Intel’s X38 Express chipset

Back in May, Intel launched the P35 Express chipset, a replacement for its popular 965P. Classified as a mid-range product, the P35 added official support for front-side bus speeds up to 1333MHz, came bundled with a new ICH9R south bridge chip, and became the first desktop chipset to support DDR3 memory. Enthusiasts lapped it up, and in the months since, we’ve seen a wave of P35 motherboards that offer great performance and tantalizing overclocking potential.

Intel’s mid-range chipsets have traditionally been excellent, so the P35’s success wasn’t a surprise. That success also raised expectations for Intel’s new high-end core logic chipset: the X38 Express. This isn’t a new trend; the all-around goodness typical of Intel’s mid-range chipsets has long made it difficult for the company’s flagship offerings to shine.

Topping the P35 Express won’t be easy, but Intel has a few tricks up its sleeve with the X38. Chief among them is next-gen PCI Express 2.0 connectivity—a first for desktop chipsets—with enough lanes for dual-x16 CrossFire configurations. As is customary for its high-end chipsets, Intel has also rolled out memory controller optimizations that promise faster performance and support for higher DDR3 memory speeds.

To find out whether these perks are enough to elevate the X38 Express over its blue-collar P35 sibling, we’ve run the first X38 boards from Asus and Gigabyte through a relentless series of memory controller, application, and peripheral performance tests. Read on to see how the X38 fares and what you can expect from the first wave of motherboards based on this new chipset.

Introducing the X38 Express

The meat of the X38 Express is in its north bridge chip, which is what sets it apart from the P35 Express. Intel’s engineers have built several optimizations into the X38’s memory controller, including what Intel calls a “flexible clock crossing architecture” and improved prefetching circuitry. Those optimizations aim to make the X38’s memory controller quicker than that of the P35, which is a tall order considering how aggressively motherboard makers push memory performance on their mid-range boards.

Still, the X38 can count official support for DDR3 memory speeds up to 1333MHz as one advantage over the P35. That chipset only supports DDR3 memory up to 1066MHz, which is, well, slower. Thanks to rather quickly ramping DDR3 memory speeds, a number of DDR3-1333 modules are on the market and ready to take advantage of the X38, as well.

In addition to supporting faster DDR3, the X38 chipset can take advantage of memory modules that feature Extreme Memory Profiles, or XMP. XMP adds additional memory timing and clock speed profiles to DDR3 memory, much like Enhanced Performance Profiles (EPP) do for DDR2 memory. These profiles make it easier for less experienced users to tune or even overclock their memory through a series of profiles rather than forcing them to manipulate individual timings and bus speeds.

DDR3 memory prices remain high, though, with 2GB kits costing more than double that of their DDR2 counterparts at the same speed grade. Fortunately, the X38 also supports DDR2, providing motherboard makers with a measure of flexibility and end users with the opportunity to save quite a lot of money when putting together a new system. You probably won’t find many X38 boards with both DDR2 and DDR3 memory slots, but we expect mobo makers to offer variants of their boards that support each. With the P35 Express, those models typically used the same basic board design, save for the DIMM slots.

X38
Express
P35 Express nForce 680i SLI SPP

Front-side bus
1333/1066/800MHz 1333/1066/800MHz 1333/1066/800MHz

Memory controller
DDR2-800/DDR3-1333 DDR2-800/DDR3-1066 DDR2-800

PCI Express lanes
32* 16 18

Multi-GPU support
CrossFire CrossFire* SLI

Chipset interconnect
DMI DMI HyperTransport

Peak interconnect bandwidth
2GB/s 2GB/s 8GB/s

Nvidia’s high-end chipset remains the nForce 680i SLI. As a result, Intel is the only chipset manufacturer offering support for DDR3 memory. The X38 Express also breaks new ground as the first to offer next-gen PCIe 2.0 support. The X38 Express north bridge offers up a total of 32 PCI Express 2.0 lanes, and those lanes can be split evenly between a pair of PCIe x16 slots for multi-GPU configurations like CrossFire. (Although the X38 could almost certainly act as a very good host to a pair of GeForce cards in SLI, Nvidia continues to lock out third-party chipsets in its graphics drivers.)

According to the PCI SIG, the standards body behind PCI Express, increasing demand for graphics bandwidth was the driving force behind PCIe 2.0. It’s no surprise, then, that gen-two PCIe lanes offer 1GB/s of bandwidth—double that of their PCIe 1.1 predecessors. This increase in bandwidth comes courtesy of faster signaling rather than wider data paths; PCIe 2.0 lanes push data at 5.0GT/s where their predecessors were only capable of 2.5GT/s.

Gen 2 PCIe has provisions to save power, as well, including support for software controls that can dynamically throttle lane speeds. Don’t think PCIe 2.0 means you have to go out and buy a new graphics card, though. PCI Express 2.0 is backward compatible in both directions, so 2.0 cards will work in 1.1 slots and 1.1 cards will work in 2.0 slots.

Thanks to the Radeon HD 2900 XT, you may already be familiar with another recent addition to the PCIe picture, the 300W card electro-mechanical (CEM) spec. This change officially blesses the new eight-pin auxiliary power connectors on cards like the 2900 XT. Those eight-pin aux power connections can deliver up to 150W of power, in addition to the 75W of juice routed through the slot itself and the 75W available in a six-pin aux PCIe power plug. Taken together, these three power sources can add up to a healthy 300W for a single PCIe x16 slot. Technically, one doesn’t need a PCIe 2.0 slot in order to make use of 300W worth of power—2900 XTs use six- and eight-pin connectors together now—but we do expect to see the Gen 2 spec and the 300W CEM used more commonly in conjunction.

As you can see in the diagram above, the X38 Express retains Intel’s DMI chip-to-chip interconnect. DMI has been around for a while, and despite offering considerably less bandwidth than the HyperTransport link Nvidia uses in the nForce 680i SLI, we’ve yet to see it suffocate system performance, mostly because Intel doesn’t divvy up PCIe graphics connectivity between two chips like Nvidia does. Outfitting the X38 Express with DMI enables it to connect to Intel’s existing ICH9 series south bridge lineup. Since the X38 is a high-end offering, you’re most likely to find it paired with the ICH9R, which stacks up against the nForce 680i SLI MCP like so.


ICH9R

nForce 680i SLI MCP

PCI Express lanes

6
28

Serial ATA ports

6
6

Peak SATA data rate

300MB/s
300MB/s

AHCI

Y
N

Native Command Queuing

Y
Y

RAID 0/1

Y
Y

RAID 0+1/10

Y
Y

RAID 5

Y
Y

Matrix RAID

Y
N

ATA channels

0
1

Max audio channels

8
8

Audio standard

HDA
HDA

Ethernet

10/100/1000
2 x 10/100/1000

USB ports

12
10

The ICH9R isn’t new—it was launched back in May with the P35 Express—so its PCI Express lanes don’t have that gen-two sheen. Not that it matters. The south bridge’s PCIe lanes will most likely be used for onboard peripherals and expansion cards slots, neither of which require the bandwidth or slot power of PCIe 2.0.

Auxiliary peripheral chips are actually a key ingredient of almost any motherboard that uses an ICH9 series south bridge. Because Intel has jettisoned ATA connectivity in favor of SATA, motherboard makers are forced to farm out ATA support (which they still deem necessary) to third-party silicon. Ethernet duties are usually shouldered by third party peripheral chips, as well. The ICH9R does have a built-in GigE controller, so it should be able to handle Ethernet on its own. However, we’ve yet to see an ICH9-based motherboard make use of the chip’s networking capabilities.

Asus’ P5E3 Deluxe WiFi-AP @n
Bursting at the seams

Manufacturer Asus
Model P5E3 Deluxe WiFi-AP @n
Price (Street) $350
Availability Now

Asus is all over the X38 chipset with the P5E3 Deluxe WiFi-AP @n, a motherboard so loaded with features that letters alone apparently aren’t enough to convey its grandness. High-end Asus offerings have always packed features galore, but the P5E3 breaks new ground with the most extensive array of extras we’ve seen on a motherboard. All this comes at a price, of course. The P5E3 isn’t yet widely available online, but the few outlets currently carrying the board are charging upwards of $350.

With scads of integrated extras to include, Asus’ board designers should be commended for managing to squeeze the P5E3 into a standard ATX form factor. They’ve done a pretty good job with the layout, too, getting little things right like putting the auxiliary 12V power connector along the top edge of the board, out of the way of the CPU socket.

The P5E3 even looks snazzy, decked out with a dark board and multi-colored slots and ports that actually make some sense. Storage-related SATA and ATA ports are red, PCIe 2.0 slots meant for graphics cards are blue, and each memory channel gets its own color. The manual even refers to these colors, which should make system assembly easier for beginners.

Like just about every other high-end motherboard, the P5E3 makes extensive use of heatpipes and passive heatsinks to cool its chipset and voltage regulation circuitry. A quartet of heatsinks surrounds the socket on all four sides, leaving little room for my thick, stubby fingers to get at processor cooler retention clips. Fortunately, the cooling fins are reasonably short. There should be enough room for larger aftermarket coolers that rise up from the socket before fanning out.

Over to the left of the socket you, can steal a peek at the P5E3’s DIMM slots. This board is equipped to handle DDR3 modules up to 2GB, allowing for total system memory configurations up to 8GB.

Moving down the board, we encounter the P5E3’s array of storage ports. The ATA connector and four SATA ports run along the edge of the board, ensuring they won’t be blocked by longer graphics cards. The low-profile south bridge cooler won’t get in the way, either.

Seven expansion slots is a lot to squeeze onto an ATX motherboard, but Asus manages to do it with the P5E3. In addition to two PCIe 2.0 x16 slots, you get a pair each of PCIe 1.1 x1 slots and standard PCI slots. Asus also throws in a first-gen PCIe x16 slot that offers one or four lanes of connectivity, depending on whether the board’s x1 slots are in use.

Pay particular attention to the space between the lower blue PCIe x16 slot and the PCI slot directly below it. There, you can see a small PCB attached directly to the board. This protrusion holds a flash memory chip Asus has loaded with a custom Linux distribution from Splashtop.

When you boot the system, you’re greeted with a mouse-driven interface that provides options to go directly to a web browser or Skype client, enter the BIOS, or proceed to whatever operating system you have installed. Choosing one of those first two options launches the embedded Linux install, sending you to Skype or a web browser within seconds.

Options are limited from there, but the web browser and Skype client are fully-functional. More importantly, the fact that this distro is stored on an onboard flash chip means that it may be possible for users to roll their own custom configurations. We can think of plenty of additional capabilities we’d like to see added to what’s essentially an instant-on operating system, including stress test options for overclockers and the ability to download and flash the latest BIOS revision. But I digress. Let’s get back to the motherboard.

Around the rear, the P5E3’s port cluster is predictably bursting with connectivity options. All the usual suspects make an appearance, including two flavors of digital audio output, eSATA, and Firewire. Curiously, though, you won’t find a PS/2 mouse port. Most mice are USB these days, so the omission isn’t a major concern. However, if you’re running an older KVM switch that lacks USB peripheral support, you’ll be out of luck with the P5E3 Deluxe.

If you happen to be looking for integrated Wi-Fi, the P5E3 is exactly what you want. The board comes with integrated 802.11n capabilities courtesy of an AzureWave AW-NA830 module that sits just above the top PCIe expansion slot.

Asus throws in a couple of Wi-Fi antennas to complement the integrated 802.11n chip. The box also includes an IR remote with several programmable buttons, a couple of chipset cooling fans intended for use with water-cooled systems that lack sufficient airflow around the CPU socket, and an I/O shield with a cushioned inner surface that removes the need for those annoying little metal tabs that always seem to bend the wrong way. Like other Asus boards, the P5E3 comes with handy little blocks for USB, Firewire, and front panel headers that make connecting enclosure wires much easier.

Gigabyte’s GA-X38-DQ6
Surprisingly sedate

Manufacturer Gigabyte
Model GA-X38-DQ6
Price (MSRP) $250-300
Availability Now

While Asus’ P5E3 Deluxe WiFi-AP @n rides the bleeding edge with DDR3 memory and extras galore, Gigabyte’s GA-X38-DQ6 is decidedly more sedate. The company has elected to target DDR2 memory with its first X38 board—a smart move considering the huge price premium that still exists for DDR3 modules. We expect to see a version of the DQ6 equipped to handle DDR3 memory soon, though.

Since the DQ6 favors less expensive memory, it’s only fitting that the board is cheaper than the P5E3. Gigabyte has set an MSRP for the board between $250 and $300, and although it’s still scarce online, initial offerings appear to be at the high end of that range. As one might expect, this lower price tag means you don’t get as many extras as with the Asus board. That’s not necessarily a bad thing, though. Plenty of folks don’t need or want to pay for onboard Wi-Fi and embedded Linux distros.

Most do appreciate a good layout, and the DQ6 delivers that in spades. I can’t think of a single major layout issue that afflicts this board, save for the fact that it’s a little difficult to remove memory modules when longer graphics cards are installed. But then that’s an issue that pops up with just about every motherboard that has more than five or so expansion slots, so it’s easy to forgive.

Of course, the DQ6 has heatpipes and passive heatsinks cooling its chipset and voltage regulation circuitry. This mass of copper pipes and fins is quite a bit more restrained than on the P5E3, and it only rings the CPU socket on two sides. That leaves plenty of room to poke around when installing larger heatsinks, although unlike Asus, Gigabyte doesn’t provide auxiliary cooling fans for use in water- or passively-cooled systems with little ambient airflow.

Flipping the DQ6 reveals additional heatsinks below the board’s CPU socket and its north and south bridge chips. These heatsinks should help to radiate heat from the back of the board, but at least around the CPU socket, they can also make cooler installation a bit tricky. New processor coolers, particularly those bundled with Intel’s retail CPUs, tend to fit a little tightly, warping the motherboard slightly. Motherboards are flexible, so a little flexing isn’t usually a problem. With a stiff metal heatsink strapped to the back of the socket, though, the DQ6 has a lot less give.

Topside, the DQ6’s storage ports are clustered in the bottom right-hand corner of the board. The ports are arranged so that longer graphics cards with wide coolers shouldn’t rob you of SATA connectivity. Color coding differentiates from SATA ports connected to the ICH9R south bridge (orange) and the board’s auxiliary SATA chip (purple). It would be more helpful if Gigabyte actually referred to this color coding in the manual, though.

Around the slot stack, the DQ6 provides a couple of second-gen PCIe x16 slots alongside a trio of old-school PCIe x1s and a couple of PCI slots. Installing a double-width graphics card will only cost you a PCIe x1 slot, and double-width CrossFire configs only cannibalize one more PCI slot on top of that, giving this board a good balance of PCI and PCIe expansion options.

Plenty of connectivity lurks in the DQ6’s port cluster, including a whopping eight USB ports and two flavors of Firewire. You also get a full assortment of analog audio ports, in addition to coaxial and TOS-Link S/PDIF outputs. eSATA ports are curiously absent, but Gigabyte has an ace up its sleeve to handle external Serial ATA devices.

Rather than arbitrarily binding eSATA ports to onboard storage controllers, Gigabyte provides users with a pair of PCI back plates (only one is pictured) that can transform up to four internal SATA ports into external ones. Users can select whether they want eSATA connectivity hanging off the board’s auxiliary storage controller, off the chipset’s ICH9R south bridge, or both. If there’s no need for eSATA, all the board’s storage controllers can be dedicated to internal drives.

BIOS options and tweaking software

Years ago, getting a good array of tweaking and overclocking options in a motherboard BIOS wasn’t guaranteed, even on a high-end enthusiast board. These days, however, motherboard makers are pretty good about exposing timing, clock speed, and voltage controls to users. Asus and Gigabyte have consistently provided feature-rich BIOSes on their enthusiast-oriented products, and their latest boards are no exception.

Despite offering similar functionality, the two boards’ BIOSes do have a few key differences. We’ve summarized the most important options below.


Asus P5E3 Deluxe
WiFi-AP @n

Gigabyte
GA-X38-DQ6

BIOS
AMI Award

Bus speeds
FSB: 200-800MHz in
1MHz increments

PCIe: 100-150MHz in 1MHz increments

DRAM: 667, 800, 835, 887, 1020, 1066, 111, 1333MHz

FSB: 100-700MHz in
1MHz increments

PCIe: 90-150MHz in 1MHz increments


Bus multipliers
CPU: 6x-10x (Core
2 Duo E6750)

MCH strap for:
200, 266, 333, 400MHz
CPU: 6x-8x (Core
2 Duo E6750)

DRAM: 2b, 2.4b, 3.2b, 2d, 2.66d, 3.33c, 4c, 2a, 2.5a, 3a, 4a

MCH strap for:
266, 333, 200, 400MHz

Voltages CPU: 1.1-1.7V in 0.00625V increments

DRAM: 1.5-2.78V in 0.02V increments

NB: 1.25-1.91V in 0.02V increments

SB: 1.05-1.2V in
0.15V increments


CPU PLL:
1.5-2.782V in 0.02V increments


FSB termination:
1.2-1.5V in 0.2V increments


Clock over
charging: 0.7-1.0V in 0.1V increments

CPU: 0.5-2.35V in 0.00625V increments

DRAM: +0.05-1.55V in 0.05V increments

PCIe: +0.05-0.35V in 0.05V increments

FSB: +0.05-0.35V in 0.05V increments

MCH overvoltage: +0.025-0.375 in 0.025V increments
MCH ref: +/- 10%

DDR ref: +/- 10%
DDR termination: +/- 0.05V
CPU GTL ref1 -3-6% in 3%
increments
CPU GTL ref2 -3 – -9% in 3% increments


Monitoring
Voltage, fan
status, and temperature monitoring
Voltage, fan
status, and temperature monitoring

Fan speed control
CPU, chassis CPU

Some of these differences, such as the range of front-side bus and PCIe speeds, are largely academic. Others, like how each controls the memory bus speed, are mostly semantic. For example, Asus separates its MCH strap control from its memory bus speed options, while Gigabyte combines the two. In the Gigabyte BIOS, DRAM multipliers each come with a corresponding letter that denotes an MCH strap for either 266, 333, 200, or 400MHz.

You’ll notice that both boards offer front-side bus and MCH strap options for effective 1600MHz front-side bus speeds. Support for 1600MHz front-side bus speeds isn’t a part of Intel’s official specs for the X38 chipset, although that may be because the company has yet to officially unveil processors with a 1600MHz FSB.

On the voltage front, the DQ6 has a distinct Vcore advantage. The board supports CPU voltages up to a heart-stopping 2.35V, giving liquid-nitrogen-cooled overclockers additional headroom over the P5E3 Deluxe. That headroom is unlikely to come in handy if you rely on traditional cooling methods, though. The P5E3’s array of overvolting options are more than adequate for the rest of us.

Asus comes out a little ahead of Gigabyte when it comes to BIOS-level fan speed control, but only because that control extends to a system fan header in addition to the CPU. Surprisingly, neither board offers the ability to set individual fan voltages or reference temperatures in the BIOS. Only Asus offers temperature target control, and then only for the system fan, not for the CPU.

Perhaps the most interesting differences between the Asus and Gigabyte BIOSes are the unique features offered by each. The Asus, for example, provides support for configuration profiles that can be saved and loaded from the BIOS interface. Gigabyte, on the other hand, offers Quad BIOS—a feature that allows for BIOS recovery from a backup chip on the board in addition to copies stored on the driver CD and system hard drive.

If you’re a little shy about poking around in the BIOS, both Asus and Gigabyte offer tweaking and hardware monitoring software for Windows.

Gigabyte wraps tweaking and monitoring capabilities into a single EasyTune application with a clumsy, almost sloppy interface. The functionality is there, but the presentation leaves much to be desired—even the color gradients are rough and pixellated. There’s no accounting for taste, of course, but I’d much prefer a clean, simple interface.

Asus does a little better on the interface front, splitting tweaking and hardware monitoring between its AI Suite and PC Probe applications. The number of overclocking and tweaking options presented is roughly on-par with EasyTune, but PC Probe offers much more extensive hardware monitoring capabilities.

Specifics on specifications

Enthusiasts can be sticklers for detail, so we’ve provided the full specifications for the P5E3 and DQ6 below. This doesn’t make for particularly stimulating reading, of course, but there are a few highlights worth noting.


Asus P5E3 Deluxe
WiFi-AP @n

Gigabyte
GA-X38-DQ6

CPU support
LGA775-based
Celeron, Pentium 4/D, Core 2 processors
LGA775-based
Celeron, Pentium 4/D, Core 2 processors

North bridge
Intel X38 Express Intel X38 Express

South bridge
Intel ICH9R Intel ICH9R

Interconnect
DMI (2GB/s) DMI (2GB/s)

Expansion slots
3 PCI Express x16

2 PCI Express x1

2 32-bit/33MHz PCI

2 PCI Express x16

3 PCI Express x1

2 32-bit/33MHz PCI


Memory
4 240-pin DIMM
sockets

Maximum of 8GB of DDR3-667/800/1066/1333 SDRAM

4 240-pin DIMM
sockets

Maximum of 8GB of DDR2-667/800/1066 SDRAM


Storage I/O
Floppy disk

1 channel ATA/133 via JMicron JMB363

6 channels Serial ATA with RAID 0, 1, 10, 5 support

Floppy disk

1 channel ATA/133 via ITE8718

6 channels Serial ATA with RAID 0, 1, 10, 5 support
2 channels Serial ATA
with RAID 0, 1 support

Audio 8-channel HD audio
via ICH9R and Analog Devices AD1988B codec
8-channel HD audio
via ICH9R and Realtek ALC888 codec
Ports 1 PS/2 keyboard

6 USB
2.0 with headers for 4 more

1 RJ45 10/100/1000 via Realtek RTL8110SC
1 RJ45
10/100/1000 via
Marvell
88E8056
1 802.11n Wi-Fi via AzureWave AW-NA830
1 1394a Firewire via Agere
L-FW3227 with header for 1 more
2 eSATA via JMicron JMB363


1 analog front out

1 analog bass/center out

1 analog rear out

1 analog surround out

1 analog line in

1 analog mic in
1 digital coaxial S/PDIF output
1 digital TOS-Link S/PDIF
output

1 PS/2 keyboard

1 PS/2 mouse

8 USB
2.0 with headers for 4 more

1 RJ45 10/100/1000 via Realtek RTL8111B

1 RJ45 10/100/1000
via Realtek RTL8111B
2 1394a Firewire via Texas Instruments TSB43AB23
with header for 1 more


1 analog front out

1 analog bass/center out

1 analog rear out

1 analog surround out

1 analog line in

1 analog mic in
1 digital coaxial S/PDIF output
1 digital TOS-Link S/PDIF
output

IDE support probably isn’t a concern for most folks putting together high-end systems; SATA optical drives are common enough that there’s no reason to use clumsy ribbon cables ever again. Still, it’s interesting that the Gigabyte board gets its IDE support from a dedicated ITE chip while the Asus settles on one from JMicron that also provides a SATA controller for the board’s eSATA ports.

There are also differences on the networking front, and not just the P5E3’s inclusion of 802.11n Wi-Fi. The Asus board splits its wired networking between Marvell and Realtek networking chips, with the latter tied to the PCI bus. We’ll see how that affects performance a little later, but keep in mind that both of the DQ6’s Realtek GigE chips have a PCIe interface.

Our final spec highlight comes when we look at Firewire, where Asus employs an Agere controller while Gigabyte uses one from Texas Instruments. Both boards support only 1394a, or “Firewire 400.” Gigabyte has dabbled in 1394b “Firewire 800” in the past, but the lack of native Windows XP support for the faster standard created performance problems without the proper hotfixes and third-party drivers. Those complications may have spooked motherboard makers, which could explain why 1394b is all but absent from newer mobos.

Our testing methods

We’ve assembled a small collection of boards for testing. In addition to X38 offerings from Asus and Gigabyte, we also have P35 boards from each company. We’ve stuck with Gigabyte for DDR2 and Asus for DDR3. EVGA’s 122-CK-NF68 is also in the mix because it’s essentially Nvidia’s reference platform for the nForce 680i SLI.

All application tests were run on each board. However, since our Intel-based DDR2 and DDR3 configs all share the same ICH9R south bridge, we’ve limited our Serial ATA, USB, PCI, and PCIe testing to the DDR2 systems.

All tests were run three times, and their results were averaged.

Processor

Core 2 Duo E6750 2.67GHz
System bus 1333MHz (333MHz
quad-pumped)

Motherboard


Asus P5K3 Deluxe
Asus P5E3 Deluxe
WiFi-AP @n
Gigabyte
GA-X38-DQ6


Gigabyte GA-P35-DQ6


EVGA 122-CK-NF68
Bios revision 0604 0201 F5a F6 31

North bridge
Intel P35 Express Intel X38 Express Intel X38 Express Intel P35 Express Nvidia nForce 680i
SLI SPP

South bridge
Intel ICH9R
Intel ICH9R
Intel ICH9R Intel ICH9R Nvidia nForce 680i
SLI MCP
Chipset drivers Chipset 8.3.1.1009

AHCI 7.6.0.1011

Chipset 8.3.1.1009

AHCI 7.6.0.1011

Chipset 8.3.1.1009

AHCI 7.6.0.1011

Chipset 8.3.1.1009

AHCI 7.6.0.1011

nForce 15.08
Memory size 2GB (2 DIMMs)
2GB (2 DIMMs)
2GB (2 DIMMs) 2GB (2 DIMMs) 2GB (2 DIMMs)

Memory type


Corsair CM3X1024-1066C7 DDR3 SDRAM
at

1066MHz


Corsair TWIN2X2048-8500C5 DDR2 SDRAM
at 800MHz
CAS latency
(CL)
7
7
4 4 4
RAS to CAS
delay (tRCD)
7
7
4 4 4
RAS precharge
(tRP)
7
7
4 4 4
Cycle time
(tRAS)
21
21
12 12 12

Audio codec
Integrated
ICH9R/AD1988B with 7.0.0.0 drivers
Integrated
ICH9R/AD1988B with 5.10.1.6110 drivers
Integrated
ICH9R/ALC899A with 1.78 drivers
Integrated
ICH9R/ALC889A with 1.78 drivers
Integrated
nForce 680i/ALC885 with 1.78 drivers
Graphics

GeForce 7900 GTX 512MB PCI-E
with ForceWare 163.69 drivers
Hard drive
Western Raptor X 150GB
OS

Windows Vista Ultimate x86
with KB936710, KB938194, KB938979, KB940105
updates

Thanks to Corsair for providing us with memory for our testing. 2GB of RAM seems to be the new standard for most folks, and Corsair hooked us up with some of its 1GB DIMMs for testing.

All of our test systems were powered by OCZ GameXStream 700W power supply units. Thanks to OCZ for providing these units for our use in testing.

Finally, we’d like to thank Western Digital for sending Raptor WD1500ADFD hard drives for our test rigs. The Raptor’s still the fastest all-around drive on the market, and the only 10K-RPM Serial ATA drive you can buy.

We used the following versions of our test applications:

The test systems’ Windows desktop was set at 1280×1024 in 32-bit color at an 85Hz screen refresh rate. Vertical refresh sync (vsync) was disabled for all tests.

All the tests and methods we employed are publicly available and reproducible. If you have questions about our methods, hit our forums to talk with us about them.

Memory performance

Memory subsystem performance doesn’t always track with real-world applications, but it’s a good place to start. Can the X38 Express’s memory controller offers tangible gains over that of the P35?

Not in these synthetic tests. The X38 offers marginally more bandwidth than its P35 counterpart in Sandra, but only with DDR3 memory. Latency scores are too close to call. It’s worth noting, though, that Nvidia’s nForce 680i SLI achieves the fastest access latencies and highest throughput.

Memory controllers don’t always handle four DIMMs gracefully, so we popped an additional two memory modules into each system for another round of tests. In these tests, we had to back off to a 2T command rate for the nForce and DDR3-equipped X38 systems. This is common adjustment for four-DIMM configurations.

With all the system’s DIMM slots populated, the DDR3-equipped X38 claws its way to the front of the class in Sandra, followed closely by a pair of P35s. DDR2 doesn’t fare so well with the X38 here, as that configuration falls to last place.

Turning our attention to latency, all the Intel boards are bunched together between 60.5 and 61.3 nanoseconds. The nForce retains a clear lead in this test.

The following latency graphs are a little indulgent, so I won’t be offended if you skip them. They show access latencies across multiple block and step sizes, painting a full picture of memory controller performance with each chipset. I’ve arranged the graphs in order of highest latency to lowest. Yellow represents L1 cache, light orange is L2, and dark orange is main memory.

These access latencies are from our two-DIMM configs, and the X38 can’t quite catch the nForce overall.

STARS Euler3d computational fluid dynamics

Few folks run fluid dynamics simulations on their desktops, but we’ve found this multi-threaded test to be particularly demanding of memory subsystems, making it a good link between our memory and application performance tests.

Here the X38 manages to top its P35 predecessor, but only just barely. The difference in performance is more prominent with DDR2 memory than it is with DDR3. Despite its strong showing in memory bandwidth and latency tests, the nForce doesn’t fare well here.

Worldbench

WorldBench uses scripting to step through a series of tasks in common Windows applications. It then produces an overall score. WorldBench also spits out individual results for its component application tests, allowing us to compare performance in each. We’ll look at the overall score, and then we’ll show individual application results alongside the results from some of our own application tests.

We ran into a curious issue with the latest version of WorldBench when testing our P35 system with DDR2 memory. In the first run, the system posted a score in the WinZip test that matched our expectations given the scores of the other systems. However, subsequent test runs registered WinZip test times that were 20% faster than the first run. This behavior wasn’t consistent with that of the other systems, despite the fact that each starts testing with a fresh hard drive image. We suspect Windows Vista’s intelligent disk caching may be responsible for the faster scores, so we’ve reported the results of the first WorldBench test run for the P35 DDR2 config below.

Overall scores are close across the board, with only two points separating the fastest system from the slowest. The X38 Express isn’t able to catch the nForce 680i SLI at the front of the pack, but it scores one point higher than the others when paired with DDR3 memory.

Individual results for Photoshop, Movie Creator, and Windows Media Encoder are all close, with only a few seconds separating the field.

The Firefox and Office tests start to favor the DDR3-equipped X38, though. In the Firefox tests, the Intel chipsets have a more distinct advantage over the nForce. Nvidia gets right back into the thick of things in the Office test.

Order is restored in WorldBench’s 3ds max tests, where rendering provides little opportunity for the chipsets to differentiate themselves.

However, the nForce 680i SLI makes a point in Nero and manages another win in, er, WinZip, ahead of the Intel field. In these last two tests, we don’t see a consistent advantage for the X38. In fact, our X38 configurations are slower than those based on the P35 in three of four matchups.

Gaming

3DMark06 scores are predictably consistent between the chipsets, but can the latest and greatest games show a performance advantage for either of these platforms?

Not really. Our systems produce near-identical scores in Bioshock and Quake Wars. The close Bioshock scores are particularly notable because we had to test it with FRAPS, manually playing through a 90-second portion of the game five times and averaging the results.

Only Team Fortress 2 separates the field. There, the X38 Express rises to the top, led by its DDR3 configuration. DDR3 is faster than DDR2 on the P35, as well, although the gap there is wider than it is with the X38.

Serial ATA performance

The Serial ATA disk controller is one of the most important components of a modern core logic chipset, so we threw each platform a selection of I/O-intensive storage tests using a Western Digital Raptor WD1500ADFD.

IOMeter

We’ll begin our storage tests with IOMeter, which subjects our systems to increasing multi-user loads. Testing was restricted to IOMeter’s workstation and database test patterns, since those are more appropriate for desktop systems than the file or web server test patterns.

Our X38 and P35 systems use the same ICH9R south bridge, so it’s no surprise to see their scores evenly matched here. The nForce 680i SLI manages to keep up, too, at least until the load scales beyond 64 concurrent I/O requests.

Response times are very closely matched until we start scaling beyond 64 I/O requests. There, the nForce starts to slow down while the X38 and P35 keep rolling.

CPU utilization is low across the board.

HD Tach

We used HD Tach 3.01’s 8MB zone test to measure basic SATA throughput and latency.

Sustained read speeds are similar across all three contenders, but the Intel chipsets have a slight advantage over the nForce in the burst and write speed tests.

The Nvidia chipset attains quicker access times, though.

CPU utilization results are within HD Tach’s +/- 2% margin of error in this test.

USB performance

Our USB transfer speed tests were conducted with a USB 2.0/Firewire external hard drive enclosure connected to a 7200RPM Seagate Barracuda 7200.7 hard drive. We tested with HD Tach 3.01’s 8MB zone setting.

USB performance is locked in a dead heat in the burst and read speed tests, but writing starts to spread things out. Interestingly, the X38 system has quicker write speeds than the P35, although both trail the nForce. The X38’s CPU utilization is a little higher than can be explained by HD Tach’s +/- 2% margin of error in that test, as well.

PCI Express performance

We used ntttcp to test PCI Express Ethernet throughput using a Marvell 88E8052-based PCI Express x1 Gigabit Ethernet card.

Throughput doesn’t vary from chipset to chipset. CPU utilization does, and our P35 config shaves 4% off versus the competition. I’ll explain why in a moment.

PCI performance

To test PCI performance, we used the same ntttcp test methods and a PCI VIA Velocity GigE NIC.

Our PCI performance testing provides some curious results. Throughput is higher for our Intel-based systems than it is for the nForce 680i SLI. The nForce suffers from higher CPU utilization, as well.

Power consumption

We measured system power consumption, sans monitor and speakers, at the wall outlet using a Watts Up Pro power meter. Power consumption was measured at idle and under a load consisting of a multi-threaded Cinebench 10 render running in parallel with the “rthdribl” high dynamic range lighting demo.

In testing, we discovered that the latest BIOS for the GA-P35-DQ6 motherboard we used for our P35/DDR2 system doesn’t correctly throttle clock speeds with either C1E or SpeedStep. Older BIOS revisions used to work, but Gigabyte broke something with its last release. That could explain the lower CPU utilization results we saw for the P35 configuration in some of our peripheral testing.

Power consumption looks reasonable for the X38 Express. The chipset is certainly more frugal than the nForce 680i SLI. Under load, there’s little difference between its power consumption in our DDR2 and DDR3 configs.

Don’t pay too much attention to the higher power consumption figures for our P35/DDR3 setup. That system uses an P5K3 Deluxe motherboard that has always been rather power-hungry. Asus is aware of the issue, and they worked to address it with the P5E3 Deluxe we used in our X38/DDR3 config. As you can see, their efforts paid off.

Overclocking

For our overclocking tests, we dropped our CPU multiplier to 6X—its lowest possible value. Memory bus dividers and MCH straps were also adjusted to keep our DIMMs running well within their limits at overclocked front-side bus speeds. Next, we turned our attention to the front-side bus, cranking it up and using a combined load of Prime95 and the rthdribl HDR lighting demo to test stability along the way.

With the GA-X38-DQ6, we sailed easily up to a 500MHz front-side bus without the need for extra voltage. 510MHz refused to POST, though, and no amount of voltage manipulation, sweet talking, or other tweaking could coax more than 500MHz from the board.

Like the Gigabyte board, the Asus flew up to 500MHz without issue. Booting at 510MHz required a little extra CPU voltage, but that was it. However, 520MHz wasn’t in the cards. We tried fiddling with voltages, adding the board’s optional chipset coolers, and even a gentle massage, but our P5E3 refused to boot with a 520MHz front-side bus.

Front-side bus speeds in the 500-510MHz range are certainly impressive, and they suggest that both boards offer more than enough overclocking headroom for most folks. However, we’ve had a number of P35 Express boards up to a hiccup-free 490MHz front-side bus, suggesting that the X38 doesn’t have a distinct advantage when it comes to overclocking headroom.

As is always the case with overclocking, your mileage may vary.

Motherboard peripheral performance

Core logic chipsets integrate a wealth of peripherals, but they don’t handle everything. Firewire and audio are farmed out to auxiliary chips, for example. Intel chipsets also rely on third party silicon for networking, and many motherboards feature additional SATA controllers to complement south bridge Serial ATA offerings.

To provide a closer look at the peripheral performance you can expect from the motherboards we’ve tested today, we’ve complied Ethernet, USB, Firewire, Serial ATA, and Audio performance results below. You’ll notice that there isn’t much variance from one board to another, but there are a few things worth pointing out. Our X38-based motherboards are highlighted and in bold to make them easier to pick out from the crowd.

NTttcp Ethernet
performance

Throughput (Mbps)

CPU utilization
(%)

Asus P5E3 Deluxe
WiFi-AP @n (RTL8169)

728.961

14.47

Asus P5E3 Deluxe
WiFi-AP @n (88E8056)

942.923

15.96

Asus P5K3 Deluxe
(RTL8187)
716.975 17.06

Asus P5K3 Deluxe
(88E8056)
938.913 15.38

EVGA 122-CK-NF68
(nForce 680i SLI 1)
946.448 21.56

EVGA 122-CK-NF68
(nForce 680i SLI 2)
835.589 21.75

Gigabyte
GA-P35-DQ6 (RTL8111B)
938.625 13.03

Gigabyte
GA-X38-DQ6 (RTL8111B 1)

940.481

15.58

Gigabyte
GA-X38-DQ6 (RTL8111B 2)

940.772

15.97

For the most part, Ethernet throughput is consistent from one board to the next. However, PCI-based GigE controllers offer substantially lower throughput than their PCIe-based counterparts. This sub-par throughput is a strike against the P5E3, whose secondary GigE chip rides the pokey PCI bus.

It’s also interesting to note that the nForce 680i SLI’s integrated networking capabilities consume more CPU resources than competing solutions from Marvell and Realtek. There was a time when Nvidia’s hardware-accelerated GigE had a CPU utilization advantage, but that appears to no longer be the case.

HD Tach USB
performance

Read burst

speed (MB/s)


Average read

speed (MB/s)


Average write

speed (MB/s)


CPU utilization

(%)


Asus P5E3 Deluxe
WiFi-AP @n

33.9

32.5

30.2

4.7

Asus P5K3 Deluxe
33.9 32.5 26.2 5.7

EVGA 122-CK-NF68
33.9 32.6 32.4 4.3

Gigabyte
GA-P35-DQ6
33.9 32.6 27.4 3.0

Gigabyte
GA-X38-DQ6

33.9

32.6

30.1

9.0

USB performance results are pretty close. However, the X38 boards offer faster write speeds than those based on the P35. They all use the same ICH9R south bridge, but something about those X38 configs makes them a little faster.

HD Tach
Firewire performance

Read burst

speed (MB/s)


Average read

speed (MB/s)


Average write

speed (MB/s)


CPU utilization

(%)


Asus P5E3 Deluxe
WiFi-AP @n

42.0

37.5

28.7

1.0

Asus P5K3 Deluxe
42.4 37.5 28.7 0.3

EVGA 122-CK-NF68
30.5 28.5 14.4 0.3

Gigabyte
GA-P35-DQ6
42.3 37.6 28.7 1.3

Gigabyte
GA-X38-DQ6

42.1

37.4

28.7

2.0

Four of five boards offer identical Firewire performance, with only the nForce falling victim to a slow Firewire chip.

HD Tach Serial
ATA performance

Read

burst

speed

(MB/s)


Average

read

speed

(MB/s)


Average

write speed

(MB/s)


CPU

utilization

(%)


Random

access

time

(ms)


Asus P5E3 Deluxe
WiFi-AP @n

133.4

78.0

101.3

3.0

8.2

Asus P5K3 Deluxe
133.4 78.0 104.7 3.7 8.3

EVGA 122-CK-NF68
128.1 78.0 94.4 2.3 7.9

Gigabyte
GA-P35-DQ6 (ICH9R)
134.5 78.0 100.5 3.3 8.3

Gigabyte
GA-P35-DQ6 (GSATA)
131.6 78.0 49.5 3.3 8.0

Gigabyte
GA-X38-DQ6

133.3

78.0

101.2

2.3

8.2

Nothing to see here. Move along.

RightMark Audio
Analyzer audio quality

Overall score

Frequency response

Noise level

Dynamic range

THD

THD + Noise

IMD + Noise

Stereo Crosstalk

IMD at 10kHz

Asus P5E3 Deluxe
WiFi-AP @n

3

5

1

1

3

1

3

3

3

Asus P5K3 Deluxe
4 5 3 3 3 1 3 4 3

EVGA 122-CK-NF68
3 5 1 3 3 1 3 3 3

Gigabyte
GA-P35-DQ6
4 5 3 3 3 1 3 4 3

Gigabyte
GA-X38-DQ6

4

5

1

1

3

1

3

3

3

Among X38 boards, RightMark Audio Analyzer prefers the DQ6 over the P5E3. Both turn in lower scores than their competitors in the noise level and dynamic range tests.

Conclusions

Generation after generation, Intel’s high-end chipsets have had a hard time living up to the buzz generated by their mid-range cousins. That doesn’t mean Intel is incapable of producing competitive high-end core logic chipset, but its mid-range offerings are so good they make it difficult to justify spending more. The fact that Intel shares a common pool of south bridge chips between its mid-range and high-end offerings is part of the reason for this dynamic. The bulk of the chipset’s integrated peripherals are found in the south bridge. Since the X38 and P35 Express can both be paired with Intel’s ICH9R, any differences between the platforms are confined to the north bridge.

There, the X38 takes a major step beyond the P35 with its 32 PCI Express 2.0 lanes, which make the X38 the first chipset to offer second-generation PCI Express, ensuring plenty of bandwidth for future graphics cards. The X38’s full 32 lanes also make it the first Intel chipset capable of supporting dual-x16 CrossFire configurations.

The X38 has other perks, too, such as support for DDR3 speeds up to 1333MHz. DDR3 memory modules have quickly scaled to 1333MHz and beyond, making support for faster memory an attractive feature. However, DDR3 still carries a hefty premium, and we suspect most enthusiasts will prefer to stick with DDR2-based X38 implementations for now.

Well, that’s a lie. We expect most enthusiasts will prefer to stick with DDR2-based P35 implementations. High-end products like the X38 Express rarely offer an attractive value proposition, and enthusiasts tend to seek out the sweet spot rather than paying a premium for flagship products that offer little in the way of improved performance. There is no doubt in my mind that the X38 Express is a great chipset. It’s easily the most forward-looking core logic around and arguably the best high-end chipset for the Core 2 Duo. But with boards costing $300 and up, it’s just not good value.

If you are in the market for a $300 motherboard, you definitely want one based on the X38. Despite its awkward name, the P5E3 Deluxe WiFi-AP @n packs an astounding array of goodies, including a third physical PCIe x16 slot, useful extras like a programmable remote and 802.11n Wi-Fi connectivity, and a very slick embedded Linux distribution. The embedded Linux distribution has perhaps the most potential, particularly if Asus or end users are able to load it with BIOS flashing, stress testing, and other applications that enthusiasts may find more useful than a web browser and Skype client. We can’t hold the P5E3’s need for pricey DDR3 memory against it, either, because we’re sure to see a DDR2-based version of the board before long. However, we have to take issue with the board’s use of a slow PCI-based Gigabit Ethernet chip. There’s simply no reason a $350 motherboard should be saddled with poor Ethernet performance, even on a secondary controller.

The GA-X38-DQ6 has PCIe Ethernet all around, so it doesn’t suffer from poor throughput. It doesn’t suffer from much at all, really, aside from a gaudy interface for its Windows tweaking software. The board is otherwise solid, well laid-out, and packed with overclocking options galore alongside subtle-but-thoughtful features like two flavors of Firewire connectivity and a smart approach to eSATA devices. Combine that with a lower price tag than the P5E3, and you’ve got an attractive alternative for those who aren’t keen on Wi-Fi or embedded Linux extras.

Questionable value keeps us from handing out an Editor’s Choice award to either the P5E3 Deluxe WiFi-AP @n or the GA-X38-DQ6. However, if you have disposable income to burn, both are excellent examples of what can be done with the X38 chipset, and they are TR Recommended.

Comments closed
    • BIGIRON
    • 12 years ago

    From what Ive read the DDR3 ram isn’t showing the advantage they had hoped for so Im not ready to run out and buy a new MB just to use it but I must say that embedded Linux is freaking sweet and I think I might build an Intel setup with this MB just to try it out and I need a 2nd gaming rig anyway so theres my excuse for my wife…Honey this ones for you baby…no really.

    • Bensam123
    • 12 years ago

    Good article. I would never recommend a board over $200 based on any of the features these have, though. If you exclude PCI-E 2.0, they really aren’t that special.

    For $350 you would have to be getting more then 2.5x the board all in one then a $125 board.

      • eitje
      • 12 years ago

      SLI mobo? 😮

    • Forge
    • 12 years ago

    The Asus name is P5E3, yes? It was P53E at one point and P5NE in the conclusion. Just typo or what?

    • deinabog
    • 12 years ago

    Heh, I just built a machine that’s powered by a good 965 board so I won’t be performing any major upgrades anytime soon. The X38 looks like a good chipset but I wonder how will the X48 improve on it aside from support for faster DDR3? Intel has already confirmed it’s release for early next year and they may even release a tweaked version of the X38.

    Seems to me that it’s best to go with a P35-powered board for now.

    • Flandry
    • 12 years ago

    I, for one, don’t find Gigabyte’s greater support of /[

      • Gungir
      • 12 years ago

      My take on the differences in voltage offerings just said to me that the P5E3 is your board if you prefer memory overclocking, and the DQ6 is the preferred CPU overclocking board. Wanna bet Nordic Hardware creates some Frankenstein of the two?

    • MasterRanger
    • 12 years ago

    Great mobo reviews as usual. One thing’s been bugging me about your testing methods recently:

    The gaming performance benchmarks are useless when run at 1024×768. A) No one runs games at that resolution any more. B) If you’re going to test at that ridiculous resolution, having the benchmarks statistically even shows they’re hitting a bottleneck. It would have been useful to know what that bottleneck was.

      • Krogoth
      • 12 years ago

      Surprisingly enough, the vast majority of the market still plays on that resolution. It is because they simply don’t have the fill-rate to handle higher resolutions with a smooth frame-rate or just don’t care.

        • Nitrodist
        • 12 years ago

        §[<http://www.steampowered.com/status/survey.html<]§ 36.66% use 1024x768 while gaming at 1440x900 or higher is relegated to less than 20%. The highest concentration of gamers (40%) plays at the odd resolution of 1280x800.

          • JustAnEngineer
          • 12 years ago

          There are a lot of suckers who bought underpowered and overpriced graphics cards because they didn’t check the reviews at Tech Report before they purchased their hardware. Steam’s hardware survey results always make the surprising revelation that folks play actually play games on this wimpy hardware.

          Tech Report has been giving a lot of love to folks with tiny monitors lately, but I suspect that the extremely low resolution was chosen for testing so that the graphics card fill rate would not limit performance, but some CPU or memory bottlenecks might become apparent.

            • ew
            • 12 years ago

            /[

            • d0g_p00p
            • 12 years ago

            Wimpy hardware will work with old games and low rez. However when I purchase new hardware i could care less about how much wattage it consumes. I mean it costs what maybe $100 extra year. I only care about not having to read the system requirements on a game box and how much noise it makes. Plus being able to play said game at max *everything*

            I never understood the wattage thing. I run my box 24×7 and when I decided to test turning it on and off everytime I used it the difference was under $5.00, over a month period. Hell most TR people have/live/make more cash than me. I never understood it.

            • Flandry
            • 12 years ago

            It’s because there’s more to power consumption than your energy bill.

            Some people also make a game out of minimizing the system power draw, just like some people like to tweak an extra 0.5% performance out of the system, regardless of the time cost in doing so.

            • eitje
            • 12 years ago

            latest attempt – 32W @ idle! 😀

            • PetMiceRnice
            • 12 years ago

            Well all I can say is that 10 years ago, I was a hardcore gamer. Now I’m more or less a casual gamer. I pick up a lot of the major releases like Tiberium Wars, etc., but I hardly spend as much time with games as I used to. That said, I can survive quite comfortably on older hardware that most gamers would have ditched long ago. Eye candy is not an important consideration to me.

            I was running a P4 Northwood C 2.8GHz with 1GB RAM and a 9800 Pro video card for over three years before buying a new system with an E6400, 2GB RAM and a 7950GT about 11 months ago. In retrospect, I felt I bought my new system a bit too soon, I should have waited until about now to make the jump. Even then, I wonder if a new video card like a 7600GS or GT and extra RAM might have been enough to keep me happy for a while. 1024×768 is perfectly alright to me, and I’d even drop to 800×600 without much hesitation.

          • Firestarter
          • 12 years ago

          that’s because a lot of the mainstream people are buying cheap 15.4″ laptops

    • someotherguy5
    • 12 years ago

    A few thoughts:

    The 2GB DMI bottleneck between the northbridge and southbridge sucks. That’s only equivalent to a x4 PCIe 2.0 lane… ah, but the southbridge doesn’t support PCIe 2.0.

    The limitation to x1 PCIe lanes on the southbridge sucks.
    The southbridge is limited to PCIe 1.1. that sucks, too.

    Few of your tests seem to max out the bus: 6 disks, Raid-5 in conjunction with high-speed ethernet could be interesting.

    Also I’d like to see you guys start to test 10gbps NICs which are coming out from Intel, Myricom, etc — think it would get some interesting results.

    Thanks for your great reviews!

      • Krogoth
      • 12 years ago

      It would be interesting to review such equipment.

      Just remember that 10Gbit Ethernet is practically useless outside of servers and backboning.

    • someotherguy5
    • 12 years ago

    I’m a little confused by this quote on the first page:

    “It’s no surprise, then, that gen-two PCIe lanes offer 1GB/s of bandwidth—double that of their PCIe 1.1 predecessors.”

    My understanding is that:

    a PCIe 2.0 (x1) link is capable of 5 Gbits/sec in both directions, simultaneously, (before overhead is subtracted). After overhead, you get perhaps 4Gbits/second/link (= 500 MBytes/second/link) in both directions, simultaneously.

    Are you adding both directions to get 1GByte/second? Or perhaps, is what you’re calling a “lane” different from a link? Thanks.

    • mako
    • 12 years ago

    Great article as always. I liked the peripheral I/O comparison.

    Typo page 5: “…clumsy ribbon cables every again”.

    • mortifiedPenguin
    • 12 years ago

    would it be unusual to say that the most interesting thing that i found about the P5E3 @n is that the I/O shield doesn’t have those annoying bendable tabs?

    • Krogoth
    • 12 years ago

    Oh, I forget to mention that X38 is literately the next 925X.

    A soild chipset that started out in very expensive boards. They were rendered obsolete by socket and VRM changes by the time boards became affordable.

      • Nitrodist
      • 12 years ago

      Doesn’t the board still take LGA775 processors? If it does, then that’s all I need.

        • Krogoth
        • 12 years ago

        Nehalem is going to on a completely different LGA socket.

        The fastest CPU that X38 will support is some high-end flavor of Yorkfield/Wolfedale.

    • Krogoth
    • 12 years ago

    X38 = P35 with PCIe 2.0 support.

    It is not worth $349 for a X38 board. I doubt PCIe 2.0 will ever be necessary as GPUs are already hitting some walls with power consumption and laws of diminishing returns.

    PCIe 1.0 is just starting to be use in non-video cards. There is no reason for extra bandwidth outside of high-end RAID cards.

      • Anomymous Gerbil
      • 12 years ago

      Haha, very droll!

      “We don’t need that extra capacity right now, and there’s no way that future tech improvements could possibly enable its use, so let’s criticise these improvements and hope the PC industry stops moving forward.”

        • Krogoth
        • 12 years ago

        I guess you were never burnt or annoyed by AGP 1x, 2x, 4x and 8x non-sense eh?

        The only reason that PCIe 2.0 is being pushed out so quickly is for power reasons to feed even more power hungry GPUs. It has very little to do with bandwidth reasons. PCIe 1.0 already provides enough bandwidth for mainstream and enthusiast for the foreseeable future.

        The catch is that GPU-guys are already getting darn close to the practical thermal and power limits of silicon. The laws of diminishing returns are also creeping up as well.

      • someotherguy5
      • 12 years ago

      Even dual gigabit NICs need more bandwidth than this.
      Intel’s dual-port gigabit NIC simply won’t work because it’s PCIe x4 (this only supports PCIe x1 on the southbridge)

      “There is no reason for extra bandwidth outside of high-end RAID cards.”

      You’re screwed if you want to run anything above PCIe 1.1 x1.
      There are plenty of cards out there that require x4.

      10gbps NICs also won’t work.
      They’re useful for running from iSCSi, among other things.
      You’re out of luck with this.

      You’re cutting yourself out of the cutting edge with this, alas.

        • Krogoth
        • 12 years ago

        Those examples are *[

          • eitje
          • 12 years ago

          are you kidding? enthusiast hardware is just last year’s server hardware. 🙂

    • Usacomp2k3
    • 12 years ago

    Reading it, I was given cause to wonder if firewire 800 works right in Vista.

    • JustAnEngineer
    • 12 years ago

    It’s a nice review, Geoff. I would have liked to have seen more comparison of memory performance for DDR2-800 vs. DDR2-1066 vs. DDR3-1066 vs. DDR3-1333.

      • flip-mode
      • 12 years ago

      I’d like that too. I’d also like to see a test with an odd number of sticks, forcing single channel, just because that sometimes happens in the real world.

      • Nitrodist
      • 12 years ago

      The problem with mobos that use the dual DDR2/DDR3 is that they’re inherently worse at memory than just DDR2 or DDR3 exclusively. I got my info from some Intel rep. He said it had something to do with the way the memory pathways were setup? Someone help me out here and find the quote 😛

    • indeego
    • 12 years ago

    Intro
    Conclusion
    comment. Doneg{

      • z00100
      • 12 years ago

      Funny thing, is I do the same thing tooo……….

      Intro,
      Conclusion,
      Comments (sometimes)

      • flip-mode
      • 12 years ago

      Intro
      Conclusion
      HW Pr0n
      Comment
      Full read when I get time
      Done!

    • ssidbroadcast
    • 12 years ago

    First. I guess I should read the article now.

    /I am annoying right now.

    Edit:

    Seems we have a cost trifecta here:
    X38 mobo price premium
    DDR3 price premium
    higher overall power consumption vs the p35 w/DDR2

    Considering the relatively tiny gains, I can’t recommend an X38/DDR3 upgrade path.

      • Sargent Duck
      • 12 years ago

      Same here. Post first, then read article.

      Of course Dissonance, I just want you to know that this article is taking precedence over my studying. If I get poor marks….

        • BoBzeBuilder
        • 12 years ago

        If I get poor marks I will cut my TR visits by 12.9% which in turn would reduce my CPU utilization and thus the overall electricity bill.

        • lyc
        • 12 years ago

        i can almost feel the collective slogging of students worldwide, oof.

        good thing we have tr for some quality procrastination time 🙂 heh, if only our exams were about benchmarking…

Pin It on Pinterest

Share This