Intel Westmere: Six-Core in 32nm in 1H/2010

Shtal

Veteran
Intel's next processor architecture Nehalem has been the subject of months in the media. Later this year it will be with Bloomfield in the desktop sector knock, but behind the scenes is already loudly on the West Shrink Mere talk. We are now a preliminary roadmap of the world's largest semiconductor producer in the hands, not only the time window of the chip sets, but also provides some technical information in store. According to Intel's 2006 imported Tick Tock model should be completed by 2008 a Tock (new architecture) and 2009 a tick (shrink). The fact that this road map but not quite its validity and it has now become very closely, proves that the notion of Nehalem architecture until the end of 2008. And according to our information, there will be Intel therefore no longer create, West Mere still - as planned - 2009 for launching, Instead, spoken in the latest roadmap is now generously from the first half of 2010.

It is interesting that most of the West Mere not a simple manufacturing change from 45 to 32 nm act, but two other seeds on the Space take a six instead of four. The six computing cores, presumably to the fed and continue existing SMT technology on a total of 12 threads. The L2 cache size will be as usual 256 KB per core; he L3 cache will be linear to the number of cores from 8 to 12 MB rise. In addition, improvements in the memory support expected Mere West supports DDR3-1600 memory. Bloomfield is on the other hand. As has been reported Fudzilla, officially only a maximum of DDR3-1066.

Preislich, the first six-core processors, as Bloomfield, at around 300 U.S. dollars start in the high-end version (Extreme Edition) up to 1000 U.S. dollars expensive. Whether in the course of 2010 even quad-core and possibly even dual-core processors in 32 nm will be remains to be seen.
12208214448b768fd679.jpg


http://www.hardware-infos.com/news.php?news=2377
 
Nice! I thyink I'll wait for this rather than going for Nehalem as soon as it launches. I didn't really feel Nehalem was enough of an upgrade over my exisiting dual core conroe to justifiy it. Especially with the talk of it not being much faster in gaming (both on a per core basis and the fact that games don't scale beyond 2 cores very well).
 
Especially with the talk of it not being much faster in gaming (both on a per core basis.

Clock for clock Nehalem is about ~20% faster then conroe core2.


-OR- Unless if this is true fact!! I would guess 30% faster clock for clock.
Intel has told that the upcomming Processors Core i7 are faster than first-tests have shown. The cause is that all tests were made with the B0-Stepping, while the final processors will have the B2-Stepping or even higher.
http://forums.vr-zone.com/showthread.php?t=323664
 
Clock for clock Nehalem is about ~20% faster then conroe core2.

Conroe or Penryn? Since Penryn is already around 5-10% faster then that would only make Nehalem 10-15% faster than an equally clocked Penryn. Not much when you consider what they have added.

I read that Intels focus with Nehalem was not desktop performance but on server performance (where it will likely wipe the floor with Penryn).

Still, if you meant Penryn and your second link is accurate then a 30% performance boost would be nice. I think I'll just wait and see how it pans out. There's certainly nothing even my modest 2.4Ghz C2D can't handle at the moment so I might even hand on for Sandybridge. AVX looks sexy :p
 
Does anyone know offhand - can Nehalem's L3 be partitioned or is it just an aggregate chunk of cache accessible by all cores?
 
Conroe or Penryn? Since Penryn is already around 5-10% faster then that would only make Nehalem 10-15% faster than an equally clocked Penryn. Not much when you consider what they have added.

One thing people are not remembering about Nehalem - its new cache hierarchy includes a new level (L3) which has latency, and is inclusive. Either they've decided "to hell with latency, we want bandwidth and threads" or their simulated cache hit rates are through the roof and effective latency is actually down so it's irrelevant.

I read that Intels focus with Nehalem was not desktop performance but on server performance (where it will likely wipe the floor with Penryn).

Still, if you meant Penryn and your second link is accurate then a 30% performance boost would be nice. I think I'll just wait and see how it pans out. There's certainly nothing even my modest 2.4Ghz C2D can't handle at the moment so I might even hand on for Sandybridge. AVX looks sexy :p

Nehalem's focus is on the future, and that means its a bandwidth monster with a multi-threading specialty, let's say.

I'm holding out for Westmere at least, unless I can afford to build a DP Gainestown system (the next skulltrail) that is :p

In my future PC universe (in my head) I can see a 2MB L2 per core, 8 core/16 thread part with 16MB or more L3. Oh, and approx. .5TFLOPs @ 4GHz too ;)

:D
 
Last edited by a moderator:
Conroe or Penryn? Since Penryn is already around 5-10% faster then that would only make Nehalem 10-15% faster than an equally clocked Penryn.

It could be Conroe or Penryn - depends on the application, sometimes extra cache on Penryn not needed; also, you could consider SSE4 not supported by older applications :)
 
I wonder, how would 6 core Westmere (2+2+2) fair against Nehalem's native 4 core design in apps that can't take advantage of >4 cores? I think early i7 adopters could be happy for a long time.
 
Hmm with all this i may just get a q9650 to replace my q6600. I want to move foward to 8 cores for video editing but that may be a long while yet
 
I wonder, how would 6 core Westmere (2+2+2) fair against Nehalem's native 4 core design in apps that can't take advantage of >4 cores? I think early i7 adopters could be happy for a long time.

Thats a very good point actually, I hadn't considered that. I wonder how the L3 cache will work then?

Thiat actually makes me more interested in picking up an i7 now. Especially since they're unlikely to scale much, if any faster than 3.33Ghz this year.
 
Back
Top