Official Desktop Penryn Discussion Thread

[duploxxx]

@GoThr3k
So, lets see. You don't believe that 4GHz 45nm C2D with 65W TDP is possible?
According to the 45nm High-K + metal gate parameters(Vcore, IdSat, etc) it provides 30% lower transistor power consumption at 20+% higher frequency.
Lets consider a 65W TDP E6700 as base in the calculation.
2.67GHz * 1.2 = 3.2GHz
65W * 0.7 = 45.5W
If for the 45nm @4GHz the Vcore remains same as for 3.2GHz, it will waste 56.875W.
(4GHz / 3.2GHz) * 45.5W = 1.25 * 45.5W = 56.875W
To 65W, there is +14.3% headroom.
65W / 56.875W = 1.143
Or you can increase Vcore to +6.9% and still be in 65W TDP.
Square root of 1.143 is 1.069.
All E6700 can OC at least 25% and be stable at default Vcore. So, its not a miracle for 65W TDP Woldfdale to have 25% higher clock at 7% higher Vcore.

gODJO

a nice calculation of very very big fanboy, that's completely rubbish, if you're calculation is true they should already brought 65nm above 3ghz and they will never do that because of TDP increase and you are talking about 4GHZ . downplay every comment and reply in an AMD thread and try to keep the heads up in an INTEL thread.

as stated before, if you don't have any decent NDA You are just another one in line that needs info from the web. try to live with it.

Looks like old info. The busses have gone up since.

Also the product launch is in 2007.

indeed it will be end oktober-begin november you will see Penryn in retail

Not really, unless you use a highend SLI/CF setup. The appealing part would be quadcores. Personally I wait for nehalem. Since K10 looks like a R600 hype aswell.

yeah sure and Nehalem is not a hype? first onboard memory controller (the old 386 does not count, that's old tech) , csi, .... it is still early dev and paperwork and offcourse fanboy's like it already.

[Shintai]

yeah sure and Nehalem is not a hype? first onboard memory controller (the old 386 does not count, that's old tech) , csi, .... it is still early dev and paperwork and offcourse fanboy's like it already.

I dont think Nehalem is hyped yet. it sure will be closer to the launch.

What about 486SL then?

Onboard memory controller is only useful for the multisockets systems as I said before. For the desktop its 100% useless. Its just a sideeffect from the server oriented design. And in my eyes its a somewhat bad one. Just look on 939->AM2->AM3 situation.

But since Nehalem is a new architecture, I would expect alittle more and better than Penryn over Conroe. And I say I wait on nehalem because I dont believe in K10 hopes (It looks great for the bigger servers tho). Nor do I believe Penryn will make my Conroe obsolete before Nehalem.

[Shadowmage]

Nehalem is pretty interesting. You would have to believe in K10 hopes though... the two are in many ways pretty similar.

[gOJDO]

@duploxxx
Why don't you come back with something intelligent?
I don't care about stupid comments on my posts, but your hanging on my branch and dangling left-right annoys me. If you have something to say, related to computers or IT, we can discuss. But don't come back with another low IQ reply, without any valid arguments, data, facts or logical explanations. For example if you think that something is wrong with my calculation, you can point at it and explain, but your tasteless comment "a nice calculation of very very big fanboy, that's completely rubbish" makes me think that you are unlettered villager.

[duploxxx]

@duploxxx
Why don't you come back with something intelligent?
I don't care about stupid comments on my posts, but your hanging on my branch and dangling left-right annoys me. If you have something to say, related to computers or IT, we can discuss. But don't come back with another low IQ reply, without any valid arguments, data, facts or logical explanations. For example if you think that something is wrong with my calculation, you can point at it and explain, but your tasteless comment "a nice calculation of very very big fanboy, that's completely rubbish" makes me think that you are unlettered villager.

well that's just the same way you fill each thread, with garbage comments

the answer is very simple, look at power consumption at load of the c2d architecture from e6300-6400 and e6600-e6700-e6800 the series are the same, only different speed binding and yet you see a higher power consumption due to higher speed. That's way you don't see a higher bin from INTEL, X6800 is at teh tdp limit, until they create a new revision so they can drop the vcore a little or go to penryn.

http://www.xbitlabs.com/articles/cpu..._11.html#sect0

And don't put your high marks on penryn at the moment, there is a reason that they only release it late this year and ES samples are no +3,0Gig. But you'll find out when the NDA lifts

[accord99]

the answer is very simple, look at power consumption at load of the c2d architecture from e6300-6400 and e6600-e6700-e6800 the series are the same, only different speed binding and yet you see a higher power consumption due to higher speed. That's way you don't see a higher bin from INTEL, X6800 is at teh tdp limit, until they create a new revision so they can drop the vcore a little or go to penryn.

Looks like they did:

http://www.hardware.fr/articles/674-...1333-ddr3.html

Of course, the main reason that Intel hasn't gone higher is because they haven't needed to and therefore decided to use the TDP headroom on quad-cores.

And don't put your high marks on penryn at the moment, there is a reason that they only release it late this year and ES samples are no +3,0Gig. But you'll find out when the NDA lifts

Even if true, I'd imagine it has to due with AMD's feeble response, especially on the desktop.

[gOJDO]

well that's just the same way you fill each thread, with garbage comments

The only garbage I see here is you. Lay off the crack dude.

[Movieman]

@gOJDO and duploxxx:
Gentleman: Cut the personal attacks.
This is XS not Family Feud!
Thank you.

[savantu]

well that's just the same way you fill each thread, with garbage comments

You should be warned for monumental stupidity.

the answer is very simple, look at power consumption at load of the c2d architecture from e6300-6400 and e6600-e6700-e6800 the series are the same, only different speed binding and yet you see a higher power consumption due to higher speed. That's way you don't see a higher bin from INTEL, X6800 is at teh tdp limit, until they create a new revision so they can drop the vcore a little or go to penryn.

Reality speaks otherwise : Intel has a 75w TDP on its C2D Xtreme.Since at 2.93 it burns 66w I wouldn't be surprised if they cram a 3.2GHz in the 75w TDP envelope.
However , with no competition why shoot yourself in the foot and not maximize yields ?

Also , the new E6850 3GHz C2D burns 59w under 2x Prime. 3.33+ Ghz in the 75w TDP ?

If you want to see which uarch is power limited look at the jump from 2.6 to 2.8GHz on k8.

And don't put your high marks on penryn at the moment, there is a reason that they only release it late this year and ES samples are no +3,0Gig. But you'll find out when the NDA lifts

That's why Intel demoed 3.33GHz 45nm Quad-Cores 2 months ago at IDF ?

6 months from now as they refine the process why not see a 3.66Ghz Quad-Core at launch ?

[alfaunits]

Also , the new E6850 3GHz C2D burns 59w under 2x Prime. 3.33+ Ghz in the 75w TDP ?

Out of curiosity, how da hell can you tell how much a CPU (ONLY the CPU) uses vs. the entire system difference?
FWIW, Xeon 5148LV (DC) is rated at 35W @ 2.33. So the numbers you ask for are very much possible - only not useful, as you said ;-)

[Shintai]

Out of curiosity, how da hell can you tell how much a CPU (ONLY the CPU) uses vs. the entire system difference?
FWIW, Xeon 5148LV (DC) is rated at 35W @ 2.33. So the numbers you ask for are very much possible - only not useful, as you said ;-)

Very hard or close to impossible without very advanced gear. However, the easy way is to measure the consumption before the VRM. But then the efficiency/loss is something you have to guess.

[adamsleath]

3.6 ghz quad








if the pleb's can squeeze another few hundred MHz out of a penryn i spose they'll be happy

[madcho]

3.6 ghz quad








if the pleb's can squeeze another few hundred MHz out of a penryn i spose they'll be happy

don't believe this. Actual Quad at 3.33ghz are 150W tdp with 30% less power in 45nm you can run a kentfield at 105watt for 3.33ghz.

Penryn is 6Mo per dual core. with more cache it will be more hard to go higher. Dual core can, i'm pretty sure of this, but quad is too hot.

3.66 quad will go out one day, but not this year.

[gOJDO]

45nm offers +20% higher clocks at 30% less power consumption. So
It would be like 3GHz * 1.2 = 3.6GHz @ 130W * 0.7 = 91W, but that is the case for a die shrink only. Although the L2 is more than 50% of Core2 transistors it consumes around 20% of the total power consumption under 100% CPU utilization. So the additional transistors for the L2 in Penryn can't make a significant difference in power consumption. There would be also another 15 millions(on the quadcore, around 7.5M per die or around 3.5M per core) of transistors included in Penryn. I expect 3.6GHz 45nm quadcore at around 100W TDP.

BTW, here are some slides about Penryn from Intel:
http://media.corporate-ir.net/media_...gerOshanew.pdf

[adamsleath]

I expect 3.6GHz 45nm quadcore at around 100W TDP.

that would be amazing...but i think it might be a bit optimistic.
even though i can see how you arrived at those numbers sort of; who knows you may be right....assuming 90W tdp at 3.33 for yf.

does anyone know what speed wolfdale or yorkfield are supposed to clock at stock? what their codings are etc? have intel actually started to churn these chips out yet? i get the feeling that intel are not sorting /binning penryn cpu's as we speak or are they?

and just how much different architecturally are they from c2d's?

and sorry if this has already been discussed.


looking at that Gelsinger link:

intel forecasts (or is targeting) 45nm shipments to equal 65nm shipments by Q3 '08.

Actual Quad at 3.33ghz are 150W tdp with 30% less power in 45nm you can run a kentfield at 105watt for 3.33ghz.

where does this come from?

if this is correct 3.6 should be a walk in the park for a yorkfield.

[Baron]

edit---------


my bad Hexus site is silly.

[adamsleath]

yep that same demo again.

[gOJDO]

that would be amazing...but i think it might be a bit optimistic.

even though i can see how you arrived at those numbers sort of; who knows you may be right....assuming 90W tdp at 3.33 for yf.

I haven't seen a Penryn CPU yet. But according to the process parameters it should be like: 30% lower power consumption at 20% higher frequencies, compared to Merom/Conroe. Too good to be true, but that are the benefits of high-K dielectrics + metal gates + 45nm shrink.

does anyone know what speed wolfdale or yorkfield are supposed to clock at stock?

They can clock more than 20%, compared to Conroe/Kentsfield, while consuming 30% less energy.

what their codings are etc?

What do you mean? what codings?

have intel actually started to churn these chips out yet?

yes, right now they are mass producing them.

i get the feeling that intel are not sorting /binning penryn cpu's as we speak or are they?

I have absolutely the opposite feeling. The first produced Penryn CPUs were able to run multiple OS-es and softwares. That is very rare and big exception. Usually are needed months and a lot of debugging before you got a CPU which can boot an OS. With the first Penryn CPUs running in Dec 2006, Intel should release Penryn earlier.

and just how much different architecturally are they from c2d's?

http://www.anandtech.com/cpuchipsets...spx?i=2972&p=1

and sorry if this has already been discussed.


looking at that Gelsinger link:

intel forecasts (or is targeting) 45nm shipments to equal 65nm shipments by Q3 '08.

Currently Intel have only one fully functional 45nm fab for mass production. In mid 2008 they'll have 3. No wonder, their 65nm production will be higher than their 45nm until mid 2008.

where does this come from?

if this is correct 3.6 should be a walk in the park for a yorkfield.

Kentsfield @ 2.93GHz has TDP of 130W. At 3.33GHz, without increasing the voltage Kentsfield should have 148W TDP(130W * 3.33GHz/2.93GHz).

[BobHarris]

Yorkfield Quad core will be 1066 or 1333? One site says 1066, but the X38 shows it will support a quad core at 1333.

[Shintai]

Yorkfield Quad core will be 1066 or 1333? One site says 1066, but the X38 shows it will support a quad core at 1333.

1333. Servers will be 1600 sofar I recall for DP.

[BobHarris]

Is 1333 for the fastest Yorkfield? Otherwise, it's going to be very hard to overclock the yorkfield without having expensive 8500 RAM.

[savantu]

Is 1333 for the fastest Yorkfield? Otherwise, it's going to be very hard to overclock the yorkfield without having expensive 8500 RAM.

Why ? To run in sync you need DDR2 5300...

[lapdog]

Intel first showed a E6700 (limit of 65w TDP) and they released the X6800. I'd expect just the same with here, they demoed at 3.33 and will release a 3.67 extreme. 3.67GHz is probably just out of the mainsteam TDP window. Well, if history repeats it's self.

When a CPU has letters like EE, X, or FX in the name TDP goes out the window.

[alfaunits]

Err EE for Intel yeah, but for AMD it's actually the low wattage version ;-) (Energy Efficient vs. Exterme Edititon)
I reckon the 3.33 is not far from the top Penryn can achieve though (mb 4GHz tops official).

[Neon Biker]

Why ? To run in sync you need DDR2 5300...

Yeah, at stock.

I wonder what multi the lowest clocked Yorkfield will have. Or maybe they'll release some cheapo 1066 FSB parts... that would be great.