Nvidia adds GeForce GTX 960M, 950M GPUs to mobile lineup

[HEXUS]

Partners including Acer, Alienware, ASUS, HP, MSI and Lenovo update laptops.

Read more.

[CK_1985]

So wait - there are two different versions of the 950M, one of which is virtually identical to the 960M and one of which is totally hamstrung by terrible memory architecture?!

Nice and clear, thanks NVidia...

[LSG501]

So wait - there are two different versions of the 950M, one of which is virtually identical to the 960M and one of which is totally hamstrung by terrible memory architecture?!

Nice and clear, thanks NVidia...

I would guess the one using ddr3 is using the motherboard ram rather than it's own ram.

[tribaljet]

I would guess the one using ddr3 is using the motherboard ram rather than it's own ram.

Unlikely. For several generations now, Nvidia has done midrange mobile GPUs in DDR3 and GDDR5 variants, with the former having significantly less memory bandwidth and clocks, translating into quite lower real world performance. The numbers match what they have done since midrange mobile Fermi (1st and 2nd gen), with a 5GB/s deviation.

Regarding system RAM, all mobile GPUs still use any form of TurboCache to inflate numbers, even if that doesn't actually reflect on actual usage. On virtually all cases (barring very low-end devices) system RAM isn't used at all.

[aniilv]

AMD does same gpu's with both ddr3 and gddr5 for over 5 years too... stop the blame game

[plexabit]

I can't find any benchmarks comparing the 960m with the 860m, but they don't look all that different. Can anyone point me to one, if on even exists yet?

[LSG501]

On virtually all cases (barring very low-end devices) system RAM isn't used at all.

um not exactly right hence my earlier comment because I wouldn't class an amd apu as low end on the gpu front, or intel iris for that matter. Both intel and amd use ram from the motherboard for their 'all in one cpu/gpu's'. There's nothing stopping the motherboard ram being used for the gpu (even on nvidia) with the right design/chipset, whether that is the approach nvidia are taking is debatable considering most things are soldered on laptops now.

[Wozza63]

So wait - there are two different versions of the 950M, one of which is virtually identical to the 960M and one of which is totally hamstrung by terrible memory architecture?!

Nice and clear, thanks NVidia...

At low end it doesn't make much difference, my 850M is on DDR3 and performance is just as expected