by Ilya Gavrichenkov
06/13/2006 | 10:44 AM
The contemporary chipset market has become very much different from what we remember. There are much fewer active players left, and those who are still there offer their solution without any real alternatives to consider. Take for instance, Athlon 64 platform. Thanks to the implementation of integrated memory controller in these AMD processors, the role of chipsets in today’s systems has become less important.
<%BANNER[article]%>In the old times of Socket A processors the chipset developers could really compete with one another using a pretty significant argument such as overall system performance. Now that the memory controller has been moved into the CPU, the influence of the system chipset on the overall platform performance has almost vanished. We can even claim that if a given chipset doesn’t suffer from any evident issues with the implementation of graphics and other busses for external equipment, then its performance will be exactly the same as that of any other alternative core logic set out there.
The second thing users may potentially care about in the system chipsets besides performance is the set of features it supports. Different chipsets offer support for different number and types of ports for hard disk drives and other external devices, may support faster or slower data transfer protocols or different work modes for the integrated controllers. However, these differences can hardly be regarded as significant, also. Finally, you can always eliminate certain chipset drawbacks with external controllers, which have become an irreplaceable part of every mainstream and high-end contemporary mainboard.
In other words, it is more important these days to pay special attention to the mainboard manufacturer’s reputation and the features of the particular mainboard model, rather than to the chipset it is actually based on. And speaking of the chipset makers, who do best of all in the today’s market, these are the ones with the smartest marketing policy, most attractive chipset pricing for the mainboard makers and on-time introduction of the new interfaces and protocols support.
Continuing our discussion of the chipset solutions for the Athlon 64 processors, we can state that the leader here remains Nvidia Corporation that has easily ousted its competitors from the arena during the transition from K7 to K8 micro-architecture.
Another player in the chipset market that managed to occupy quite a bit of a share there is ATI, which started aggressive marketing promotion of its core logic sets not so long ago. Even though the ATI solutions for hardware enthusiasts appeared a short while back, we have every right to state that they make the main alternative to Nvidia’s nForce these days.
There is one more reason explaining the success of GPU developers in the Athlon 64 chipset market, which is so beloved by extreme gamers and overclockers. The introduction of SLI and CrossFire technologies that allows combining a few graphics accelerators into a single high-performance graphics subsystem also played a significant role in ATI’s and Nvidia’s success in the system chipset market. The selectiveness of the ATI and CrossFire configurations in terms of the chipset used as a basis for the system mainboard limits the available chipset choices for the user. In other words, since a pair of Nvidia graphics cards can only work in SLI mode in an nForce based mainboard, as well as a pair of ATI graphics cards can only work in CrossFire mode in an ATI chipset based mainboard, no discussion of better alternatives is even possible.
Therefore, the user’s graphics preferences are the major factors that determine the chipset choice for the Athlon 64 system mainboard today (if the user intends to enable multi-GPU graphics solutions, of course). Otherwise, the choice of the Athlon 64 mainboard chipset doesn’t really matter, because the chipset will hardly affect the mainboard performance, stability and overall impression it makes. But what does matter, it is the mainboard designers and their level of expertise.
In fact, this is where we could actually end our story, because we have just drawn all the logical conclusions about the chipset choice for the Athlon 64 platform. However, we dare continue our study in order to find out what solutions ATI and Nvidia offer these days for the new Socket AM2 platform. We will discuss the main features of the new solutions and will try to put together a list of recommendations for those users who disagree with us and still address the chipset choice as an important and responsible matter.
First of all I would like to point out that new AMD processors intended for Socket AM2 mainboards do not feature any innovations that should require new sets of core logic. In other words, the new CPUs use the same HyperTransport interface with 1GHz frequency and 16bit width in each direction to connect the CPU and the chipset. As for the DDR2 SDRAM support that has been introduced in the Socket AM2 platforms, it has nothing to do with the chipset, because the memory controller is built into the processor. It means that the manufacturers can still use the old chipsets to design new mainboards for the recently announced Socket AM2 CPUs.
Nevertheless, both – ATI and Nvidia – decided to tie the innovations in the core logic solutions to the recent launch of the first Socket AM2 processors. As a result, most mainboards for Athlon 64 processors with DDR2 SDRAM support will be based on the new chipset models.
Although the ATI core logic for Socket AM2 platform is called CrossFire Xpress 3200, the name that has already been used before, the Canadian manufacturer introduced a new SB600 South Bridge for this chipset. I have to stress that this is a significant improvement of the previous chipset solution, because the previous-generation ATI South Bridges have received a lot of criticism for the quality of some busses and protocols implementation. Moreover, ULi Company that has recently been acquires by ATI’s main rival, Nvidia Corporation, can no longer supply their South Bridge chips for CrossFire Xpress 3200 based mainboards. Therefore, looks like SB600 South Bridge chip is coming in the right time.
As for the features of this microchip, they have changed a lot since the days of SB450. The new South Bridge features an updated Serial ATA and USB 2.0 controller. As a result, firstly, the new implementation of CrossFire Xpress 3200 is free from earlier problems with High Speed USB bus speed and has more USB 2.0 ports now: the total of 10. Secondly, the new ATI chipset has fully-fledged Serial ATA II support, i.e. higher interface bandwidth of 3Gbit/s and NCQ support. As for the RAID functionality, the new South Bridge supports array 0, 1 and 0+1, and later on should also acquire RAID array 5 support, too. Note that the number of Serial ATA II ports supported by the chipset remained the same and equals 4, although there is a tendency toward getting support for more ports, of course.
Speaking of other features offered by the new ATI SB600 South Bridge I would like to point out that it retained the High Definition Audio (Azalia) support.
The networking functions of the ATI SB600 chip have been removed in order to encourage mainboard developers to use fast gigabit PCI Express x1 network controllers on their platforms. By the way, Intel also uses the same tactics in their LGA775 solutions, so no wonder that ATI resorted to it, too.
Just like in Intel’s South Bridges, ATI reduced the number of supports Parallel ATA ports. SB600 features only one port like that, i.e. it allows connecting maximum two Parallel ATA-133 devices.
SB600 South Bridge retained the traditional hardware monitoring controller that is absent by other manufacturers’ solutions and is usually implemented as a separate chip on the mainboard PCB.
The ATI SB600 South Bridge is connected with the chipset North Bridge via the A-Link Xpress II bus that boasts the same architectural design as PCI Express x4. As a result, the maximum bandwidth of the bus between the North and the South Bridge of the CrossFire Xpress 3200 chipset is 2GB/s. You should understand that this connection is no bottleneck here, because all the hi-speed PCI Express busses are implemented in the RD580 North Bridge of the chipset.
In other words, it is the chipset North Bridge that is responsible for two PCI Express x16 busses for CrossFire configurations as well as PCI Express x1 bus of the CrossFire Xpress 3200 chipset.
Just like ATI, Nvidia scheduled the launch of the new chips solutions together with the Socket AM2 platform announcement. The new chipsets also got new names and started Nvidia nForce 5XX series giving you to understand that these solutions are different from the previous-generation ones in terms of supported features.
Nvidia nForce 500 series includes four different products targeted at different price groups. All of them are put together from the same building blocks: MCP chip (Media and Communications Processor) and SPP chip (System Performance Processor). MCP is actually none other but a fully-fledged single-chip core logic set (the corresponding product is called nForce 570). SPP adds the support of additional PCI Express x16 bus for SLI configurations. This way Nvidia can create completely different functional solutions using one or both chips. The chipset with the richest features set is Nvidia nForce 590 SLI, which we are going to review in our today’s article.
Nvidia nForce 590 SLI is a dual-chip solution with the SPP chip inserted between the CPU and the MCP connected via the HyperTransport bus. As a result, nForce 590 supports two PCI Express x16 busses and thus can be used as a basis for SLI platforms. However unlike ATI CrossFire Xpress 3200, the two PCI Express x16 busses of the Nvidia nForce 590 are implemented in two different controllers connected with the HyperTransport bus.
The support of additional graphics bus is the only role of the SPP chip. All other functions are implemented in the MCP chip that can also be used as a chipset controller without any additional companion-chips if the platform doesn’t have to support a pair of PCI Express x16 graphics interfaces.
Besides the support of PCI Express x16 graphics bus and PCI Express x1 bus for additional equipment, MCP also offers a complete set of features typical of every chipset South Bridge. I have to stress that Nvidia chipsets have always been known for extremely rich set of features, and our today’s hero is no exception either. Unlike the previous generation nForce 4, the MCP of nForce 500 series acquired a second gigabit controller, the long-awaited High Definition Audio (Azalia) support and more Serial ATA II ports. Moreover, redesigned software (including new drivers, of course) allowed Nvidia engineers to claim the support of additional technologies. And thanks to the aggressive effort of the company marketing department these technologies are very often regarded as revolutionary innovations, even though they are sometimes not.
Take for instance the built-in network controller that has become dual-port in the current MCP implementation. Besides the 1Gbit/s bandwidth provided by each network port of the pair, the driver allows combining the potential of the two ports into a single connection to double its data transfer rate and increase reliability. Nvidia marketing people named this feature DualNet .
Moreover, the network controller built into the new Nvidia chipset (its driver, to be more exact) supports another technology called FirstPacket . It allows assigning high priority to network packets generated by certain applications. It serves to reduce the network delays during online gaming with nForce500 based platforms as well as to ensure flawless VoIP operation.
The updated Serial ATA II controller of the new Nvidia nForce 500 chipset family can also boast some innovations. Now it can support 6 SATA II ports with the bandwidth of 3Gbit/s, NCQ support and plug-n-play function. The number of Parallel ATA channels supported by the chipset not remained the same. With a lot of chipset developers give up parallel interfaces, nForce 500 MCP features only one PATA channels. Note that Nvidia chipset allows creating RAID arrays 0, 1, 0+1 from Serial ATA as well as Parallel ATA hard disk drives. Note that the RAID functionality of nForce 500 series is united under the MediaShield banner.
Besides the special technologies that serve to enrich the functionality and features of Nvidia nForce 500 controllers and interfaces, there are also a few “general-purpose” technologies. Here I would first of all like to point out LinkBoost – the simplest optimized algorithm that allows improving the performance of the graphics subsystem in case there are two graphics cards working in SLI mode. The idea behind this technology is to overclock by 25% the PCI Express x16 “graphics” busses and that part of the HyperTransport bus that lies between the SPP and MCP chips. This trick allows Nvidia to speed up the data transfer rate between the graphics cards in SLI configurations.
Another technology Nvidia loves talking about in relation to their new chipsets is SLI-Ready Memory . In fact, this feature has nothing to do with chipsets. SLI-Ready Memory technology implies that the mainboard BIOS supports SPD specification extension suggested by Nvidia – Enhanced Performance Profile (EPP) . It allows to set as default not only the major memory module timings, but also a few more indepth parameters, including the voltage and secondary latency settings.
The full list of features of all four representatives of the new nForce 500 chipset family is given in the following table:
Nvidia nForce 570 SLI and Nvidia nForce 570 Ultra are based on a single MCP chip without any additional microchips involved. The only difference between these two solutions is that the first one allows splitting the PCI Express x16 bus into two PCI Express x8 busses, so that it can serve as a basis for mainstream SLI platforms.
Nvidia nForce 550 chipset is a cut-down version of Nvidia nForce 570 Ultra. One of its network controllers is disabled and it only supports 4 SATA II ports.
Now that we have briefly introduced to you the major ATI and Nvidia chipsets for Socket AM2 platform, let’s compare their features and specifications side by side:
ATI CrossFire Xpress 3200 | NVIDIA nForce 590 SLI | |
Processor interface | 16-bit/1GHz | 16-bit/1GHz |
Chipset (bridge chips) | North Bridge: RD580 | MCP: MCP55XE |
Internal bus between chips | A-Link Xpress II | 16-bit/1GHz |
PCI Express lines | 36 (2 x 16, 4 x 1) | 46 (2 x 16, 8, 6 x 1) |
Multi-GPU support | CrossFire | SLI |
Serial ATA ports | 4 | 6 |
Serial ATA 2.0 | Yes | Yes |
RAID support | RAID 0, 1, 0+1 | RAID 0, 1, 0+1, 5 |
Parallel ATA channels | 1 | 1 |
Integrated sound | High Definition Audio | High Definition Audio |
USB 2.0 ports | 10 | 10 |
Integrated network | None | Dual Gigabit Ethernet |
Supported PCI devices | 6 | 5 |
Integrated hardware monitoring | Yes | None |
I would like to point out another curious thing that didn’t get into the table above. I am talking about the algorithms used to make several graphics cards work in SLI or CrossFire configurations. ATI and Nvidia use different approaches to the roles of the chipset North and South Bridges. As a result, Nvidia has two different chips responsible for the implementation of two PCI Express x16 busses, while in ATI’s solution the chipset North Bridge deal with both graphics busses.
In this case the solution from Canadian engineers seems to be more reasonable. The absence of any intermediate busses between the graphics cards ensures efficiency from the latency prospective.
And although Nvidia uses a high-speed HyperTransport bus to connect the chips with one another, it still increases the latency when the GPUs of different graphics cards communicate. Even LinkBoost technology that should speed up the connection between the SLI graphics cards cannot resolve this issue.
However, despite everything we have just mentioned, Nvidia nForce 590 SLI seems to be a more functional solution with richer features set. It means that in order to design a mainboard with the comparable set of features, the mainboard developer who has selected an ATI chipset will have to integrate more onboard controllers. But it shouldn’t affect the final price of the product. ATI offers its CrossFire Xpress 3200 at a slightly lower cost than Nvidia nForce 590 SLI, so the end-price of the ready solutions levels out. Both mainboards will be selling for over $150, and the price will mostly depend on the manufacturer’s ambitions and accessories bundle.
In this chapter we are going to introduce to you those mainboards that were used in our ongoing test session of the ATI CrossFire Xpress 3200 and Nvidia nForce 590 SLI chipsets.
To test the ATI CrossFire Xpress 3200 features ATI Technologies provided us with a reference mainboard known under the codename of Sturgeon.
Although we try to avoid testing reference mainboards in our lab, because they usually have very little to do with what the end-user is going to get from the store, we made an exception for Sturgeon. The thing is that this fish will most likely surface in the market as a Sapphire product very soon.
The features of ATI CrossFire Xpress 3200 chipset have been expanded with two additional PCI Express Serial ATA II RAID controllers SiI3132 that add another four Serial ATA II ports to the already existing four, with a Marvell Yukon PCI Express gigabit network controller, and an IEEE1394 controller. Moreover, the mainboard also features a fully-functional POST controller with a two-digit POST codes indicator.
The mainboard carries 2 fully-fledged PCI Express x16 slots, two PCI Express x1 slots and a PCI slot. It is interesting that the primary PCI Express x16 slot is not really close to the chipset, but has been moved to the left side of the mainboard PCB instead. When you install a graphics card with a dual-clot cooling system into it, you will no longer be able to use the only PCI slot on the mainboard.
As for the mainboard design in general, it has a few very unpleasant drawbacks, unfortunately. In particular, the location of the SATA ports is not very convenient, as they have been placed on the left-hand side of the PCB. Also, the BIOS Setup jumper that serves to reset the BIOS settings is in a very hard to reach place: it sits in front of the PCI Express x16 slot, so you will have hard times reaching for it if there is a graphics card with a massive cooler onboard.
For testers’ convenience ATI provided its Sturgeon mainboard with two micro-buttons: PowerOn and Reset. Also, the mainboard has a lot of diagnostic LEDs not only indicating the powering status of different mainboard knots but also showing the activity on the Serial ATA controllers. Sturgeon has another interesting peculiarity, which we haven’t yet seen in our lab before: the PCB has several rubber pads on the reverse side, so that you can put it on the desk without damaging the polished surface :)
The CPU voltage regulator on ATI Sturgeon mainboard is designed according to four-channel schematics. All MOSFETs in this regulator are equipped with passive heatsinks. The capacitors come from the Japanese Nichicon Company and are among the best quality ones out there.
The chipset North and South Bridges are covered with pretty simple passive aluminum heatsinks to better dissipate the heat. They are more than enough, since the RD580 and SB600 are manufactured with 0.11micron technology and hence dissipate very little heat. It is one of the trumps of the new ATI CrossFire Xpress 3200 that can be used in quiet and economical systems.
As a reference solution for their new chipset, Nvidia nForce 590 SLI, Nvidia is offering a Foxconn mainboard. As far as we understand, Foxconn C51XEM2AA mainboard design was developed under thorough supervision of Nvidia engineers and represents exactly what Nvidia would like to see as an end-product on their new nForce 590 SLI.
However, we decided to test in greater detail another retail product on nForce 590 SLI instead. It is going to be ASUS M2N32-SLI Deluxe mainboard.
This solution reveals all the features of the Nvidia nForce 590 SLI chipset that have been expanded with a few additional controllers. Among them are: IEEE1394 controller and SiI3132 PCI Express Serial ATA II RAID controller. I would like to point out that one of the SATA ports implemented in this controller has been laid out on the rear panel of ASUS M2N32-SLI Deluxe mainboard, so that you can easily connect hard drives to the platform without even getting inside the system case.
As we see, solutions based on the new Nvidia chipset do not need the whole lot of additional microchips besides the chipset. And the mainboard we are talking about right now proves this fact greatly.
ASUS M2N32-SLI Deluxe features two fully-fledged PCI Express x16 slots, PCI Express x4 slot, and three PCI slots. They are located in such a way that the graphics cards with dual-clot cooling systems onboard will completely block one PCI and the PCI Express x4 slot.
As for the overall PCB design, this solution is also not flawless. Since the memory slots are located along the right-hand side of the PCB and since there is pretty bulky chipset cooler installed, the external devices connectors are fit into the available empty spots without much logical thinking to it. For example, the SATA 2.0 ports appeared in the area in front of the secondary graphics card, which can cause potential problems. The additional ATX power supply connector is also in far not the best place: it is practically in the middle of the PCB behind the first PCI Express x16 slot.
However, ASUS M2N32-SLI Deluxe has a few indisputable strengths. First of all, it is high-quality 8-channel processor voltage regulator that can generate high-power current and at the same time retain clear signal.
All MOSFETs are covered with copper heatsinks with very large cooling surface.
More than that, the heat dissipation on ASUS M2N32-SLI Deluxe is worth our special attention. The thing is that MOSFET heatsinks located on both sides of the Socket AM2 also serve to dissipate the SPP and MCP heat. To make this entire thing work, ASUS engineers used two heatpipes to connect the small chipset heatsinks with the large heatsinks in the CPU voltage regulator circuitry. As a result, ASUS didn’t need any fans on the mainboard: the necessary airflow is created by the CPU cooling fan.
I would like to stress that although there is a snowflake drawn on one of the heatsinks, the entire heatpipe construction gets really hot. The hottest part of this chain is the chipset MCP that is made with 0.13micron technology and is usually equipped with an active cooler on most nForce 500 based mainboards.
The developers of high-end chipsets have recently started claiming that their solutions are specifically optimized for overclocking. The two products we are talking about today are also no exceptions.
It is quite natural that the developers care about overclocking fans that much: computer enthusiasts is exactly the user group that will always buy the most expensive hardware for their needs. Therefore, ATI and Nvidia do their best to get overclockers interested in their new solutions.
At the same time, it is important to understand that this notorious optimization doesn’t really mean anything. The new chipset can affect the overclocking results only indirectly. The major factors that are essential for successful overclocking are processor, graphics card and system memory. And as for the mainboard and chipset, the best they can do is not to impose any limitations. This should be exactly what the chipset developers are trying to say here: their solutions will not be a bottleneck during overclocking experiments, they will allow to reveal the processor, graphics card and memory potential to the full extent. In other words, ATI and Nvidia guarantee that the boards based on ATI CrossFire Xpress 3200 and NVIDIA nForce 590 SLI will allow you to easily increase the clock generator frequency, memory bus frequency and maybe even HyperTransport and PCI Express bus frequencies. However, it is also very important to understand that the time and effort spent by ATI and Nvidia engineers may be very easily diminished by some mainboard developers, who may not do a good job adding the necessary overclocking friendly options into the BIOS Setup and implementing them properly.
Luckily, the mainboards we reviewed this time turned out to be very well and thoroughly designed. They both boasted great set of overclocking friendly functions, which proved that ATI CrossFire Xpress 3200 and NVIDIA nForce 590 SLI are really well-designed to satisfy the overclocking needs of some hardware enthusiasts.
Both mainboards, ATI Sturgeon and ASUS M2N32-SLI Deluxe, allowed changing the clock generator frequency in very wide range; setting higher processor Vcore, Vmem and Vchipset; managing all latencies and memory settings; adjusting the clock frequency multipliers for the CPU and HyperTransport bus. In other words, both mainboards proved ideal for overclocking. And our practical experiments proved it once again. When we set the HyperTransport frequency multiplier to 3x, the maximum clock generator frequency without losing the system stability reached 338MHz for ATI Sturgeon and 340MHz for ASUS M2N32-SLI Deluxe. It should be more than enough for successful overclocking of any Socket AM2 CPU out there.
|
|
Moreover, the screenshots above indicate that both mainboards also allow very impressive absolute CPU overclocking, so that the processors end up running at very high frequencies.
Note that when Nvidia mentions special overclocking friendly functions they also imply their LinkBoost technology and SLI-Ready Memory with EPP. However, these technologies cannot really improve overclocking by themselves, and they can only make it easier. The first technology mentioned above offers to automatically increase the performance of the bus between the chipset bridges and the PCI Express x16 bus, while the second technology sets the most optimal memory timings (from the manufacturers standpoint) for the installed memory. So, these opportunities should be of primary interest to commencing overclockers, while real enthusiasts will definitely prefer to adjust all the settings on their own.
The software supplied by the chipset developers is also intended to make overclocking simpler. Both, ATI and Nvidia, offer special utilities for their chipsets that will make it easier to adjust the major system parameters, such as clock generator frequency, major voltages and memory latencies.
By ATI CrossFire Xpress 3200 this is ATI System Manager tool. Besides the rich overclocking options available from Windows, this software also offers hardware monitoring tools as well as intellectual fan rotation speed management system.
A similar utility has been developed by Nvidia for their nForce 500 chipset family. It is nTune 5.0, which you should already know if you have ever come across an Nvidia based mainboard. This time it has been advanced with a new interface and got free from a number of related issues. From the functional standpoint, nTune 5.0 offers practically the same things as ATI System Manager.
I would like to stress that nTune 5.0 boasts more user-friendly interface than ATI System Manager. Also Nvidia’s software features a built-in stability test and an automatic overclocking tool that should help first-time overclockers to get a nice start.
To test the performance of the new Socket AM2 chipsets we used the following hardware:
We tested both mainboards with identical BIOS settings adjusted for maximum performance.
We decided to start with the least interesting results: the performance of ATI Sturgeon and ASUS M2N32-SLI Deluxe based on ATI CrossFire Xpress 3200 and Nvidia nForce 590 SLI chipsets respectively in traditional benchmarks.
No wonder that the testing participants based on different chipsets demonstrate very similar results. Contemporary chipsets for Athlon 64 systems do not contain any critical units that can affect the general performance scores. The memory controller is in the CPU, that is why the slight difference between ATI Sturgeon and ASUS M2N32-SLI Deluxe in some applications can be explained by different secondary settings of the memory controller.
In most cases ASUS M2N32-SLI Deluxe mainboard on Nvidia nForce 590 SLI chipset is faster than the competitor, which is not surprising at all, as this is a mass production mainboard made by one of the market leaders. Of course, the developers paid more attention to optimizing this product, so it showed slightly better performance than the reference solution from ATI.
Now let’s move from general performance tests to more interesting discussion of various chipset subsystems. This is where we could really see the difference between ATI CrossFire Xpress 3200 and Nvidia nForce 590 SLI.
We tested the Serial ATA 2.0 controller integrated into the chipsets by connecting one of the today’s fastest hard drives, WD Raptor X (WD1500AHFD) to the tested mainboards. Although this HDD doesn’t support 3Gbit/s data transfer speed, it is a way better than most HDDs out there in terms of all other parameters. Thus, Raptor X supports NCQ, features a 16MB buffer and boasts unprecedented spindle rotation speed for a Serial ATA hard drive – 10,000rpm. We tested the controller with a popular PCMark05 testing suite.
The obtained results demonstrate certain advantage of the Serial ATA controller built into the Nvidia nForce 590 SLI. It can be actually explained by the more advanced driver that caches all typical operations thus ensuring the above mentioned advantage. This supposition is also backed up by the fact that in case the operation is more sensitive to sequential data transfer rate, Nvidia’s controller loses its superiority.
So, we have every reason to state that the SATA controller of the new Nvidia nForce 500 chipset family is more advanced than the solution introduced in the new chipsets from ATI not only from the formal standpoint. For some reason we encountered driver issues that wouldn’t let us test the SATA controller of ATI CrossFire Xpress 3200 chipset in the RAID mode. The RAID array we created on ATI Sturgeon mainboard from the RAID page of the BIOS Setup wasn’t recognized by the Windows drivers for some reason. Therefore, we do not have any comparison of RAID arrays performance to offer you this time.
Low performance with High-Speed USB devices used to be a serious bottleneck of previous South Bridges from ATI. This was actually one of the main reasons why the mainboard makers were eager to replace the ATI’s own South Bridge with the ULi’s. Now that Nvidia owns ULi, ATI can only count on themselves. Let’s see what has happened to the USB 2.0 bus speed in the new ATI SB600 South Bridge. We will also check the bus speed provided by the new Nvidia MCP – MCP55XE.
We used an external Maxtor OneTouch 250GB hard disk drive to tests the practical bandwidth of the USB 2.0 bus.
As we see, ATI managed to resolve the issues with the USB 2.0 performance. The USB 2.0 controller in SB600 South Bridge not only supports more USB ports, but is also working much faster having finally caught up with the performance rates of other controllers. The only remaining issue is relatively high CPU utilization when USB 2.0 bus is loaded heavily, however, it is not very noticeable during work in real applications.
ATI CrossFire Xpress 3200 chipset doesn’t have its own network controller. ATI followed into Intel’s footsteps and suggests that the manufacturers should use external PCI Express x1 gigabit network controllers. The most popular solution here is Marvell Yukon. Therefore, we are going to compare the performance of the integrated Nvidia nForce 500 gigabit network controller against Marvell Yukon used on ATI Sturgeon mainboard. To measure the performance we resorted to NTttcp utility from Microsoft Windows NT DDK.
This test session didn’t reveal any surprising results. The performance of the gigabit network controller integrated into Nvidia nForce 590 SLI chipset is almost identical to that of the Marvell Yukon controller on ATI Sturgeon mainboard.
By the way, it is really interesting to see that the CPU utilization on Nvidia based platform using special hardware resources for efficient TCP/IP packets processing is nearly as high as in case of an external controller. It proves beautifully pure marketing origin of a lot of wonderful technologies of the Nvidia nForce 500 chipset family, doesn’t it?
As we have already said above, Nvidia introduced High Definition Audio (Azalia) in its new solutions fro enthusiasts from nForce 500 chipset family. Let’s compare the performance of this solution with the one implemented in ATI CrossFire Xpress 3200 chipset. First of all we will look at the CPU utilization chart:
According to these results from the 3DMark03 audio-test, the new HAD solution from Nvidia loads the CPUs slightly heavier than the integrated HAD solution in CrossFire Xpress 3200.
As for the sound quality that is determined not only by the chipset parameters and driver quality but also by the vocal tract implementation on each particular mainboard model, we obtained the following results during our tests:
ATI Sturgeon | ASUS M2N32-SLI Deluxe | |
Frequency response (from 40Hz to 15kHz), dB: | +0.26, -0.31 | +0.11, -0.03 |
Noise level, dB (A): | -83.6 | -82.4 |
Dynamic range, dB (A): | 83.4 | 82.1 |
THD, %: | 0.0065 | 0.0048 |
IMD + Noise, %: | 0.019 | 0.021 |
Stereo crosstalk, dB: | -84.3 | -82.2 |
IMD at 10 kHz, %: | 0.021 | 0.022 |
Summing up I can conclude that both mainboards boast very high quality of the sound tract implementation.
When we discussed the features and design of ATI Sturgeon and ASUS M2N32-SLI Deluxe mainboards we pointed out that the North and South Bridges of ATI CrossFire Xpress 3200 based board are equipped with simple aluminum heatsinks, while the Nvidia nForce 590 SLI based platform is equipped with a high-tech cooling system featuring heatpipes and copper heatsinks with a lot of ribs. Looks like the heat dissipation of the new ATI and Nvidia chipsets for Socket AM2 platform should differ dramatically.
Knowing the direct connection between the heat dissipation and power consumption we decided to measure the amount of consumed power by the similar systems built around different sets of core logic. We measured the total power consumption of the complete platform solution except for the monitor in several work modes.
First of all we measured the platforms power consumption in idle mode. Note that we didn’t activate Cool’n’Quiet technology during out tests.
The first results indicate that there is a difference between the two chipsets from the power consumption point of view. In idle mode Nvidia nForce 590 SLI based platform consumes 20W more power. And these 20W can definitely be assigned to the chipset, because the mainboard was the only thing different about the two platforms in question.
Now let’s check out the results of our second experiment when the CPU was fully loaded by the S&M utility.
The difference in power consumption of the two platforms is less dramatic this time: 10W.
The third experiment was carried out when the graphics subsystem was loaded to the full extent by the 3DMark06 testing suite.
In this case the power consumption difference is even more remarkable and reaches 30W.
So, we can state that Nvidia nForce 590 SLI chipset is considerably less economical than ATI CrossFire Xpress 3200. and if you are looking for the quiet and cool system, the Nvidia chipset may not be the right choice to make. ATI CrossFire Xpress 3200 based platform boasts much lower power consumption and heat dissipation.
the detailed testing of the new ATI CrossFire Xpress 3200 and Nvidia nForce 590 SLI based platforms proved that both these solutions are very high-quality products with no significant anatomical or functional issues. In other words, we didn’t reveal any serious problems with either of them that could force us to give any negative feedback. It means that the statements made in the introductory part to this article should be taken really seriously: the choice of Socket AM2 chipset is a secondary task today. In most cases, the question about the chipset choice is answered once you make up your mind about the graphics subsystem.
The question about which chipset is the best one is appropriate only if you are not going to use SLI or CrossFire graphics configurations and do not even consider the possible upgrade in this direction in the future. However, even in this case it is really hard to give any recommendations about the best chipset choice, because ATI CrossFire Xpress 3200 and Nvidia nForce 590 SLI have different strengths and drawbacks, which we have already discussed in great detail today.
And although we get the impression that Nvidia solution offers broader range of features and higher level of performance, and hence should be regarded as a better choice, we cannot agree to that. The majority of Nvidia nForce 500 family’s advantages have pure marketing origins and in reality this solutions doesn’t have any unique outstanding features.
Even though ATI CrossFire Xpress 3200 seems to have fewer attractive features, the mainboard manufacturers can easily avoid all the potential drawbacks by integrating additional onboard controllers. This way, it should be not the chipset but the mainboard manufacturer with the particular platform implementation that should determine your choice of a mainboard solution. Especially, since it is the mainboard manufacturer who is in the end responsible for the end-product features, including overclocking friendliness and stability.
And this is where Nvidia nForce 590 SLI based solutions look more attractive at this time. Mainboards based on this chipset are already available in retail, while ATI can only offer a reference Sturgeon mainboard for reviews. As for Nvidia nForce 590 SLI, the ASUS M2N32-SLI Deluxe mainboard may become an excellent basis for a Socket AM2 system.
In other words, it is still too early to draw any final conclusions. We will make them when we get a chance to see more retail mainboards on ATI CrossFire Xpress 3200 and Nvidia nForcd 590 SLI.
