Intel P45 Express - Preview - BeHardware
>> Motherboards

Written by Marc Prieur

Published on June 4, 2008


Page 1

P45 Express

Intel took advantage of the Computex to announce the P45 Express, the successor to the very popular P35 Express. What’s new?

The P45 Express
Engraved in 65nm versus 90nm for the previous versions, this chipset is supposed have lower power consumption. Voltage has been reduced to 1.1v versus 1.25v for the P35. Otherwise, in terms of functions, Intel was content to provide the strict minimum. Thus, for the MCH the FSB stays limited to 1333, the remaining 1600 being reserved for the X48, while DDR3 support goes from DDR3-1066 to DDR3-1333 with of course DDR2-800 included. Note that in practice, this mode already functioned on P35 motherboards and some manufacturers even made the effort to advertise this.

So in short, it’s more in terms of PCI-Express that we should expect the real innovation of the P45. Indeed, up until now only the X38/X48 enabled having a second high performance graphic port via the support of two PCI-E x16 2.0 ports. Actually, the P35 only had support for a single x16 port in PCI-E 1.0 while the second integrated to some motherboards was cabled in x4 and relayed to the ICH. The P45 is limited to 16 PCI-Express lanes; however, they adhere to the 2.0 norm, thereby doubling the bandwidth, and can be used in two ports cabled in x8 or the same bandwidth as an x16 port in the 1st version of PCI Express.

As for the ICH, there was no big change to the ICH10. In its standard version it supports six SATA in AHCI, while the classic ICH9 only supported four SATA without AHCI. Otherwise, we had to opt for the more expensive ICH9R to have six SATA in AHCI. The ‘’R’’ version of the ICH10 is of course still in the catalogue and, like its predecessor, it offers RAID support.

Page 2
In practice, Memory, PCI-E

In practice
So how does this chipset fare in practice? To answer this question, MSI lent us a P45D3 Platinum which we compared to this same brand’s P35D3 Platinum. With its imposing cooling system based on heatpipes, the motherboard has all of the capabilities of the P45 while also adding an e-SATA (JMB362 chip), two SATA and a PATA (JMB363 chip) ports, as well as FireWire 400 (JMB381 chip). HD Audio is handled by a Realtek ALC888 CODEC while the Gigabit network is provided by a PCI-E Realtek RTL8111C chip.

Note that the CPU power block uses a DrMos-type component, in other words, a driver-MOSFET is integrated to the same unit in order to decrease its size. The chips, developed by Renesas, are announced as being able to attain up to 89% efficiency. This motherboard also features GreenPower technology. It is MSI’s name for the possibility of automatically changing the number of phases used by the motherboard for CPU, chipset and DDR power in order to find the most efficient use of energy depending on required levels. Otherwise, the motherboard comes with a PCB which fits between your power supply and the motherboard’s ATX block (relayed to the motherboard). It enables displaying the power used on each line in addition to the power used by the CPU via the ATX12V connector.

Despite the array of functions, we obtained rather mixed results, although we could not determine the exact cause: the beta bios, a pre-version of the motherboard, an ‘’ES’’ (sample) chipset? Indeed, on the first and then a second model something rather strange happened.

After our usual tests in FSB increases on the motherboard with a quad core (something much more demanding than dual core FSB increases) consisting of raising the frequency and if necessary calculated increases in the various voltages, the quality of the chipset seemed to have been altered. Afterwards, it didn’t even function at FSB1333!

For this reason, we couldn’t finish all of our tests on the P45. Moreover, the overall reliability of this solution seems to be in jeopardy even if specify again that this situation repeatedly happened with pre-series material.
We were first interested in evaluating the memory controller by using ScienceMark 2 to obtain speeds and latency. The framerate was also measured in Crysis and file compression times were obtained with WinRAR 3.7. These two ‘’practical’’ applications were chosen because they are noticeably affected by the speed of the memory subsystem, something that isn’t always the case.

As you can see, there are slight gains in latency but much bigger improvements in bandwidth with the MSI P45 motherboard compared to the MSI P35. In practice, the impact is variable. It’s significant in WinRAR and barely noticeable in Crysis, though always in favor of the P45.
In practice, what does improved PCI Express support add to the P45? This is what we wanted to determine with the help of Crysis, the latest game from Crytek, and which has the most advanced graphic engine at this time. In fact, in standard resolution, 1680*1050, high quality mode causes problems for 256 MB cards and they are required to use central memory to compensate for this. Therefore, for this test we used two PCI-E 2.0 cards, 8800 GTs equipped with 256 and 512 MB of video memory. We started with exclusive connection to the first graphic port and then we tested the second port with another card being placed in the first.

With a 256 MB card the result is clear and the benefits of the P45 are obvious. On the first port, the advantage of 16 PCI-E 1.0 lanes versus 16 PCI-E 2.0 lanes is already significant. Moreover, when comparing 8 PCI-E 2.0 lanes to 4 PCI-E 1.0 lanes by using the card on the second port, there is really no comparison.

With a 512 MB card, the differences are much less, to the extent that the advantage on the first port is very small. On the other hand, on the second port, the advantage is clearly in favor of the P45. However, note that there isn’t really a difference for the P45 between 16 2.0 lanes (the first port used by itself) and 8 2.0 lanes (the case when using two cards). For this reason, the interest of the X38/X48, which has support for twice as many PCI-E 2.0 lanes, is very minimal.

Page 3
Overclocking, conclusion

Unfortunately, we couldn’t test frequency increases on the P45 with four memory modules because our two motherboards failed before this test. On the P35D3 Platinum, note that it was able to synchronously function up to FSB1560/DDR3-1560 with two modules and up to FSB1400/DDR3-1400 with four. With the P45, Intel only makes the DDR3-1333 support official, something which was already well established in practice on the P35.

For the FSB, with a QX6850 (which is our quad core with the highest FSB wall), we were able to attain 440 MHz (FSB1760) on the P35 at the initial voltage of 1.25V and 470 MHz by increasing the chipset voltage to 1.45V and the VTT voltage from 1.2 to 1.35V. On the P45, at the starting 1.1V we attained 430 MHz (FSB1720) and then 460 MHz by increasing to 1.4V. These were all stable frequencies validated by a 30 minute session of Prime95. Note that the P45 used for this test stopped functioning properly after we tried to stabilize it at 470 MHz.
On paper as in practice, the P45 doesn’t really have anything revolutionary as was the case of the P35 compared to the P965. Of course, DDR3-1333 support is now official (but it was something already effective for the P35) while the ICH only evolves in its standard version.

Actually, it is especially the improved support of PCI-Express compared to the P35 which is of interest. This is either due to new PCI-E 2.0 support or the 16 lanes that can now be shared between two graphic ports. It enables Intel to offer a mid-level chipset capable of properly running two graphic cards independently or in CrossFire. Otherwise, this meant previously having to opt for the more expensive X38 or X48.

So, the P45 will not necessarily motivate you to change your motherboard, all the more so that like all Socket 775 models, it will not be compatible with the next generation of Intel Nehalem processors which will use a new Socket. Moreover, this compatibility is unfortunately impossible given the number of changes on the chipset level in the transition from the Core 2 to the Nehalem: different processor bus, the memory controller moves from the chipset to the CPU, etc.

Finally, like with each new chipset we should keep in mind that it’s the first buyers that will deal with unresolved bios issues. For this reason, some will be tempted to wait for user feedback…or opt for a P35 Express which overall doesn’t have much to fear from the P45. As for us, we were particularly unlucky in this test due to two successive motherboards that could not handle their chipsets being pushed too hard in overclocking. Is this MSI’s fault and related to the motherboard or does this fall on Intel and the fact that the chipset was in its pre-series stage? We can’t answer this question and even if we hope that the latter is true, user feedback on commercialized versions should be closely monitored...

Copyright © 1997-2015 BeHardware. All rights reserved.