Asus has decided to do something about that requirement in rather inventive fashion. The CT-479 is an adapter card that plugs into the processor socket of certain Asus Socket 478 motherboards and converts them to a Socket 479 configuration capable of accepting a Pentium M. This adapter neatly solves both the price and performance issues, the latter by supplying a dual-channel memory configuration unavailable in the Intel 855GME laptop chipset.
How well does the CT-479 adapt the Pentium M for desktop use? We’re about to find out.
A look at the CT-479
Asus sells the CT-479 separately, allowing you to choose the motherboard for yourself. A list of the motherboards currently supported can be found here. Since support for the CT-479 requires a BIOS upgrade, it’s unlikely that you’ll be seeing CT-479 support on any non-Asus motherboards in the future.
The BIOS upgrade could prove to be an interesting problem. If your board doesn’t have a recent BIOS version, you will likely need a Socket 478 CPU long enough to update the BIOS.
Here is a view of the bottom of the CT-479 board. There’s not much to see here except 478 pins. (Uhh, yeah, I counted them…) The board itself is pretty small, since it has to fit inside the Socket 478 heatsink retention bracket and can’t protrude past the arm on the motherboard socket.
The view from the top is a little more interesting. In addition to the laptop socket, there is a floppy-style power connector for the circuitry on the board, as well as jumpers to choose between a 400MHz and 533MHz front-side bus. The kit includes a Molex to floppy adapter cable as well.
The CT-479 also comes with its own heatsink. It’s considerably lighter and smaller than most Northwood coolers, let alone the gargantuan Prescott models. The cut-out allows space for the jumpers and power connector on the CT-479.
Installing the CT-479 is a piece of cake. Lift up the arm on the motherboard’s socket, and the CT-479 just drops right in:
I left the arm up to give you a better idea how the board fits into the confines of the socket area. Once you’ve lowered the arm, you’ll need to install a Pentium M or Celeron M processor into the CT-479 socket, and then install the heatsink.
Here’s the finished install. The heatsink clamps hook into the same holes used by a Socket 478 heatsink.
The majority of the overclocking fun takes place on the page shown below. You can set front-side bus speed, CPU multiplier (this is supported on the Pentium M but not the Celeron M, and the multiplier can only be adjusted down from its original setting) and lock the PCI and AGP busses. Voltage adjustments are available for CPU, DDR and AGP.
It should be noted that BIOS features are implemented not on the CT-479 but on its host motherboard. In addition, supported features vary from motherboard model to motherboard model, so you’ll want to research as best you can to ensure that any advanced features you want (such as CPU voltage control) are available in the latest BIOS of the motherboard you plan to use.
For example, Asus sent along a P4C800-E Deluxe motherboard to go with our CT-479 review unit. This is an Intel 875P-based board that certainly doesn’t skimp on the features. You can read more about it here. The first P4C800-E BIOS with CT-479 support was missing out on the ability to adjust the voltage of the CPU. The most recent BIOS (1022) took care of that problem, and took care of it well.
Unlike previous Pentium M desktop solutions we’ve seen, which capped the CPU voltage at the processor’s upper range, the P4C800-E BIOS allows the voltage to go up to 1.6V. Considering that the maximum recommended voltage for our test chip (a 2.0GHz Dothan) is 1.34V, Asus has given you plenty of rope with which to hang yourself.
A few brief notes, because things got a little complicated this time around:
Obviously we wanted to see what effect dual-channel memory would have on the Pentium M’s performance, so testing against DFI’s 855GME-MGF was a must. Since this review also marks a brief return to a socket and chipset of yesteryear, we decided to throw in the hottest chips (both figuratively and literally) ever to run in Socket 478, a trio of 3.4’s: the Pentium 4 “Northwood” 3.4GHz, Pentium 4 “Prescott” 3.4E, and the Pentium 4 Extreme Edition 3.4GHz.
We considered testing the Pentium M configurations on a 400MHz front-side bus, but due to limitations of the P4C800-E Deluxe motherboard, the Asus configuration would have been limited to DDR266 speeds. We wanted a single vs. dual-channel test with the same RAM speed on both the DFI and Asus boards. Since a 533MHz front-side bus is the future of the Pentium M, and 533MHz parts are already available, we therefore used the overclocking features of each motherboard to test our Dothan chip at speeds approximately equivalent to a Pentium M 770, which runs at 2.16GHz on a 533MHz front-side bus.
Why “approximately?” Well, the BIOS on the Asus board has some counting difficulties. Specifically, when you tell it to run at 133MHz front-side bus, it actually runs at 135. We tried turning it down to 132MHz, but the DDR333 memory speed isn’t available below 133MHz. We also tried turning it up to 134MHz, but that threw the RAM divider off. Therefore, we ran both the Asus and DFI boards with a 135MHz front-side bus and a 180MHz memory speed, which matched them up in terms of both front-side bus and memory clocks.
The Asus board’s higher CPU voltage options allowed me to push the Pentium M farther than on the DFI board, and I wound up with a maximum stable speed of 2.57GHz. The chip basically hit a wall here. 2.57GHz required only a slight voltage boost, but even a 10% overvolt wouldn’t keep the Dothan stable at 2.7GHz. I wasn’t willing to go any higher.
As ever, we did our best to deliver clean benchmark numbers. Tests were run at least twice, and the results were averaged.
Our test systems were configured like so:
|Processor|| Pentium M 755 2.0GHz@2.16GHz
Pentium M 755 2.0GHz@2.43GHz
| Pentium M 755 2.0GHz@2.16GHz
Pentium M 755 2.0GHz@2.57GHz
|Pentium 4 3.4GHz
Pentium 4 3.4GHz E
Pentium 4 Extreme Edition 3.4GHz
|Front-side bus||540MHz (135MHz quad-pumped)||540MHz (135MHz quad-pumped)|
|Motherboard||DFI 855GME-MGF||Asus P4C800-E with CT-479 adapter|
|Chipset drivers||INF Update 22.214.171.1247||INF Update 126.96.36.1997|
|Memory size||1024MB (2 DIMMs)||1024MB (2 DIMMs)|
|Memory type||Corsair XMS3200 DDR SDRAM at 360MHz||Corsair XMS3200 DDR SDRAM at 360MHz|
|Corsair XMS3200 DDR SDRAM at 400MHz|
|RAS to CAS delay||2||2|
|Hard drive||Maxtor DiamondMax 10 250GB SATA 150|
|Graphics 1||GeForce 6800GT with 71.84 drivers|
|OS||Microsoft Windows XP Professional|
|OS updates||Service Pack 2, DirectX 9.0b|
Thanks to Corsair for providing us with memory for our testing. If you’re looking to tweak out your system to the max and maybe overclock it a little, Corsair’s RAM is definitely worth considering.
The test systems’ Windows desktops were set at 1024×768 in 32-bit color at an 85Hz screen refresh rate. Vertical refresh sync (vsync) was disabled for all tests.
We used the following versions of our test applications:
- Cachemem 2.65MMX
- SiSoft Sandra 2005 SR1
- DOOM 3 1.3 with trdelta1 demo
- Unreal Tournament 2004 v3323 with trdemo1
- Far Cry v1.32
- Sphinx 3.3
- LAME 3.96.1 (build from mitiok.cjb.net)
- Xmpeg 5.0.3 with DivX Video 5.21
- Cinebench 2003
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.
The Pentium 4’s Netburst architecture and its pre-fetching capabilities win this benchmark, but remember that Sandra is a relatively synthetic measure of memory bandwidth. Remember also that the Pentium 4 configurations used DDR400 while the Pentium M was stuck at DDR333. Speaking of the Pentium M, the dual channels of DDR333 on the Asus board allowed it to post significant gains over the single-channel DFI board.
Cachemem typically presents a more realistic portrayal of memory bandwidth. Here, the Prescott’s improved pre-fetch takes top honors, but note that the Asus Pentium M configurations maintain their healthy lead over the DFI configs.
While the Asus board beats the DFI hands down in terms of bandwidth, latency is another story, as the Asus Pentium M configurations come in dead last by a significant margin.
Unreal Tournament 2004
It’s benchmarks like this that started people clamoring for a Pentium M desktop solution in the first place, as the slowest Pentium M still comes in ahead of even the Pentium 4 Extreme Edition. The added memory bandwidth of the CT-479 system benefits UT slightly at 2.16GHz. Doom 3
The Pentium M configurations extend their lead over the Pentium 4 in the Doom 3 benchmark, and once again the Pentium M 2.57GHz configuration takes the lead by a relatively large margin. Far Cry
The Extreme Edition’s 2MB of L3 cache helps it to take down at least one of the Pentium M configurations here. Memory bandwidth seems to matter quite a bit in Far Cry, as the dual-channel CT-479 system at 2.16GHz manages to outdo even the overclocked DFI configuration.
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 that needs the latest computer hardware to run at speeds close to real-time processing. We use two different versions, built with two different compilers, in an attempt to ensure we’re getting the best possible performance.
Memory bandwidth is exceedingly important to Sphinx, and it shows here. The DFI configurations and their single channel of memory simply can’t keep up with the Asus board’s dual-channel configurations. Among those dual-channel configurations, the Pentium M at 2.57GHz is fastest. LAME MP3 encoding
We used LAME to encode a 101MB 16-bit, 44KHz audio file into a very high-quality MP3. The exact command-line options we used were:
lame –alt-preset extreme file.wav file.mp3
CPU performance is everything to LAME. Note that the two Pentium M 2.16GHz configurations tied exactly, while the performance of the overclocked configs tracks very well with increasing clock speed. DivX video encoding
One area where the Pentium 4’s Netburst architecture has always excelled is media encoding, an an excellent example of this is the XMPEG benchmark. Dual memory channels help the Pentium M only slightly here. Cinebench 2003
Hyper-Threading can be a real asset if an application can take advantage of it, as we see here. The Prescott lags considerably behind the two Northwood configurations, but with multithreading it still manages to beat the fastest Pentium M. When restricted to a single thread, however, the fastest Pentium M configuration manages to come out on top. Yonah, anyone?
We measured the power consumption of our entire test systems, except for the monitor, at the wall outlet using a watt meter. The idle results were measured at the Windows desktop, while load results were obtained after firing up one instance of Prime95 for the Pentium M and two instances for the Pentium 4 processors with Hyper-Threading.
Here we see another reason why people are clamoring for Pentium M on the desktop. The Pentium M’s power consumption is dramatically lower overall. Also, while the Pentium 4 chips tack on as much as 114W going from idle to full load, the Pentium M adds only 23W at most. Clearly, DFI’s use of a laptop chipset pays some dividends here, as well. If you’re wondering how the DFI board can overclock a processor by 300MHz without consuming any more power either at idle or under load, well…. so am I. I reran the tests to be sure, but they always turned out the same. My current theory on the subject revolves around little green men, but it’s still a work in progress.
Compared to other available Pentium M desktop solutions, the CT-479 is compelling. Currently, the CT-479 is available for approximately $42, and compatible motherboards are available for as little as $80-85. For that price, you get dual channels of DDR333 memory and the ability to raise the CPU voltage until the blue smoke escapes. Neither of these options is available with dedicated Pentium M motherboards, and the CT-479 is cheaper than those motherboards. The combination of price and features makes the CT-479 difficult to beat. If anything, the CT-479 solution suffers from the age of its host motherboards, since the use of the Intel 865PE and 875P chipsets means that you’re stuck back in AGP and PCI land.
For nearly all consumers interested in the Pentium M, the CT-479 offers both better performance and a lower cost than other solutions, and that’s really all that needs to be said. If you’re interested in a quiet, Watt-sipping Pentium M desktop, the CT-479 is as good as I’ve seen.