Penryn notebook chips offer marginal battery and speed boosts

[AppleInsider]

A new family of Penryn-based Core 2 notebook chips formally introduced by Intel on Monday boost battery performance by as much as 16 percent and encoding operations by as much 40 percent when pitted against their Merom predecessors in a set of mobile benchmark tests.

The five new chips -- detailed previously -- largely run at clock frequencies available from the Intel's current Merom-based notebook processors but leverage improvements from the chipmaker's new 45-nanometer (nm) transistor technology to lower voltage, which directly translates into reduced power consumption and longer notebook battery life.

In its quest to gauge the battery performance of Penryn-based notebook systems, AnandTech landed two Dell Latitude D630 notebooks from Intel, identically configured with the exception of the processor. One employed a Core 2 Duo T7800 -- the Merom-based chip currently offered as a build-to-order option on Apple's high-end MacBook Pro systems -- running at 2.6GHz with an 800MHz font-size bus (FSB) and a 4MB Level 2 cache. The other came equipped with a brand new 45nm Penryn Core 2 Duo T9500 -- which will make its way into Apple's upcoming MacBook Pro refresh -- also running at 2.6GHz on a 800MHz FSB but with a larger 6MB L2 cache.

When run on both machines, the Windows-based CPU-Z system profiler application indicated that the Penryn-based notebook was running at 1.15V compared to the 1.225V of the Merom-based system. This reduction in voltage for the Penryn notebook immediately signals reduced power consumption and, hence, improved battery life.

To quantify the battery improvements, AnandTech turned to MobileMark 2007, the latest battery life measurement tool from BAPCo.*The benchmarking suite consists of three individual benchmarks: a Productivity 2007 test, a Reader 2007 test, and a DVD playback test.

In effect, the Productivity 2007 test emulates a SYSMark 2007 Productivity test, but does so with a focus on how long the battery will last, not how fast the system can perform.*In this test, the Penryn-based notebbok lasted an extra 55 minutes, or an increase of almost 16.5 percent in battery life over the identically configured Merom system.

The results of the Reader 2007 test, which simulates mostly idle time when a notebook user would be reading a digital document, were similar but more marginal. In this test, battery performance improved 5.6 percent or a total of 23 minutes on the Penryn notebook compared to the older Merom-based system.

Mobile Penryn vs. mobile Merom battery tests | Source: AnandTech.

Again, and as would be expected, the Penryn notebook outperformed the Merom system in a DVD playback test, offering up to 18 minutes longer battery life for an approximate 7 percent increase in battery performance.

As noted by AnandTech, these increases in battery life will come at no additional charge to the end user, as Intel has priced the new mobile Penryn chips in line the the Merom-based models they replace. PC manufacturers like Dell and Apple are expected to adopt the chips as part of their mid-winter notebook refreshes, which will similarly arrive at price points that are either in line or cheaper than the price points of the systems they replace.

In general processor performance, the Penryn-based notebook was also marginally faster than the Merom-based system, boasting a MobileMark Productivity 2007 score that was 6.4 percent higher. Results of several SYSMark 2007 tests yielded similar results, with the Penryn system besting the Merom system by a range of 1 to 8 percent depending on the test.

Where the mobile Penryn really outshines its predecessor is in support for applications written to take advantage of Intel's Supplemental Streaming SIMD Extension 4 (SSE4) instruction set, which is making its debut as part of Intel's broader Penryn platform.

Mobile Penryn vs. mobile Merom battery tests | Source: AnandTech.

Running Intel's TMPEGEnc benchmark, which does take advantage of SSE4 during a movie transcode, AnandTech was able to witness a wide performance gap. The Penryn-based system, because of its SSE4 support, completed the encoding test in 31 percent less time than the Merom system that lacked such support.

This performance gap reported widened even further to 40 percent when a SSE4-optimized VirtualDub 1.7.2 DivX encode was run. Unfortunately, these sorts of heavily tuned applications are expected to be few and far between for the considerable future.

Apple is widely expected to announce its own plans for implementing Penryn-based mobile chips in its MacBook lines next Tuesday at the Macworld Expo in San Francisco. The chips, however, are not yet readily available, meaning any associated Apple products would not ship until several weeks later.

[ View this article at AppleInsider.com ]

[mdriftmeyer]

Programming Reference Library from Intel on what SSE4 adds:

http://softwarecommunity.intel.com/i...0Reference.pdf

[mciarlo]

"Apple is widely expected to announce its own plans for implementing Penryn-based mobile chips in its MacBook lines next Tuesday at the Macworld Expo in San Francisco. The chips, however, are not yet readily available, meaning any associated Apple products would not ship until several weeks later."

I need a laptop for school and unfortunately cannot wait several weeks. I hope this isn't true...

[Splinemodel]

Originally Posted by mciarlo

"Apple is widely expected to announce its own plans for implementing Penryn-based mobile chips in its MacBook lines next Tuesday at the Macworld Expo in San Francisco. The chips, however, are not yet readily available, meaning any associated Apple products would not ship until several weeks later."

I need a laptop for school and unfortunately cannot wait several weeks. I hope this isn't true...

If you're really dead-set on Penryn, get a clunker off eBay that will last a few months. You can probably sell it for about what you will pay. Most schools have plenty of clusters, too. Honestly, I think I could have gone without a computer while I was at college, although that's become a longer and longer time ago.

As for me, I see a 2.5 GHz iMac in my future. I have a nice space on the desk for it next to my iMac G5, which will be relegated to server and occasional 3D modeling duties (until I decide to shell out $4000 for Intel compatible versions of the 3D software I use).

Yeigh.

[mciarlo]

Well, I will be doing a lot of design, programming, and possibly video editing. If the battery life and performance gains are accurate, then Penryn is what I hope to get. I would just hate to purchase a new MBP knowing in a month it will be updated.

Thus is the wave of technology.

[thebeat]

i also need this new penryn laptop for school, can't believe its been taking this long. I am in to 3rd semester and this is the semester the school REQUIRES me to get one before February. This is very bad.

[bobmarksdale]

I would say that 16-40% is a little more than marginal...

[xander0985]

Wait 'til Tuesday, see what happens and buy from there. At the very least, you'll get a better deal on a now-priced-down MBP if you must have it by Feb 1st. Just don't buy now if you don't need it tomorrow. See what happens after MacWorld and don't loose any sleep over it. Enjoy your winter break.

[monstrosity]

is it me or has the speed increase of processors really started to slow down lately?

I'm sure the old IMB chips used to double with each release. Or certainly kept up with moores law.(except towards the end when they were frankly rubbish)

Whats the score with mores law nowadays, it still relevant?

[Splinemodel]

Originally Posted by monstrosity

is it me or has the speed increase of processors really started to slow down lately?

I'm sure the old IBM chips used to double with each release.

The advance of manufacturing techniques has slowed down to the extent that, to maintain a given power dissipation (wattage), speed is not doubling. However, the number of transistors per chip is still holding to "Moore's Law," which actually doesn't specify anything other than that, the number of transistors per chip.

[fstop]

Originally Posted by bobmarksdale

I would say that 16-40% is a little more than marginal...

The article was probably written by Dvorak...

[Vox Barbara]

Originally Posted by mciarlo

Well, I will be doing a lot of design, programming, and possibly video editing. If the battery life and performance gains are accurate, then Penryn is what I hope to get. I would just hate to purchase a new MBP knowing in a month it will be updated.

Thus is the wave of technology.

As a rule of thumb: buy when the need is there. If you need it NOW,
buy it NOW. If you just spent some time considering to buy, then the
need to buy is not that strong. That simple.

best

[Vox Barbara]

Originally Posted by bobmarksdale

I would say that 16-40% is a little more than marginal...

Yes I was just mull over the very same question. So what is
the definition of marginal? A marginal increase in something
(in example performance) is something you may or may not perceive,
is it not? But 16-40% performance increase is pretty much something
you'll perceive. Imho. However I actually doubt that even 40% performance
increase has a noticeable impact concerning the Finder issue . Anyway.

Originally Posted by fstop

The article was probably written by Dvorak...

Also true. At least Dvorak doesn't write run of the mill stuff, though he is very
prone to run off the rails. My take.

best

[aplnub]

How much better is battery life and performance vs. the first MBP CD?

[g3pro]

Who used the word "marginal" needs to look up the definition of the word before trying to sound smart by including in the article?

[JeffDM]

Originally Posted by bobmarksdale

I would say that 16-40% is a little more than marginal...

But that's the max. Typical battery life improvement looks to be about 5%, which might not be noticeable. To get the performance gains of SSE4, you need to have a program that can use that feature. As it is, from numbers people tell me on other forums, Mac media encoding software in anything other than FCP's Compressor (and maybe some other high-end software) is pretty sluggish. Quicktime's encoding speed left a lot to desire for me. I think there's plenty of speed to be gained with better programming before having to use SSE4. Software that doesn't even use it won't get that speedup.

[freakboy]

Don't you guys feel just a little bit guilty regurgitating someone else's hard work? Maybe just put a link to their article with a few sentence summary of what they found. Including their figures, etc. Makes you guys look really bad.

[Programmer]

Originally Posted by monstrosity

is it me or has the speed increase of processors really started to slow down lately?

Process technology really ran into a wall at about 90nm, which was 5-6 years ago. The clock rate free ride ended rather abruptly, and at that point Intel started its course change from the Pentium4 to the Core series (basically clock rate vs. power efficiency).

I'm sure the old IMB chips used to double with each release. Or certainly kept up with moores law.(except towards the end when they were frankly rubbish)

Whats the score with mores law nowadays, it still relevant?

Moore's Law is a statement about the rate at which the number of transistors on a chip increases over time. From the time Moore noticed this pattern of improvement (in the 70s, I think?) until about 2002 this translated fairly directly to proportional clock rate improvements, and those translated roughly proportionally to performance improvements. And then it start to translate to super-linear power consumption increases too, and that's where things went off the rails. Basically at 90nm and smaller sizes the rate at which power leaked from the on-chip traces between the transistors at high clock rates began to increase very dramatically and that became the limiting factor. Woops.

Since then the emphasis has been two-fold. First, do as much as you can as efficiently as you can each clock cycle. Second, have more than one core running in parallel. Unfortunately parallelism impacts the software, and software is slow to change and thus takes a while to catch up with the new hardware reality. It doesn't help that the problem of concurrent software is really hard.

Hope that helps. Don't expect a return to the old glory days any time soon. Or maybe ever.

[Marcus]

I currently use a 2.16 Core Duo MBP with the bto 256mb Radeon X1600 (2 gig RAM). Only issue is, it is looking *really* tatty, and starting to play up.

I was planning to use it until it died, but beginning to wondering how much quicker these Penryn chips would be compared to the original Core Duo...

[McDave]

Originally Posted by JeffDM

But that's the max. Typical battery life improvement looks to be about 5%, which might not be noticeable. To get the performance gains of SSE4, you need to have a program that can use that feature. As it is, from numbers people tell me on other forums, Mac media encoding software in anything other than FCP's Compressor (and maybe some other high-end software) is pretty sluggish. Quicktime's encoding speed left a lot to desire for me. I think there's plenty of speed to be gained with better programming before having to use SSE4. Software that doesn't even use it won't get that speedup.

A similar argument to not obsessing about the graphic chipset too much - GPUs aren't really engaged in decode/encode functions except for corevideo special effects playback.

Quicktime is probably the prime suspect and maybe prior knowledge of SSE4 is why Apple have kept the encoding CPU-centric. If you pause/unpause an AppleTV it gives a hint why they've kept decoding on the CPU and not shifted to GPU. That should speed up practically all video & audio encoding and probably a lot of the playback with a single software update.

Intel gets it's Altivec at last

McD