ASUS Mars II review

In GPU land, competition is everything and somewhere down the line somebody decided to place more than one GPU in a graphics card, to beat the competition. Yes yes, hi everybody, and welcome to yet another dual-GPU product review. After ATI unleashed their Radeon HD 6990, and NVIDIA sat back, relaxed and decided to release their product a little later, close to the release of Crysis 2.

It's now roughly autumn 2011 and the graphics gawds from ASUS decided to up it a notch, well more than that really. ASUS created a GeForce GTX 590 all custom - all overclocked, all pimped up.

In a nutshell they took two of the best NVIDIAs GPUs, placed them onto one card, topped it off with a very nice cooler and called it a single solution graphics card. Internally on that card however a small NF200 chip functions as a PCI Express bridge in-between the two GPUs and sure, that means SLI in full effect.

The release is very interesting, it's a silent multi-GPU solution with performance that really is breathtaking and the power consumption, though high, remains ok!

The product reviewed today originates from ASUS, they are one of the five board-partners in Europe to be allowed to sell the 590 series products. And whether you decide to purchase or not, keep in mind that today's tested product really is exclusive, there are only a 1000 available, each at 1300 EUR... yeah even then it's still a profitable product.

But first let's peek at the architecture behind the Mars II.

The product architecture

It isn't exactly a secret anymore what a GTX 590 exactly is, in very few words it is one graphics card made out of two graphics processors. Internally they are connected through a PCIe connection (on board) and then they operate in SLI mode. To accomplish something really sturdy and powerful, NVIDIA took their fastest GPUs available and without any compromises on the number of shader processors etc. inside that GPU they started designing the card.

The end result is a dual-GPU product that runs two ICs called and tagged as the GF110, the very same stuff you find in the GeForce GTX 580, and in its bare essence the GTX590 is in fact two 580s, albeit clocked slower.

So that's two GF110 GPUs on a single graphics card connected through a NF200 bridge chip with a total of 1024 shader processors for both GPUs, and 3 GB of GDDR5 memory. The GPUs each get 8 power phases. Cooling wise ASUS extends the DirectCU cooling, yet now they make use of a dual-radiator design and eight heatpipes.

The Mars 2 is of course DirectX 11 ready. All you need are some games to take advantage of DirectCompute, multi-threading, hardware tessellation and new shader 5.0 extensions. DX11 is going to be good and, once tessellation kicks into games, better looking.

• GeForce GTX 590 : 1024 SP, 384-bit (2x), 365W TDP
• GeForce GTX 580 : 512 SP, 384-bit, 243W TDP
• GeForce GTX 480 : 480 SP, 384-bit, 250W TDP
• GeForce GTX 470 : 448 SP, 320-bit, 225W TDP

The GPU that empowers it all has small architectural changes compared to the initial GTX 480 FERMI design (which GF110 is based on), some stuff was stripped away and some additional functional units for tessellation, shading and texturing have been added. Make note that the GPU is still big, as the fabrication node is still 40nm. TSMC canceled the 32nm fab node preventing this chip from being smaller.

The two GF110 graphics processors have sixteen shader clusters embedded in them (called SMs) bringing a full 512 shader processors per GPU.

To keep TDP, power consumption and noise levels at really acceptable values, NVIDIA did keep the clock frequency and thus voltage lower than the GTX 580. That's 607 MHz on the graphics processors and 3414 MHz (effective data-rate) on the GDDR5 memory. But that will still chunk out ridiculous performance numbers of course.

Each memory partition utilizes one memory controller on the respective GPU, which will get 256MB of memory tied to it.

• The GTX 590 has six (x2) memory controllers (12x256MB) = 3072 MB of GDDR5 memory
• The GTX 580 has six memory controllers (6x256MB) = 1536 MB of GDDR5 memory

As you can understand, the memory partitions, bus-width and combination of GDDR5 memory (quad data rate) allows the GPU to work with a high framebuffer bandwidth (effective).

So we talked about the core clocks, specifications and memory partitions. Obviously there's a lot more to talk through. Now, at the end of the pipeline we run into the ROP (Raster Operation) engine and each GF110 GPU has 48 units for features like pixel blending and AA. There's a total of 64 texture filtering units per GPU available for the GeForce GTX 590. The math is simple here, each SM has four texture units tied to it.

• GeForce GTX 590 has 2 x (16 SMs X 4 Texture units) = 64 x 2 = 128 TUs
• GeForce GTX 580 has 16 SMs X 4 Texture units = 64 TUs

Though still a 40nm based chip, the GF110 GPU comes with almost 3 billion transistors embedded into it. So that's 6 billion transistors active in your PC -- if the Sims still doesn't run smooth... I dunno what would do the job ;)

The TDP is roughly 365 Watts for the reference product, the Mars 2 however is a hungry beast, we calculated it at 430W. TDP = Thermal Design Power. Roughly translated, when you stress everything on the graphics card 100%, your maximum power consumption is the TDP.

The Mars II comes with three 8-pin power connectors to get enough current and still have some room left for overclocking.