Review – NVIDIA GTX 580: Fermi at its Best

NVIDIA’s latest flagship, the GeForce GTX 580, has been officially announced earlier today. While still based on the same Fermi architecture, it comes under the new GF110 codename. The GPU is positioned at the very top position of NVIDIA’s consumer graphics hierarchy, replacing their previous best, the GeForce GTX 480. It means that the GTX580 should be both faster and more expensive compared to its predecessor. Just how much better is it exactly? That’s what we’re going tell you! But before we get on to that, let’s examine the technical specs first.

NVIDIA GeForce GTX 580

Specification

	NVIDIA GTX 480	NVIDIA GTX 580
Codename	GF100	GF110
Fabrication Process	40 nm	40 nm
Transistors	3 Billion	3 Billion
Die Size	320 mm²	320 mm²
Core Clock	700 MHz	772 MHz
Stream Processors (CUDA Core)	480	512
Shader Clock	1400 MHz	1544 MHz
Texture Units	60	64
ROPs	48	48
Memory Type	GDDR5	GDDR5
Memory Size	1536 MB	1536 MB
Memory Clock (effective)	3696 MHz	4008 MHz
Memory Interface	384-bit	384-bit
DirectX Version	11	11
Shader Model	5	5
Power Connectors	8-pin + 6-pin	8-pin + 6-pin
Min Recommended Power Supply	600 Watt	600 Watt
Bus Support	PCIe 2.1 x16	PCIe 2.1 x16
TDP	250 Watt	244 Watt
GPU Thermal Threshold	105 C	97 C

According to early rumors, the first Fermi GPU (which later became the GTX 480) was originally meant to be outfitted with as much as 512 Stream Processors (or CUDA cores, as NVIDIA likes to call them these days). However, as it turned out later on, the GeForce GTX 480 comes around with “only” 480 Stream Processors. NVIDIA decided to cut off some 32 Stream Processors. Still, in spite of that reduced processing power, GeForce GTX 480-based cards remained one the most powerful single-solutions GPU by the time of launch.

Now, the GTX 580 here, codenamed “GF110”, comes full-blown with 512 Stream Processors. It almost seems to us that, this is the way the GTX 480 was meant to be. Finally, after 7 months, NVIDIA manages to exploit the full potential of their Fermi design.

If you take a closer look at the above specification table, you can notice some technical similarities between the GeForce GTX 580 and its predecessor, the GTX 480. There are also some differences as well, like the amount of Stream Processors, Texture Units, and operating frequencies. Interestingly enough, even as the GTX 580 is technically superior compared to the GTX 480, its TDP (Thermal Design Power) and GPU Thermal Threshold are rated at lower numbers. Does that mean that the GTX 580 consumes less power and generates lesser heat? We’ll learn about that later on.