 | |||
|
|
|
|
|
 |
NVIDIA nForce 780a/ASUS M3N-HT Deluxe Performance Preview May 06, 2008 Brandon Sandman Bell |
|
 |
|
| |
Summary: With support for HybridPower, NVIDIA's nForce 780a chipset for the AMD platform is designed to mix SLI performance in games with IGP power draw at idle. But that's not the only feature the new chipset supports. Join us as we go over all the new features introduced in nForce 780a as well as the first board we've received based on the chipset: the ASUS M3N-HT Deluxe!
 Introduction | Page:: ( 1 / 12 ) |
While these issues weren’t the fault of AMD’s processors, their CPUs themselves were rock-solid, the platform was perceived as being buggy and less than 100% reliable in comparison to Intel.
Enter NVIDIA. Their nForce chipsets were the first to introduce cutting edge features which we now take for granted such as dual-channel memory architecture and dual Gigabit Ethernet. In addition while they weren’t always perfect, the nForce platform as a whole has a much better track record than VIA when it comes to compatibility and system stability, and it also doesn’t hurt that the overclocking crowd has had a lot of success with nForce motherboards and scaling.
[image]
| <% print_image("01"); %> | <% print_image("02"); %> |
In more recent years, NVIDIA’s trump card has been SLI. Those three magic letters, S-L-I, have helped NVIDIA sell more graphics cards, and very quickly got NVIDIA a Pentium 4 bus license. If you recall, for years NVIDIA’s chipsets were exclusive to the AMD platform. This was due in part because Intel and NVIDIA couldn’t agree on a price for Intel’s bus license. You see, unlike AMD who licenses their bus for free (including HyperTransport), Intel enjoys royalties from the sale of any chipset that uses their bus to communicate to the processor. Once SLI debuted at the end of 2004 on the AMD platform with the original nForce 4 chipset, NVIDIA and Intel quickly came up with a broad cross-license agreement that has lived to this day.
Since then however, AMD and NVIDIA’s relationship has slowly soured. The nForce 590 chipset was launched just a couple of months ahead of AMD’s stunning acquisition of ATI in July 2006. Since that announcement, NVIDIA hasn’t released another high-end chipset for the AMD platform, skipping the nForce 6 generation (and 3-Way SLI) entirely so they could focus on the Intel platform, where they’ve released nearly a dozen chipsets since nForce 590.
With ATI’s platform group taking on chipset development for AMD, who historically hasn’t had much of a fulltime chipset division, NVIDIA has in some ways been put on the backburner: AMD’s 790FX chipset launched last year to positive reviews. 790FX supports high-end features like HyperTransport 3.0, 4-Way CrossFireX, PCI Express 2.0, and support for AMD’s powerful Overdrive overclocking utility. In fact, enthusiasts have found that the 790FX chipset scales to very high speeds with Athlon 64 X2 5000+ Black Edition CPUs.
In light of AMD’s new 7-series chipsets, NVIDIA’s venerable nForce 590 SLI (and the rest of the nForce 500 family) looks downright antiquated and in desperate need of replacement.
[image]
| <% print_image("03"); %> | <% print_image("04"); %> |
Today NVIDIA is finally getting around to updating their nForce line of chipsets for the AMD platform, and they’ve come up with an entirely new 7-series of chipsets of their own ranging from the nForce 780a SLI on the high-end, down to the 730a chipset for single GPU users. Today we’ll be taking a look at the new 780a platform and ASUS’ upcoming M3N-HT Deluxe.
| Meet the new nForce lineup | Page:: ( 2 / 12 ) |
GeForce Boost is designed to give low-end value systems with integrated graphics (IGP) a boost in performance when the IGP (which is based on the same architecture as the GeForce 8400 GS with 16 stream processors) is paired with a GeForce 8400 GS or 8500 GT discrete graphics card. HybridPower and ESA are probably the features most enthusiasts would be most interested in.
With HybridPower, discrete GPUs can be powered down when they’re not being heavily taxed, say for instance when you’re browsing the Internet or typing up a Word document. In these cases, the discrete GPU can be powered down, allowing the IGP to handle these 2D tasks, and since the IGP supports NVIDIA’s PureVideo HD, it can also be used for video playback duties as well. The only caveats are that a Phenom CPU (and its HyperTransport 3.0 link) is required when playing HD videos and that noise reduction must be disabled when using the IGP (Windows Vista’s Aero interface is also disabled when HD video is being played).
The IGP is capable of driving one single-link DVI or HDMI display (limiting the max resolution to 1920x1200), and with its 300MHz RAMDAC one VGA display with a max res of 1920x1440 at 75Hz.
ESA, NVIDIA’s Enthusiast System Architecture, is designed to provide a communications ecosystem for controlling and monitoring all the components inside the PC. This includes all fans within the system, as well as the CPU, GPU, power supply, RAM, water cooling, etc. With ESA the end user can manage and monitor the status, voltages, and temps of all these components.
This is just a quick refresher on the new features NVIDIA has announced inside the new nForce chipsets, but there are other changes. The following chart summarizes the key features found inside NVIDIA’s new 750a and 780a SLI chipsets:
| nForce 7-series SLI Comparison | ||
| NVIDIA nForce 780a SLI | NVIDIA nForce 750a SLI | |
| CPU Support | AMD Phenom, Athlon 64 FX, Athlon 64 X2, Athlon 64 | AMD Phenom, Athlon 64 FX, Athlon 64 X2, Athlon 64 |
| Memory Support | DDR2-1066 | DDR2-1066 |
| HybridPower Technology | Yes | Yes |
| SLI-Ready Memory | Yes | Yes |
| PCI Express 2.0 | Yes | Yes |
| NVIDIA SLI Technology | Yes 3-Way (x16, x8,x8) | Yes 2-Way (2x8) |
| PCI Express Lanes | 35 lanes | 19 lanes |
| PCI Express Links | 7 links (Configuration 2x16 or 4x8 or 1x16, 2x8 3x1) | 4 links (Configuration 2x8, 2x1 or 1x16, 3x1) |
| GPU-to-GPU Direct Link (PWShort) | Yes | Yes |
| Broadcast Support | Yes | Yes |
| ESA Technology | Yes | No |
| SATA/PATA Drives | 6/2 | 6/2 |
| SATA Speed | 3Gbps | 3Gbps |
| RAID | 0, 1, 0+1, 5 | 0, 1, 0+1, 5 |
| NVIDIA MediaShield | Yes | Yes |
| Native Gigabit Ethernet Connections | 1 | 1 |
| NVIDIA FirstPacket Technology | Yes | Yes |
| USB Ports | 12 | 12 |
| PCI Slots | 5 | 5 |
| Audio | HDA (Azalia) | HDA (Azalia) |
| NVIDIA Control Panel Utility | Yes | Yes |
| NVIDIA System Monitor | Yes | Yes |
| Onboard Display | RGB, DVI, HDMI | RGB, DVI, HDMI |
| NVIDA PureVideo | PureVideo HD | PureVideo HD |
| ||
Notes
Based on the chart above, one of the key differentiators between the nForce 780a SLI and 750a SLI is the number of lanes supported by the chipset – 35 in nForce 780a SLI versus just 19 in nForce 750a SLI. These extra lanes allow the nForce 780a SLI chipset to support 3-Way SLI (two PCIe cards run in x8 mode, one x16) while the nForce 750a SLI is limited to just 2-Way SLI, and even in conventional 2-Way SLI, the nForce 750a SLI chipset is limited to running both cards in x8 mode (16 PCIe lanes are sent to each graphics card under the same 2-Way scenario with nForce 780a SLI).
So how did NVIDIA manage to add the extra lanes to nForce 780a? Take a look at the block diagram of the nForce 780a versus the nForce 750a:
[image]
| <% print_image("05"); %> | <% print_image("06"); %> |
As you can see, the nForce 200 chip makes a return in the nForce 780a chipset. If you recall, nForce 200 was previously used to add PCIe 2.0 support to the nForce 780i SLI chipset for the Intel platform. Apparently NVIDIA has recycled it again to add PCIe lanes to yet another 780 chipset, in this case the nForce 780a SLI chipset.
Other than the difference in PCIe configuration, the 750a and 780a chipsets are very similar. Both chipsets support HybridPower, PWShort, and Broadcast Power (which we’ll discuss on the next page), although only the nForce 780a chipset supports ESA.
The nForce 730a chipset is NVIDIA’s chipset for gamers looking for a single GPU motherboard. Here’s the block diagram for the 730a chipset:
[image]
| <% print_image("07"); %> |
In terms of pricing, nForce 780a motherboards are expected to sell for $250 and up, while nForce 750a motherboards will start around $120+. Pricing on nForce 730a motherboards is expected to start at $80 and go up from there.
| nForce 780a Features/HybridPower (cont’d) | Page:: ( 3 / 12 ) |
PWShort
One feature first launched in the nForce 780i SLI chipset, and now carried over to the nForce 780a SLI and 750a SLI chipsets is Posted-Write Shortcut (PWShort). PWShort provides direct GPU-to-GPU communication without having to rely on the link between the PCIe and memory controller. In previous chipsets, data sent from one GPU to another had to go through this link, but the PCIe controller inside the nForce 780a chipset can send the data directly to its destination. This feature reduces the amount of latency for traffic among GPUs and helps to reduce the amount of congestion on the Memory-to-PCIe controller link.
Broadcast
This is another feature introduced with nForce 780i SLI. In systems with multiple GPUs, the CPU often sends the same data to all the GPUs. Instead of having to replicate this same data to each GPU individually, one broadcast is sent out from the CPU to the front-side bus to the chipset, which replicates it in parallel to all the GPUs.
This feature reduces the amount of data that has to go out across the front-side bus and reduces the amount of latency for CPU-to-GPU commands.
More On HybridPower
The nForce 780a SLI, nForce 730a, and 750a SLI chipsets will be the first NVIDIA platforms to hit the market with HybridPower support. Other chipsets that support the technology include the GeForce 8100/8200/8300 and the nForce 720a chipset. Besides the right motherboard/chipset, you will also need the right GPU in order to take advantage of the technology. Right now the GeForce 9800 GTX and 9800 GX2 are the only graphics processors with support for HybridPower. NVIDIA will also be providing a list of certified power supplies on slizone.com.
[image]
| <% print_image("08"); %> |
In order for HybridPower to work, the monitor must be connected to the motherboard, not the GPU. This allows the system to use the IGP when the discrete graphics cards are powered down. When multiple displays are attached to the IGP and the discrete graphics card, the motherboard picks the primary display based on settings you select inside BIOS.
HybridPower Modes
HybridPower runs in one of three different modes: Save Power, Boost Performance, and Additional Displays.
[image]
| <% print_image("09"); %> |
All three modes function exactly how they sound. In “Save Power” mode, the motherboard shuts down the discrete GPU(s), using the IGP to handle all graphics tasks. By shutting down the discrete GPU(s), this allows the system to consume less power, as well as run cooler and quieter. This feature really comes in handy as you add more GPUs. With previous chipsets these cards ran powered up at all times, but under HybridPower’s Save Power mode, these cards are completely shut off.
[image]
| <% print_image("10"); %> | <% print_image("11"); %> |
In “Boost Performance” mode, all GPUs, both integrated and discrete, are powered on to increase performance. In this mode, all graphics data is passed from the discrete GPU frame buffer through the PCIe 2.0 interface to system memory, where it is read by the motherboard GPU and sent to the display. This can result in a performance hit, although in the words of NVIDIA, the performance hit is small, less than 5%. The exact amount varies based on the screen resolution, the application being used, and the frame rate delivered.
Users who find this performance hit unacceptable will have to disconnect the display from the motherboard GPU, hooking it up directly to the discrete graphics card instead.
Hybrid SLI supports up to two displays. If the user wishes to run more displays than this, the system will have to run in the “Additional Displays” HybridPower mode in order to run the additional monitors (since the motherboard’s IGP is limited to one digital output, you will also have to run in the additional displays mode if you wish to connect two DVI displays to the system as well).
When the Additional Displays mode is selected, Save Power and Boost Performance modes are disabled. In this mode, all GPUs (including the motherboard and all discrete GPUs) are turned on and work independent of each other in order to drive all displays on both the motherboard and the discrete GPUs.
| ASUS M3N-HT Deluxe Motherboard | Page:: ( 4 / 12 ) |
[image]
| <% print_image("12"); %> | <% print_image("13"); %> |
For starters the board features ASUS’ Express Gate feature that was first launched with the P5E3 Deluxe WiFi-AP@n last year. Express Gate refers to the motherboard’s built-in OS, which consists of a Linux distribution from Splashtop. The OS is loaded onto a flash memory chip which rests underneath the primary PCI Express graphics slot.
When booting up the M3N-HT Deluxe motherboard, you’ll be presented with the ASUS Express Gate splash screen. From there you have the option to load up BIOS, boot into the OS on your hard drive, reboot or shut down the computer, or you can click the “Web”, “Skype” and “Chat” icons to load the Splashtop OS. It only takes a few seconds to boot the Splashtop OS, providing you near instant access to the Internet via the Splashtop web browser. The browser is pretty good, we were able to load up popular sites like Youtube, Myspace, CNN, ESPN, etc without any problems, and it supports features like tabbed browsing. Express Gate is a really neat feature that sets ASUS motherboards apart from others.
Another neat feature that the M3N-HT Deluxe supports is Mempipe. Mempipe is short for ASUS’ Memory Heatpipe. With this feature you can connect copper heatpipes to cool your system memory, providing heatpipe cooling for your system RAM. The Mempipe adapter fits directly above the heatpipe cooler that rests on the chipset North Bridge, all you’ll need is a small Philips screwdriver to mount the Mempipe module to the motherboard. Once mounted, you can then adjust the Mempipe module to fit your specific memory modules.
[image]
| <% print_image("14"); %> |
Mempipe is compatible with single and double-sided memory, as well as memory modules with or without heat spreaders. The width of the Mempipe modules is completely adjustable so you can fit it with practically any memory module. The only downside to Mempipe is that you’re obviously limited to cooling just two memory modules. With RAM prices so low, many enthusiasts are picking up four 1GB memory modules.
The M3N-HT is built with all-solid capacitors, and features an 8-phase power design for the CPU. On top of that, an additional 2-phases are devoted for the integrated memory controller and HyperTransport, for a total of 10 phases.
[image]
| <% print_image("15"); %> | <% print_image("16"); %> | <% print_image("17"); %> |
On the backplane of the motherboard, ASUS includes one VGA adapter and one HDMI display output which are tied to the IGP. ASUS then includes an HDMI-to-DVI adapter for driving DVI-equipped displays. You’ll also find digital audio outputs, 4 USB ports, and a Gigabit Ethernet port among the I/O options available on the back of the board.
BIOS
Officially the BIOS of the M3N-HT Deluxe is still a work in progress, so we aren’t going to comment too much on this topic until we see the final BIOS release (our tests were conducted with the 0701 BIOS) that ships with motherboards. Already though it looks like ASUS will be including all the settings enthusiasts will want for OC’ing. HyperTransport frequencies from 200-600MHz are available in 1MHz increments, while Black Edition CPU owners will have CPU multipliers available ranging from 5.0 to 25.0 in increments of 0.5x. PCIe frequencies are also adjustable and the BIOS provides memory speeds up to 1066MHz.
Those of you with B2-stepping Phenom processors will be glad to hear that the TLB patch can be disabled from within BIOS. This is how we performed all of our tests with the Phenom 9600 Black Edition. The motherboard’s BIOS also has settings for SLI-Ready Memory OC’ing as well as ASUS AI Overclock settings that allow you to OC the CPU by set percentages of 3, 5, 8, or 10%. Voltage adjustment for HyperTransport (1.2V-1.5V in increments of 0.02V), North Bridge (1.1-1.7V in increments of 0.02V), and most importantly the CPU (0.775V-1.55V in increments of 0.0125V) are also available. In fact, you can also adjust the voltage of the nForce 200 chip also.
| System Setup | Page:: ( 5 / 12 ) |
AMD Phenom 9600 Black Edition
ASUS M3N-HT Deluxe
Gigabyte GA-MA790FX-DS5
4GB Corsair TWIN2X2048-6400C4
150GB Western Digital Raptor
Windows Vista Ultimate 64-bit w/Service Pack 1
Benchmarks
Company of Heroes 1.71
Call of Duty 4
Lost Planet
Crysis
Cinebench 10
Windows Media Encoder
DivX
LAME MT
| Media Benchmarks | Page:: ( 6 / 12 ) |
| Company Of Heroes | Page:: ( 7 / 12 ) |
Company of Heroes – Direct3D
| Call of Duty 4 | Page:: ( 8 / 12 ) |
Call of Duty 4 – Direct3D
| Lost Planet | Page:: ( 9 / 12 ) |
Lost Planet – Direct3D
| Crysis | Page:: ( 10 / 12 ) |
Crysis – Direct3D
| Power Consumption | Page:: ( 11 / 12 ) |
| Conclusion | Page:: ( 12 / 12 ) |
NVIDIA’s HybridPower technology is the first of its kind in the industry. The user selectable power and performance modes available are long overdue in our opinion. The menu options are easy to understand, so novices shouldn’t have a problem getting everything setup, although power users would obviously love to have the option of setting everything up automatically depending on usage or pre-selected profiles (HybridPower is supposed to switch automatically between Boost Performance and Save Power mode automatically within Windows, but we had problems getting it to work with our testbed).
Ever since SLI was first conceived, one of its biggest weaknesses has been power usage at idle. HybridPower’s Save Power mode finally addresses this important issue. In our testing with a pair of 9800 GTX cards running in SLI, HybridPower saved 107 watts at idle. When compared to a 3-Way 9800 GTX system, HybridPower saved 173 watts!
Enthusiasts will likely be disappointed to hear about the performance hit you’ll take running in BoostPower mode. If this is the case for you, you will have to disable Hybrid mode in BIOS. Fortunately this can be easily accomplished with just a few keystrokes, but it is annoying having to reboot to accomplish this task. If we’d had more time we would have tested the exact performance hit out with a few different testbed configurations. According to NVIDIA the performance hit is small, although if you’re running Windows Vista 32-bit with 2GB of RAM you will see an additional performance hit that is independent of Hybrid SLI.
We also would like to see NVIDIA integrate support for two digital displays (including one dual-link DVI) into future nForce products supporting Hybrid SLI. After all, the type of user who will spend $250+ on a motherboard and multiple graphics cards likely isn’t relying on a VGA display anymore.
Regarding the ASUS M3N-HT Deluxe, our early impressions are that this looks like another slam dunk product for ASUS. The board is loaded with features, we’re particularly fond of Memlink and Splashtop, and the BIOS looks like it’s going to have all the settings an enthusiast would need to OC their Black Edition Phenom or Athlon 64 X2 CPU to new heights. The board layout is also top notch, with passive cooling for the chipset (although the end user can add an external ASUS fan which is included with the motherboard if he wishes) and plenty of space between the PEG slots for running 2-Way SLI. ASUS expects the board to sell for $249 when it hits shelves later this month.
But the M3N-HT Deluxe isn’t the only nForce 780a motherboard ASUS is releasing. Also under development is an even higher-end second-generation Crosshair II board which should retail for $289. Considering how powerful the original nForce 590 SLI Crosshair board was, many of you will probably want to wait to see how this motherboard pans out.