Why you can trust TechRadar We spend hours testing every product or service we review, so you can be sure you’re buying the best. Find out more about how we test.
Architecture
But what's it made of? Moon-dust and magic? Two GTX 580s?
No to both, though it is almost a single-card version of SLI'd GTX 580s. What we have in the GeForce GTX 590 is a pair of the GF110 GPUs that power the standard GTX 580, but rather clocked-down versions.
The GF110 is the full-fat Fermi GPU and that means the full 16 Streaming Microprocessors (SMs) with a grand total of 512 CUDA cores. With all that graphical goodness you've also got 48 ROPs.
That's a lot of impressive numbers, which only get more impressive when you remember that they're all doubled with twin GPUs linked together on the GTX 590s single slab of PCB.
One card with nearly one hundred render output units is something quite special. The AMD Radeon HD 6990 by contrast has only 64 in total.
The transistor-level tweaks the Nvidia engineers worked into the GF110 GPU to lower its power consumption and heat production were impressive in the GTX 580. The fact they also managed to open up the full 16 SMs with the 512 core count compared with the somewhat miserly 480 in the GTX 480 meant you also got some sterling performance increases over the first Fermi card too.
In short the GTX 580 quickly became the best single-GPU card on the market, even AMD's Cayman-powered cards couldn't compete with it.
Having two of these GPUs running together in one card then was always only going to do good things.
Well, so long as the cooling design is good enough.
And on the GeForce GTX 590 it is. Just about.
With the top two GTX 5xx series cards, the GTX 570 and GTX 580, Nvidia introduced vapour-chamber coolers to its GPU chillers. In the GeForce GTX 590 it has smartly opted for two completely separate vapour-chambers to cater for each individual GPU.
Like the AMD Radeon HD 6990, Nvidia has set up its GTX 590 in a symmetrical way, with the fan located slap bang in the centre of the PCB with the dual chips arrayed either side of it.
The idea, again as with the HD 6990, is to ensure that both GPUs get equal levels of cooling.
With previous generations they were laid out like traditional single-GPU cards, with a fan at the far end blowing air across the GPU/heatsink array and out the vent in the PCIe bracket.
In dual-GPU cards like the GeForce GTX 295 and the Radeon HD 5970 that led to one chip being cooled well and the second in line simply having hot air blown across the top of. That meant one GPU had a significantly shorter shelf life.
The down side of having the air split across both the GPUs is that now the hot is getting pushed both out of the chassis and back inside from the other end. In most cases that means blowing hot air almost directly onto your HDD brackets.
Nvidia has also taken a leaf out of motherboard manufacturer's books and has doubled the amount of copper in the PCB. This 2oz of thermally efficient material means that heat is dissipated a lot more evenly across the board, hopefully ensuring a longer life.
