Real Life Mobile Penryn vs Merom Benchmarks

[MacRumors]

Anandtech offers a direct comparison between the existing Merom processors that currently power the MacBook Pro line, and the just-released Mobile Penryn processors. Apple is rumored to be working on MacBook Pro revisions as early as Macworld Expo next week that use these new Intel processors.

Anandtech was able to provide a direct comparison between the two processors:

Intel sent us two Dell Latitude D630 notebooks, identically configured, with one variable: the CPU. In one D630, we had a Core 2 Duo T7800, which is a Merom based chip running at 2.6GHz with an 800MHz FSB and a 4MB L2 cache. The other D630 came equipped with a new 45nm Core 2 Duo T9500, also running at 2.6GHz/800MHz but with a larger 6MB L2 cache.

With just the processor change alone, the new Penryn laptop offered 5-10% more battery life on their benchmarks. Meanwhile, the new processor saw 1 - 8% speed boosts on common tasks, and up to 40% improvements in applications that support the SSE4 instruction set.

Apple's MacBook Pro is currently available in 2.2GHz, 2.4GHz and 2.6GHz speeds. The last major revision of the MacBook Pro was in June of 2007.

Article Link

[CWallace]

Here is hoping new MacBook Pros at Mac World, followed by new MacBooks and iMacs next summer.

Here is also hoping Photoshop adds SSE4 support for common filters.

[InkMaster]

40%? drool

[thejadedmonkey]

10% on 3 hours.. not shabby.

[theman]

Originally Posted by thejadedmonkey

10% on 3 hours.. not shabby.

3 hours... I get 5.

[gnasher729]

Originally Posted by thejadedmonkey

10% on 3 hours.. not shabby.

Keep in mind that this was the savings comparing two processors running at full speed. The Mobile Penryn runs about six percent faster, so you get about sixteen percent more work done until the batteries run out.

[Rocketman]

To me this chip revision is more about the 45 nm technology, with its associated chemistry change, and an incremental improvement in power management, and the SSE4 language for bleeding edge apps, which do not effect most people, are the geek-kicker. Pro apps.

The notable thing beside better power usage is far better likelihood of deeper back-end price drops due to far smaller die sizes. Given Apple uses long term supply agreements with back end discounts, this product was designed more for manufacturability than even for features, which are strong enough indeed.

Apple is winning the vendor-supplier game, and Intel is reaping the benefits.

Rocketman

[CRIIM44]

now im no expert but werent the macbook pro's updated to santa rosa chips in june? or is merom another name for santa rosa.


http://www.macrumors.com/2007/06/05/...th-santa-rosa/

[pianoplayer1]

yea santa rosa is a subgroup of the merom processors.

Penryn is a new 45nm chip.

[tjugo]

Originally Posted by CRIIM44

now im no expert but werent the macbook pro's updated to santa rosa chips in june? or is merom another name for santa rosa.


http://www.macrumors.com/2007/06/05/...th-santa-rosa/

SantaRosa is a platform. A platform includes a set of MCHs, ICHs and CPUs.

Merom and Penryn are CPU families. It is up to intel to continue calling SantaRosa platform the new combination of Penryn + MCH + ICH.

Cheers,

[David G.]

Originally Posted by MacRumors

Meanwhile, the new processor saw... 40% improvements in applications that support the SSE4 instruction set.

What applications/types of applications take advantage of this?

[ChrisA]

Originally Posted by David G.

What applications/types of application take advantage of this?

The SSE instructions are of most use to software that process large streams of audio or video media. Doing things like changing color spaces, scaling pixels or encoding a ripped CD to MP3

Apple might modify Core Image or other "core" libraries to use SSE4 and then software that uses these libraries would be able to take advantage without need to be changed. Right now I think Apple is the biggest user of these libraries with programs such as Preview, iPhoto, Aperture Garage Band and so on. I think Adobe uses their own image processing code.

Use of SSE4 in software would have to wait until the SSE4 hardware is widely available or else how could Apple run a beta test?

[bobcb]

Originally Posted by David G.

What applications/types of application take advantage of this?

So far the benchmark Intel provided Anand for that article, although it can be assumed more will

[mattvolp]

"...40% improvements in applications that support the SSE4 instruction set."

Damn. That's not bad. Wait, what's SSE4 Instruction set?

[mattvolp]

Oh, my bad, just read all the posts. Wednesday hump!

[theLimit]

What kind of battery is in those Dells?!
4.5 to 7 hours of run time seems like a dream!

[gnasher729]

Originally Posted by David G.

What applications/types of applications take advantage of this?

Nothing right now. There are no Macs available yet with these processors (I guess the first reader receiving a MacPro will post here).

These things take time. SSE4 has two major components:

One is a more complete set of vector operations, which makes automatic vectorisation possible. What that means: Instead of the programmer having to write code specifically for the vector unit, he or she writes ordinary code and the compiler translates it into vector instructions. However, that fails when something that the code does is not available as a vector instruction. SSE4 is much more complete, so a lot more code can be translated automatically. That is something that applications like Photoshop would benefit from.

The other component is highly specialised instructions for video encoding. The most time consuming part of video encoding by far is "motion prediction", where the encoder takes a tiny bit of one frame and tries to find a similar image in a previous frame. For example, there is one instruction that does the following: Lets say you have four pixels abcd. And another eleven pixels ABCDEFGHIJK. This instructions calculates the difference between a and A, b and B, c and C, d and D and adds them up. Then it calculates the difference between a and B, b and C, c and D, d and E. Then between a and C, b and D, c and E, d and F and so on. The result is eight sums of four differences, and each of these 32 differences is the absolute value of the difference between two values. Without a vector unit, that would be 32 subtractions, 32 absolute values, and 24 additions - 88 operations in total. Probably a dozen vector instructions without SSE4. With SSE4 it is one instruction. That makes motion prediction a lot, lot faster.

You will benefit from this eventually, but not right now.

[MikeTheC]

Well, I want to buy a new iMac, so I hope it gets upgraded to Penry-class CPUs. C'mon Steve-O !!!

[acslater017]

i'm not really familiar with this...does anyone know what applications support this instruction set? do any of the major Apple apps?

[asdfTT123]

Do NOT expect 40% performance increases or you'll be very dissapointed with Penryn. There are currently little to no applications that have SSE4 support and there probably won't for quite some time. The reason is because the vast majority of machines out there do not support SSE4 because it's coming new with Penryn. Software usually lags behind hardware in terms of advancement. If you get a Penryn notebook, expect about 1-8% during normal usage, which honestly, for normal tasks, it's barely neglegible. For Intel, Penryn is nothing more than a "refresh" of the Core 2 line. If we see any serious breakthroughs it won't be until the end of the year or beginning of 2009 with Nehalem. Penryn's advantages come more from it's battery life.

[aliquis-]

Originally Posted by MacRumors

With just the processor change alone, the new Penryn laptop offered 5-10% more battery life on their benchmarks. Meanwhile, the new processor saw 1 - 8% speed boosts on common tasks, and up to 40% improvements in applications that support the SSE4 instruction set.

Who cares about processors, it's all about custom chips anyway, MHz myth and race suck

What really suck in the MBP is that I got 128MB vram, which is ****ing retarded. The 1440x900 resolution is lame, TN-panel sucks ass, what is this bright spot which seems to be behind my pixels?!, bigger and 7200 rpm harddrive would be nice aswell.

256MB vram, IPS-panel and 1680x1050 would be nice, a faster CPU isn't.

Originally Posted by CWallace

Here is also hoping Photoshop adds SSE4 support for common filters.

Since Adobe is to lazy to even type the ****ing filenames right so CS 3 works on HFSX+ I wouldn't count on or hope on to much .. But it should be done.

Originally Posted by Rocketman

To me this chip revision is more about the 45 nm technology, with its associated chemistry change, and an incremental improvement in power management, and the SSE4 language for bleeding edge apps, which do not effect most people, are the geek-kicker. Pro apps.

SSE4 will in deed be nice where it makes a difference, thought we already have SSE, SSE2 and SSE3 and it's not like it makes a huge difference on average per generation.

Originally Posted by CRIIM44

i know what penry is. Though santa rosa chips are already 45nm and use the high k method that im guessing was carried over to penry as well

I doubt that, I'm very confident they are 65nm.

Originally Posted by rikers_mailbox

What OS? Windows Vista I presume?
Is Vista fully 64bit?
Is Vista optimized to handle multiple processors/cores?
Can it throttle processor utilization and power consumption?

Sorry, I'm fulla questions and sorta feel like 10.5.2 will address some of these issues.
Penryns will show even better improvement with OS X.

Windows?
Probably
Yes, I think so.
Of course it is.
Yeah.

OS X whatever may be compiled with support for SSE4 instructions I guess. I don't know if one can compile with support for it without requiring it thought. I know compile stuff with GCC for a lowest common instruction set say i386 and then add optimizations for a higher one say i686 where available. I'm not sure the same can be done for SSE instructions. I guess it should be possible but if not this won't help much at all.

Originally Posted by David G.

What applications/types of applications take advantage of this?

Whatever stuff will have use for the new SSE4 instructions =P, SSE are mostly used for graphics and gaming stuff and such, as someone have already said.

Originally Posted by Stridder44

>45nm chips means less heat yes (among other things)?

As always I guess.

Originally Posted by ChrisA

The SSE instructions are of most use to software that process large streams of audio or video media. Doing things like changing color spaces, scaling pixels or encoding a ripped CD to MP3

Apple might modify Core Image or other "core" libraries to use SSE4 and then software that uses these libraries would be able to take advantage without need to be changed. Right now I think Apple is the biggest user of these libraries with programs such as Preview, iPhoto, Aperture Garage Band and so on. I think Adobe uses their own image processing code.

Use of SSE4 in software would have to wait until the SSE4 hardware is widely available or else how could Apple run a beta test?

Do you know if one can build binaries for whatever SSE generation instruction set but with additional optimizations for a higher generation one if availble with no extra work so to speak?

Originally Posted by theman

3 hours... I get 5.

I get around 1 thanks to flash ****** performance on OS X. I always tens of tabs open in my browser and that makes my load around 100% since flash suck ass.

Originally Posted by shamino

They are used extensively by many parts of Mac OS X, including the desktop and related UI components.

Additionally, any code using the Accelerate framework (instead of direct SSE calls), should be able to take advantage of the new capabilities, as soon as Apple releases an update with the capability. (For those who don't know, the Accelerate framework wraps a wide variety of SIMD-type operations, so code can use them on both PPC and Intel systems, mapping to either AltiVec or SSE instructions, as appropriate.)
Not necessarily. Apps that currently use the Accelerate framework should be able to take advantage as soon as Apple updates Mac OS X. They shouldn't have to be recompiled.

Code that directly makes SSE/2/3 calls, of course, will have to be updated, but there may not be that many apps in this category. Accelerate has existed since Mac OS X 10.3, and Apple has been encouraging its use since then. I suspect that Adobe will be one of the few major app-suppliers that will need to update their code for SSE4.
Mac OS uses the Accelerate framework for those subsystems that use SSE. They should all start using the new instructions as soon as Apple updates the framework.

Well, doesn't that require new accelerate instructions in many cases? =P So same problem anyway
Some new instructions may be used by old functions I guess.

Originally Posted by gnasher729

Nothing right now. There are no Macs available yet with these processors (I guess the first reader receiving a MacPro will post here).

These things take time. SSE4 has two major components:

One is a more complete set of vector operations, which makes automatic vectorisation possible. What that means: Instead of the programmer having to write code specifically for the vector unit, he or she writes ordinary code and the compiler translates it into vector instructions. However, that fails when something that the code does is not available as a vector instruction. SSE4 is much more complete, so a lot more code can be translated automatically. That is something that applications like Photoshop would benefit from.

The other component is highly specialised instructions for video encoding. The most time consuming part of video encoding by far is "motion prediction", where the encoder takes a tiny bit of one frame and tries to find a similar image in a previous frame. For example, there is one instruction that does the following: Lets say you have four pixels abcd. And another eleven pixels ABCDEFGHIJK. This instructions calculates the difference between a and A, b and B, c and C, d and D and adds them up. Then it calculates the difference between a and B, b and C, c and D, d and E. Then between a and C, b and D, c and E, d and F and so on. The result is eight sums of four differences, and each of these 32 differences is the absolute value of the difference between two values. Without a vector unit, that would be 32 subtractions, 32 absolute values, and 24 additions - 88 operations in total. Probably a dozen vector instructions without SSE4. With SSE4 it is one instruction. That makes motion prediction a lot, lot faster.

You will benefit from this eventually, but not right now.

I can see use for the later part. Is that a part which is accelerated in GPU for H.264 or not? Because if it is I see no use for it, if it is it will probably make a huge difference. Thought I doubt the amount of instructions are dozens vs 1 since they said themself 40% when using SSE4. May be an average and not a best of SSE4 instructions thought.

Originally Posted by tristan

That's like comparing a dude who graduated from h.s. yesterday with one who just started college today and wondering why the college kid isn't a lot smarter. Expect the 45nm processors to get smaller and faster and cheaper in the next coming months. Also, the next gen of CPUs wouldn't even be possible without 45nm. Give the mfg process some time to mature.

Smaller? Why? Because they will use less transistors? Yeah right .. They will get smaller with next manufacturing advancement.
Faster? Yeah, if heat allows it, which it probably does. Say 2.8 and eventually 3 GHz.
Cheaper? Then the old series? Doubt so, cheaper than they are as new? Yes, of course, but Apple doesn't lower prices anyway.

Originally Posted by RichP

I think that battery life improvement is quite impressive. Lets face facts..the screen and the hard drive are the big power eaters.

You have to realize, a 5-10% OVERALL improvement means the processor is actually 20-40%? more efficient..as it is not the only thing using power in the system.

During idle load yes... During full load I assume the CPU to take a much bigger chunk.

Originally Posted by acslater017

i'm not really familiar with this...does anyone know what applications support this instruction set? do any of the major Apple apps?

None now of course, if Apple haven't planed for it for Leopard or whatever. With time hopefully anything which can get an improvement from it.

[tristan]

Originally Posted by aliquis-

Smaller? Why? Because they will use less transistors? Yeah right .. They will get smaller with next manufacturing advancement.
Faster? Yeah, if heat allows it, which it probably does. Say 2.8 and eventually 3 GHz.
Cheaper? Then the old series? Doubt so, cheaper than they are as new? Yes, of course, but Apple doesn't lower prices anyway.

Yeah, why did I say smaller? Maybe my brain is smaller after waiting so long for Penryn. They say it shrinks with age.

What i meant to say is that as the mfg process ramps up Intel gets better yields and greater production capacity and is then able to introduce new speeds, cut prices, and bring out other products based on that manufacturing technology. We as Apple customers probably won't get an immediate price decrease when it happens - instead we get speed bumps, new products, etc. So 45nm will probably look better in 6 mos is what I really mean.

[dante@sisna.com]

These are not compelling performance numbers to me.

1 to 8%. I have seen these estimates before and this seems accurate.

This is not a noticeable performance increase.

Glad I bought the MBP 2.6 17" high res last month when I needed it.

I think MacWorld will be interesting, can't wait, but my gut tells me it will be interesting for many products besides the MBP update.

Just my gut. I could be wrong on this.

[winterspan]

I made this for myself but I thought I'd share to help people out..

[shamino]

Originally Posted by winterspan

I made this for myself but I thought I'd share to help people out..

This is overly simplified. For instance, the Santa Rosa Centrino platform has been (or soon will be) refreshed to support Penryn chips.

Laptop designers are not going to have to wait for Montevina in order to use a Penryn processor.