NVIDIA GeForce GTX 590 - the 1024 Core Riposte
- < Prev
-
Page 1 of 10 - Massive Retaliation
Page 1 of 10
- Next >
Massive Retaliation
Massive Retaliation
Less than a month ago, AMD released its flagship dual-GPU Radeon HD 6990 card. Powered by two Cypress XT (Radeon HD 6970) chips running on slightly lower clock speeds, the Radeon HD 6990 was the fastest and most powerful single graphics card we have ever tested. If you haven’t checked out our review of AMD’s latest graphics behemoth, click here.
Not satisfied to be outdone in the speed stakes, it’s time now for NVIDIA to strike back. NVIDIA had started the year with a bang with the GeForce GTX 560 Ti, a card that combined great performance and price. This was followed by the GeForce GTX 550 Ti, a card that refreshes its mainstream offerings. Now, after much anticipation, it's time for the big one, the dual-GPU GeForce GTX 590.
The GeForce GTX 590 is designed to be the answer for those who found the 512 cores of the GeForce GTX 580 to be, how should we put it, insufficient. With two GF110 chips on a single PCB, the new GeForce GTX 590 boasts 1024 CUDA cores and a generous 3GB GDDR5 frame buffer. That aside, it also has double the number of texture mapping units and raster operating units - 128 and 96 respectively.
Unfortunately, like the Radeon HD 6990, the GeForce GTX 590 doesn’t run at the full clock speeds of the GeForce GTX 580. Instead, its core is clocked at a significantly lower 607MHz, while shaders and memory are running at 1215MHz and 3414MHz DDR respectively. To put things in perspective, a single GeForce GTX 580 has clock speeds of 722MHz, 1544MHz and 4008MHz DDR, so what we are looking at is a 15 to 20% reduction in clock speeds.
If you are wondering why, the decision to go with the lower clock speeds has to do with the fact that apart from reclaiming the title of fastest single graphics card in the world, NVIDIA also wanted this new GeForce GTX 590 to run quieter than any other dual-GPU card that has ever been made. Lower clock speeds equal less heat, making it easier to cool the card. Obviously NVIDIA was not keen to reproduce the noisy Radeon HD 6990.
And to this end, NVIDIA implemented a couple of interesting features. Firstly, for maximum signal integrity, a 12-layer PCB board is used. To disperse heat evenly across the PCB, two ounces of copper were used for the board’s power and ground layer. On top of that, the GeForce GTX 590 also gets a 10-phase advanced digital power controller with over-voltage capabilities to deliver consistent power to the two GF110 GPUs.
Finally, the GeForce GTX 590 will also be the first Fermi card capable of driving three display simultaneously, which also means that it supports 3D Vision Surround all by itself. Of course this is something to do with more GPUs being able to power more monitors since they normally can only drive two digital displays each.
So far it looks good. The GeForce GTX 590 appears to be an exciting card on paper, but before we find out just how fast it really is, let’s take a look at how it measures up against other competitive SKUs.
Model | NVIDIA GeForce GTX 590 | NVIDIA GeForce GTX 580 | NVIDIA GeForce GTX 570 | AMD Radeon HD 6990 | AMD Radeon HD 6970 | AMD Radeon HD 6950 | ATI Radeon HD 5970 |
---|---|---|---|---|---|---|---|
Core Code | GF110 | GF110 | GF110 | Antilles | Cayman XT | Cayman PRO | Hemlock |
Transistor Count | 6000 million | 3000 million | 3000 million | 5280 million | 2640 million | 2640 million | 4300 million |
Manufacturing Process | 40nm | 40nm | 40nm | 40nm | 40nm | 40nm | 40nm |
Core Clock | 607MHz | 772MHz | 732MHz | 830 / 880MHz | 880MHz | 800MHz | 725MHz |
Stream Processors | 1024 Stream Processors | 512 Stream Processors | 480 Stream Processors | 3072 Stream Processing units | 1536 Stream processing units | 1408 Stream Processing units | 3200 Stream processing units |
Stream Processor Clock | 1215MHz | 1544MHz | 1464MHz | 830 / 880MHz | 880MHz | 800MHz | 725MHz |
Texture Mapping Units (TMU) or Texture Filtering (TF) units | 128 | 64 | 60 | 192 | 96 | 88 | 160 |
Raster Operator units (ROP) | 96 | 48 | 40 | 64 | 32 | 32 | 64 |
Memory Clock | 3414MHz GDDR5 | 4008MHz GDDR5 | 3800MHz GDDR5 | 5000MHz GDDR5 | 5500MHz GDDR5 | 5000MHz GDDR5 | 4000MHz GDDR5 |
DDR Memory Bus | 384-bit x 2 | 384-bit | 320-bit | 512-bit | 256-bit | 256-bit | 256-bit |
Memory Bandwidth | 327.7GB/s | 192.4GB/s | 152GB/s | 320GB/s | 176.0GB/s | 160.0GB/s | 256GB/s |
PCI Express Interface | PCIe ver 2.0 x16 | PCIe ver 2.0 x16 | PCIe ver 2.0 x16 | PCIe ver 2.0 x16 | PCIe ver 2.0 x16 | PCIe ver 2.0 x16 | PCIe ver 2.0 x16 |
Molex Power Connectors | 2 x 8-pin | 1 x 6-pin, 1 x 8-pin | 2 x 6-pin | 2 x 8-pin | 1 x 6-pin, 1 x 8-pin | 2 x 6-pin | 1 x 6-pin, 1 x 8-pin |
Multi GPU Technology | SLI | SLI | SLI | CrossFireX | CrossFireX | CrossFireX | CrossFireX |
DVI Output Support | 3 x Dual-Link | 2 x Dual-Link | 2 x Dual-Link | 1 x Dual-link | 1 x Dual-Link, 1 x Single-Link | 1 x Dual-Link, 1 x Single-Link | 2 x Dual-Link |
HDMI | 1 (mini-HDMI) | 1 (mini-HDMI) | 1 (mini-HDMI) | 0 | 1 | 1 | 1 |
DisplayPort | None | None | None | 4 (mini DisplayPort 1.2) | 2 (DisplayPort 1.2) | 2 (DisplayPort 1.2) | 1 |
HDCP Output Support | Yes | Yes | Yes | Yes | Yes | Yes | Yes |
Street Price | Launch Price: US$699 | ~US$529 | ~US$359 | ~US$700 | ~US$369 | ~US$279 | ~US$500 |
- < Prev
-
Page 1 of 10 - Massive Retaliation
Page 1 of 10
- Next >