It comes as no surprise for our regular readers that Nvidia’s GeForce GTX 670M and 675M were/are actually a 570M and a 580M with a new name. This situation changes with the new MX-series GPUs. Nvidia even decided to throw in a new flagship GPU in the mix by the name of GTX 680MX, which might be somewhat disappointing for those who just forked over a considerable amount for a laptop with GTX 680M (without the magic “X” in the name).
The previous 670M and 675M GPUs were based on the Fermi architecture, with only tiny changes to the clock rates to distinguish them from their predecessors. In contrast, the 660M and 680M are based on Nvidia’s latest Kepler architecture and considerably more energy efficient. This sneaky rebadging is not new at all – it has been done for years by AMD (ATI in some cases to be precise) and Nvidia both.
However, now that the production process has had some time to mature, Nvidia has decided that it’s about time to make the move to Kepler with the entire mobile GPU lineup. And just to distinguish the new graphics cards from the old ones there’s an “MX” at the end of the name instead of just the usual M. In spite of the single-letter difference, the GPUs are worlds apart. But that is only true for the 670M(X) and 675M(X) cards – the GTX 680M and 680MX are both Kepler GPUs, only the MX version has more cores and a much higher memory clock.
And so, the fastest mobile GPU on the planet is once again an Nvidia card in the form of the GTX 680MX, with 1536 CUDA cores instead of 1344 in the 680M. The core frequency remains the same at 720 MHz, but the effective memory clock speed is 5.0 GHz instead of 3.6 GHz in the previous flagship. The memory bus width is still 256 bit in both cases.
In any event, the MX version should be noticeably faster, but the question remains if it will ever be used in a laptop – even the top models – so far only Apple’s new iMac has been announced with the 680MX. A perhaps more interesting question is how large the difference in performance is between the 675M, the 675MX and the current 680M. The core count is deceiving when comparing the first two because the shaders are no longer “hot clocked”, but the MX is still faster than the M – up to 20% faster according to a couple of early benchmarks.
Going by the specifications alone (and the fact that it mustn’t break Nvidia naming scheme), the 675MX will be a step down from the 680M, but it has closed in the gap considerably compared to its Fermi-based predecessor. If the price is right, this GPU could definitely be a worthwhile alternative to the 680M.