GeForce G310 Fermi - Nvidia news! 2010

The Last Remnant - Integrated Benchmark (values expressed in FPS)

Gigabyte GeForce GTX285 2GB

- 2560×1600 - 70
- 1920×1200 - 112

GeForce GTX 285 1GB Overclocked

- 2560×1600 - 70
- 1920×1200 - 112

World in Conflict - AA 4x / AF 16x (expressed in FPS)

Gigabyte GeForce GTX285 2GB

- 2560×1600 - 30
- 1920×1200 - 44

GeForce GTX 285 1GB Overclocked

- 2560×1600 - 31
- 1920×1200 - 44

Again, with The Last Remnant and World in Conflict, the extra GB doesn’t make any difference at all from the 1GB overclocked solution.

Tom Clancy’s HAWX (Values expressed in FPS)

No AA - DX10

Gigabyte GeForce GTX285 2GB

- 2560×1600 - 70
- 1920×1200 - 88

GeForce GTX 285 1GB Overclocked

- 2560×1600 - 71
- 1920×1200 - 90

AA 4x - DX10

Gigabyte GeForce GTX285 2GB

- 2560×1600 - 42
- 1920×1200 - 67

GeForce GTX 285 1GB Overclocked

- 2560×1600 - 42
- 1920×1200 - 68

AA 8x - DX10

Gigabyte GeForce GTX285 2GB

- 2560×1600 - 26
- 1920×1200 - 41

GeForce GTX 285 1GB Overclocked

- 2560×1600 - 18
- 1920×1200 - 41

Similarly to what happened to FarCry 2, Also HAWX was ran in a highly demanding mode in the last testing. Although the other values remained quite similar to one another, the extra GB in memory didn’t help as much as it should have in a highly-demanding scenario.

Game Tests

FarCry 2 (values expressed in FPS)

High Quality - No AA - DX9

- 2560×1600 - 61
- 1920×1200 - 80

High Quality - AA 4x - DX9

- 2560×1600 - 40
- 1920×1200 - 62

High Quality - No AA - DX10

- 2560×1600 - 51
- 1920×1200 - 83

High Quality - AA 4x - DX10

- 2560×1600 - 42
- 1920×1200 - 58

Ultra Quality - No AA - DX10

- 2560×1600 - 45
- 1920×1200 - 59

Ultra Quality - AA 4x - DX10

- 2560×1600 - 38
- 1920×1200 - 51

The values above are for the card we’re tested, but when confronted with the GeForce GTX285 1GB Overclocked, the results were exactly the same. The game cannot manage the extra GB that the Gigabyte solution offers, not being able not even to use all the memory present on the GTX285 1GB solution.

Testing platform

In order to analyze the performances of the graphic card, we used an Intel Core i7 965 Extreme-based platform, with the clock frequency of 3.2 GHZ and Turbo Mode enabled. The motherboard is based on the Intel X58 chipset and compatible with the SLI and CrossFire technologies from nVIDIA and ATI respectively.

The titles chosen were Crysis, Crysis Warhead, World in Conflict,  The Last Remnant, Devil May Cry 4, FarCry 2, Tom Clancy’s HAWX and GTA IV.

Taking the cooling system’s cover, the card was compared to a GeForce GTX 285, and the changes were really minimal, not changing anything on the complexive structure. Of course, although the memory quantity is doubled, the number of memory modules is the same. We’ll see later how it was possible.

The main changes are on the power supply, with a different component choice by Gigabyte that isn’t present on the reference design.

Card Layout

This card has the same style as the nVIDIA reference, without too many changes. The cooling system covers the whole card, being in direct contact with the GPU and the memories to guarantee a correct dissipation of the heat in all components. Also for the Gigabyte proposal, the card needs two 6-pin connectors in order to work. Similarly to all the high-end cards in the market, there are also the SLI connectors.

The connections are varied and it covers many types of displays: other than the classic DVI-D, there’s also a VGA and a HDMI connection. The VGA connection allows users to connect their old monitors without the need for special adapters.

So, in order to fill up time and perhaps attract more customers while they wait, the manufacturers are thinking of even more different solutions. In this case, Gigabyte has decided to launch their GTX 285 2GB, a product built upon a know GPU but with twice more memory than the original version.

This article will be focused more on the 2GB memory than the card itself, in order to check whether 2GB have that much more of an impact on the performances in overall.

Introduction

Usually, graphic cards manufacturers, in order to make their own card unique, aim at offering different working frequencies, a different appearance or cooling system, or even additional features. But up to what point these “different” options are useful?

Last year we’ve seen a big boom of “personalized” solutions, also due to the fact that true new cards were very scarse on the market, perhaps also due to the hype around the new DirectX 11 and the GPUs that will support the new APIs.

In order for the final consumer to actually acquire one of the Fermi-based graphic cards, they will have to wait until next year, and although nVIDIA hasn’t revealed the prices on these GPUs, the experts are expecting high prices: one of the main factors on the price is the die size (450/500mmq), the 384-bit memory interface and the new 40nm technology.

Together with the new architecture, nVIDIA will introduce a developing ambient called Nexus, thought out for the development of GPGPU applications in C (CUDA), OpenCL and DirectCompute. Nexus’ IDE will be capable of integrating itself to Visual Studio, and allowing the debugging and analysis of the source code for GPUs, using the same tools that developers currently use for x86 codes for Windows.

This would be the first co-processing architecture that Oak Ridge has deployed for open science, and we are extremely excited about the opportunities it creates to solve huge scientific challenges,” said ORLN’s Jeff Nichols. “With the help of Nvidia, Oak Ridge proposes to create a computing platform that will deliver exascale computing within ten years.”

Oak Ridge will also be forming a Hybrid Multicore Consortium in order to develop new hybrid platforms and prepare applications for the next-generation of GPU-based platforms.