Introduction

Even if the high segment concerning the sales figures far behind the mid-range and the low end market must give, to fight the two graphics chip giant ATi and nVidia just in this price territory particularly hard, as many buyers like to transfer the pecking order in the high range to the slower models. The Canadians have two very good cards on offer that can score both in various disciplines with the Radeon X 1900 XTX and Californians with the GeForce 7900 GTX in the high-end segment.

If even these maps still bring not the desired performance, the customer remains still the handle to an SLI or CrossFire system, in which two graphics cards be used for the joint creation of images. nVidia went one step further and presented at the CeBIT “Quad SLI” in which a total four GPUs do the work. For this, nVidia developed a completely new card layout that has the width of a dual-slot card and includes two PCBs, on which each a G71 GPU is used. It includes two such graphics cards on the name “GeForce 7900 GX2″ listen together, is a Quad-SLI system.

But the cards have several major drawbacks. So, an enormous length measures alone the PCB and fits into it not in many cases. In addition, each GPU required an own PCIe connector and the cards are therefore not only freely available in the retail market. nVidia Quad SLI makes available only the “system builders”, say, you will encounter a team only in expensive complete PCs. Gave up nVidia but not satisfied with these disadvantages and presents the “GeForce 7950 GX2″.

This principle is nothing more than the GeForce 7900 GX2, makes much but still better. The PCB has been completely renewed and is now much shorter than with the predecessors. Also the fan system was reworked and requires only a power plug per card. There is still a little bad news: the GeForce is 7950 GX2 in the retail market at least only as single ticket sold; Quad-SLI is only the manufacturers of complete systems available also with these cards. We are confident, however, that this will change in the future.

nVidia a GeForce provided us 7950 GX2 for a test ready, which has to prove whether the SLI technology on a map against a single GPU may be kindly. Also, with this article, we dedicate a our new test system, as well as some new test methods. We ask in particular the latter note, as we enter a new realm with the selected quality settings.

Bookmark

• GeForce 7600 GT, 7900 GT and 7900 GTX
• nVidia GeForce 7800 GTX 512 SLI
• nVidia GeForce 7800 GTX 512
• nVidia GeForce 7800 GTX (SLI)
• nVidia GeForce 7800 GT
• nVidia GeForce 7800 GS
• nVidia GeForce 7600 GS
• nVidia GeForce 7300 GS
• ATi Radeon X 1900 XTX and X 1900 CF Edition
• ATi Radeon X 1900 GT
• ATi Radeon X 1800 XT CrossFire Edition
• ATi Radeon X 1800, X 1600 and X 1300
• ATi Radeon X 1800 GTO
• Radeon X 1000 vs. GeForce 7

Technical data

< div >

Technical details

Technically the nVidia GeForce offers some finesse 7950 GX2, as it previously has not given a card with two PCBs and two GPUs for the desktop market. But only once before: The two GPUs are normal G71 chips that are already used in the GeForce 7900 series. Accordingly, 24 pixel pipelines with two shader units, which each can calculate a MADD, available are the buyer of the chip. As well, each G71 can access 24 texture mapping units and 16 ROPs and eight vertex shader. Because on the GeForce 7950 GX2 two complete G71 chips are installed, you can therefore quite quickly by 48 pixel pipelines, 96 shader units one gigabyte of memory talk, where we want to refrain from this however as the SLI technology depending on the real not everything twice can be used.

So you can double – and nothing else uses the GeForce 7950 GX2, even if nVidia eschews the word “SLI” on the double card – not just the memory known with SLI technology. Although, a GPU can draw on 512 MB of GDDR3 memory, which is why on a GeForce 7950 GX2 are 1024 MB of memory. It would be but wrong to speak of usable 1024 MB. In SLI mode always same data must exist in every store, because it may be that both GPUs have to rely on identical information. The VRAM to 512 MB is limited by this characteristic.

The VRAM is connected with a 256-bit wide memory interface to the GPU. This added 600 MHz so far below the level of a GeForce 7900 GTX, which goes with rich 800 MHz to works. Similar repeated nVidia in the core clock, which has a frequency of 500 MHz, while the GeForce 7900 GTX 650 MHz. Now you can occur naturally on the idea that nVidia the G71 chips artificially slowing and soberly considered it behaves exactly so. But we must note that the heat on the GeForce is a much larger problem 7950 GX2, because the two PCBs are mounted directly side by side. The same clock on the dual-GPU card each clock domain of G71 (shader domain, ROP domain, geometric domain).

Individually seen the GeForce approaches 7950 GX2 therefore not on the power of a GeForce 7900 GTX. But enables SLI mode – now by the way called “Multi-GPU mode” on a map; only in Quad-SLI, opens a window with the name “SLI mode”-, see the other GeForce models no country. The GeForce 7950 GX2 is then the “old” flagship in all respects superior. Also a Radeon X 1900 XTX from ATi is then possession. Only the shader performance can keep up with the 7950 GX2.

Gives a little further in the show, these are more interesting details. For example, 7950 GX2, as in the previous GeForce uses a PCIe switch that distributed the 16 PCIe lanes of boards at the two G71 GPUs 7900 GX2 nVidia on the GeForce. A total of 48 lanes available are this switch. 16 are needed for the contact between PCIe bus and PCIe switch. 16 more are led to the “GPU 1″, while the remaining 16 PCIe lanes the “GPU 2″ is supplied. But this PCIe switch the advantage that both cores are provided with the full number of lanes is not alone. So he gives the mainboard in by its function, that there is only a graphics card.

For this reason, it is possible to operate the GeForce 7950 GX2 in each available PCIe motherboard – a SLI certified chipset is not necessary! But, it may be that a BIOS update for the motherboard is necessary. The Board currently working with nVidia to the appropriate BIOS versions, if these are not already compatible with the dual-GPU card.

&Lt; div > in the following a list of all currently compatible mainboards, which increases however daily. < div >

• ABIT AA8 DuraMAX
• Abit AW8
• Abit AN8 SLI
• Abit K8N SLI
• Albatron K8SLI
• ASUS A8N32-SLI Deluxe
• ASUS A8N5X
• ASUS A8N-E
• ASUS A8N-SLI Premium
• ASUS A8R32-MVP Deluxe
• ASUS A8V-E Deluxe
• Asus M2N32-SLI Deluxe
• Asus P5LD2
• Asus P5LD2-VM
• ASUS P5ND2-SLI
• Biostar i945 G-M7
• DFI infinity NF4 SLI
• ECS KN1 SLI Lite
• ECS nForce4-A939
• Foxconn NF4SLI7AA-8EKRS2
• Foxconn C51XEM2AA-8EKRS2H
• Gigabyte GA-8I945P Pro
• Gigabyte GA-8N-SLI Quad Royal
• Gigabyte GA-K8N Pro-SLI
• Gigabyte GA-K8N51GMF
• Gigabyte GA-K8N51PVM9-RH
• Gigabyte GA-K8NF-9
• Gigabyte GA-K8N-SLI
• Gigabyte GA-K8NXP-9
• Gigabyte GA-K8NXP-SLI
• Intel D955XBK
• Intel D975XBX
• Intel SE7525GP2
• iWill DK8EW of 59102
• MSI K8N diamond plus
• MSI K8N Neo4 Platinum v1.0
• MSI K8N Neo4-F
• Supermicro X6DA8
• Tyan S2895 D/T

Away from the PCIe switch, the GeForce 7950 GX2 models through an internal interface can communicate with each other. A special connection is placed on both PCBs, connecting the boards using a small SLI bridge. This internal bandwidth between the PCBs increases further, without which otherwise could lead to shortages.

The usual SLI bridge on the video card is only on a Quad-SLI configuration of importance. There, the first card is connected to the second GeForce 7950 GX2. The attentive observer will be noticed already that there are a further difference to the GeForce 7900 GX2. This has two SLI bridges per card and two bridges have to be used in Quad-SLI. Because nVidia card layout has developed, a second SLI bridge according to the California is no longer necessary. The range should not be decreased despite the absence of second Verbindungsweges.

In addition, the GeForce 7950 GX2 with a small detail can score: the HDCP encryption. This is important for future HD video playback, because only with fully HDCP-compliant graphics card can play the videos in the full resolution. All GPUs are compatible with HDCP already for some time, it is still an additional EPROM needed, that contains the HDCP key to decrypt the data. This has been lacking on almost all video cards. nVidia the GeForce 7950 GX2 from support as standard with a HDCP key, and thus can easily play HD videos in the maximum resolution.

According to nVidia 142 Watts, the maximum power consumption of the GeForce is 7950 GX2. With regard to the rendering modes with SLI, nothing has changed in comparison with two traditional single cards in SLI mode. In Quad-SLI mode, the GeForce works 7950 GX2 identical to the GeForce 7900 GX2.

Impressions

nVidia GeForce 7950 GX2

The nVidia GeForce 7950 GX 2 is an impressive piece of hardware. Something so previously still did not exist. Two high end GPUs are working on a map where this consists of two PCBs, in principle, each representing an own graphics card for themselves. So of course the question arises, whether the GeForce 7950 GX2 is still a single map, or whether you should treat this not as a SLI system? This is in our view not really have to answer as the high-end model is somehow a proper single, nor a right dual card. For this reason, we will let stand the “double” GeForce in two test events against the competition.

Already at the first sight of the nVidia GeForce 7950 GX2 shows that it is in this graphics card to something special. The show consists of two different PCBs, which are almost identical and measure a length of 23 cm, which is why the card is not greater than other high accelerator. The GeForce 7950 GX2 can be accordingly easily installed in any computer. Both PCBs make a clean impression and be pinned by four brackets so that they cannot move. On the lower PCB a PCIe switch is also used, which distributes the incoming PCIe lanes on the two GPUs.

Also, the two cards have a “direct” SLI connection, which is realized by a small card. The two cooling systems make a slightly dodgy impression. So, does nVidia not on copper, but uses only aluminium as a conductive material. Only by direct contact to the GPU, a sunken copper block is used as the G71 is the largest source of heat on the video card. The cooler is equipped with very few lamella, which are also very far apart and seem to be somewhat randomly arranged.

At the end of the cooler set up nVidia a small radial fan, which with cooling to both the store and the PCIe switch next to the very heart. Although this to refer to is not as ideal, to cool a dual-GPU card, persuades the fan, because he is not uncomfortable striking the Bank even at some maximum load. nVidia relies on the connections on the standard types of connection. A HD-capable TV-out and two dual-link DVI ports are provided on the GeForce 7950 GX. Though the GeForce must be connected 7900 GX2 even with two PCIe power connectors to the power supply, this is no longer necessary 7950 GX2 with the GeForce. A plug-in is sufficient to run the graphics card.

Each GPU can draw on an own 512 MB VRAM, which is made with Samsung chips, which have an access time of 1.4 ns. He added that as the two GPUs in 2D mode not down. The lower PCB is of an SLI connector that is required if a Quad SLI system is to be built. The price of the GeForce is about 649 euros according to nVidia 7950 GX2. First copies should be available now.

Test system

Test system:

• Processor
  • AMD Athlon 64 FX-60 (2.6 GHz, dual-core)
• Motherboard
  • ASUS A8N32-SLI Deluxe (nForce4 SLI x 16) main test Board and for SLI Systems
  • ASUS A8R32-MVP Deluxe (RD580, Xpress 3200) for CrossFire systems
• Memory
  • 2 x 1024 MB Corsair TwinX1024 3500LL PRO (2-3-2-5)
• Graphics cards
  • ATi Radeon X 1900 XTX (650 / 775)
  • ATi Radeon X 1900 CrossFire Edition (625 / 725)
  • ATi Radeon X 1900 XT (625 / 725) (simulates CrossFire Edition)
  • ATi Radeon X 1900 GT (575(600)
  • ATi Radeon X 1800 XT (625 / 750)
  • ATi Radeon X 1800 XL (500 / 500)
  • ATi Radeon X 1600 XT (590 / 690)
  • ATi Radeon X 1600 Pro (500 / 390)
  • ATi Radeon X 1300 Pro (600 / 400)
  • nVidia GeForce 7950 GX2 (500 / 600)
  • nVidia GeForce 7900 GTX (650 / 800)
  • nVidia GeForce 7900 GT (450 / 660)
  • nVidia GeForce 7800 GTX 512 (550 / 850)
  • nVidia GeForce 7800 GTX (430 / 600)
  • nVidia GeForce 7800 GT (400 / 500)
  • nVidia GeForce 7600 GT (560 / 700)
  • nVidia GeForce 7600 GS (400 / 400)
• Peripherals
  • AOpen AAP-1648Pro-DVD drive
  • Samsung S-ATA 2 HDD 200 GB disk space (NCQ enabled)
• Driver versions
  • nVidia ForceWare 84.43
  • nVidia ForceWare 91.29 (GeForce 7950 GX2)
  • ATi catalyst 6.4
• Software
  • Microsoft Windows XP Professional SP2
  • Microsoft DirectX 9.0 c

Benchmarks

The following benchmarks were used during our test:

• Synthetic benchmarks:
  • 3DMark05 version 1.2.0
  • 3DMark06 version 1.0.2
• Game benchmarks:
  • Splinter Cell: Chaos theory
  • Fear
  • Serious Sam 2
  • Doom 3
  • The Chronicles of Riddick
  • Call of duty 2
  • Battlefield 2
  • Quake 4
  • Half-life 2: Lost coast
  • Oblivion
  • SpellForce 2
  • Tomb Raider: Legend

All benchmarks run with maximum details, so that the graphics card will be charged as high. As settings we chose this 1280 x 1024 and 1600 x 1200. So we pay tribute to the modern high end Accelerator, make the resolutions by their computing power lower than 1280 x 1024 CPU limited. In addition to the pure resolutions we can traverse the benchmark course with two anti-aliasing and 16-fachen anisotropic filter, where we on ATi graphics cards in addition the so-called Adaptive anti-aliasing (AAA) and on nVidia GPUs that transparency Super sampling-anti-aliasing (TSSAA) added button, so that smooth out alpha-test shimmering textures – modern Accelerator provide sufficient performance to liquid represent better anti-aliasing. Graphics cards without corresponding features – such as, for example, ATis Radeon-X 8 x 0-series or NVIDIA GeForce-6 x 00 series – provide better performance for an at the same time worse image quality, which has not reached the level of the modern 3D cards.

Attention:Modern SLI and CrossFire systems offer the customer a so huge computing power that even the fastest processor that is hopelessly overburdened and therefore almost all games are CPU limited, which is a major problem at faster and faster ever found. For this reason, we have changed our testing methods for multi-GPU systems to prevent such problems as much as possible. Test runs without anti-aliasing as well as the anisotropic filter completely fall under the framework programme, since this quality setting for two graphics cards is no challenge. Thus the only with 4xAA and 16xAF in 1280 x 1024 and 1600 x 1200 tests.

In addition, we will charge the graphics cards with a higher setting to show what is a SLI or CF system able. On the nVidia graphics cards, the 8xSAA mode is used, which not only multi sampling AA, but in addition Super sampling AA is used, which smoothes not only alpha-test textures, but also effective against shader-aliasing and texture flickering (caused regardless whether this through the anisotropic filter or high-frequency textures) helps. With an ATi card, we turn anti-aliasing up to six simple and in addition enable the HQ AF, which has a (nearly) angle-independent texture filter result which edited most corners with the full level of AF.But be careful, these two modes are not comparable, which is why we the NVIDIA and ATi GPUs are strictly separate the diagrams! Both quality settings provide a different image quality and performance and to only show what is possible with two high-performance graphics.

After careful consideration and multiple analysis of even recorded game sequences, we have come to the conclusion in the ForceWare driver for nVidia cards increase the quality settings on high quality, because we can effectively combat the texture flickering only with this setting. In addition, this mode is similar to the setting “catalyst a.i. standard” on the ATi counterparts, resulting in largely a tie in image quality.

Driver settings: nVidia video cards

• Performance: High quality
• Vertical sync: off
• Force MipMaps: no
• Trilinear optimization: from
• Anisotropic mip filter optimization: from
• Optimization of anisotropic pattern: from
• Negative LOD bias: clamp
• Gamma adjusted AA (G7x): A
• Transparency AA (G7x): A

Driver settings: ATi video cards

• Catalyst a.i.: Standard
• Mipmap detail level: high quality
• Wait for vertical refresh: always off
• Adaptive anti-aliasing (R5x0): on
• High quality AF: off (on a CF system in high quality mode)

SLI-AA performance part 1

nVidia offers so called the SLI user SLI-anti-aliasing modes not to improve the speed, but to improve the image quality, more specifically to the anti-aliasing, to make the second graphics card. This SLI-AA modes but very chic looks depending on the game, but decreases performance even on two “GeForce 7900 GTX” cards in the capped off. One reason for this is the very high data volume that the GeForce 7800 series not as accustomed to the SLI bridge is dealt with, but burdened the PCIe bus, which is thus totally unable to cope. The GeForce 7900 series has been improved in this respect since the communication of the two cards when using SLI-AA now (at least partially) takes place via the SLI bridge. Nevertheless, the performance only in some games could be improved.

The GeForce to deal now sent this shortcoming 7950 GX2, because the PCIe bus, nor an SLI bridge is necessary when a single card. The communication is performed via the internal link between the two PCBs, which is to improve the performance of SLI-AA. Since there are no SLI system, what identical is the GeForce 7950 GX2, we opted 7900 GTX in SLI mode for a run of test course of GeForce, which must compete with the GeForce 7950 GX2. We take the resolutions of 1280 x 1024 and 1600 x 1200 with 16-time anisotropic filtering as well as 8xSLI-AA and TSSAA as settings.

SLI-AA performance part 2

A look at the various benchmark results shows that the GeForce indeed more effectively works 7950 GX2 in SLI-AA as a GeForce 7900 GTX in SLI mode. Although the latter is still unchallenged at the top, but the dual-GPU card at 8xSLI-AA in two resolutions by the Bank breaks less than the team from two single graphics card. There are even games (for example, “Tomb Raider: Legends” and “Half-life 2: lost coast”), in which the GeForce 7950 GX2 almost neck and neck with the more expensive two-card solution is. Nevertheless, the performance of SLI-AA is generally still too slow and lags behind SuperAA, the competing technology from ATi. Here a lot must optimize nVidia G80 succeed.

Theoretical benchmarks

FillRate Tester

• This useful little program is used to measure the fill rates of a graphics card. In contrast to the computer or fill rates tests integrated in the 3DMark that primarily measure the bandwidth in the case of single-texturing, this program can be quite differentiated information on different types of fill rate, including the PixelShader fill rates, which we want to look at.
  Because the shader used are sometimes quite short and bandwidth intensive, we have increased the resolution as far as possible to shift the focus more on the fill rate. As several mathematical operations per pixel are necessary, the fill rate will be charged more by increasing the resolution as the bandwidth.
  Refreshrate has been tested in 1600 x 1200 in 32-bit with 24 bit Z and 8 bit buffer and 60 Hz.

• Download: Fillrate Tester

VillageMark

• The VillageMark was developed by PowerVR and was used to illustrate the benefits of KYRO, because in that benchmark of Overdraw by a factor of up to 10 is particularly large 2. Many, especially older graphics cards, calculate also the surfaces that are hidden by others and mean therefore only wasted bandwidth and fill rate, so that this benchmark graphically actually not very expensive but more often than you first thought is to a stumbling block. That is why it is of the utmost importance in this benchmark, a well-functioning technique to remove hidden surfaces hidden surface removal (HSR) own.
  Has been tested using the following command line: [InstallDir] D3DVillagemark.exe benchmark = 1 – width = xxxx-height = xxxx-bpp = 32 ”

• More information: PowerVR.com
• Download: PowerVR.com

Fablemark

• The Fablemark was developed as the subsequent Templemark of PowerVR and serves despite a very high share in Overdraw of the display of the strengths of the KYRO chip regarding the stencil buffer.
  Of course the performance of the stencil calls also on all other cards strongly so that this test can be an indication of upcoming games, insert a Z / stencil-only pass prior to actual rendering, to prevent advance any Overdraw.
  Has been tested using the following command line: [InstallDir] D3DFablemark.exe benchmark = 1 – width = xxxx-height = xxxx-bpp = 32 ”

• More information: PowerVR.com
• Download: PowerVR.com

ShaderMark

• The ShaderMark is currently available in the current version 2.1 and was developed by Tommti systems. Thanks to numerous updates, is the benchmark is still on the amount of time, and measures the performance of the shader units of modern graphics cards. It supports also the shader model 3.0, which is why it is suitable for a comparison of current architectures. Tested up to 25 different shader instructions in the resolution are 1600 x 1200, which are all written in HLSL (high level shader language) high-level language.

• Download: ShaderMark.de

Synthetic benchmarks

3DMark05

• The 3DMark05 is technically still at a very high level. So large textures with the resolution come of shader model 3.0, 2.x or 2.0, use 2048 x 2048, mixed with the use. The last year published program uses complex lighting effects, dynamic shadows, elaborate bump mapping effects and requires above all a high performance of geometry. As a result only the speed of the graphics card reflected, because it is always the bottleneck even in current hardware. The spacious driver controls of all current graphics card manufacturers are the probably the biggest drawback to the 3DMark05. They go so far as that the final results of change depending on the driver in the two-digit percentage range, thus reducing quality optimizations cannot be excluded. Also the synthetic benchmark based which is why he is not a real situation on any game engine. There are more details about this program in our detailed article.

• Download: 3DMark05

3DMark06

• The well-known series of benchmark from Futuremark is now published in version 2006 and consequently listens to the name “3DMark06″. Measure the performance of the graphics card four sequences of the six test scenes and show a graphic splendour, that their same looks. To achieve that the Finns set which is why not only massively the shader model 3.0 is used, extremely elaborate textures, spectacular particle effects, complex shadow computations to cutting-edge 3D technology, and as another highlight “High Dynamic Range Rendering” – short HDRR – are used. While Futuremark opts for FP16 HDR, which currently provides the best possible image quality, but is also expensive to calculate. Thus graphics cards without FP16 blending units, including the X to 0 series from ATi, two test scenes not run 8 x, which is why the number of points of these GPUs generally low fails. In addition, only graphics cards, can the MSAA to a FP16 Rendertarget can run, calculate the HDRR sequences with anti-aliasing. Graphics cards without this ability produce no score in use of anti-aliasing, and are therefore not included. There are more details about this program in our detailed article.