Intel's 6-core Xeon and Nehalem CPU info leaked

Intel's had its new processor plans slipped out to the public thanks to Sun, according to DailyTech. Details on the 6-core (!) Xeon Dunnington, as well as the kinda-sorta hush-hush Nehalem were apparently leaked out onto Sun's public web server over the weekend, including plans for the new Xeons to overtake the company's Tigerton CPU line. The Dunnington processors will have a 16MB L3 cache shared by all six cores, and will be pin-compatible with the Tigertons, thus making integration with your Clarksboro chipset slightly less painful... by being possible. The Nehalem also got the spy treatment, with news that it will not only replace the Penryn line in Q4 '08, but will also be the first time in 18 years that Intel includes on-die memory controllers. If this sort of thing is important to you (and we think it may be) hit the read link and get all the juicy details.

[imacmatt09 @ Feb 25th 2008 12:06PM]

Where the hell do they come up with these weird ass code names?

[AutoTom @ Feb 25th 2008 12:30PM]

crystal meth, its got this nation by the balls man

[Mr.Tech @ Feb 25th 2008 1:07PM]

I thought they were going to release an 8 Core processor. I guessed wrong huh? Would been nice though if they did create an 8 Core processor.

[w00fy @ Feb 25th 2008 1:30PM]

Intel mostly names their projects in the region where the design center is located. Mobile/Core/Core2 were developed in Israel so you have Middle Eastern names such as Banias, Dothan, Nehelam. Chipsets are usually in California or Pacific Northwest. Most of the mobile platforms are named for regions in Northern California such as Napa, Sonoma, and Santa Rosa. Special projects usually comes out Oregon so you have names like Yamhill (64bit extensions). Itanium/IA64 shares technology from what DEC in Massachusetts so you have projects such as Tanglewood.

I don't believe this method is written in stone but the codenames do give a guide on what may be coming down the pipeline.

[SimonRichards @ Feb 25th 2008 12:07PM]

These will go very nicely with my skulltrail mobo :D

(unless I have missed some glaringly obvious compatibilty issue)

[kal326 @ Feb 25th 2008 1:04PM]

@SimonRichards
Uh, yeah you did miss one little detail. These chips are socket 604 not socket 771 that the skulltrail uses.

[SimonRichards @ Feb 25th 2008 1:34PM]

Ah, damn. Well never mind. Thx though, I didn't realise.

[Mr. Gremlin @ Feb 25th 2008 12:10PM]

Yeah... last time I checked, we are still waiting on the latest penryns for the desktop, so I wouldn't hold your breath on this one...

[NONE @ Feb 25th 2008 12:13PM]

I'm replying to this with an E8400 machine.

[Mr. Gremlin @ Feb 25th 2008 12:26PM]

Yeah, I know they released a bunch of mid-low end dual-core chips (under $250) and one uber high-end (over $1000) quad-core chip, but I was referring to everything in between (like the Q6600 replacement) ...

[hp540 @ Feb 25th 2008 1:07PM]

you might be posting this on an E8400 machine...but for every happy E8400 user, there are many more who can't get their grubby hands on one.

[Abuzar @ Feb 25th 2008 2:51PM]

Ahh, I COULD get an E8400 for 270+, but I can get a Q6600 for 200 and OC it to 3.8-4 Ghz!

I believe the time for Quad cores has come!

[Crayola @ Feb 25th 2008 12:13PM]

When will Intel ever run out of its product names, they should put a little marketing zest into those dowdy Jewish old testament misogynistic names, try Pocahontas or something...

[Mike @ Feb 25th 2008 12:15PM]

I don't know about anyone else, but I don't want a chip called "Pocahontas" under the hood.

[SimonRichards @ Feb 25th 2008 12:25PM]

Id prefer to see:

Powered by Jesus

on my PC

[kal326 @ Feb 25th 2008 1:07PM]

I'm pretty sure Eatshitanddieamd would be a nice code name, a bit long though.

[Yubastard @ Feb 25th 2008 1:19PM]

"Powered by Jesus"!?!? that must be the funniest thing I've heard in all week.

[Tits @ Feb 25th 2008 6:29PM]

I prefer pocahontas... gayer than a gay chimpanzee!!

[Liam Billington @ Feb 25th 2008 12:30PM]

Beat that AMD with your Triathlon processors!

[HaloZero00 @ Feb 25th 2008 12:33PM]

Its kind of crazy how much Intel is pushing more and more cores onto a chip.

Eventually all programs will just have to design their code based on the criteria of a finite unknown number of cores.

[Andir3.0 @ Feb 25th 2008 2:23PM]

That will be the day...

Poor Amiga/Be. You would have been spectacular.

[Ruben @ Feb 25th 2008 12:33PM]

There was also a story on ZDnet about this. They estimated the performance against AMD's next coming CPU (also estimated).

Poor AMD if this ends up being true.


http://blogs.zdnet.com/Ou/?p=1025

[Matt @ Feb 25th 2008 12:33PM]

Doesn't Liv Tyler say 'nehalem' while casting that weird elf river spell in the Fellowship of the Ring?

[GameboyRMH @ Feb 27th 2008 9:51AM]

Whoa, there. A certain amount of geekiness is cool, but there's a freakin' limit. This isn't Slashdot :P

[J @ Feb 26th 2008 12:57PM]

Isn't that also an old everclear song about what I assume is a town in CA?

[AutoTom @ Feb 25th 2008 12:35PM]

wow, talk about dumb.. the only native quad core cpu out there is by AMD and thats an abortion.

step back intel, we dont care how many cores there are just yet.
give us 3.5ghz the ability to address more than 3.5gb of ram in 32bit windows ( yes, i know im blaming the wrong people... i just dont know who to blame for that.. is it the kernel or the architecture? )

oh and Christ, does it take a rocket-scientist?
remember Pentium 4 HT ( hyper threading )?

cant you just un-hyper thread so the 2/4/6 cores appear as one and have the cpu delegate tasks?

[Ruben @ Feb 25th 2008 12:47PM]

First of all, ive never seen anyone so angry about technical advancement. Native quad core is good, but if you can get excellent performance on two duals or tris mated together, what the hell does it matter to you if they are physically or "logically" quads?

Second, 32bit can only reference that amount of memory. Its a limit at the size of address that can be stored in 32 bits. An OS maker can fake access, but why bother. Just switch to 64 as it has become alot more mainstream that it was when XP64 was released.

Third, the OS is now being built with multi cores in mind, so theres little need for Intel to take care of controlling access to 4 cores. As soon as you scale that up, then you start to run into bottlenecks. As soon as Graphics cores are introduced and a core carries graphics as well as integer/fp units within it, then embedded access and delegation controllers will be part of the chip, which is what the OS will communicate with. But we really dont need that now. There are not enough cores, nor are any programs built natively to work with a single core that delegates tasks.

Remember, the programs run on the Cell have to be written with that in mind.

[Oinquer @ Feb 25th 2008 1:13PM]

win XP pro, Vista 32Bits... can handle 4 Gigs of ram...activate the PAE (Physical address extension)...look in microsoft website its well explained there...
If youre linux dude just switch to 64 bits...
if you're mac? i really don't know and don't care, but there should be many people that will help you in here.

just google before you start sending useless posts about 32 bits architecture...it can only address 3.2gb ram cause of physical or mathematical facts period.

[stupid math @ Feb 25th 2008 6:18PM]

Blame math. 2^32 = 4GB.

[TX30000 @ Feb 25th 2008 6:32PM]

More cores are not always advantageous though. I attended a seminar offered by one of the engineers who worked in the Niagara processor for Sun last fall. She explained that anywhere about 8 to 12 cores the performance would actually drop off with current hardware. The reason is memory synchronization issue, with many different cores having the possibility of trying to access the same piece of data, you get contention between cores. Currently the way to solve it is by putting locks, similar to thread synchronization but for CPU this time. However, as one core is processing with certain piece of data, the piece of data is locked and other cores will have to wait to in order to execute. This leads to a performance drop off.

Where as in a small number of cores, different cores could be completing different tasks (or different threads) at the same time; in the case with a high number of cores, they are almost guaranteed to be at some point trying to work with the same piece of data at the same time. Some of the novel solution is to have things like transactional memory.

So that is pretty much my "simplified but flawed" explanation as to why more cores may not always be better. In conclusion, they need changes to the hardware platform in regards to memory access in order to push for more cores on a die, which In my very wild guess is what Intel is trying to do when they try to implement an on-die memory controller.

[wrabbit @ Feb 25th 2008 12:41PM]

Do we really need two more cores?! We still can't take full advantage of four - just seems like a waste of resources. They should be concentrating on making their current CPUs use even less power, generate even less heat, and be very very cheap. In most cases for games the bottleneck is the graphics card anyway, and most other software can't take full advantage of today's CPUs.

[Ruben @ Feb 25th 2008 12:49PM]

We dont need it, but then these are not marketed for us. These are Xeons, and are for Servers. The more processors a server has without having to introduce a new processor package or increase the power usage, the better.

The Nehalems probably wont be 6 core initially.

And even if we dont need it, why not be a step ahead, if the cost isnt much more than a quad core?

[eggothewaffle @ Feb 25th 2008 1:28PM]

If you've ever tried doing anything with audio, video, or 3D rendering for money, love, or anything else, then yes.

The more cores there are (and the better written the software is), the less time we have to wait. The less time we have to wait, the more awesome stuff we can make.

Let's not forget that gamers aren't the only ones who use computers for more than just Firefox and iTunes.

[Andir3.0 @ Feb 25th 2008 2:27PM]

"current CPUs use even less power, generate even less heat, and be very very cheap"

They are... and they are putting more of those on the same die. So in a way... you are getting your way. It's just that your being forced to buy more of them.

[gbfluteman @ Feb 27th 2008 2:00AM]

In regards to focusing on heat and performance instead of more cores, I believe that might be where VIA's Isaiah chip comes in, if the thing is as good as I'm hearing it is, with as little power consumption and heat output as it's supposed to have. It would make Isaiah comparable to this chip, since it doesn't have the issue or possibility of multiple cores trying to access the same piece of data (those guys were thinking ahead with their architecture). Could be wrong, since we don't have official benchmark numbers yet (a lot of people thought Apple Air would be great, until benchmarks showed otherwise), but it's certainly possible, based on what I've read.

[stankychicken @ Feb 25th 2008 12:49PM]

so is this similar to the ps3 cell chip?

[Ruben @ Feb 25th 2008 12:51PM]

No.

The Cell has a controller core with x number of small calculation cores.

The individual cores dont fetch for data, but are fed data.

These are 6 individual controller "like" cores. They can do their own thing, or can be programmed to work together.

[kal326 @ Feb 25th 2008 1:11PM]

Your average joe may not be able to take advantage of 4 cores, but these are Xeon business chips meant for servers and workstations. I use a VMWare box that makes very good use of its 8 cores.

[T-Bone @ Feb 25th 2008 2:41PM]

I guess I'm a little above average, but the other night I was ripping 2 DVDs simultaneously while converting a 3rd DVD image to an iPhone-compatible format. 4 cores would have been very nice to have.

[lifehacker @ Feb 25th 2008 1:28PM]

This is a big deal, probably the most significant change in the Intel architecture since the original Core chips were released. Nehalem introduces a direct CPU-to-memory bus (so memory accesses do not have to share bandwidth with I/O), and the shared bus is replaced by QuickPath Interconnect (QPI), a new switched interconnect similar to AMD's HyperTransport, which means the bus will no longer be the bottleneck. These should yield massive improvements in performance for systems with more than 2 cores, but will also require completely new motherboards.

If you want to buy a PC, it makes a lot of sense to wait for Nehalem, which is why Intel is not spilling too many beans on it because they don't want sales of late Penryn versions to evaporate.

[TMM @ Feb 25th 2008 1:35PM]

12 Core MacPro? Something made poink in my head.

[Mr Deeds @ Feb 25th 2008 3:16PM]

A 6 Core CPU will probably cost more money than running 2 quads in a Skull Trail... Not going to 8 Cores on 1 CPU was a dumb move, in my opinion. Well, I guess they will only be selling to companies with more than 2 physical CPU's anyway's?

[Cranky Media Guy @ Feb 26th 2008 2:18AM]

"Where the hell do they come up with these weird ass code names?"

Nehalem is the name of a town and river on the north Oregon coast. Whether that figures into the reason for the code name, I do not know.

[Special K @ Mar 4th 2008 10:40AM]

In my Systems Analyst point-of-view, you people are missing the big picture.

Server virtualization!

Running 10+ virtual servers on a single box would suddenly became much more feasible. Now, if metaram will ever come out...

I run an E8400 on my game rig. Seriously, how many SLI/Crossfire GPUs would you need to not bottleneck a 6 core CPU?

[Xerox @ Mar 18th 2008 4:03PM]

Not Cell processor.. Intel just haven't get to there, they are planning on but, it's called the Larrabee Architecture for GPU media et all, put together with AVX, or Advanced Vector Extensions for compilers that will increase performance in floating point, media, and processor-intensive software. Support for it will not appear until the "Sandy Bridge" family of processors is released in 2010. so after all "yes" cell id signaling the way