Official Desktop Penryn Discussion Thread

[terrace215]

http://www.tgdaily.com/content/view/32008/135/

We recently learned that Wolfdale, the desktop variant of Penryn, has reached final pre-release status and in fact may be sent out to most system builders for testing purposes and early reviews next month.

...

there is every indication that Intel is pushing hard to get this processor out to system builders, analysts and journalists as quickly as possible.

...

fully assembled, commercial-grade production units in LGA775 packaging are making the rounds within the company.

...

[WillyWonka]

K10 better deliver or AMD is toast.

[safan80]

Wolfdale = dual core
YorkField = quad core

I'm waiting for the quad core not the dual core.

[terrace215]

Wolfdale = dual core
YorkField = quad core

I'm waiting for the quad core not the dual core.

The Quad is made from 2 duals, in an MCM. And QC (low volume, high ASP) is what Intel will release first.

45nm DC parts will follow as 45nm volumes ramp. It's the same thing AMD is doing with their 65nm K10 ramp: QC first, then DC when the volume ramps.

[MarlboroMan]

I wanna see some 45nm benches badly
my next mobo is P35 to be sure that i would be able to upgrade to 45nm when intel release it

[nn_step]

K10 better deliver or AMD is toast.

Penyrn is a Die shrink and minor enhancements, not exactly a revolution.

[vitaminc]

Penyrn is a Die shrink and minor enhancements, not exactly a revolution.

radix 4 to radix 16 divider is a pretty boost in floating point. and floating point has historically been the weakest link in intel's design.

[safan80]

The Quad is made from 2 duals, in an MCM. And QC (low volume, high ASP) is what Intel will release first.

it's still a quad core that will be better than kentsfield.

[WillyWonka]

Penyrn is a Die shrink and minor enhancements, not exactly a revolution.

A +10% per clock improvement (conservative) is not minor given that Conroe is only +20% per clock faster than K8.

[gdogg]

Penyrn is a Die shrink and minor enhancements, not exactly a revolution.

Guess all those new sse4 instructions don't count either lol

[Theli]

Is this when we will start seeing overclocks on Penryn processors?
Perhaps sub-10 seconds 1m Superpi benchmark results will become commonplace.

[Mad1723]

Is this when we will start seeing overclocks on Penryn processors?
Perhaps sub-10 seconds 1m Superpi benchmark results will become commonplace.

4.5 Ghz will easily be attainable on Air I guess

6 GHz will be attained, I'm sure of it, quote me on that one

[theteamaqua]

Penyrn is a Die shrink and minor enhancements, not exactly a revolution.

yes that is sooo true .. its basically Conroe with 6MB L2 cache with SSE4 i think
and Penryn is going to have 2 of that in a single package...

like what they did with Pentium D 800's to 900's , die shrink = larger cache... that was bigger cuz cache size went from 2MB to 4MB ... twice as much .. . this time only 50% more 4MB to 6MB

if Barcelona has really good clock for clock performance it might have a chance ... but AMD is just wayy behind schedule , i think Opteron is going ot launch at Dec as dailytech suggest, but not sure about the desktop part

AMD Claims Quad-core Desktop CPUs Ready for Christmas
http://www.dailytech.com/AMD+Claims+...rticle7232.htm

or is it summer?? not sure the faster they get it out the better...

[iterations]

Penyrn is a Die shrink and minor enhancements, not exactly a revolution.

Penryn is a die shrink, SSE4, vastly improved Radix divider, shuffle engine, 50% increased L2 cache, better thermals, higher clocks, and ~10% improved IPC over the already dominant Conroe core.

Sounds great, and now it sounds like we'll be getting non NDA'd ES soon, and release by August.

4.5 Ghz will easily be attainable on Air I guess

6 GHz will be attained, I'm sure of it, quote me on that one

Yeah, you can take that to the bank.

[Emerica]

Intel has a lot to live up to with their new processors. Conroe was a hit and if penryn/wolfdale doesn't live up to it people are going to call it crap.

[Cronos]

radix 4 to radix 16 divider is a pretty boost in floating point. and floating point has historically been the weakest link in intel's design.

Quite the contrary, floating point always was and is the strongest point of Intel CPUs. Even then Intel struggled with P4 Prescott, blamed by everybody for high power consumption and relatively low performance -still its P4 had better floating point than AMD's K8.

Core2 has twice floating point peak performance than K8 -at the same clock speed. K10 should be able only to catch Core2 on peak, and given its (much?) lower clock speed, will have hard time to compete against C2 here.

[Qkjhfhaiguihfma]

The Quad is made from 2 duals, in an MCM. And QC (low volume, high ASP) is what Intel will release first.

45nm DC parts will follow as 45nm volumes ramp. It's the same thing AMD is doing with their 65nm K10 ramp: QC first, then DC when the volume ramps.

i think it's a bit different since k10 is native quad core while yorkfield is 2 x wolfdale on a single package.

[freecableguy]

Intel has a lot to live up to with their new processors. Conroe was a hit and if penryn/wolfdale doesn't live up to it people are going to call it crap.

This line of reasoning makes absolutely no sense....

[Shadowmage]

Penryn is a die shrink, SSE4, vastly improved Radix divider, shuffle engine, 50% increased L2 cache, better thermals, higher clocks, and ~10% improved IPC over the already dominant Conroe core.

Sounds great, and now it sounds like we'll be getting non NDA'd ES soon, and release by August.



Yeah, you can take that to the bank.

It's not +10% IPC, then all the others. It's +10% IPC *DUE* to the other enhancements you listed.

However, a certain program that many of you guys are obsessed about may have +50% improvement, since it uses many divides and things like that.

[iterations]

It's not +10% IPC, then all the others. It's +10% IPC *DUE* to the other enhancements you listed.

However, a certain program that many of you guys are obsessed about may have +50% improvement, since it uses many divides and things like that.

Where did I say that IPC was an independant variable?

Of course the better IPC comes from the actual architectural improvements. Just like higher clock and better thermals come from the die shrink and material science/fab improvements. It would be rather silly to think IPC is a simple thing that architecture designers can simply "add more of". I doubt most people that read these forums are that misinformed.

[red]

FLG, I think terrace meant quad core, period, was low volume relative to dual core, rather than MCM vs native.

Divide.. Super PI?

[ted3]

Quite the contrary, floating point always was and is the strongest point of Intel CPUs. Even then Intel struggled with P4 Prescott, blamed by everybody for high power consumption and relatively low performance -still its P4 had better floating point than AMD's K8.

Not really, for most FP intensive apps K8 at 2.4GHz was better than 3.4GHz P4. FP benchmarks are multithreaded and benefits substantially from HT, not a correct reflection of real world performance with singlethreaded apps. Compare X2 to non-HT Pentium D and X2 takes a clear lead.

[terrace215]

Actually... the MCM design means HIGH volume since the binning processis easier and more efficent. With MCM they just have to have 2 cores (one die) perform to a given speed, then find a similar Dual Core package to "mate" with it.

I meant MARKET volume. The QC market is much lower volume than the DC market, and has higher ASPs, so you use your DC dies to make MCM-based QCs first, then sell DCs when you've ramped more volume.

I was not bashing MCM yields vs "native" QC yields. Anyone who isn't a moron realizes that the MCM approach has a yield advantage.

[Cronos]

Not really, for most FP intensive apps K8 at 2.4GHz was better than 3.4GHz P4. FP benchmarks are multithreaded and benefits substantially from HT, not a correct reflection of real world performance with singlethreaded apps.

Only poorly optimized. And only poorly optimized apps. really benefit from HT.
On the other hand, almost all windows apps. are poorly optimized, so you are somewhat right.

By the way, 2.4GHZ K8 has 4.8GFLops peak, while 3.4Ghz P4 -6.8GFlops.

[theteamaqua]

4.5 Ghz will easily be attainable on Air I guess

6 GHz will be attained, I'm sure of it, quote me on that one

yeha and i think that Wolfdale, the dual core one should OC higher than Penryn .. unless intel cripples it or do something to force people buy quad core... its very likely that wolfdale will clock higher, if a 65nm conroe can do 3.5 to 4GHz on air .. i dont think its too much to assume that the 45nm wolfdale can achieve 4.5+GHz ... on air

kinda liek the transition we had from Pentium D 800 to 900 ... lots of people with 930/940 were able to get 4GHz on default vcore ... and 4.5GHz 24/7.. while D 800's can barely achieve 3.7GHz ...