instructions and data tagged in the common cache. So, processes that are single threaded, will completely utilise the cache of a core (or if two cores share an L2 cache (as in Kentsfield/Yorkfield)).

IMO, HT is good way of making sure the core is fully utilised. And please don't bring the Presler-HT into this argument. Sure it was dual-core + HT but a lot of other factors went against it performing well.

 

[VroomBang]

30% better performance, on a clock-for-clock basis, when compared to Core 2, in a heavily-multithreaded environment -

Promising for those who multi-task or encode/decode.

If they start coding games for quads, this Nehalem is definitly something to look forward to, in my case, depending on the price/performance ratio compared to other cpu's of course.