Guide to Intel's Penryn and other Current Notebook Processors

by Dustin Sklavos

INTRODUCTION

The splash that Intel's Centrino branding made back when it was introduced has had some profoundly lasting effects on the notebook industry. While on the one hand it's dramatically improved performance and battery life hand in hand in modern notebooks and provided AMD with more competition that they know what to do with, the whole Centrino brand and label has sort of missed its original mark.

At the beginning, Centrino meant three things: Pentium M processor, Intel Wireless 2200BG, and Intel 915 mobile chipset. The branding worked so well people were calling the processor itself "Centrino" and notebooks were advertised as "Centrino 1.5GHz."

Now, there are multiple different kinds of processors that merit the Centrino branding, three different network adaptors, and three different chipsets. And while the last two don't matter that much in the grand scheme of things (wireless support for A and/or N is still fairly irrelevant as the rest of the world has pretty much standardized on G), that first one is leading to a world of confusion.

It shouldn't be that bad. Every time Intel has announced a new line of mobile processors, their lineup has been for the most part pretty simple, yet once chips start going to OEMs, all kinds of random crap starts popping up. When the Pentium Dual Cores first materialized, I was actually able to ask an Intel rep at the Game Developers Conference (GDC) about them, and he had no idea what I was talking about. He had to refer me to someone else. If Intel can't keep this straight, how can they expect us to?

It gets really ugly when you look at how inconsistent Intel is with their branding. The shift from Yonah (Core Duo) to Merom (Core 2 Duo) cores somehow merited an increment in the brand name, yet the shift to the new Penryn cores apparently still just means Core 2 Duo, except they're differentiated in the model number, which is supposed to clarify things but winds up just feeling exactly like what it is - an abstract that requires a decoder ring to comprehend.

With the shift to Intel's new Penryn-based Core 2 Duos comes a need to explain just what the heck all this stuff actually means. That's where I come in.

SPECIFICATIONS

In order to understand the differences between the individual models between the cores, there are a couple of things you need to understand.

Process (45nm or 65nm): The process refers to how a processor is manufactured. Without boring you with the details, it's basically a matter of scaling the individual components of the processor. A smaller process typically means a chip that requires less power and generates less heat, as the electricity that runs through the chip has shorter distances to traverse. It's important to keep in mind that the heat generated by a chip is basically wasted electricity that leaks out of gaps in the processor design. Before anyone goes "why doesn't someone just plug the gaps," it bears mentioning that it doesn't work that way, and there's no point in being an armchair engineer. The basic lesson here is that a smaller process is generally better.

L2 Cache: Most of you know what a hard drive and memory are, but it bears explaining the reason why memory exists. Hard drives are pitifully slow, far far slower than is useful for a processor, so information from the hard drive gets cached into memory, or RAM. Memory has a much faster access speed than a hard disk (roughly 130 times faster) and a closer connection to the processor. But wouldn't it be faster if the memory had no distance to travel to the processor, if it were just on the processor?

That's why processors have their own cache. While notebooks with 3GB of memory are starting to normalize, processors have a much smaller amount that is built directly into the chip which runs substantially faster. Current dual core processors top out at 6MB of L2 cache, or roughly 1/500th the typical higher end notebook's memory. Data gets loaded into this L2 cache (called L2 short for Level 2; chips even have a smaller L1 cache that runs at still faster than the L2) and run through the processor.

So the big question becomes: why not just load up a processor with L2 cache? A couple reasons: first, a processor simply may not be fast enough to utilize boatloads of L2 cache and second, L2 cache already takes up a substantial amount of space on a processor.

It bears mentioning that the L2 cache on Intel's chips only seems to make a notable difference in performance when it drops below 2MB. 2MB and above tends to be plenty for efficient performance.

Front Side Bus: The front side bus does two things - it helps define the actual clock speed measured in MHz/GHz of the processor, and it represents the speed at which the processor communicates with memory. As such, a higher front side bus is always better for performance, but with the higher speed comes higher battery life. As you'll see later on in this article, Intel has some impressive power saving features to be found here.

Clock Speed: This is the last and most obvious characteristic of a processor's performance. This is what's measured in MHz/GHz, and it's what everyone's typically familiar with, going back to the Pentium 4 and prior days when chips were known by this instead of model numbers. This is the primary factor in determining an Intel processor's performance.

These four factors are chief in defining the performance characteristics of an Intel processor, but it bears mentioning that they are not necessarily as applicable to AMD processors, which have radically different architecture. AMD processors, for example, have no front side bus.

The processors I'm going to be talking about are all dual core - effectively two processors. Note that this doesn't mean that a 2GHz dual core processor effectively performs like a 4GHz single core; it performs like two 2GHz processors. Dual core chips offer smoother performance in multitasking, and Intel's architecture is efficient enough that their current dual core chips see power consumption not dissimilar from their older single core chips while offering substantially better performance. Such is progress.

So, working our way from the bottom rung to the top...

PENTIUM DUAL CORE

Intel's Pentium Dual Cores all share a 533MHz Front Side Bus, 1MB of L2 Cache, and were made on a 65nm process.

Intel's Pentium Dual Core mobile processor line is basically their cut rate line; their "value" line. In that respect, they're great finds, as they offer dual core performance on the cheap.

There are six Pentium Dual Core mobile chips in circulation; three that are 64-bit capable and three that aren't. The 64-bit capability is important for using the 64-bit version of Windows Vista, which allows you to use 4GB or more memory in your notebook. At present, this isn't that important, but in the future it may be. Mercifully, the non-64-bit Pentium Dual Cores have been quietly going extinct.

The only differentiating characteristic between Pentium Dual-Core chips, outside of the 64-bit capability, is the clock speed.

64-Bit Pentium Dual Cores:

• T2370 - 1.73 GHz
• T2330 - 1.6 GHz
• T2310 - 1.46 GHz

Non-64-Bit Pentium Dual Cores:

• T2130 - 1.86 GHz
• T2080 - 1.73 GHz
• T2060 - 1.6 GHz

It does bear mentioning that the non-64-bit Pentium Dual Cores are based on Intel's older Core Duo architecture, and as such, are somewhat less efficient than the 64-bit chips.

For the few that it matters to, Pentium Dual Cores do not support hardware virtualization. If you don't know what that means, it's not important and you don't need it.

CORE 2 DUO

Traditional Core 2 Duos vary WILDLY between specs, and this is largely due to OEM processor specs being added fairly regularly. To me, at least, it's frustrating that Intel has no less than twenty different Core 2 Duo models, not counting the new Penryn chips. It's unnecessary.

The most desirable parts come from Intel's T7000 series. Beyond the usual clock throttling control in all Core 2 Duos - sacrificing speed for battery life - these chips can dynamically change the clock speed of their Front Side Bus to increase power savings.

The vast majority of Core 2 Duos support hardware virtualization, excepting the T5450, T5500, T5550, and T5750. Again, if you don't know what this does, you probably don't need it.

These processors are all excellent performers, but what's frustrating is that Intel's already overlong spec sheet is missing parts that are showing up in OEM notebooks with alarming regularity. The T5450, a chip that was even a subject of contention on our forums and is very common in the wild, doesn't appear anywhere on their sheet. There's just no need for this level of confusion.

Wikipedia's article here breaks things down pretty thoroughly; all you really need to know is that the ideal Core 2 Duo chips are in the T7000 series, but the T5000s are still excellent choices. You're really choosing between better and best here.

CORE 2 DUO (PENRYN)

Why these chips weren't given a new name (Core 3 Duo comes to mind) is a mystery to me, as there's a decent difference between these and the older Core 2 Duos. Penryn chips are identifiable by being members of the T8000 and T9000 Core 2 Duo lines, and offer on average a 10% performance increase over equivalently clocked 65nm Core 2 Duos. I say 65nm, because Penryn-based Core 2 Duos are made on a 45nm process, reducing power consumption and heat dissipation.

Penryn Core 2 Duos all share an 800MHz Front Side Bus, and have either 3MB or 6MB of L2 Cache. I'd hesitate to quibble over this; with 4MB of L2 Cache having netted diminishing returns over 2MB in the last generation, it's hard to imagine 6MB being that important. 3MB is probably a good sweet spot for performance and price.

It goes without saying that these new chips are the most desirable of the lot, offering modest battery life increases with their modest performance improvements. Allegedly the big addition is the "SSE4" instruction set these chips support, but since the only software seen thus far to support SSE4 has been the DivX Pro encoder, it's hard to get really excited about what winds up being largely a checkbox feature. By the time SSE4 matters, there's a good chance the market will have normalized to include this instruction set.

Mercifully there are only four models of Penryn in the wild right now, and they are:

• T8100 2.1GHz, 3MB L2 Cache
• T8300 2.4GHz, 3MB L2 Cache
• T9300 2.5GHz, 6MB L2 Cache
• T9500 2.6GHz, 6MB L2 Cache

Any one of these will be more than enough for the average user, but if you have to make a decision between one of these and one of the older 65nm Core 2 Duos and the price difference is an issue at all, you're probably fine with one of the older ones.

CONCLUSION

It's important to remember that just because one thing is better than another, that doesn't make the other "less good." The new Penryn chips are notable, sure, but they don't just automatically make the older Core 2 Duos suck, and if you have one or are looking to buy, you really aren't missing the boat here.

What this does bring to light, though, is frankly how awful Intel's naming scheme has become. It's not due to the model numbers necessarily, which for the most part do their job, but the fact that Intel just releases too many chips in too many grades. This doesn't do anyone any favors. I'm not even sure it does the OEMs any favors.

I'm not about to suggest to Intel how to do their business - they're a multinational corporation and I'm some voice on a technology site - but it just seems wise to drastically simplify their lineup by discussing with the OEMs exactly what they want. These days, the models they announce and list don't even really matter to anyone but the five percent of users (and I'm being generous) who build whitebox notebooks or upgrade the processors in their existing notebooks, and it's always the models that weren't announced that randomly show up in notebooks and lead everyone to ask questions (the Pentium Dual Core being one such disaster).

So ranting aside, Penryn is honestly not a big deal, and maybe that's why it wasn't given a new name. If you just bought a 65nm Core 2 Duo, or are weighing between one and a new Penryn chip, the Penryn isn't worth a major premium. However, Penryn is progress, and progress is good and nice to see in a processor market that's presently stagnant competitively. Hopefully Intel will normalize the market on Penryn chips pretty quickly (you'll know it when there are seventeen different chips based on it), but if you didn't get one, don't chuck your new notebook.

It's evolutionary, not revolutionary.

The revolution comes late this year when Intel rolls out Nehalem, their next-generation processor with an integrated memory controller in the CPU core.

 

I never got a specific date when Nehalem would actually see the AC inside a laptop. Do you have any information about the period when it will be released (Q3, Q4)?

 

Finally... some light shed on this subject!
Thank you very much!

 

Sticky anyone? It's at least worth a rep.

 

Can't give rep because it says I already gave some to him recently. Fantastic job Pulp!

I really don't think we'd see Nehalem this year, manufacturers are just rolling out Penryn and Montevina is to come out later this year. To launch Nehalem as well and have three updates from Intel in a year is going to get far too cluttered! Penryn is already going to become "obsolete" in a matter of months!

 

Nitpick: at the beginning Centrino actually meant Pentium M + Intel Wireless 2100B + Intel 855 chipset.

 

Wow!!! Pulp thanks for that. It was nice to learn a bit without all the fat in there for a change. Excellent article!!!

 

Agreed, good job.

Montevina is scheduled for May, I believe.

Someone correct me if I'm wrong.

 

Good stuff nice and simple for the masses to understand

Minor stuff, 915 chipset wasn't the 1st ones...there was the 855 and others before that But so true in the fact that how many people thinks that Centrino is a line of CPU surely this must cear a few things up for them!

@matt_is_pro... you are right (Q2 2008, so the ntoebooks should be rolling out by may! )

Nehalem is also the biggest change in the Intel CPU architecture in a very very long time...


Good job Pulp...rep for you!

 

Second the call for sticky, I'll be sure to refer to this when I do laptop shopping.

 

A nice sober overview of the current state of the Intel processor market, great for those shopping and trying to figure it out. At least Intel is changing the game and we have to keep up with things, you could have written an article like this for AMD two years ago and it wouldn't need to be updated. You're right that there's a lack of competition.

 

Got a kick out of this line

 

That's a whole lotta wiki! Thanks Pulp, rep to you

 

Very good read! Thank you for the enlightening information!

 

Who told Andrew that Pulp was sober when writing this? Did I miss something?

okok..just a little pun. Again :wink: Great Article!!:wink:

 

Just got a few minor corrections and additions. (Surprised? )

=> roughly 1,000,000 times faster, is probably a more accurate figure. (Seek time ~10 ms vs ~30-40 ns)
If you ignore seeks, harddrives do a bit better like you said, but seek time is a better basis for comparisons in this case (that's what the CPU relies on, generally)

And if you want a figure for the speed difference between RAM and L2 cache, it's something like 10-20x.

Fun fact: Roughly half the space on a modern CPU die is already used for L2 cache.
So the processor *is* loaded up with L2 cache.

Lower, right?

Anyway, nice article. Very informative

 

Ok..That was a good read but im going to ask a question. I just bought one of the new Gateway P-6831 Fx laptops. It has a 1.67 Centrino and an 8800GTS video card. I was all ready to upgrade the chip but now im even more confused. I play C.O.D 4 and simulator games and stuff like that. What would I or should I even upgrade to or am I wasting money. This is a socket "P" setup. Thanks guys

 

Yes, it can support Penryn CPUs (or at least that's what I think) and you can upgrade the CPU (if not to Penryn at least to Santa Rosa refresh CPUs, such as T7700, T7800).

 

Some nice information, I learned some new stuff so thanks!

 

That was quite an interesting read. Thank you.

Wow - I'd no idea half the die is comprised of L2 cache.

 

I was referring mainly to the MB/sec throughput.

I think I was sort of in that ballpark.

Derp. I'm many kinds of retarded.

LEAVE ME ALONE YOU DON'T COMPREHEND MY GENIUS

 

Maybe Intel is saving Core 3 Duo for Montevina?

 

Noooo.........

Monteviena will use Peryn, slightly different ones...a few Extreme models and lower voltage standard models...

 

I know, but Montevina is a really huge upgrade that will be around for almost a year.

 

Well with any luck it might be Nehalem compatible... ...oh i wish...knowing it wont happen lol! I am buying a notebook this summer

 

It can't happen. Nehalem has an integrated memory controller.

 

Alright alright I know! I was just having a moment of hope ok!?!? you have just killed it... great! thanks alot! ..... ehehhehe jokes

 

I share your pain man . . . if only . . .

 

But then again why not wait for 2nd or 3rd gen Nehalem...it is something brand new, so might as well wait and see!, buy one in the summer nehalem out early 2009 (hopefully) and I'll get another notebook in 2010/11 if I have the money!

 

Truthfully, we don't know if Nehalem will be any good. It's a major architectural change for Intel, more so than the Core Duo/Core 2 Duo line was.

Anyone who remembers the transition from Pentium III to Pentium 4 (or Northwood to Prescott for that matter) will remember that Intel's not immune to taking one step forward and two steps back.

 

Everybody seen this? Nehalem FTW!:

 

Unwieldy doesn't even begin to describe that design, but honestly I'm not a fan of integrating the GPU anyhow.

 

Well, I think I'll save myself the upgrade, and see how Nehalem does. If it shows signs of crappiness (which it better not!) then I'll just get the Penryn then - woo, it will be cheaper too! I'm happy with whatever it was that came before Santa Rosa (I forget the name, asfadj)

And Pulp, great stuff as usual

 

If this isn't stickied, I think it deserves a spot in the general hardware sticky =]

very nicely done!

 

Thanks for the guide!

I have my eye on a T9300, but I would like to get confirmation that a Quad won't work in regulat socket P first.

 

A quick question. I am getting my laptop soon with q6600. will i be able to, at one point, upgrade it to a penryn?

 

I'll stick to dedicated graphics and a Clarksfield (Gilo) mobile quad core. Auburndale will have what essentially represents the successor to the Intel GMA line and only two cores....

That design actually does look quite unwieldy. I'll pass, thanks...

 

You complain a lot about Intel flooding the market with a load of different models. Maybe I'm in the minority, but I prefer to have many choices. It really bothers me to have one manufacturer forcing a scarce amount of options down my throat.

They're available as configurable options in T(and R)61 Thinkpads now. An upgrade from the T7250 to the T8300 is just $10 with their current sale.

Now is a really good time for options!

 

nice work!

 

I'm getting the "Penryn 45nm T9300 2.5GHz, 6MB L2 Cache" in my Sager config. What do you guys think of it compared to the rest of the available Core 2 Duos, taking into consideration price, performance, etc. ?

 

Good work. Would be nice to know exactly how "modest" Penryn's battery saving expected to be. (instead in NBR's next article we get Penryn Extreme overclocking LOL)

 

Here is the reason why Core 2 Duo name wasnt changed with Penryn - it wasn't an architectural update. Core Duo to Core 2 Duo even though small, was an architectual update (EMT64, etc). If you're confused, think about the Pentium III when it changed from the Slot 1 to socket, and also when it changed from 512k off die to 256k on die cache and changed process (nm) and voltage. Also the Pentium 4 changed sockets, process and cache sizes too. The underlying architecture didn't change - thus no name change.

 

Expensive, but if you have the money and need the performance, go for it.

 

I think the lesson is with a given budget, choose RAM or video card upgrade over CPU upgrade.

 

Excellent Breakdown. I just have a few things to add.

The common misconception about the FSB is that it is a communice only to the memory. Notso. In the Intel chispet architecture, the FSB allows the processor to communicate with the Northbridge first, and every other component on the system communicates to the processor through the southbridge which in turn is connected to the Northbridge before it gets to the processor. Higher Bus speed means better overall communication between the processor and the components of the computer.

The "process" or fabrication process refers to the diameter of a single transistor on the processor. Currently they are at the 45nm process, which is very small, and thus a die shrink.

Although a processor has alot of built in instruction, the L2 cache really only stores commonly used data and non built in instructions. It would be very inefficient if it had to constantly swap information, although it is very fast.

 

I say wait a couple of months, until Montivena is released and a) prices will drop b) you can get a Montivena laptop instead.

 

Thanks a lot! Rep points!

 

Nice article Pulp! Well written, informative, accurate, and a great help to a lot of users I'm sure!

It's nice to see someone writing this stuff up for others, as not everyone knows this stuff off the top of their head (or wants to) but may occasionally need an accurate reference without having to dig for it.

This fits the bill nicely.

 

Adding my voice to the chorus of praise for the article. Next, I'll have to read through all the comments.

 

Agreed with pretty much everyone else...excellent guide, very clear and easy to understand