Logo   Banner   TopRight
TopUnder
Transparent
NVIDIA Quad-SLI with the ASUS EN9800GX2
Transparent
Date: Mar 25, 2008
Section:Graphics/Sound
Author: Marco Chiappetta
Transparent
Introduction and Related Information


Last week, NVIDIA formally launched the dual-GPU powered GeForce 9800 GX2 graphics card and nForce 790i SLI Ultra chipset.  We evaluated both of the products in depth in our coverage of the launch, but had to leave out one important detail per NVIDIA's embargo agreement - that being Quad-SLI.

It wasn't that Quad-SLI needed some more time in the oven, however.  In fact, we had the drivers in hand to do Quad-SLI testing well before last week's launch.  The delay in GeForce 9800 GX2 Quad-SLI coverage was simply due to a timing issue and commitments NVIDIA had made with a few of their OEM customers, who will likely be announcing Quad-SLI powered system sometime today.
 

 
NVIDIA Quad-SLI Technology: The nForce 790i SLI Ultra with two GeForce 9800 GX2 cards


As its name suggests, Quad-SLI consists of a quadruple GPU configuration.  And while today's version of Quad-SLI is new, it is not a brand new technology.  As many of you probably know, NVIDIA had previously launched Quad-SLI with the GeForce 7950 GX2 in mid-2006.  Due to the timing of its arrival, GeForce 7-series Quad-SLI was married to Windows XP.  Ultimately, difficulties getting many games to scale led to diminished support from NVIDIA.  With Windows Vista and its completely new driver model and a new version of DirectX right around the corner, however, it's clear why NVIDIA took some focus away from the GeForce 7-series Quad-SLI solution, as distasteful as it is that existing customers were left out in the cold.

The situation today is somewhat different though.  At this moment in time, both NVIDIA's and AMD's flagship graphics cards are powered by dual-GPUs.  Windows Vista has also had some time to mature and a few of the roadblocks that prevented good scaling with three or four GPUs on Windows XP have been eliminated.  Properly implementing a quad-GPU configuration is still immensely difficult, but the focus on multi-GPU configurations has never been more acute with both major players in the GPU space seemingly committed to supporting the technology in the long term.

    

    
ASUS EN9800GX2

We're going to evaluate 2008's version of Quad-SLI technology with a pair of retail-ready graphics cards from ASUS, dubbed the EN9800GX2.  ASUS' take on the GeForce 9800 GX2 is much like the reference design we showed you in our launch coverage, save for the custom decal affixed to the card's enclosure.  And as is usually the case with their graphics cards, ASUS includes a robust accessory bundle with the EN9800GX2 complete with a CD / DVD wallet, a user's manual, a driver CD, another disc with a copy of Futuremerk's 3DMark06, and a full version of the DX10 game Company of Heroes: Opposing Fronts. In addition to the software, ASUS also thrown is DVI to VGA adapter, a dual-Molex to 6-pin PCI Express power adapter, a 6-to-8-pin PCI Express power adapter, and an S/PDIF audio cable to take advantage of the card's HDMI output.

Model

EN9800GX2/G/2DI/1G

Graphics Engine
 
GeForce 9800 GX2
 

Video Memory

1GB DDR3

Engine Clock

600 MHz

Memory Clock

2GHz (1GHz DDR3)

Memory Interface

256-bit + 256-bit

DVI Max. Resolution

2560 * 1600

Bus Standard

PCI Express 2.0

DVI Output

DVI-I * 2

HDMI Output

Yes

HDCP compliant

Yes


The rest of the ASUS EN9800GX2's specifications are essentially identical to NVIDIA's reference design.  The card's GPUs are clocked at 600MHz and its 1GB (512MB per GPU) of GDDR3 frame buffer memory is clocked at 1GHz (2GHz effective).
Transparent
NVIDIA Quad-SLI Technology


To implement Quad-SLI with a pair of GeForce 7950 GX2 cards, NVIDIA developed a rendering mode designed to leverage the power of the four GPUs powering the cards. 





Typically, multi-GPU scaling is achieved by using either AFR, for Alternate Frame Rendering, or SFR, for split frame rendering.  For older GeForce 7950 GX2 Quad-SLI configurations NVIDIA combined these two modes of operation and created AFR of SFR.  What this mode did was take alternate frames and split the frames for rendering across the four GPUs.  SFR, however, doesn't scale as well as true alternate frame rendering and SFR isn't universally compatible either.  SFR also usually has more communications overhead and applications with a heavy vertex workload don't benefit as much.


 

With a pair of GeForce 9800 GX2 cards, however, NVIDIA is now able to use AFR in many games for better scaling across the GPUs.  AFR is the preferred mode because all of the 3D rendering workload is parallelized (pixel fill, raster operation, vertex transform, etc.) and each GPU can work on its own frame.  AFR is not universally compatible with every application, but when it does work, performance is usually increased significantly.

Transparent
Our Test Systems and 3DMark06


HOW WE CONFIGURED THE TEST SYSTEMS: We have two sets of benchmark tests to discuss in this article.  The first round of tests consist of five different motherboards compared using identical supporting hardware.  In addition to the motherboard tests, we also tested all of the graphics cards used in this article on either an Asus nForce 790i SLI Ultra based Striker II Extreme motherboard (NVIDIA GPUs) or an Asus P5E3 Premium (ATI GPUs) powered by a Core 2 Extreme QX6850 quad-core processor and 2GB of low-latency Corsair RAM. The first thing we did when configuring these test systems was enter their respective BIOSes and set all values to their "optimized" or "high performance" default settings. Then we manually configured the memory timings and disabled any integrated peripherals that wouldn't be put to use. The hard drive was then formatted, and Windows Vista Ultimate was installed. When the installation was complete we fully updated the OS, and installed the latest DX10 redist and various hotfixes, along with the necessary drivers and applications.

HotHardware's Test Systems
Intel and NVIDIA Powered


Hardware Used:
Core 2 Extreme QX6850 (3GHz)

Asus Striker II Extreme
(nForce 790i SLI Ultra chipset)

Asus P5E3 Premium
(X48 Express)

Radeon HD 3870 X2 (x2)
Radeon HD 3870
GeForce 9800 GX2 (x2)
GeForce 8800 GTS 512 (x2)
GeForce 8800 GTX

2048MB Corsair DDR3-1333 C7
(2 X 1GB)

Integrated Audio
Integrated Network

Western Digital "Raptor" 74GB
(10,000RPM - SATA)


Relevant Software:

Windows Vista Ultimate

NVIDIA Forceware v174.53
ATI Catalyst v8.3

Benchmarks Used:
3DMark06 v1.0.2
Unreal Tournament 3
Crysis
Half Life 2: Episode 2
Enemy Territory: Quake Wars

Futuremark 3DMark06
Synthetic DirectX Gaming


3DMark06

3DMark06 is the most recent addition to the 3DMark franchise. This version differs from 3Dmark05 in a number of ways, and includes not only Shader Model 2.0 tests, but Shader Model 3.0 and HDR tests as well. Some of the assets from 3DMark05 have been re-used, but the scenes are now rendered with much more geometric detail and the shader complexity is vastly increased as well. Max shader length in 3DMark05 was 96 instructions, while 3DMark06 ups that number to 512. 3DMark06 also employs much more lighting and there is extensive use of soft shadows. With 3DMark06, Futuremark has also updated how the final score is tabulated. In this latest version of the benchmark, SM 2.0 and HDR / SM3.0 tests are weighted and the CPU score is factored into the final tally as well.



According to 3DMark06's default benchmark, the various Radeon HD 3870 X2 CrossFireX configurations have an advantage over the GeForce 9800 GX2 and Quad-SLI in general.  Why that is the case with this benchmark reveals itself if we look at the individual test results.






3DMark06's Shader Model 2.0 test is essentially CPU bound with all of the multi-GPU configurations, but the ATI powered configurations held onto a slight lead.  In the Shader Model 3.0 / HDR test though, the ATI-powered rigs scaled to a larger degree, which boosted the overall results.


 

Transparent
Half Life 2: Episode 2

Half Life 2: Episode 2
DirectX Gaming Performance


Half Life 2:
Episode 2

Thanks to the dedication of hardcore PC gamers and a huge mod-community, the original Half-Life became one of the most successful first person shooters of all time.  And thanks to an updated game engine, gorgeous visual, and intelligent weapon and level design, Half Life 2 became just as popular.  Episode 2 offers a number of visual enhancements including better looking transparent texture anti-aliasing. These tests were run at resolutions of 1,920 x 1,200 and 2,560 x 1,600 with 4X anti-aliasing and 16X anisotropic filtering enabled concurrently.  Color correction and HDR rendering were also enabled in the game engine as well.  We used a custom recorded timedemo file to benchmark all cards in this test.




The ASUS EN9800GX2 Quad-SLI setup showed significant scaling in our custom Half Life 2: Episode 2 benchmark.  At both resolution the Quad-SLI rig was significantly faster than a single GX2, and it was the fastest NVIDIA-based configuration.  At 1920x1200, scaling wasn't terribly dramatic, but at 2560x1600 the delta was huge.  Also note, the dual-Radeon HD 3870 X2 quad-CrossFireX setup held its own and took the lead at 1920x1200, but it fell behind by quite a bit once the resolution was increased to 2560x1600.

Transparent
Unreal Tournament 3

Unreal Tournament 3
DirectX Gaming Performance


Unreal Tournament 3

If you're a long-time PC gamer, the Unreal Tournament franchise should need no introduction.  UT's fast paced action and over the top weapons have been popular for as long as Epic has been making the games.  For these tests, we used the latest addition to the franchise, Unreal Tournament 3.  The game doesn't have a built-in benchmarking tool, however, so we enlisted the help of FRAPS here.  These tests were run at resolutions of 1,920 x 1,200 and 2,560 x 1,600 with no anti-aliasing or anisotropic filtering enabled, but with the UT3's in game graphical option set to their maximum, with color correction enabled.


Unreal Tournament 3 did scale on the ASUS EN9800GX2 Quad-SLI setup, but not by a very large degree.  The Radeons on the other hand did very well here, and bested the Quad-SLI setup at the higher resolution.  We did have a problem, however, with the quad-CrossFireX configuration.  No matter what we tried, the game would crash at 1920x1200.  But curiously, it worked fine at 2560x1600.

Transparent
Enemy Territory: Quake Wars

Enemy Territory: Quake Wars
OpenGL Gaming Performance


Enemy Territory:
Quake Wars

Enemy Territory: Quake Wars is Based on id's radically enhanced Doom 3 engine and viewed by many as Battlefield 2 meets the Strogg, and then some.  In fact, we'd venture to say that id took EA's team-based warfare genre up a notch or two.  ET: Quake Wars also marks the introduction of John Carmack's "Megatexture" technology that employs extremely large environment and terrain textures that cover vast areas of maps without the need to repeat and tile many small textures.  The beauty of megatexture technology is that each unit only takes up a maximum of 8MB of frame buffer memory.  Add to that HDR-like bloom lighting and leading edge shadowing effects and Enemy Territory: Quake Wars looks great, plays well and works high end graphics cards vigorously.  The game was tested with all of its in-game options set to their maximum values with soft particles enabled in addition to 4X anti-aliasing and 16x anisotropic filtering.


The dual ASUS EN9800GX2 powered quad-SLI rig kicked some serious butt in our custom Quake Wars: Enemy Territory benchmark.  In this game, the quad-SLI rig scaled very well, and bested every other configuration by a large margin.  Conversely, this game does not scale at all in the three and four-GPU CrossFireX configuration, hence the similar scores for all three Radeon powered data points.

Transparent
Crysis v1.2

Crysis v1.2
DirectX 10 Gaming Performance


Crysis

If you're at all into enthusiast computing, the highly anticipated single player, FPS smash-hit Crysis, should require no introduction. Crytek's game engine visuals are easily the most impressive real-time 3D renderings we've seen on the computer screen to date.  The engine employs some of the latest techniques in 3D rendering like Parallax Occlusion Mapping, Subsurface Scattering, Motion Blur and Depth-of-Field effects, as well as some of the most impressive use of Shader technology we've seen yet.  In short, for those of you that want to skip the technical jib-jab, Crysis is HOT.  We ran the SP demo with all of the game's visual options set to 'High' to put a significant load on the graphics cards being tested.


The quad-SLI test system showed some scaling in Crysis, and was the fastest system overall, but its victory over the next fastest system - powered by a pair of GeForce 8800 GTS 512 cards in SLI mode - wasn't dramatic.  The Quad-SLI rig did, however, blow way past and of the Radeon CrossFireX configurations which didn't scale past three GPUs in this game.

Transparent
Total System Power Consumption


Before we bring this article to a close, we'd like to cover a few final data points. Throughout all of our benchmarking and testing, we monitored how much power our test systems were consuming using a power meter. Our goal was to give you all an idea as to how much power each configuration used while idling and under a heavy workload. Please keep in mind that we were testing total system power consumption at the outlet here, not just the power being drawn by the motherboards alone.

Total System Power Consumption
Tested at the Outlet



Looking at these power consumption numbers, one thing is clear - if you're going to slap three or four GPUs into a single system, its going to sap some serious power.  The dual ASUS EN9800GX2 powered quad-SLI rig consumed over 300 watts while idling at the Windows desktop (with Aero enabled) and over 550 watts when running with a full 3D rendering workload, which is about 40 and 135 more watts, respectively, than the GeFroce 8800 GTS 512 SLI configuration while idling and under load.  Quad-SLI's power profile doesn't look so daunting next to CrossFireX, however.  The Radeon-powered CrossFireX configurations consumed less power while idling, but under load the three GPU configurations consumption shot up to 511 watts, and the four GPU configuration to 648 watts.

Transparent
Our Summary and Conclusion


Performance Summary: A pair of ASUS EN9800GX2 cards running in a quad-SLI configuration showed increased performance in every application we tested versus a single GeForce 9800 GX2.  Overall, performance was extremely good and in the majority of tests the Quad-SLI configuration was the fastest of the bunch, occasionally by a large margin.




This most recent incarnation of NVIDIA's Quad-SLI technology is vastly superior to 2006's version.  A single GeForce 9800 GX2 is fast, but a pair of them running in Quad-SLI mode is extreme, albiet not for everyone.  Although we're certain Quad-SLI won't scale in every game, it did scale with every application we threw at it here today and its performance was top notch.  And in the near term, we expect NVIDIA to further improve Quad-SLI's performance scaling as well with upcoming driver releases.  What the long term future holds remains to be seen, however, because it is getting increasingly hard to leverage multi-GPU configurations with more advanced rendering techniques.

Quad-SLI with a pair of GeForce 9800 GX2 cards also consumes less power than the competition's quadruple GPU configuration.  While idling AMD's implementation used less power, but once put under load, NVIDIA's Quad-SLI was clearly more efficient - not only was it faster, but it used less power.

Of course, this kind of power comes at a price.  A pair of GeForce 9800 GX2 cards will currently set you back a cool $1200, provided you've already got the necessary nForce chipset-based platform to install them in.  Cutting edge PC hardware has always been an expensive proposition, however, so we're sure none of you are surprised by this.

In the end, we're impressed with GeForce 9800 GX2 Quad-SLI and hope NVIDIA continues to improve the technology.  Scaling is good, power consumption is high, but in line with expectations, and it actually costs less than a pair of GeForce 8800 Ultras did at launch.

 

  • Great Performance
  • Most Powerful PC Graphics Configuration
  • Good Scaling
  • Scaled with DX9, DX10, and OpenGL
  • Expensive
  • Won't Scale All The Time



Content Property of HotHardware.com