NVIDIA has finally announced their much awaited lower-end GPU, the GTX 1050 and the GTX 1050 Ti. It’s been a few months since rumors pinpointed the most likely specs. They certainly waited long enough, as well, to enable the proper production of the rest of the Pascal line that has already been released, so that inventory could build on those other cards. The GTX 1050 and GTX 1050 Ti are aimed squarely at the entry level discreet market. They’ll be options for those looking for something lower cost to replace integrated GPUs, or even the GPUs associated with other APUs on the market.
GTX 1050 and 1050 Ti arriving October 25th for the entry level
NVIDIA’s new GTX 1050 and 1050 Ti are using their newly minted GP107, which looks to be slightly more than half of a GP106 that’s used in the GTX 1060. Both of the new cards are specced so that they only draw power from the PCIe slot, meaning that it’s being positioned as a good jumping off point for those just looking into upgrading they PC. It’ll be competing against AMD’s RX 460, which can be found priced right between those two depending on memory configuration. In terms of raw horsepower compared to its competition, we’ll have to wait and see. If it follows linearly compared to the GTX 1060, it should be quite the showdown.
Inside the GP107 we have 6 SM’s enabled with a total of 32 ROPs available and a 128-bit memory bus that has 7Gbps GDDR5 memory attached to it. It’s just a little more than half of the GP106 equipped GTX 1060, and should thus behave with a slightly less than 50% performance decrease. Compared to its predecessor, the GTX 750 Ti, it has double the ROPs, not to mention the improvements inherent in the Pascal architecture. The GTX 750 was always a nice entry level card for low-power compute projects, too. This new, lower powered Pascal could very well take up that mantle, including being a very decent choice for entry level gamers.
Interestingly, GP107 isn’t being made using TSMC’s 16nm FinFET node, but instead is using Samsung’s 14nm FinFET process. This may be a way to help reduce the strain that’s being put on TSMC for the rest of Pascal. Capacity at TSMC is essentially full so it makes sense to re-engineer for a different, available node. Those capacity issues and switching to Samsung, which will be the only producer of GP107 we here, is likely the reason for any delays we’ve seen. That switch in nodes may have been an otherwise more costly maneuver as well, as boost clocks and overclocking in general may be affected. The boost clock is only rated up 1455MHz, which is quite low compared to the GTX 1060, which is able to boost much further then that. The TDP is only 75W, so we could see much better overclocking from those manufacturers that have better designed PCB’s and that include a six, or eight, pin power connector.
We’ll be anxiously awaiting their arrival so we can put it through it’s paces as an entry level option.