what's diff between Yorkfield and Kensfield?

[nyker96]

besides 45nm vs 65nm process? I mean any core differences? I do realize York will be paired with a new board capable of 1333/1066 FSB.

and how will York be compare to AMD K8L? from the spec, K8L is only running about 3Ghz while York will be well into 3.5-4 range with much more cache on board. On papaer York seems to be a lot faster than K8L.

 

[stevty2889]

Not really enough info available on yorkfield, or K8L to make any kind of comparison..

 

[Viditor]

Originally posted by: nyker96
besides 45nm vs 65nm process? I mean any core differences? I do realize York will be paired with a new board capable of 1333/1066 FSB.

and how will York be compare to AMD K8L? from the spec, K8L is only running about 3Ghz while York will be well into 3.5-4 range with much more cache on board. On papaer York seems to be a lot faster than K8L.

Any speeds you've read are pure speculation by fanbois at this point...not even Intel or AMD know as final silicon hasn't even been made yet.
K8L's core should (on paper) be equivalent to C2D and equivalents, only of course it will also have HT and ODMC. But as stevty points out, we just don't know until samples start coming out...

 

[The-Noid]

Yorkfield has 12 megs of cache and is a real QuadCore Processor. It is also based on 45nm. Clock speeds have yet to be announced. There is also bloomfield which is a true quad core version of Kentsfield that comes out before that though. I don't really see Intel loosing the performance crown while they are using the Core Marchitecture, but things have been known to change. We have heard nothing about K8L as they are much better on the NDA side of things.

 

[Viditor]

Originally posted by: Yoxxy
Yorkfield has 12 megs of cache and is a real QuadCore Processor. It is also based on 45nm. Clock speeds have yet to be announced. There is also bloomfield which is a true quad core version of Kentsfield that comes out before that though. I don't really see Intel loosing the performance crown while they are using the Core Marchitecture, but things have been known to change. We have heard nothing about K8L as they are much better on the NDA side of things.

We actually do have quite a bit of info so far on K8L architecture, what we don't have is a chip to test (I assume that's what you meant).
Any who are interested should look at this pdf from AMD in June...some interesting stuff there.

 

[nyker96]

I was also thinking that if York uses Core architecture then it certainly cannot loose to A64 architecture on a per core basis. However, if K8L features a major step forward in a new A64 core design, it could shift the balance of power. On one hand we got Intel's 45nm advantage which from look of things should go higher than 3.5 considering 65nm processes of same cores allow that much headroom, 45nm I'd assume will be higher. While AMD could come out blasting with the new K8L core architecture using "just" 65nm process. But I have to say at this point the odds in my mind is against AMD. running 65nm their core architecture have to be quite a bit more efficient to compete against Intel's 45nm's clock potential.

Of course all this just speculation. Since none of us can gauge the effect of chipset designs - Intel's Bearlake nor AMD's HTT 3 boards. There's even more unknown here. It looks to me that by raising this question now, we dragged out more questions than answers on K8L vs. York.

 

[Viditor]

Originally posted by: nyker96
I was also thinking that if York uses Core architecture then it certainly cannot loose to A64 architecture on a per core basis. However, if K8L features a major step forward in a new A64 core design, it could shift the balance of power. On one hand we got Intel's 45nm advantage which from look of things should go higher than 3.5 considering 65nm processes of same cores allow that much headroom, 45nm I'd assume will be higher. While AMD could come out blasting with the new K8L core architecture using "just" 65nm process. But I have to say at this point the odds in my mind is against AMD. running 65nm their core architecture have to be quite a bit more efficient to compete against Intel's 45nm's clock potential.

Of course all this just speculation. Since none of us can gauge the effect of chipset designs - Intel's Bearlake nor AMD's HTT 3 boards. There's even more unknown here. It looks to me that by raising this question now, we dragged out more questions than answers on K8L vs. York.

65nm on dual core allow for that much headroom...

 

[RaynorWolfcastle]

Well AT's and other reviews from around the net seem to find that 3.2-3.3 GHz is possible at stock voltage with the stock cooler on the Kentsfields they reviewed so > 3.5 GHz on 45 nm is really not much of a stretch at all.

 

[Brunnis]

Originally posted by: RaynorWolfcastle
Well AT's and other reviews from around the net seem to find that 3.2-3.3 GHz is possible at stock voltage with the stock cooler on the Kentsfields they reviewed so > 3.5 GHz on 45 nm is really not much of a stretch at all.

Exactly. This means that K8L will have to be able to clock considerably higher than what the initial products are rumored to be OR be significantly more efficient than the Core arch.

Personally, I don't really think AMD will achieve either. At least not at release. I'm really hoping that they prove me wrong, though.

 

[Furen]

Originally posted by: Yoxxy
Yorkfield has 12 megs of cache and is a real QuadCore Processor. It is also based on 45nm. Clock speeds have yet to be announced. There is also bloomfield which is a true quad core version of Kentsfield that comes out before that though. I don't really see Intel loosing the performance crown while they are using the Core Marchitecture, but things have been known to change. We have heard nothing about K8L as they are much better on the NDA side of things.

Errr, Yorkfield is a 45nm dumb shrink of Kentsfield. It will have more cache (12MB total) and SSE4 but that's it. It will also be a dual-die solution. All the "fields" are, in fact, dual-die parts. Intel's first "real" quad core will be Nehalem.

Viditor: Intel's 3.46-3.7GHz clock speed estimates seems reasonable enough, unless yields at 45nm suck. Most current Kentsfields can hit 3.2GHz without problem (though power draw is quite high), so the problem for Yorkfield will be attaining those clocks at 130W TDP.

 

[harpoon84]

Originally posted by: Furen
Errr, Yorkfield is a 45nm dumb shrink of Kentsfield. It will have more cache (12MB total) and SSE4 but that's it. It will also be a dual-die solution. All the "fields" are, in fact, dual-die parts. Intel's first "real" quad core will be Nehalem.

Viditor: Intel's 3.46-3.7GHz clock speed estimates seems reasonable enough, unless yields at 45nm suck. Most current Kentsfields can hit 3.2GHz without problem (though power draw is quite high), so the problem for Yorkfield will be attaining those clocks at 130W TDP.

How is it a 'dumb shrink' if it has more cache and SSE4? Was Northwood a 'dumb shrink' of Willamette? Was Prescott a 'dumb' shrink of Northwood? AFAIK the only dumb shrink by Intel in recent years is Prescott -> Cedar Mill.

Also, Yorkfield will have a faster 1333MHz FSB compared to 1066MHz for Kentsfield.

 

[Furen]

Cedar Mill a dumb shrink? It added VT! There are always tweaks whenever there's a new silicon spin. The D stepping for A64s tweaked the hell out of the memory controller but I'd still consider it a dumb shrink. Intel always adds cache whenever it does a shrink, and about SSE4... well, it could be in Conroe for all I know, since Intel also likes throwing stuff into a core and activating it until later, not to mention that its nothing earth-shattering about adding new instructions.

Prescott was a significant redesign of the architecture, not even close to a dumb shrink. Northwood was just a shrink of Willamette with additional cache, a higher frontside bus and Hyperthreading. I'm not absolutely certain about this but I believe HT was actually in Willamette, it just wasn't enabled 'till Northwood. Willamette performed horribly because it was insanely bandwidth starved, so both the cache and the FSB helped there. Northwood also introduced Intel's 130nm process and copper interconnects to the Netburst family. Remember that Conroe came out this summer, any changes we see within the next few months will only be relatively minor tweaks. I'm actually a bit skeptical about whether even Nehalem will be completely different architecture... It just seems to me like it'll be a native quad-core Conroe with a northbridge thrown into it (kind of like how the K8 was to the K7 ).

Cloverton will also use a 333MHz FSB, that's nothing particularly exciting, especially considering that Yorkfield will also, supposedly, run on DDR3-1333-supporting motherboards.