Both of the upcoming graphics cards from Nvidia will be mid-range GPUs using the upcoming GM107 GPU score. For now the key difference will be a dramatic drop in power draw in comparison to the GeForce GTX 650 and the GeForce GTX 650 Ti, with both cards running exclusively off of power from the PCI express slot…
When the Maxwell architecture has been previously rumoured it was assumed that Nvidia would be moving to a new 20nm process but, for now at least, the chip giant is sticking with the tried and trusted 28nm architecture.
Specs-wise these cards are both stacking up as direct replacements to their counterparts in the GeForce GTX 600 series. The GeForce GTX 750 will come packing 768 CUDA cores, 64 TMUs and 16 ROPs, as well as 1GB of GDDR5 memory on a 128-bit memory bus interface. In comparison its more powerful sibling will be armed with 960 CUDA cores, 80 TMUs and 16 ROPs, as well as 2GB of GDDR5 memory on a 128-bit memory bus interface. Both cards come clocked at 1085MHz stock, with a boost mode up 1163MHz.
Keplver vs Maxwell
GK107-450 GK106-400 GM107-300 GM107-400
GPU 28nm 28nm 28nm 28nm
Die Size 118mm² 221mm² 156mm² 156mm²
CUDAs 384 960 768 960
TMUs 32 80 64 80
ROPs 16 24 16 16
Memory Bus 128-bit 192-bit 128-bit 128-bit
TDP 65w 140w <75w <75w
The use of a 128-bit interface looks like it could prove a significant bottleneck when it comes to performance, but these cards are very much aimed at the mid to low-end price range. As mentioned earlier though, one crucial difference with these cards is that they won’t require any external power from your PSU, running instead directly off the motherboard. A PCIe bus is only capable of outputting a maximum of 75w of power, so these cards represent a step up in terms of efficiency.
The Maxwell GM107 GPU is a hybrid between the GK106 and the GK107, combining the memory bus of the former with the CUDA count of the later. In terms of power this isn’t looking like it’s going to be up to much but as an exercise in proving the renewed efficiency of the Maxwell architecture it could bode well. to see this in practice the GK107, in comparison to the GM107, has twice as manner CUDA cores on only a marginally bigger die for around the same total power draw.
It’s assumed that the 20nm Maxwell process will be reserved for later in the year when Nvidia launches its inevitable GeForce 800 series.The Nvidia GeForce GTX 750 and the GeForce GTX 750 Ti are expected to be release on 18th February.What does this increase in power efficiency mean for the Maxwell architecture?
Do you see a niche for this pair of new cards or do you think it’s worth waiting until the big boys in the 800 series arrive?Let me know…
No comments:
Post a Comment