First Penryn (wolfdale) Numbers

[MAS]

new s-pi champion

expect 6 seconds (1M) for 5,5-6 GHz penryn

[JuanFlaiter]

No. Core Multiplexing just disables/enables the cores, leaving just one enabled. For example, core multiplexing will disable 3 cores on a Kentsfield, 1 core on a Conroe/Penryn, etc.

Also, I really don't think 14.x seconds for SuperPI 1M is possible at 2.66GHz stock speed for a Penryn family processor.

SuperPi loves cache, and disabling a core means the used core can use all 6MB, and that is the reason - maybe - of that score.

[nn_step]

SuperPi loves cache, and disabling a core means the used core can use all 6MB, and that is the reason - maybe - of that score.

you have a great point

[Shintai]

SuperPi loves cache, and disabling a core means the used core can use all 6MB, and that is the reason - maybe - of that score.

One core would get all the cache anyway in a singletreaded application even when both cores are enabled. Thats the bonus of being shared and dynamicly allocated.

[Movieman]

Just like with the AMD's, it looks to be an interesting year ahead.
Whether Penryn is 10% or 25% faster than C2D isn't an issue, 10% faster than an already great chip is a huge increase.
Look at what is going on with the new G0 stepping 65nm chips and now from the looks of things Penryn will be even better!
Enough to give an old guy wood!

[MAS]

Movieman, i see only 5-8 per cent penryn average advantage over conroe

where did you find 10-25 per cent he-he

[Movieman]

Movieman, i see only 5-8 per cent penryn average advantage over conroe

where did you find 10-25 per cent he-he

From the different comments I've seen. If it's 5-8%, I won't argue, I still feel the same, any increase over what is already a great cpu(C2D) is a plus..
I can't wait until someone on the site gets one in hand and sees what they can do in real world numbers.
I grew up with slide rules and adding machines so that might explain why you see the excitement of a little kid in me when I see what these can do so please excuse my enthusiam.
When I stop to think that when I went to school there was no such thing as a PC, calculator, color TV, cable TV, internet, disc brakes, seatbelts,etc, it really boggles the mind to see the advances.
I was 9 years old when Alan Shepard made his 15 minute trip into space and watched it on a 19" B+W TV with rabbit ears for an antenna in a classrom with 30 others glued to the screen.
Like I said, amazing advances and now I sit here with not one, but 2- 8 core clovertown machines in my house and it really makes one sit back in awe at times.
To appreciate what we have today, one has to look back a generation and see what was.

[Gautam]

I'd put some cash forward on that pi run being at 3.2GHz+.

[p8ntslinger676]

wow, looks good hopefully get lower 11's in 1m s-pi @ 4.0ghz for daily use

[informal]

I'd put some cash forward on that pi run being at 3.2GHz+.

Me too.

[metro.cl]

No. Core Multiplexing just disables/enables the cores, leaving just one enabled. For example, core multiplexing will disable 3 cores on a Kentsfield, 1 core on a Conroe/Penryn, etc.

Also, I really don't think 14.x seconds for SuperPI 1M is possible at 2.66GHz stock speed for a Penryn family processor.

Shot with unknown at 1969-12-31

his is what im talking about

[Shadowmage]

As for core multiplexing, Shintai is right for once.

[URL=http://imageshack.us]
his is what im talking about

EDAT can't be it, since it's a laptop technology. The chip in question is a Wolfdale.

[MoTiG]

It has obviously lower multi comparing to c2d, thus it's not assured it will clock that high.

[dexman]

It has obviously lower multi comparing to c2d, thus it's not assured it will clock that high.

depends on what clocks will Penryn start off. lets say there will be 333*6 Penryn, then it takes some mobo to overclock it to 3.5 GHz.
i´m also curious to see where low-multi Penryns hit fsb walls.

[ineedaname]

I know its not possible now but it would b cool if they could make it so you can overclock each core individually. Would bring a whole new level to overclocking. Well more like quad levels

[xsbb]

impressive

[dinos22]

i dont really think that 1M run was at 2.66GHz

it wasn't
Penryn is slower than that

[K404]

it wasn't
Penryn is slower than that

I did wonder if Intel would have to drop the 45nm clocks until their process matures- is that what you mean?

A question re the 65nm G0 stepping chips...are the improvements in the microcode or is the silicon tweaked? If its a silicon improvement, will they be able to use it in 45nm?

[Brother Esau]

Just like with the AMD's, it looks to be an interesting year ahead.
Whether Penryn is 10% or 25% faster than C2D isn't an issue, 10% faster than an already great chip is a huge increase.
Look at what is going on with the new G0 stepping 65nm chips and now from the looks of things Penryn will be even better!
Enough to give an old guy wood!

Dave...You're Slaying me brother ....P.S....I still use a Stick Ruler in the Wood Shop as my preffered everyday Measuring Device ...Much more accurate than a Tape Measure ever will be!!!!

[dinos22]

I did wonder if Intel would have to drop the 45nm clocks until their process matures- is that what you mean?

A question re the 65nm G0 stepping chips...are the improvements in the microcode or is the silicon tweaked? If its a silicon improvement, will they be able to use it in 45nm?

no that is not what i mean

the SuperPI time at those clocks should be slower

[kiwi]

Who in the right mind posts SP time without clock speed?

[DoubleZero]

From the different comments I've seen. If it's 5-8%, I won't argue, I still feel the same, any increase over what is already a great cpu(C2D) is a plus..

Well comments are cool and all, but in everest wolfdale lags behind conroe in every bechmark but the one that uses sse4.

Thank good it's a Intel chip, if it was a AMD chip this thread would be useless by now thanks to the trolling.

[Shintai]

Well comments are cool and all, but in everest wolfdale lags behind conroe in every bechmark but the one that uses sse4.

Thank good it's a Intel chip, if it was a AMD chip this thread would be useless by now thanks to the trolling.

Maybe because its tested in several real world applications on real chips. So not much to argue about.

Nobody here can run around headless and claim 50% overall performance here on an extrapolated estimate out from 1 synthetic bench

Intel have proven superb with Conroe and Penryn how to kill hype before ti starts. Along with getting alot of free PR for basicly no effort.

[TarTheDark]

I wouldn't rely too much on the Everest benches.

Results vary from version to version - as is pointed out by Everest itself.

[audiofreak]

The radix and shuffle engine can help on any version of SSE. But I dont see superpi as one.

They can help with x87 floating point as well so they can help with SuperPI.

There is no compiler that I'm aware of that can optimize code very well by using any extended instructions, all of them do job very poorly.

That means you are not aware of Intel C++ Compiler 10.0.025 which was out two weeks ago and which can already generate SSE4 optimized code for Penryn.