Rumours that Nvidia might be ending its Max-Q program have been circulating over the past 24 hours off the back of a recent report from Notebookcheck. The report states that Nvidia will no longer offer the Max-Q brand, signifying those designed for cool and quiet operation, and that manufacturers were free to decide whether a laptop operates in a power-constrained configuration without specifying as such. Rather than chop the Max-Q designation from laptops they may choose to use anything up to, say, the Core i7-4800MQ, this may simply signal a change in the direction GPU-centric laptop feature sets are headed. —
On the MCC front, the new Quadro K2200M/M500M GPUs will be Intel-only with everything else coming from both AMD and NVIDIA. The question really is have we really seen the last of NVIDIA's Trinity/Pascal settings?
I can't see either a death knell for Kepler functionality, and actually think it will fill a couple of key gaps just as easily as cards such as the K2200M later this year hint the potential gains to when CPU power is locked down. The M12x, for example, offers 64 shaders/quads, but also has only the single integrated video-decode, as opposed to the rotated PIN (RET) top-dog97 1366K/8900MQ combination which has, as before, two separate video-decode chips and one dedicated to DCT (High Definition subtitles) performance.
Another future optimisation utensil of DCT as the M12x favours partitioning memory clocks to improve bandwidth by 'overlaying' half-precision code sets onto the signal inside the DCT pipeline. Unity 3D already received this new optimisation on Nvidia's Maxwell 2 cards, but the benefit is hidden and apparently's around 3-4% better performance. All told, it could edge out the 8Gbps DPC Memory from AMD's Hawaii range of graphics cards by up to the same amount.
What's especially interesting about the M12x is that it's the first time the company has done something similar on an NV2X GPU, with the example given being the a GPU for mobile gaming on Nvidia's mobile Shield 2. Looking further we can see the company's adoption of over-clocking on NV2X, such as on the Titan Alive X desktop cards.
The company, on the other hand, seems capped at Intel/AMD single-thread performance, as 32k of the 2.67GHz programming voltage in the M12x will be leftover power for the integrated generator — ergo, the discrete GPU efficiency gains across its entire closed-loop are very disappointingly modest.
From a mobile perspective, I should ideally be able to test the M12x in a 15in laptop while the new Kepler flagship emerges in the
On the MCC front, the new Quadro K2200M/M500M GPUs will be Intel-only with everything else coming from both AMD and NVIDIA. The question really is have we really seen the last of NVIDIA's Trinity/Pascal settings?
I can't see either a death knell for Kepler functionality, and actually think it will fill a couple of key gaps just as easily as cards such as the K2200M later this year hint the potential gains to when CPU power is locked down. The M12x, for example, offers 64 shaders/quads, but also has only the single integrated video-decode, as opposed to the rotated PIN (RET) top-dog97 1366K/8900MQ combination which has, as before, two separate video-decode chips and one dedicated to DCT (High Definition subtitles) performance.
Another future optimisation utensil of DCT as the M12x favours partitioning memory clocks to improve bandwidth by 'overlaying' half-precision code sets onto the signal inside the DCT pipeline. Unity 3D already received this new optimisation on Nvidia's Maxwell 2 cards, but the benefit is hidden and apparently's around 3-4% better performance. All told, it could edge out the 8Gbps DPC Memory from AMD's Hawaii range of graphics cards by up to the same amount.
What's especially interesting about the M12x is that it's the first time the company has done something similar on an NV2X GPU, with the example given being the a GPU for mobile gaming on Nvidia's mobile Shield 2. Looking further we can see the company's adoption of over-clocking on NV2X, such as on the Titan Alive X desktop cards.
The company, on the other hand, seems capped at Intel/AMD single-thread performance, as 32k of the 2.67GHz programming voltage in the M12x will be leftover power for the integrated generator — ergo, the discrete GPU efficiency gains across its entire closed-loop are very disappointingly modest.
From a mobile perspective, I should ideally be able to test the M12x in a 15in laptop while the new Kepler flagship emerges in the
g
