ASUSTeK CT-479: Socket 479 on Socket 478 - BeHardware >> Motherboards
Written by Marc Prieur
Published on April 13, 2005
URL: http://www.behardware.com/art/lire/565/
Page 1
The Pentium M, the CT-479
Since the end of last year, the Pentium M processor, which until then was only dedicated to laptops, entered the desktop market with DFI and AOpen motherboards. We wrote an article on this subject last January and, at the time, we were unimpressed by the motherboard price and its limitations, whether connected to the chipset or not. Following the success of this platform, ASUSTeK has decided to release the CT-479. This adaptor provides an opportunity to use a Pentium M with a Socket 478 motherboard!
The Pentium M in brief Even though we already addressed this subject in our previous article, it is good to take another look at the ups and down-sides of the Pentium M architecture. Its main interest lies in its extremely efficient processing units. Thanks to these units, this processor has an excellent performance/thermal dissipation ratio. This ratio is one of Intel’s major problems since the release of the Pentium 4 " Prescott ". It is also possible to use the Overclock Pentium M " Dothan " to 2.4 GHz without increasing its voltage.

Of course, the overall picture isn’t perfect because a simple Pentium M 740 clocked at 1.73 GHz (possible Overclocking to 2.4 GHz) costs about $255 in shops. Also, even if the Pentium M provides better performances for games, the Pentium 4 is clearly better with applications using SSE image vectoring such as video encoding.
Finally, the Pentium M doesn’t support the HyperThreading, and it isn’t as smooth for intensive multi-task use (for example if you use your computer during a 3D rendering or video encoding).
The CT-479 How is it possible? Quite simply because the Pentium M and Celeron M use the same type of bus as the Pentium 4: a bus GTL+. It works in double data rate for the address bus and quad data rate for the data bus. Based on this fact, the development of an adaptor was quite easy as the CT-479 design shows.

The CT-479 is powered externally by a specific plug which connects to the Molex 4 pins supply connector and provides 5 and 12 Volts. It includes two jumpers to select the installed processor’s FSB.

As well as the power supply cable and Chinese/English instructions, the CT-479 is sold with a radiator on an aluminium base. It is designed so as not to get in the way of the power supply cable. On top of the radiator is a 70mm radiator. It spins at 3000 Rpm and provides 32.5CFM for 30 dBA at 12V.
Page 2
Hardware installation
Hardware installation The CT-479 is very easy to install. It plugs in like a processor: you need to lift the Socket 478 arms, then plug it in and lower the arm. After placing the Pentium M on the Socket 479 you can lock it with a flat screwdriver.
  
Then, if necessary, you adjust the jumpers and connect the CT-479 power supply. If you have removed the CPU cooler thermal pad (and we advise you to do so), then you just need to apply a little thermal paste on the processor. After this, you need to place the CPU cooler held together by two fastening devices and locked to the usual Socket 478 retention system.
  
Initially, only two motherboards were compatible: the P4P800 SE and the P4P800-VM, but recently ASUSTeK has added the P4C800-E Deluxe to the list. In practice, the support of these motherboards isn’t equivalent and other motherboards support the CT-479, as you will see on the following page. Another important point: without an adequate bios, the configuration simply doesn’t boot. While waiting for the motherboards in shops to be sold with the right bios, you need to flash the motherboard bios with a Pentium 4 before the CT-479 installation!
Page 3
Motherboards compatible
P4P800 SE
  
This motherboard is one of the first to support the Pentium M. It has actually supported the Pentium M since the November bios 1007 made for the previous Pentium M " Banias " generation. The right bios for the Dothan is the 1008: however it doesn’t have a processor power supply adjustment (from 1.225 to 1.6V with steps of 0.0125V). The beta 1009.001 fixes this problem and this adjustment is back again!
There is nothing more to add except for this point. You should know, however, that with FSB400 (100 MHz), we have access to the 3:4 FSB: DRAM ratio and so to the DDR266. From 133 MHz of FSB in the bios (FSB533), the ratios 1:1 and 4:5 (DDR266 and 333) are available, and with FSB800 the ratios are 3:2 (DDR266), 5 :4 (DDR320) and 1:1 (DDR400).
We find it unfortunate that it isn’t possible to change the multiplication coefficient in the bios. This problem, coupled with the fact that some FSB: DRAM ratios like the 1:1 are only accessible from 133 MHz, tends to make the Overclocking less flexible. We also noticed, and it is an unfortunate habit of ASUSTeK’s since we also came across it on other motherboards: at 133 MHz the bios doesn’t take in account the FSB:DRAM ratio indicated and uses the 3:4 ratio (for " DDR354 "). You need to set the frequency to 134 MHz to change this.
P4C800-E Deluxe
  
The P4C800-E Deluxe supports the CT-479 from the 1021 bios. This time, the processor voltage improvement isn’t available. Except for this point, it behaves in the same way as the P4P800 SE.
P4P800-E Deluxe
  
The bios beta 1007.001 enables a beginning of compatibility with the CT-479. There again, the processor voltage isn’t accessible but the coefficient is. Unfortunately, the modification of this coefficient doesn’t have any effect because the bios doesn’t take it into consideration.
P4GD1
  
The P4GD1 is a very interesting motherboard because it is the only Socket 478 motherboard based on the i915P chipset and PCI Express compatible. It also provides all the ICH6 advantages, such as the Matrix RAID and still allows the possibility of using your current DDR RAM memory modules… and it is compatible with the CT-479 with the 1005.006 bios beta. The processor coefficient isn’t accessible but the voltage is.
Also, the FSB: DRAM ratios available are very interesting. We have access to the 4:5 and 4: 6 which corresponds to the DDR250 and DDR300 mode. The same ratios are accessible in FSB533, the DDR333 and 400 modes, whereas in FSB667 (166 MHz) the 6:5 (DDR276) and 1 :1 (DDR332) and even 4 :6 (DDR498) ratios are available. These ratios are also available in FSB800 (DDR333, 400 and 600).
In practice, however, we noticed that the ratio management is completely deficient with this bios beta. In FSB400, the first adjustment, in practice gives access to the DDR200, whereas we are in DDR160 with the second. In FSB533 the first configuration gives access to the DDR212 and the second one crashes the computer. The FSB800 doesn’t work with this motherboard, and when we clocked in at 166 MHz we accessed the DDR266 (5:4) with the first adjustment and DDR333 (1:1) with the second (the DDR498 mode doesn’t work).
The PCI & PCI-Express simply don’t work. Of course in FSB400 and FSB533 (and it should be identical with the 800), PCI and PCI-E are clocked in at 33 and 100 MHz, but these frequencies aren’t fixed and they increase with the FSB. In FSB167 for example, the PCI is clocked in at 41.7 MHz and it can be a bit of a problem for some of the peripherals using this bus.
Other motherboards Just to see how it would behave, we also installed the CT-479 on three i865PE motherboards:
- ABIT IS7-E2 - Gigabyte GA-8IPE1000-G - MSI 865PE Neo2 PFISR
Like with the motherboard ASUSTeK before the bios flash, these attempts resulted in a black screen. We can’t expect all makers to work on their bios to make them compatible with the CT-479, but we hope that the situation will improve for ASUSTeK. As well as an improved support for compatible motherboards, with processor voltage and coefficient modifications, it would be better if older motherboards could also be compatible, if this is technically possible: P4G8X, P4B533 ...
Page 4
Silence & Overclocking
Silence & Overclocking You probably already grasped that the only current motherboard with a satisfying bios (and because we don’t have any other choices) is the P4P800 SE. So we mainly used that one for the last part of the article. It is important to point out two very useful utilities: EIST and ClockGen, developed by Franck Delattre.

The first one gives access to processor SpeedStep adjustments under Windows and a possibility of immediately changing (without rebooting) the voltage and multiplying coefficients. The second one offers the possibility of changing the bus processor’s speed.
Other software can also be useful. The SpeedStep that initially did not work with a Desktop motherboard, RMClock allows the possibility of changing frequency and voltage (within an interval defined by the user) according to the load. The thing is that, even if this function is very useful on a laptop in order to save some energy, it is less useful with a desktop, especially with a processor that is easy to cool down silently even at full load. Another one Speedfan will offer the possibility of reducing the fan speed with a motherboard such as the P4P800 SE.
 
It is important to point out that at its initial speed, without being really noisy, it isn’t really discreet either. SpeedFan gives the opportunity of reducing the fan rotation speed to 25% and with this reduction we have to admit that we couldn’t hear the fan’s blowing sound …but the engine sound unfortunately was still quite audible. So, because of this noise, such a reduction is useless and it is best to increase it to approximately 40% to have an equivalent fan and engine sound level. Here are a couple of temperatures measured with via ASUS Probe in load under Prime95 with the fan voltage at 12V / 5V.
P-M 2.0 GHz (1.34V) : 44 / 48°C - P-M 2.0 @ 2.4 GHz (1.34V) : 47 / 51°C - P-M 2.0 @ 2.6 GHz (1.45V) : 50 / 55°C
Even in Overclocking and with a lower voltage, the fan is able to keep the temperature low. For the Overclocking and despite the possibility of increasing the processor voltage, we were unable to reach much higher frequencies with our Pentium M 755 (2.0 GHz / FSB400) already used for the DFI motherboard test: 2.5 GHz with the initial 1.34 V and 2.6 GHz with 1.45V. This Overclocking has been stabilized and validated in use.

The other interesting point is the processor FSB increases with the platform. With the Pentium M 755, we were unable to reach beyond an FSB of 180 MHz with a stable configuration, whereas with the 740 we managed to reach 215 MHz. So it seems to confirm that the Pentium M FSB533 is more inclined to work with higher FSB.
We were able to reach this FSB under Windows via Clockgen, but from 200 MHz upwards we had inexplicable 3D graphic bugs with some of the games, even when the AGP frequency was locked in. Furthermore, the system didn’t want to boot with this FSB directly configured via the bios.
Page 5
Platform performances
Platform performances To begin with we compared the platform performances with identical FSB and frequencies : 20x100 MHz (2.0 GHz). We made two tests, one with a compression time under WinRAR expressed in seconds and the other at the average framerate under UT2004. All tests were made with a P4P800-SE using the same test configuration as during the DFI motherboard test- results were then comparable to those previously obtained.
 The first thing we noticed is, that with an equivalent memory configuration, DDR266 on one channel, the i865PE+CT-479 platform provided good results with a 52% improvement under WinRAR and 2.5% under UT2004. Changing to DDR333 for the DFI motherboard produced better results than the CT-479 solution with single channel, but it got back its lead for dual channel performances with, respectively, +7% and +5%.
The second part to be studied was the FSB and memory speed influence on performances. We changed these values and kept the final processor frequency of 2 GHz:
 Results are interesting and show that the Pentium M architecture still has some room for improvement and is clearly castrated by official use. So compared to the FSB533 / 2 DDR333 channels adjustment which is the maximum officially supported by the Pentium M in DDR-1 (via i915-GM), the FSB800 / 2 DDR400 channels is still 13.6% faster under WinRAR and 9.4% under UT2004. The best adjustment is respectively 36.4 and 24.8% faster than the worst.
Here are the performance gaps observed between all our CPU tests in FSB400 and DDR333 with the DFI motherboard and FSB400 and DDR266 on two channels with the duo P4P800-SE + CT479.:
 The result was predictable, for applications relying chiefly on the processor’s raw performances, like 3D rendering, mathematic calculations, the platform changes’ impact was low. When applications were a little more sensitive to the memory sub-system performances, the increase was noticeable.
We have already spoken of the Pentium M applicative performances in this articleand you can compare these figures directly as the platforms are identical. Just for your information, here is a direct comparison between results obtained with a Pentium M 2.4 GHz in 185/185 with CT-479 andP4P800 SE compared to an Athlon 64 FX-55 and a Pentium 4 Extreme Edition. Of course, this comparison is supplied only for information, because on the one hand, there is an Overclocked processor and, on the other, high end processors with standard frequencies. It is interesting, though, because these processors feature a low Overclocking margin in air cooling (2.7 GHz for the FX-55, +4%, 4 GHz for the P4 EE, +7.1%), the Pentium M is able to reach 2.6 GHz (+8.3% of margin improvement compared to the frequency tested).

The Pentium M provides honourable performances and is even in the lead for scientific calculations with Mathematica. Compared to the Pentium 4, we have noticed that it is slow with particularly optimised video encoding software Netburst and also with TMPGEnc. However, it provides much better performances with games and is even faster than the A64 with Pacific Fighter.
Page 6
Conclusion
Conclusion The use of the Pentium M for a desktop computer, so far only restricted to a few enthusiastic users, might now reach larger market shares with the release of the ASUSTeK CT-479. This adapter offers a possibility for resolving problems linked to the native Pentium M motherboards: as well as a relatively restricted design, they are quite expensive and their availability is low.
So if we had to spend 230€ for this type of motherboard, the CT-479 available at the end of the week will cost about 50€. You will also need to spend approximately 100€ for a compatible motherboard if you don’t already have one. That is the good side of the CT-479: it opens up new upgrade possibilities for your computer, which are rather restricted for the low clocked Northwood owners as the Northwood Socket 478 are almost unobtainable. Prescott Socket 478 processors also have a tendency to reach high temperatures and are restricted to 3.2 GHz for sales versions.
However, we think it is unfortunate that ASUSTeK hasn’t pushed the concept further and extended as far as possible its list of compatible motherboard for the P4C800 and P4P800 non " -E ", P4B533, P4G8X. On the other hand, before enlarging the list, it would be best to perfect the bios of the current compatible motherboards. Some have bugs and others don’t yet provide all functionalities. ASUSTeK told us that they are actively working on these problems. Also the CPU cooler quality could have been higher with, for example, a copper insert at the base level and a less noisy engine for the fan.
Of course it is important to keep in mind that the Pentium M isn’t the ideal processor even if it features an unbeatable performance / thermal dissipation ratio. From a technical point of view, the Pentium 4, for example, is much more comfortable if you use your computer mainly for video encoding. The price of an easily Overclockable 1.73 GHz P-M 740 is around 240€. This price also corresponds to an Athlon 64 3500+. With the 90 nm fabrication process and the Cool’n’Quiet the AMD Athlon 64 processors, even if they are a little bit more power greedy, they also provide an interesting alternative for low noise configuration.
Without being the perfect Desktop solution, the ASUSTeK CT-479 makes the Pentium M more affordable and efficient, in spite of a couple of downsides due to the early days of this solution. We can only congratulate ASUSTeK for this intelligent initiative which reduces the interest of dedicated motherboards down to zero. We do hope, though, that bios updates will be quickly released.
Copyright © 1997-2008 BeHardware. All rights reserved. |