06/30/2008 | 06:41 PM
The first association that comes to mind when we speak of Nvidia is graphics cards, however, it is also a pretty big developer of mainboard chipsets. Just recall the remarkable success of Nvidia chipsets when they first came out. The first nForce generation chipsets were not very popular, although they attracted our attention due to a number of interesting features. nForce 2 chipset, however, made a real statement. VIA’s seemingly unshakable position as of the leading chipset manufacturer melted away and every Socket A processor was better off with an Nvidia nForce2 based mainboard. nForce3 chipset didn’t create a remarkable resonance like that, however, it was mostly due to the fact that very first Socket 754 processors were very rare and expensive. Besides, many users were looking forward to Socket 939 CPUs. nForce4 chipset again experienced great success, but after that Nvidia’s core logic solutions slowed down their victorious pace. The transition to the fifth nForce generation and Socket AM2 platform required replacing DDR with DDR2 SDRAM so it didn’t happen quickly. Moreover, Intel Core 2 Duo processors came out just about that time and a lot of users switched to a higher performing platform.
Yes, speaking of Nvidia chipsets’ success we implied only AMD core logic sets, because Nvidia didn’t manage to compete on equal terms with Intel solutions. The very first mainboards on Nvidia nForce4 SLI Intel Edition chipsets surprised and disappointed us with their weak overclocking potential. Besides, their reputation suffered noticeably from the Smithfield support issues. Once our attention span moves from AMD processors to Intel Conroe, we kind of left out Nvidia nForce5 chipset series. However, since there wasn’t anything special reported about these mainboards over time, it looked like we didn’t really miss much. However, we did review a good number of Nvidia nForce 680i SLI and 650i SLI based mainboards instead. Although there was nothing particularly interesting there either: same overclocking issues, same CPU support issues, only this time with quad-core Yorkfield processors. Besides, they revealed some performance problems at high frequencies (FSB Strap), non-operational frequency intervals (FSB Hole), “falling apart” RAID arrays and lost HDD data.
No wonder that Nvidia based mainboards for Intel processors are not very popular. The Intel CPU owners and potential buyers have been waiting for the fourth generation of Intel chipsets to come out, so that one day they could get an affordable mainboards with PCI Express 2.0 support. However, there already exist more affordable mainboards than those based on Intel X38 Express and Intel X48 Express.
Our today’s article will discuss a solution like that – ASUS P5N-D mainboard. but before we talk about the mainboard, let’s take a closer look at the Nvidia nForce 750i SLI chipset it is based on.
The official specifications of Nvidia nForce 750i SLI chipset are summed up in the table below:
If we compare it with the Intel P35 Express chipset features, which are pretty similar, we will see that Nvidia nForce 750i SLI has only 8 USB ports instead of 12 and only 4 Serial ATA ports instead of 6 by the Intel solution. However, Nvidia nForce 750i SLI retained fully-fledged Parallel ATA supporting up to 4 devices, while Intel chipsets haven’t had it for a long time now and it can only be implemented with additional controllers. Overall, Intel P35 Express features seem to be more up-to-date, although Nvidia nForce 750i SLI is not critically behind, also.
However, you can get the best idea of the Nvidia nForce 750i SLI chipset features from the flow-chart below. You immediately notice a three-chip combination instead of a traditional dual-chip design with a North and South Bridges or even a single-chip design Nvidia has already used before. Now there is the third East Bridge:
Of course, the third components is Nvidia nForce 200, however, it is the primary feature distinguishing this core logic set from its predecessors. It is a bridge-chip that provides PCI Express 2.0 support. MCP51 or nForce 430 stands for the South Bridge. It appeared back in 2005 and we are very well familiar with it already. We saw it in discrete and integrated Nvidia chipsets for Intel and AMD processors. Namely, this particular South Bridge was used in Nvidia nForce 650i SLI chipset. The North Bridge of Nvidia nForce 750i SLI is codenamed C72P.
The interesting thing is that upon system boot-up ASUS P5N-D mainboard doesn’t deny that it uses MCP51 South Bridge. However, it claims that its North Bridge is a C55 chip. You should remember that it is exactly the chip that was used in Nvidia nForce 680i SLI and Nvidia nForce 650i SLI chipsets. The combination of a C55 North Bridge with MCP51 South Bridge produces Nvidia nForce 650i SLI and by adding an Nvidia nForce200 we turn it into Nvidia nForce 750i SLI.
Of course, you can check that the specifications of the Nvidia nForce 650i SLI and Nvidia nForce 750i SLI are identical (except the PCI Express 2.0 support). However, there is one more important difference beside the PCI Express 2.0: the new core logic set is declared to fully support 45nm Penryn processors including quad-core Yorkfield as well as dual-core Wolfdale. The Nvidia nForce 6 was initially supposed to support Yorkfiled processors, but then they called this feature off. In other words, the new Nvidia nForce 750i SLI chipset is made from the old Nvidia nForce 650i SLI with eliminated drawbacks and an Nvidia nForce 200 chip providing PCI Express 2.0 support.
By the way, Nvidia nForce 780i SLI chipset was made following the same principle from the combination of Nvidia nForce 680i SLI and Nvidia nForce 200. The two chipsets, Nvidia nForce 780i SLI and Nvidia nForce 750i SLI, created as modifications of the old ones, seem to be some sort of intermediate solution on the way to a really new chipset like Nvidia nForce 790i SLI designed according to a traditional dual-chip schematics.
Well, it is time for us to take a closer look at ASUS P5N-D mainboard. We will start with its official specifications:
If we compare these specs with the official specifications for Nvidia nForce 750i SLI chipset, we will see that ASUS laid out only one Parallel ATA connector of two supported by the chipset, which means we can only connect two devices of four possible. However, it is more than enough for today’s needs and many manufacturers out there do the same thing. Even Nvidia nForce 650i SLI based mainboards often had only one Parallel ATA connector.
There is another difference that seem pretty interesting to me, though. According to the official specifications of the Nvidia nForce 750i SLI chipset, ASUS P5N-D mainboard supports PCI Express 2.0 x16 mode for a single graphics card. As for two graphics cards working in an SLI configuration, they put it in a pretty strange way: “hardware ready for x16, x16”. What did ASUS actually mean? We all know that NVIDIA nForce 750i SLI based mainboard can formally work only as PCI Express 2.0 x8 + PCI Express 2.0 x8. Did they imply that the bandwidth of PCI Express 2.0 x8 equals that of PCI Express 1.0 x16? Of they are driving at the fact that there may appear an enhanced nForce 750i SLI supporting PCI Express 2.0 x16 + PCI Express 2.0 x16? Or maybe they were trying to say that NVIDIA nForce 200 bridge-chip supports PCI Express 2.0 x16 + PCI Express 2.0 x16 mode and the limitations are only made in the drivers, but not in the actual hardware?
The last version seems to me the most probable. The North Bridge of NVIDIA nForce 680i SLI chipset provided only 16 PCI Express lanes. The South Bridge added another 16 lanes necessary for PCI Express x16 + PCI Express x16 mode implementation. When they modified the chipset into NVIDIA nForce 780i SLI, 16 PCI Express 1.0 lanes from the South Bridge got assigned only for the third graphics card. However, Nvidia nForce 200 chip turned 16 PCI Express 1.0 lanes from the North Bridge into 32 PCI Express 2.0 lanes to be distributed between two PCI Express 2.0 x16 slots. So, why not do the same thing again with NVIDIA nForce 750i SLI chipset? It should be technically possible, but Nvidia has to clearly distinguish between the top 780i SLI chipset and its younger brother 750i SLI.
Of course, the question is how NVIDIA nForce 200 communicates with the? What interface it uses and at what speed? On the outside there are two PCI Express 2.0 x8 slots by NVIDIA nForce 750i SLI and two PCI Express 2.0 x16 slots by NVIDIA nForce 780i SLI. And what is inside? Could it really be PCI Express 1.0 x16, PCI Express 2.0 x1 or even PCI Express 1.0 x1? In this case, the current implementation of PCI Express 2.0 in NVIDIA nForce 750i SLI and NVIDIA nForce 780i SLI seems to be just a formality and it should be fairly easy to expand the functionality of nForce 750i SLI to supporting PCI Express 2.0 x16 + PCI Express 2.0 x16. In the meanwhile the owners of NVIDIA nForce 750i SLI based mainboards can only be certain about PCI Express 2.0 x8 + PCI Express 2.0 x8 support.
Well, let’s put all suppositions and guesses aside and discuss the ASUS P5N-D mainboard in detail.
The box should first of all protect the mainboard against physical damage, and all packaging copes with this task quite well, I should say. At the same time, the packaging should attract potential customers with its design and inform them about the peculiarities and advantages of the product inside.
The design of ASUS P5N-D mainboard package will hardly attract that many users, although theoretically, there is everything a package should have to be appealing: company name, mainboard model name, logos. However, in my humble opinion, the front side of it turned out pretty gloomy and dark.
We were also disappointed with the information on the back of the box. There is only a small mainboard photo, specifications and supported technologies are listed as icons, and two of them – ASUS EPU and Precision Tweaker 2 – deserved one phrase each in seven languages. By the way, ASUS EPU (Energy Processing Unit) that allows changing the number of active phases in the processor voltage regulator circuitry depending on the workload is also mentioned on the front of the box.
While we were working on this review we had to go back to our last year’s article called Nvidia nForce 650i SLI Chipset and only $130: ASUS P5N-E SLI Mainboard Review. Turned out that its box was designed in exactly the same way. The list of bundled accessories also barely changed since then. ASUS P5N-D mainboard comes with the following items:
I would like to draw your attention to the fact that two SATA cables are L-shaped and the other two – regular, so the user can pick the ones he likes. The same attention to small but useful conveniences has already been demonstrated by Gigabyte and discussed I our article called Gigabyte GA-EP35-DS4 Mainboard: a Short Step Away from Ideal.
ASUS P5N-D mainboard boasts another feature of the sort. From the description of ASUS Q-Shield technology I couldn’t understand what “fingers” they were talking about and what the idea behind it was. In fact, all I had to do was look at the icon on the right:
Anyone who has assembled a system at least once will understand everything right away. The rear panel I/O Shield has a few “fingers” on the back of it that provide electrical contact with the connectors thus taking off the EMI. However, these “fingers” need to be bent accordingly otherwise they tend to catch into the connectors. ASUS Q-Shield solves the problem. It has no “fingers” but its entire back surface is covered with porous material topped with a conducting layer.
You no longer have to bend and watch all those “fingers”, assembly gets way simpler, as the shield just needs to be put in place. The conducting layer will provide electrical contact and porous material will ensure that it is pressed firmly against the connectors. Nice and easy.
And finally, the last tidbit of the accessories bundle. ASUS P5N-D North Bridge is covered with a low-profile but very wide passive heatsink:
Why this shape? It is very simple: besides the actual North Bridge, this heatsink also covers Nvidia nForce 200 chip providing PCI Express 2.0 support. It is certainly cool, but Nvidia chipsets have always been famous for high heat dissipation and this heatsink has to deal not only with a hot North Bridge but also with a bridge-chip, which may get even hotter. No matter how big the heatsink is, its surface will not be enough: additional cooling is a must here.
In our recent Intel DX48BT2 mainboard review we gave Intel credit for a rare but very useful accessory they included with their mainboard: a plastic frame for attaching an additional fan to the chipset North Bridge heatsink. ASUS didn’t stop there. They included not just the retention frame, but also a 70 x 70 x 10mm fan, which is certainly a great thing to do, considering that not every computer store carries these fans.
The fan attaches in a very simple way: just clip the frame hooks onto the heatsink and connect the fan to the nearest mainboard connector:
Although ASUS is very optimistic and suggests installing this additional fan only if you have liquid or passive processor cooler in your system, I would advise to install it anyway. Even in nominal operational mode this heatsink gets so hot that you can barely tough it. And during overclocking you will be increasing the voltages, so the thermal conditions will be even worse.
Unfortunately, things were not completely perfect. According to the mainboard monitoring data, the fan rotates at ~3800RPM and it is way too loud. ASUS Q-fan rotation speed control system allows only three settings in the BIOS: Performance, Optimal and Silent. In Performance mode the fan rotation speed drops down to about 3400RPM, which is still too loud, as well as 3000RPM in Optimal mode. In Silent mode the speed drops to 2500RPM and you can no longer hear the noise from the fan airflow, although you can hear the fan motor working.
The discussion of ASUS P5N-D accessories wouldn’t be complete without mentioning its user’s manual, brief assembly guide in several languages, a system case sticker and two CD disks with Windows and Linux drivers, ASUS utilities and additional software.
Among the additional software titles we can list Norton Internet Security 2007, Intervideo DVD Copy 5 Trial, Corel Snapfire Plus SE 1.2, as well as DirectX 9.0c and Adobe Acrobat Reader for included electronic versions of the guides and manual.
ASUS P5N-D mainboard makes a great first impression, we don’t see any design flaws:
First of all you notice a gigantic North Bridge heatsink in the very center of the PCB. However, since we have already discussed it in the previous chapter let’s move on to the top part of the PCB. The power connectors are in relatively acceptable spots. The quad-phase processor voltage regulator uses contemporary components and looks very decent. The only thing that puzzled us here was the four-pin ATX12V processor power supply connector instead of an 8-pin one.
As usual, there are more components on the lower part of the PCB, but they are also placed very nicely. ASUS P5N-D mainboard features two PCI Express x16 slots, a pair of PCI Express x1 slots and two PCI slots. Chipset South Bridge is topped with a small heatsink:
IDE connector is conveniently turned sideways. Four Serial ATA ports are placed along the same line as the PCI slots, so they will not be blocked even if there are two graphics cards installed in the system. There was enough room for an additional IEEE1394 controller – VIA VT6308P. There are front panel connectors, USB connectors, IEEE1394 and FDD connectors and audio along the lower edge of the PCB.
The mainboard back panel boasts a full set of connectors, which you do not see that often any more these days: PS/2 connectors for keyboard and mouse, coaxial and optical S/PDIF, COM and LPT, IEEE1394, RJ45, four USB ports and six audio jacks (Realtek ALC883):
In conclusion to our layout discussion I would like to repeat that ASUS P5N-D is designed very thoroughly and boasts a wide range of features.
ASUS P5N-D mainboard uses BIOS based on Phoenix-Award code. It looks pretty common for an ASUS BIOS and consists of several large sections: Main, Advanced, Power, Boot, Tools and Exit. Each section may include a few sub-sections. First we get to the Main section offering only a few major functions, such as setting system time and date, selecting the language:
Advanced section has much more sub-sections and features:
Let’s check out the JumperFree Configuration sub-section first. By setting AI Tuning to AI Overclock we can use the automatic overclocking system:
However, you can overclock your CPU by only 20% maximum, so we will change AI Tuning to Manual to get full access to the board’s overclocking friendly functions.
Despite the attractive name, the “System Clocks” page has only one parameter: NB PCIE Frequency. It allows changing the PCI Express bus frequency in the interval from 100MHz to 131MHz. So, we will not stay here for long and move on to the next page called Voltage Control.
Vcore Voltage parameter can be adjusted from 0.83125V to 1.6V with 0.00625V increment. Other voltages can be changed with 0.02V increment in the following intervals:
These are pretty wide ranges, the increment is quite small and the parameters can be changed very easily. For example, just select one of the parameters and you will immediately see the supported range in the right-hand side of the screen.
For the CPU Vcore you will see not only the supported voltage interval but also a special warning. It will remind you of a well-known ASUS mainboard issue: any changes to the processor Vcore when it is not longer set to Auto mode will disable all power-saving technologies.
It is great that they admit to having this issue and honestly warn the user about it, but I wish they had resolved it instead.
FSB & Memory Config page from the JumperFree Configuration section offers us to enjoy Nvidia’s chipset’s rich functionality in setting the FSB to Memory ratio. It is way more flexible here than what Intel chipsets have to offer.
We can select Linked mode when the memory frequency is tied up to the FSB speed with a few dividers. This operational mode is absolutely identical to the way the memory frequency is set in Intel chipsets.
However, there is also Unlinked mode that offers almost unlimited choices for FSB to Memory frequency. Just set the desired FSB speed in the interval from 533MHz to 3000MHz QDR (they use 4x FSB frequency, i.e. 133MHz to 750MHz interval in conventional measuring units) and the desired memory frequency in the interval from 400MHz to 2600MHz (!). You don’t have to know or remember anything about dividers: the mainboard will find the correct one on its own so that the resulting memory frequency could be as close to the desired one as possible.
In any way Actual FSB (QDR) and Actual MEM (DDR) parameters will report the expected actual FSB and memory frequencies.
Advanced section also offers a page called CPU Configuration, where we can adjust the processor clock frequency multiplier and set up different processor technologies.
Chipset sub-section allows changing the frequency of HyperTransport bus connecting the chipset North and South Bridges. Also you can access the memory timings page from there.
There is a pretty long list of parameters that can be adjusted here. It is very convenient that you can selectively change just a few desired parameters and leave all the other at defaults.
The functionality of other pages in the Advanced section is evident from their names. For example, Onboard Device Configuration sub-section offers tools for managing hard disk drives, sound, network and IEEE1394 controllers, COM and LPT ports.
The most interesting thing in the Power section is Hardware Monitor with great functionality:
We can control processor fan rotation speed only if it has a four-pin connector. You can also adjust the rotation speed of both case fans at the same time. Only the rotation speed of a fan connected to the PWR_FAN connector cannot be changed. Besides, processor Vcore we can monitor all the main voltages from the system PSU - 3.3V, 5V and 12V. We can also monitor processor and system temperatures and rotation speed of all four fans that can be connected to the mainboard.
As for the remaining BIOS Setup sections, you should already be familiar with them from our previous ASUS mainboards reviews. Boot section, as you may have already guessed from its name, allows setting up mainboard boot-up procedure: boot-up devices order, boot-up logo, etc. Tools section has only two sub-sections: ASUS O.C. Profile and ASUS EZ Flash 2. The former offers to save two full BIOS settings profiles in the system memory or as a file on any available media. Later on you can load this profile at any time instead of adjusting all the BIOS settings manually. Of course, only two profiles may not always be enough, but it is not a big issue. Too bad you cannot name them or add a description, in case you would forget what they are for. However, a truly unique advantage is that you can save them to a file and then load on another mainboard.
As for ASUS EZ Flash 2, it is a very convenient built-in utility with graphics interface for quick and easy BIOS updating. We used it to update our default BIOS version 0302 to the latest available version 0502 at the time of tests. We didn’t notice any significant differences between them, although we suspect that JumperFree Configuration section no longer had SLI-Ready Memory parameter in the new BIOS version.
So, the BIOS of ASUS P5N-D mainboard has everything one may need for successful overclocking. However, before we get to check this functionality out in practice, we would like to point out a few BIOS drawbacks that are pretty typical for all ASUS mainboards.
ASUSTeK is a recognized leader in the mainboard business. They design and manufacture mainboards with excellent features. Some of them are truly unique and some boast terrific overclocking potential. Unfortunately, there are a few serious drawbacks that keep migrating from one BIOS version to another and may be typical of all ASUS mainboards. I believe there are three major problems:
If the latter two issues pointed out are definitely drawbacks, then the first one is actually arguable. Of course, “smart” ASUS BIOS that can change frequencies and voltages on its own to ensure system stability is great help to commencing overclockers. I have very often come across the following messages in the forums: “I have just increased the FSB frequency and everything works!”. However, the newbie has absolutely no idea how much work the mainboard has actually done for him in this case: it reduced the memory frequency, increased the processor Vcore, Vmem and Vchipset, and corrected the timing settings. The first-time overclocker doesn’t even have to know about it: “smart” ASUS BIOS will do everything for him – a really great achievement.
However, no matter how smart the BIOS is, it cannot know the actual potential f the computer hardware that an experienced overclocker knows. Parameters that suit perfectly for average overclocking may be insufficient or excessive in some specific cases. Together with the other drawbacks it hinders overclocking on ASUS mainboards a lot and sometimes may affect the buying decision not in ASUS’ favor.
To back up my words, I would like to offer you an example. Not so long ago I put together two systems on two different mainboards with two different Conroe-2M processors. Totally by coincidence both processors could overclock to the same maximum frequency of 3150MHz with identical Vcore settings. However, the results of this overclocking experiment turned out different for both processors.
The first system was built with ASUS Commando mainboard on Intel P965 Express and Intel Core 2 Duo E4300 processor (1.8GHz, 200MHz FSB). If you didn’t know or forgot that it is an excellent mainboard, then check out our article called Asus Commando: First Look at a Dream Mainboard. It has a well-known drawbacks, though: FSB Strap that causes performance to drop starting at 401MHz FSB. However, at 400MHz and below this old mainboard can compete with any contemporary one just fine. However, we were not worried about FSB Strap, because we only needed to increase the FSB frequency to 350MHz in order to push our Intel Core 2 Duo E4300 to 3150MHz with a high clock frequency multiplier of x9. So, how far could we overclock our CPU on ASUS Commando mainboard? – To 2.52GHz.
Of course, the mainboard as well as the CPU could do better than that, but the reasons that led to this result lined up as follows. Since the mainboard cannot control the rotation speed of a processor fan with a three-pin connector (drawback No.3), we used a very powerful but very quiet cooler, even at maximum rotation speed. So, to ensure that the cooler can cope with its task, we decided to overclock without raising the nominal processor core voltage. However, starting with 281MHz FSB the “smart” ASUS BIOS (drawback No.1) started raising the Vcore on its own, so we had to stop at 280MHz FSB, which prevented us from overclocking any further. It is also impossible to set the CPU Vcore at a fixed rate, because in this case processor power-saving technologies get turned off (drawback No.2).
Combination of these three drawbacks didn’t let us overclock the CPU to the maximum and cost us 630MHz of processor speed. It is a pretty significant loss, far not every processor can actually overclock that far and in our case this practically guaranteed overclocking result was taken away from us. However, when we overclocked our second system built on MSI P35 Platinum mainboard with Intel Core 2 Duo E6300 processor (1.86GHz, 266MHz FSB), there were no problems at all.
I am not going to idealize MSI mainboard based on this result and when I said we didn’t have any problems, I actually embroidered the reality a little bit. MSI P35 Platinum mainboard has very efficient chipset cooling system and a number of drawbacks that haven’t gone anywhere since we posted our review of this solution last year (for details see our article called Experience the Roller Coaster: MSI P35 Platinum Mainboard Review).
The mainboard can unexpectedly increase the NB voltage when the FSB frequency increases. During overclocking we need to reset the jumpers to trick FSB Strap and avoid performance drop, but it will inevitably catch up with us once we hit 514MHz FSB. However, we were not worried about this unique drawback typical only for MSI P35 Platinum and P35 Combo mainboards, because we only needed to increase FSB to 450MHz in order to overclock our Intel Core 2 Duo E6300 to 3150MHz frequency. Besides, the mainboard also cannot control the rotation speed of a three-pin processor fan, so our overclocking was limited by the nominal CPU Vcore and the cooler rotation sped was reduced to minimum using an external fan rotation speed controller.
I mentioned far not all the drawbacks of MSI P35 Platinum mainboard here. Timing settings are also very inconvenient: you can either set all of them to Auto or have to set all of them manually. It is hard to connect more than two hard disk drives because two chipset Serial ATA connectors exist in the rear panel as eSATA ports, and the other two are blocked by the graphics card cooler. However, once we resolved all the issues, we ended up with a very quiet system where the CPU was working at its maximum possible frequency in these conditions of 3150MHz and had all its power-saving technologies up and running, so that the CPU Vcore and clock multiplier could be lowered in idle mode.
So, what mainboard is better for overclocking: excellent ASUS Commando, or not very convenient to work with MSI P35 Platinum? I would have immediately chosen the former, but it turns out it is in fact the latter. All they really need to do is fix the following three issues to take the last chance away from the competitors:
And while they haven’t done it yet, let’s see how well ASUS P5N-D mainboard will cope with CPU overclocking.
Our CPU overclocking experiments were performed in the following test platforms:
I have to say that we haven’t heard anything inspiring about Nvidia nForce 750i SLI based mainboards. They were claimed to have lower maximum stable FSB frequency than Intel based mainboards, to have performance hits and non-operational frequency intervals. At first we decided to find that maximum FSB frequency when the mainboard would remain stable. We lowered the processor clock frequency multiplier, set the memory into synchronous mode, increased the processor Vcore, Vchipset, Vmem and HyperTransport voltage. Just in case we lowered the HyperTransport bus frequency.
We failed to get the board to boot at 450MHz FSB. AT 425MHz FSB we loaded Windows, but the system hung right after the CPU-Z launch. And even at 400MHz FSB the mainboard didn’t seem to work confidently. However, everything changed once I started increasing the FSB frequency instead of reducing it. The board passed all Prime95 short-term tests at 425MHz and then at 450MHz and even 475MHz FSB. We had to increase NB voltage from 1.4V right to 1.6V only to ensure stability at 500MHz FSB. However, it turned out to be the maximum: the mainboard wouldn’t stay stable neither at 525MHz nor at 510MHz FSB.
Well, 500MHz FSB is not enough for overclocking the youngest CPUs with low clock frequency multipliers, so it is not the best result we could have achieved. However, we obtained it in very sparing conditions, when the CPU frequency didn’t exceed the nominal, since we lowered the clock frequency multiplier. So, will this mainboard be able to overclock a dual-core processor?
The results of our Intel Core 2 Duo E8400 overclocking experiments performed on different mainboards vary between 4.05GHz and 4.1GHz. In other words, for the nominal x9 frequency multiplier the FSB speed should be pushed to 450-455MHz. The system failed at 455MHz FSB: the mainboard started but didn’t boot Windows. However, at 450MHz FSB the system remained stable in Prime95 load test for some time.
Partially satisfied with this preliminary result I decided to continue next day. However, in the morning the mainboard refused to boot at all, although nothing has been changed since last night. I recalled that last night I managed to get to pretty high FSB frequencies by starting low. So, I lowered the FSB setting. The mainboard worked for some time, warmed up and started just fine at 450MHz. As for 455MHz, I still couldn’t get the board to work at this FSB frequency.
I can hardly imagine an overclocker who would be willing to “warm up” the mainboard at low FSB speed, like a car in winter, in order to later on enjoy the advantages of maximum overclocking. Looks like ASUS P5N-D cannot really boast much here.
I had to check out one more rumor about Nvidia nForce 750i SLI chipset claiming that somewhere between 450-480MHz FSB there is FSB Strap, i.e. the performance drops dramatically. We already know the parameter settings with which ASUS P5N-D mainboard is operational up to 500MHz FSB frequency. So, we used these settings, lowered the processor clock frequency multiplier, locked the memory timings. We tested the memory subsystem in Everest program starting with 400MHz FSB and moving up with 10MHz increments in order to detect at what frequency the performance will drop. Up to 440MHz performance grew gradually and at 450MHz… No, no drop, the mainboard just stopped working.
I would have also stopped working with an unstable board like that, but I had to check out its performance during quad-core processors overclocking. If it couldn’t overclock a dual-core CPU, then it should definitely fail the quad-core. So, the obtained results were not surprising at all. I used an Intel Core 2 Quad Q9300 that is capable of working at 475MHz FSB on good mainboards. My attempt to start the system with this processor failed at 440MHz FSB, as well as at 430MHz. I decided not to go any lower, because it didn’t really make any sense already.
I could probably get very frustrated with overclocking-unfriendly ASUS P4N-D mainboard. However, these mainboard tests performed a purely control function. A few days earlier I checked out MSI P7N SLI Platinum mainboard based on the same Nvidia nForce 750i SLI chipset. The results of our overclocking experiments were very disappointing: maximum stable FSB frequency for a dual-core processor was only 485MHz, and a quad-core CPU didn’t overclock at all. That is why I decided to check out ASUS P5N-D mainboard to see if it was MSI’s fault or Nvidia’s weak chipset. Looks like a lot of problems lie in the chipset. No wonder, since it is built on not very new Nvidia nForce 6 core logic.
However, in order to make any final conclusions I had to test a reference mainboard and luckily, I got this opportunity. I believe “FTW” in the name of EVGA nForce 750i SLI FTW mainboard stands for “Engineered For The Win”. Or maybe the slogan was offered for the abbreviation, it doesn’t really matter. What matters is the mainboard’s inability to overclock.
The board looks very attractive, but is hardly an overclocker product. ASUS mainboards have really smart BIOS. It doesn’t interfere during not very aggressive overclocking and then starts increasing the processor and chipset voltages little by little. The BIOS of EVGA nForce 750i SLI FTW mainboard is not that smart. It raises the voltage even during most insignificant overclocking disabling power-saving technologies in this case. It is even worse than by ASUS mainboards. Besides, with the Intel Core 2 Duo E8400 processor multiplier lowered to x6 the mainboard could only start and boot the OS at 425MHz FSB, not any higher. So, no wonder that ASUS an MSI preferred to develop their own mainboard design, although they still failed to cope with the tricky Nvidia nForce 750i SLI chipset.
Mainboards based on the same chipsets usually perform equally, as well. If we compare systems with mainboards on different chipsets, the performance difference may be pretty dramatic. It is especially interesting in comparisons like that to ensure extreme working conditions during overclocking. In this case you can clearly see the advantages and drawbacks of the mainboard and chipset.
At first we were going to compare the performance of ASUS P5N-D mainboard against some mainboard on Intel P35 Express during maximum overclocking of Intel Core 2 Duo E8400 CPU to 4.05-4.1GHz. Unfortunately, we couldn’t do it because of unstable overclocking performance of our ASUS P5N-D mainboard. Luckily, the mainboard turned out capable of working normally during overclocking to 400MHz FSB. Do not be surprised that I said “luckily” here. MSI P7N SLI Platinum, for instance, refused to work stably in the same conditions although it could overclock the CPU to much higher speeds. Looks like each Nvidia nForce 750i SLI mainboard has its own peculiarities.
We couldn’t set Command rate to 1T because in this case the system would very quickly report errors. We couldn’t do it for abit IP35 Pro either (we used this mainboard for comparison), although for a different reason. The mainboard worked perfectly fine with 1T Command Rate setting, but none of the utilities except MemSet saw it. All other utilities reported Command Rate of 2T. It was very easy to find out, who was right here: the performance of abit IP35 Pro remained the same at 1T. It meant that the setting indeed remained at 2T no matter what was set in the mainboard BIOS. abit IP35 Pro mainboard is known to overclock processors brilliantly, but when it comes to working with memory it yields to many other solutions out there. And we have once again seen proof of that.
So, we overclocked our Intel Core 2 Duo E8400 to 3.6GHz on both mainboards. The memory divider as set at 1:1, so it worked as DDR2 800 with 4-4-4-12-2T timings. Here are the obtained results:
We can see significant advantage ASUS P5N-D demonstrates in read speed and latency, however in other tests there is no advantage to be noticed. Overall, both boards are equally fast taking turns in taking the lead. Of course, Nvidia nForce 750i SLI based board will lose to an Intel P35 Express based one that can work with memory a little better than abit IP35 Pro and can adjust Performance Level parameter that affects the performance a lot.
I was very surprised that ASUS pays so much attention to their new ASUS EPU power-saving technology in general and ASUS P5N-D mainboard in particular. Even the marketing slogan for this board says: “The best platform combining powerful performance with great energy efficiency". ASUS EPU (Energy Processing Unit) technology is mentioned on both sides of the mainboard package, in the brief list of specifications and in the detailed spec sheet on the company web-site. This technology allows changing the number of active phases in the processor voltage regulator circuitry depending on the current workload thus saving power. You can even watch a funny video about gardens blooming on our planet thanks to all the saved energy. And when we saw the first results of ASUS P5K-SE with ASUS EPU technology and Gigabyte GA-EP35-DS3L using similar DES (Dynamic Energy Saver) technology and both mainboards kind of failed, the companies responded with indignant press-releases.
ASUS can certainly doubt the credibility of the test results performed by their competitor. There may appear alternative comparisons or it may even get to the next level – a legal level. I personally am not so interested in the outcome of this duel, although I do understand the importance of power-saving technologies and am eager to encourage their promotions and development.
Each of these two competing technologies has its cons and pros. As you remember, Gigabyte DES is more flexible in changing the number of active phases, while ASUS EPU uses a two-step adjustment algorithm. However, DES requires a special utility to be installed and running constantly, because, this technology doesn’t work with an overclocked processor. So, for me any pretty much any other overclocker Gigabyte DES is useless. You can read more about Gigabyte DES in our article called Do We Really Need Intel X48 Express? - Gigabyte GA-X48T-DQ6 Mainboard Review.
We have also discussed ASUS EPU in one of our previous articles, namely Asus P5E Mainboard Review and as far as I remember, we were not very enthusiastic about it either. Of course, this technology does work and does save power, although there was no significant result. They keep stressing the advantages of ASUS EPU technology, while keeping quiet about typical drawbacks of most ASUS mainboards eliminating the power-saving gain.
As you know, power-saving technologies do not work on Intel based ASUS mainboards if the processor core voltage has been increased above the nominal. As a result, if you change the processor voltage during overclocking, any ASUS mainboard will consume more power than a similar mainboard from another manufacturer with Intel power-saving technologies up and running and lowering the voltage and multiplier in idle mode.
That is why it is interesting to compare ASUS EPU against Gigabyte DES only from the marketing prospective. It may also be interesting from a legal standpoint, too, if the fight goes too far. However, mainstream overclockers do not really care about these technologies, even though they were initially targeted for them. That is why I wouldn’t even have thought of checking out the power consumption if it hadn’t been for ASUS’ insistent promotion of EPU’s advantages. For that reason and also since I had three different mainboards on the same chipset, it was very interesting to see how their power consumption rates compare.
For our tests we used Extech Power Analyzer 380803. The device is connected before the PSU, so it detects the power consumption of the entire system without the monitor. It even detects the losses in the PSU. It may be a disadvantage for ASUS EPU efficiency estimates, because it only works for the processor voltage regulator circuitry and any insignificant power savings may simply get lost against the background of the entire system. However, it is power consumption of the entire system and not that of its individual components that matters for us, end-users.
Our testing methodology was very simple. We only replaced the mainboards and CPUs in our system case. We loaded the most optimal BIOS settings, without any additional changes. The only thing we did was disabled the FDD, because there was non in our testbed, and enabled power-saving technologies if they were disabled by default. Once Windows Vista loaded I waited until the system became idle and recorded the power data. We used FPU-test from S&M 1.9.1 utility to measure the system power consumption under heavy workload:
As you can see, the reference mainboard on NVIDIA nForce 750i SLI chipset was the most economical. MSI P7N SLI Platinum and Asus P5N-D mainboards consumed about the same amount of power, but ASUS mainboard still required more. We saw the same with a different processor. For example, here are the results taken with Intel Core 2 Duo E4300 CPU:
CPUs drop their clock frequency multiplier to x6 and their core voltage to 1.1V in idle mode that is why the systems with different processors showed pretty similar power readings in these operational conditions. Although Intel Core 2 Duo E4300 CPU works at only 1.8GHz compared to a 3.0GHz Intel Core 2 Duo E8400, the system equipped with the former processor consumes more power, since its Vcore is higher: 1.325V against 1.225V by E8400.
It is hard to tell why reference mainboard turned out the most economical. All three mainboards have very similar formal specifications. For example, all of them have an additional IEEE1394 controller. The only difference is passive North Bridge cooler with heatpipe technology on MSI board. Maybe it is the absence of a fan that determined the slight difference in ASUS and MSI results.
Anyway, we don’t see ASUS P5N-D being economical. ASUS stresses that EPU technology is a hardware solution, which means it doesn’t require any drivers or special utilities. However, you will still need them in order to adjust the power-saving modes manually. So, let’s install ASUS EPU driver and AI Suite tool with the AI Gear 3 utility that we need.
Pretty gloomy AI Suite allows controlling frequencies, voltages, temperatures and rotation speeds of the monitored fans.
AI Suite includes Q-Fan 2 utility that allows controlling the rotation speed of connected fans. AI Booster utility overclocks the system, and AI Nap – switches the system into sleep mode. However, we are only going to talk about AI Gear3 utility at this time.
Once we completed the installation there appeared a tray icon corresponding to the selected power-saving mode that reported the current level of power consumption. That’s all this icon does. It cannot even launch AI Gear3. The only thing you can do with it is disable the pop-up windows.
AI Gear3 utility interface looks as follows:
At first you have to calibrate by clicking the system case and magnifying glass icon in the upper right corner. The utility will test through all modes real quick. The wrench icon opens Settings page where you can set the time when the system should switch to AI Nap mode.
The airplane icon indicates the default High Performance mode. It is the regular operational mode for your system CPU: in idle mode the clock multiplier will drop to x6 and under workload it will go back to its default value. It is really strange that a common mode like that got such a loud name. There is one drawback though: Intel power-saving technologies do not work as they should in this case. Namely, the clock frequency multiplier will be lowered, while the voltage will remain unchanged.
A car icon stands for Medium Power Saving. In this case the mainboard acts tricky: it lowers the FSB frequency and voltage below the nominal. We tested out this mode with an Intel Core 2 Duo E4300 processor: its default 200MHz FSB was lowered to 190MHz. for Intel Core 2 Duo E8400 the 333MHz FSB dropped to 316MHz. everything else remained the same: the frequency multiplier dropped to x6 in idle mode and then returned back to the nominal setting of x9.
Only in Max. Power Saving mode marked with a pedestrian the system consumed very little power. In this case the FSB frequency remained low, the voltage was set a little below the nominal value and the CPU is forced to stay in its most economical mode: the multiplier is set at x6 and doesn’t increase even under workload. However, you save the power at the expense of performance. We didn’t buy a CPU to have it running half-way.
But the funniest thing was Max. Performance mode marked with a rocket. In this case the FSB frequency increases to 210MHz for CPUs with 200MHz FSB and to 350MHz for those with 333MHz FSB and they call it “Turbo-mode”. The only reasonable mode is marked with a star and is called Auto. As you may have already figured out, the system will set the most appropriate power-saving mode depending on the CPU workload.
So, ASUS EPU technology replaced standard power-saving technologies from Intel and expanded their functionality by dropping the voltage below the nominal. It works in exactly the same way as the utility for Gigabyte DES. However, it could be a nice software replacement for standard power-saving technologies from Intel, since they get disabled on ASUS mainboards during overclocking anyway. Even though AI Suite, EPU driver and Nap driver will be always on. Unfortunately, AI Gear3 utility shuts down during overclocking, just like Gigabyte DES. So, it is absolutely useless for overclockers.
Attention: In fact, you’d better not install AI Suite at all or use some uninstaller tool. You can delete EPU and Nap drivers, but the aaCenter.exe file from AI Suite will stay in the memory forever and will remind of itself by the following error messages:
The power consumption difference between MSI and ASUS mainboards remains small only if the systems are not overclocked or overclocked just a little bit. Once we overclocked Intel Core 2 Duo E4300 processor from 200MHz to 350MHz FSB without even adjusting the voltage, the situation changed dramatically:
MSI P7N SLI Platinum mainboard sticks to the overclocking settings: 151W in idle mode, 216W in burn mode. However, on ASUS P5N-D “smart” BIOS jumps in. it increases the processor Vcore beyond 1.4V, so that the mainboard starts consuming 161W in idle mode and 234W in burn mode. Even locking the voltage at the nominal 1.325V doesn’t save the situation. Under heavy workload the mainboards run neck and neck in terms of power consumption. However, in idle mode only the frequency multiplier lowers on ASUS board, but not the voltage, so the board again turns out the most power-hungry.
It is hard to estimate the value of ASUS EPU technology. You can notice it working only if you use AI Gear3 utility and the default processor voltage is lowered. But it is a hardware technology and should work automatically, shouldn’t it? In this case, if the savings are so low that we can barely notice them, then the technology is a failure. But maybe it does work and without it ASUS mainboard would be consuming even more power? In this case, the technology is a success. However, in any case, ASUS P5N-D mainboard cannot really boast any “great energy efficiency”. What the users care about most is the outcome, and it shows that ASUS P5N-D mainboard is still the most resource-hungry with or without EPU technology.
So where do the claimed saving rates of 58.6% and even 80.23% come from? Of course, different mainboards will have different power savings, but how did they actually calculate the percentage? The answer to this question is available in ASUS marketing materials. Namely, the number 80.23% is mentioned in a video made for EPU on ASUS P5E3 Premium mainboard. You can even see how they came up with this impressive number:
However, you have to look not at the numbers but at the comments in small gray font. They compared the power consumption of ASUS P5E3 Premium mainboard with that of some other mainboard with disabled Intel power-saving technologies, and this way prove the efficiency of the ASUS EPU. Of course, a mainboard with the CPU clock frequency multiplier and Vcore remaining unchanged all the time will lose to any other board with these technologies activated. And no ASUS EPU is necessary here.
We all understand very well that marketing materials will never get to the root of things, will never study all cons and pros of the promoted solution. Marketing should highlight some features of the product so that it could cover all possible drawbacks and issues. However, in this case they seem to be misleading us deliberately, even though the numbers are indeed correct. According to this marketing promotion, we will buy a mainboard that will be practically as efficient as any other good mainboard on the same chipset. Only if we use additional software we will see the difference, because the core voltage will be reduced below their nominal value and the performance will be lowered in maximum power saving mode. But we will certainly never be talking bout 58.6% or even 80.23% if we compare the boards in fair conditions. And if we overclock our processor, then ASUS mainboard will consume much more power because of the typical issues pointed out above and of the “smart” BIOS. That’s for sure.
The last straw that broke the camel’s back was the power consumption table where we compared mainboards on different chipsets. Let me say a few words about the testing conditions and participants.
All power consumption tests were performed in the same version of Windows Vista OS, only the mainboards on NVIDIA nForce 750i SLI chipset and processors were different. Then I confirmed that ASUS P5N-D mainboard was stable with Intel Core 2 Duo E8400 processor overclocked to 400MHz FSB. But before I started the performance tests I reinstalled the OS, without changing any settings. I installed Service Pack 1, all necessary drivers and applications. The systems were theoretically setup identically, but our tests showed that ASUS P5N-D mainboard consumes 9W less power when tested on a freshly installed Windows Vista SP1 OS.
Of course, system power consumption depends on voltage settings. we have also proven experimentally that it depends on temperatures, too: the higher the temp, the more power is actually consumed. But I have never known that I can save power by simply reinstalling the operating system. Power consumption is not a constant value. It keeps changing all the time, so we can certainly disregard the numbers past decimal point. But you cannot disregard an 8-10W difference. It remained in idle and burn modes, at nominal and overclocked speeds.
abit IP35 Pro mainboard on Intel P35 Express chipset was tested only on the freshly installed OS. ASUS P5N-D was tested on both, just for your reference, although we can only make any conclusions from comparing the results from the same OS version.
Very interesting: at nominal as well as overclocked CPU speeds, the system on Intel P35 Express chipset with 100% CPU utilization by S&M consumes almost as much power as Nvidia nForce 750i SLI based mainboard in idle mode. Even less. So, ASUS P5N-D mainboard cannot be called economical at all, no matter what technologies or features it supports. Simply because it is based on Nvidia nForce 750i SLI.
So, our tests showed that ASUS P5N-D’s slogan "The best platform combining powerful performance with great energy efficiency" is simply not true. The board boasts average performance, has slightly higher power consumption in nominal mode than other boards on the same chipset, noticeably higher power consumption in overclocked mode than the same boards and much higher power consumption than Intel P35 Express based boards. Moreover, it is unstable during overclocking just like other Nvidia nForce 750i SLI based mainboards.
We could have stopped here. Why do we actually need Nvidia nForce 750i SLI based mainboards, if they have already announced Intel P45 Express and the mainboards based on it are already appearing everywhere? They also support PCI Express 2.0 and the preliminary reports claim excellent overclocking potential of these solutions. This is true and most users may actually forget about Nvidia chipsets for Intel processors. But there is a group of users out there who will not be able to give them up: the users of SLI (Scalable Link Interface) configurations.
Nvidia allows SLI support only on Nvidia based mainboards. I didn’t quite understand this decision at first, because this way they sort of limited the graphics cards sales. I am sure there would be a lot of people who could add another Nvidia graphics card onto their Intel or AMD based mainboard… Now that we have checked out three Nvidia nForce 750i SLI based mainboards, I can totally understand why it happened. The company is doing pretty well in the graphics card market, while the limited SLI support forces people to buy Nvidia based mainboards and helps their chipset division stay afloat.
I personally do not like multiple-VGA configurations. This extensive way of increasing performance stumbles upon a lot of issues. Instead of giving us one powerful graphics card, they encourage us to buy a few of them. Of course, it is primarily beneficial for the manufacturer. However, I am still upset that there are no decent Nvidia chipsets for Intel platform. For example, if Intel chipsets had some real competition to combat, they wouldn’t be selling for $70 a piece. And If Nvidia chipsets were really good there would be no need to restrict the SLI support.
There is one more thing: I am very concerned with the growing VGA power consumption, so it would be really interesting to check out Nvidia Hybrid SLI technology. Of course, not the part called GeForce Boost, when a weak graphics card is combined with a weak integrated graphics core and creates something half-weak. It is the second part, HybridPower, that is of greatest interest. It switches between a discrete card and an integrated chipset graphics core to save power and reduce noise. So far, there are a lot of obstacles to overcome yet, but the idea is really interesting.
Nvidia graphics cards supporting HybridPower technology have been out there for some time now. However, there are no Nvidia chipsets for Intel platform with Hybrid SLI support, no mainboards, though I wouldn’t mind getting one for my home system. I only wish the overclocking potential of Nvidia based mainboards improved. Besides, they have to lower their power consumption, too, otherwise the power savings from the graphics card will be simply eaten by the power-hungry mainboard. So, I sincerely wish Nvidia to start making truly reliable, stable and overclocking friendly chipsets, so that all enforced restrictions could be removed and we all lived happily ever after :).