Intel’s Pentium M 760 versus AMD’s Turion 64 ML-44

AMD’s answer to the Pentium M is the Turion 64, an Athlon 64-based design which rides in a 754-pin socket. The Turion 64’s attributes include an on-die memory controller, a 1GHz HyperTransport link, and of course the ability to run 64-bit code. The model we’ll be looking at today is the ML-44, a new chip with a 2.4GHz operating frequency and 1MB of L2 cache.

We have been watching the mobile processor space for a while now with some interest, wondering whether the Turion 64 really matches up well against the Pentium M. The Pentium M’s power-saving features and performance per watt are the stuff of modern legend. AMD has done well with Opteron in servers and the Athlon 64 in desktops, but surely AMD’s K8-derived mobile competitor doesn’t match up with the likes of the Pentium M. Does it?

Well, we wanted to find out before the game changes again. Intel’s recently announced Core Duo marks the beginning of a transition to multicore processors in the mobile space, and AMD has stated its intention to make a dual-core version of the Turion, as well. Core Duo laptops are just starting to arrive on the market, although volumes are still somewhat limited. With that in mind, let’s take a quick look at the two most widely available mobile processors today: the Pentium M versus the Turion 64.

The combatants
We’ll start off with an up-close and personal view of the two contenders:

Being mobile chips, neither the Pentium M 760 nor the Turion ML-44 has the protective “heatspreader” cap that’s become standard on desktop processors. Instead these processors look more like an Athlon XP or an old-school Pentium III.

The Turion 64 ML-44
The ML-44 is the latest addition to AMD’s Turion line of mobile processors. On the surface, the Turion has much in common with Socket 754 Athlon 64 chips. It has an on-die memory controller capable of addressing a single channel of DDR400 memory, as well as the ability to run 64-bit code—not even the new Core Duo can do that. Depending on model, the Turion features either 512KB or 1MB of L2 cache. Turions also feature support for the NX bit, which can help prevent malicious exploits based on buffer overflows.

Being a mobile line, low power consumption is a key feature of the Turion 64. All Turions support AMD’s PowerNow! technology, which dynamically scales clock speed and voltage over a wide range in response to CPU loads. Additionally, although Turion 64s bear more than a passing resemblance to the Socket 754 Athlon 64, the Turion can operate at significantly lower voltages relative to desktop chips, which cuts down significantly on power consumption.

Finally, AMD has segmented the Turion 64 into two model lines, one focusing on performance, the other on battery life. This product segmentation shows itself in the TDP—or thermal dissipation rating—of the two lines: The ML line of Turions has a TDP of 35W, while the MT line has a TDP of 25W.

The ML-44 we’re testing today is AMD’s fastest model, featuring 1MB of L2 cache and a maximum operating frequency of 2.4GHz. With PowerNow! enabled, the system will scale the processor’s operating frequency from 2.4GHz all the way down to 800MHz depending on CPU load. The voltage scales from 1.0V at 800MHz to 1.35V at 2.4GHz.

The motherboard we’re using for Turion 64 testing is MSI’s RS482M-IL. This microATX motherboard is based on ATI’s Radeon Xpress 200M mobile chipset, and its integrated graphics should be similar to those on a Turion 64 laptop with the same chipset.

The Pentium M 760
The Pentium M 760 is at the upper end of Intel’s mobile line, and is an example of how the Pentium M has evolved over time. While the first Pentium M processors had 1MB of L2 cache and ran on a 400MHz bus, newer examples such as the 760 have a giant 2MB of L2 and a 533MHz front-side bus. Finally, like the Turion 64, the newer Pentium M chips add support for the NX bit.

In addition to these changes, memory bandwidth has improved considerably over time thanks to newer chipsets. The first chipsets could only address a single channel of DDR333, putting the Pentium M at a disadvantage relative to the competition. The newer 915 chipset we’re looking at today, however, can address dual channel DDR2 at 533MHz, a huge increase in memory bandwidth. While some of this bandwidth will likely go untapped thanks to the relatively slow 533MHz front-side bus, it’s better to have an excess of memory bandwidth, especially when using on-board graphics with shared memory.

The Pentium M 760 has a maximum operating frequency of 2.0GHz. Using Intel’s SpeedStep technology, the system will dynamically clock the processor anywhere from 800MHz up to the maximum 2.0GHz. The voltage scales from 0.714V at 800MHz to 1.315V at 2.0GHz.

Unlike the Turion 64, the Pentium M can’t run 64-bit code. Given that most laptops can’t accept more than 4GB of memory, that limitation may not be much of an issue for the Pentium M. 64-bit computing does have tangible benefits in certain scenarios, but those arguably won’t matter for mobile computing for some time yet.

For our tests, the Pentium M 760 will be sitting in an AOpen motherboard, the i915Ga-HFS. This ATX board features the Intel 915GM mobile chipset. Like the MSI board for the Turion, it should give us a reasonable approximation of a laptop based on the same chipset, in terms of both power consumption and graphics power. Not only that, but both this Pentium M mobo and our Turion 64 motherboard are strong candidates for use in quiet computing applications like home theater PCs, as well.

The match-up
Ours is not a perfect world, and we acknowledge that this is not a perfect match-up, at least not in terms of price. The problem is that AMD’s Turion ML-44, priced at $354 by AMD, lands in the middle of a large gulf in Pentium M pricing. Our two closest options for comparison were the Pentium M 760 at 2.0GHz, with a price of $294, and the Pentium M 770 at 2.13GHz and a price of $423. The $60 price difference between the 760 and the ML-44 was slightly less than that the $69 separating the 770 from the ML-44, so we based our comparison on the former.

Our testing methods
While we ran the majority of tests with integrated graphics, we wanted to get an idea of how these systems would respond with a discrete mobile graphics solution. Therefore, the gaming benchmarks were performed both with the integrated graphics as well as an NVIDIA 6600GT plugged into each board’s PCI-E x16 slot.

Because these are mobile processors, we tested with power management features like SpeedStep and PowerNow! enabled throughout our benchmark suite, with the lone exception of CPU-Z’s memory access latency test. That test reports its results in terms of CPU cycles, and it could be thrown off by processor clock speed changes.

As ever, we did our best to deliver clean benchmark numbers. Tests were run at least three times, and the results were averaged.

Our test systems were configured like so:

Thanks to Crucial for providing us with memory for our testing. Their products and support are both far and away superior to generic, no-name memory.

Also, all of our test systems were powered by OCZ PowerStream power supply units. The PowerStream was one of our Editor’s Choice winners in our latest PSU round-up.

The test systems’ Windows desktops were set at 1024×768 in 32-bit color at a 75Hz screen refresh rate. Vertical refresh sync (vsync) was disabled for all tests.

We used the following versions of our test applications:

• SiSoft Sandra 2005 SR3 10.10.69
• CPU-Z 1.31
• Compiled binary of C Linpack port from Ace’s Hardware
• POV-Ray for Windows 3.6.1
• Cinebench 2003
• LAME MT 3.97a
• Windows Media Encoder 9
• Sphinx 3.3
• picCOLOR 4.0 build 563
• Unreal Tournament 2004 v3369 with trdemo1
• 3DMark05 v120
• WorldBench 5.0

The tests and methods we employ are generally publicly available and reproducible. If you have questions about our methods, hit our forums to talk with us about them. Update 02/15/2006: After the publication of this review, two issues with the testing were found and corrected. See this link for more information.

Memory performance
These synthetic tests don’t always mirror real-world performance, but they can tell us some interesting things about the CPUs and their memory subsystems, so we’ll start here.

With double the number of channels and 33% higher memory speed than the Turion, it’s no surprise that the Pentium M wins handily here. In fact, it’s likely that the disparity would be significantly larger were it not for the Pentium M’s pokey 533MHz bus.

This Linpack benchmark is always useful for showing the size and speed of a processor’s caches and the speed of its main memory. Unfortunately, running this test on the Pentium M is sort of like sending Joe Montana into a high school football game—it’s just not fair. With a massive L2 cache that spans the entire range of test matrix sizes, the 760 blows away the Linpack test, then makes rude references to its parentage involving hamsters and elderberries.

While the Pentium M definitely has the edge in terms of raw memory bandwidth, you don’t tug on Superman’s cape, and you don’t mess around with an on-die memory controller. The Turion crushes the Intel chip on this one, with access latencies nearly half those of the Pentium M.

Gaming performance

Unreal Tournament 2004
Reasoning that newer games like F.E.A.R. are going to have the frame rate of a PowerPoint presentation on integrated graphics, we stuck with older benchmarks on the gaming side of things. As a reminder, on this test we benchmarked not only with the integrated graphics but also with a GeForce 6600GT card.

The Turion ekes out a win here, although neither integrated solution even makes it to 30 frames per second. More interesting, perhaps, is the fact that, graphics being equal, the Pentium M comes in ahead of the Turion 64. That’s right, Intel wins a gaming benchmark! This is not your father’s Netburst chip.

3DMark05
As with UT2004, we tested 3DMark05 with both integrated graphics and the 6600GT.

Neither integrated solution is, err. . . competitive in 3DMark05, but that’s hardly shocking. If the integrated graphics are in the cellar, however, the ATI Radeon Xpress 200M is at least halfway up the stairs relative to the Intel GMA 900, for whatever that’s worth.

As with UT2004, however, the Pentium M comes out ahead of the Turion 64 when integrated graphics is taken out of the equation.

The individual gaming tests mimic the overall score, with the Turion 64 system with ATI graphics consistently beating the Pentium M with Intel graphics. With the discrete GeForce 6600 GT, The Pentium M outruns the Turion in all tests, though the margin of victory varies considerably.

While the Pentium M takes the gaming tests, the Turion roars back to win (or in one case tie) all of 3DMark’s CPU tests.

WorldBench overall performance
WorldBench uses scripting to step through a series of tasks in common Windows applications and then produces an overall score for comparison. More impressively, WorldBench spits out individual results for its component application tests, allowing us to compare performance in each. We’ll look at the overall score, and then we’ll show individual application results alongside the results from some of our own application tests.

The Turion 64 ekes out a win here, though it’s not a huge win on a percentage basis.

Audio editing and encoding

LAME MP3 encoding
We have results for two different versions of LAME from different compilers, one from Microsoft and one from Intel, doing two different types of encoding, variable bit rate and constant bit rate. We are encoding a massive 10-minute, 6-second 101MB WAV file here, as we have done in our previous CPU reviews.

The Turion 64 goes four for four in the LAME benchmarks, though the Pentium M at least stays competitive.

MusicMatch Jukebox

Although the overall WorldBench score showed a modest win for the Turion, there’s still plenty of drama in the individual tests. Here the Turion 64 hands the Pentium M 760 a convincing defeat.

Video editing and encoding

Windows Media Encoder Advanced Profile
We asked Windows Media Encoder to convert a gorgeous 1080-line WMV HD video clip into a 320×240 streaming format using the Windows Media Video 8 Advanced Profile codec.

Windows Media Encoder

Adobe Premiere

VideoWave Movie Creator

The margin of victory varies, but the Turion 64 beats the Pentium M across the board in our video editing tests.

Image processing

Adobe Photoshop

ACDSee PowerPack

Though the victories aren’t as convincing as some of the video tests, the Turion 64 still wins both of the WorldBench image processing tests.

picCOLOR
picCOLOR was created by Dr. Reinert H. G. Müller of the FIBUS Institute. This isn’t Photoshop; picCOLOR’s image analysis capabilities can be used for scientific applications like particle flow analysis. Dr. Müller has supplied us with new revisions of his program for some time now, all the while optimizing picCOLOR for new advances in CPU technology, including MMX, SSE2, and Hyper-Threading.

Scores in picCOLOR, by the way, are indexed against a single-processor Pentium III 1GHz system, so that a score of 4.14 works out to 4.14 times the performance of the reference machine.

The picCOLOR test is tight; the two chips trade positions in the individual tests. When the dust settles, however, the Pentium M 760 comes out on top.

Multitasking and office applications

MS Office

WorldBench’s Office test involves switching between the various components of the Office suite, which are all running at once. This one is basically a tie.

Mozilla

Mozilla and Windows Media Encoder

Apparently after the tie in MS Office, the Turion 64 wanted to make a statement, and that’s exactly what it did on both of these tests.

Other applications

Sphinx speech recognition
Ricky Houghton first brought us the Sphinx benchmark through his association with speech recognition efforts at Carnegie Mellon University. Sphinx is a high-quality speech recognition routine. We use two different versions, built with two different compilers, in an attempt to ensure we’re getting the best possible performance.

Call this one a tie. Move along, nothing to see here.

WinZip

Nero

It’s not a huge margin of victory, but the Pentium M manages to add a couple of checks to the win column.

3D modeling and rendering

Cinebench 2003
Cinebench measures performance in Maxon’s Cinema 4D modeling and rendering app.

The Turion 64 looks like the better processor for 3D rendering, at least with Cinema 4D.

The most notable result here is the Pentium M’s commanding lead in the hardware OpenGL test. This may be the due to problems with ATI’s OpenGL drivers.

POV-Ray rendering

The Turion 64 shows its floating point prowess in this rendering test, as well, beating the Pentium M by a significant margin.

SiSoft Sandra
Next up is SiSoft’s Sandra system diagnosis program, which includes a number of different benchmarks. The one of interest to us is the “multimedia” benchmark, intended to show off the benefits of “multimedia” extensions like MMX and SSE/2. According to SiSoft’s FAQ, the benchmark actually does a fractal computation:

This benchmark generates a picture (640×480) of the well-known Mandelbrot fractal, using 255 iterations for each data pixel, in 32 colours. It is a real-life benchmark rather than a synthetic benchmark, designed to show the improvements MMX/Enhanced, 3DNow!/Enhanced, SSE(2) bring to such an algorithm.

The benchmark is multi-threaded for up to 64 CPUs maximum on SMP systems. This works by interlacing, i.e. each thread computes the next column not being worked on by other threads. Sandra creates as many threads as there are CPUs in the system and assignes [sic] each thread to a different CPU.

The “Integer x16” version of this test uses integer numbers to simulate floating-point math. The floating-point version of the benchmark takes advantage of SSE2 to process up to eight Mandelbrot iterations at once.

The Pentium D owns this test among desktop processors, but the crown switches hands in the mobile world. The Turion 64 bests the Pentium M on both the integer and floating point tests.

Power consumption
We measured the power consumption of our entire test systems, except for the monitor, at the wall outlet using a Kill A Watt watt meter. The test rigs were all equipped with OCZ PowerStream 520W power supply units. The idle results were measured at the Windows desktop, and we used Prime95 to load up the CPUs.

As I mentioned in our testing methods section, both processors were tested with their power-saving features—SpeedStep for the Pentium M and PowerNow! for the Turion 64—enabled. Because of a bug in the BIOS of the MSI motherboard used for the Turion 64 tests, RMClock was used on that configuration to match the maximum voltage to the official spec of the processor.

The power consumption tests are actually very interesting. While it’s true that the Pentium M beats the Turion 64 by a substantial margin under load, the fact is that typical laptop usage leaves the system at idle the vast majority of the time. At idle, the Turion 64 system actually comes in with slightly lower power consumption than the Pentium M rig. This result suggests that, depending on usage patterns, the Turion could prove a worthy alternative to the Pentium M in terms of battery life.

Conclusions
From a performance perspective, it’s clear that the Turion 64 is the winner. By my count, the Pentium M was faster in only five of the tests, and one of those (the hardware OpenGL test in Cinebench) was probably due to graphics drivers. The rest were either a toss-up or a win for the Turion 64. The other thing that struck me about the results was that even in the tests the Pentium M did win, its margin of victory was fairly small. A number of the Turion 64 wins, however, were by an impressively large margin.

From a power consumption perspective, the Turion 64 surprised me. Yes, our Turion 64 test system consumed 19W more power than the Pentium M system at 100% CPU load, but unless you’re using your laptop to crunch that Folding@Home work unit on the plane, maximum power consumption isn’t usually all that important. For typical use, it seems likely the Turion 64 would be reasonably competitive with the Pentium M on the battery life front, as well.