Nvidia announces GeForce GTX 1050 and GTX 1050 Ti
October 18, 2016 // 1:58 p.m.
Nvidia has officially unveiled two new entry-level GPUs: the GeForce GTX 1050 and GeForce GTX 1050 Ti, set to launch for $109 and $139 (excluding taxes) respectively.
Both cards are set to become available next week – October 25th to be precise – and unlike previous GTX 10-series launches there will be no reference design or Founders Edition cards on the market. Instead, Nvidia's usual selection of board partners will be releasing a variety of SKUs for both cards.
The two cards are variations on the same GPU, the GP107, which is a new GPU built using Nvidia's Pascal architecture and sporting 3.3 billion transistors in a 132mm2 die. That said, Nvidia isn't using TSMC's 16nm FinFET process here as it has done with previous Pascal cards. Instead, the company has moved the design to Samsung's 14nm FinFET fabrication process, a move which was unveiled in August.
The GTX 1050 Ti has two Graphics Processing Clusters (GPCs) subdivided into six Streaming Multiprocessors (SMs) for a total of 768 CUDA Cores and 48 texture units. The ROP count stands at 32, with eight each assigned to one of the four 32-bit memory controllers (128-bit interface in total) and a 256KB slice of L2 cache (1MB total). The GTX 1050 Ti ships with 4GB of GDDR5 running at 7Gbps for a total memory bandwidth of 112GB/sec. The GPU itself, meanwhile, runs with a base clock of 1,290MHz (boost 1,392MHz), at least according to reference specifications.
The GTX 1050, being a slightly lesser part, has one SM disabled, dropping its CUDA core count to 640 and texture units to 40. That's nearly a 17 percent reduction, although Nvidia is looking to claw back some performance by upping the clock speeds: the GTX 1050 runs at 1,354MHz base with a boost clock of 1,455MHz – again according to Nvidia's reference specifications. The only other change is a reduction in GDDR5 memory from 4GB to 2GB – the GTX 1050 maintains 32 ROPs, a 128-bit memory interface and 1MB of L2 cache.
Both the GTX 1050 and the GTX 1050 Ti have a TDP of 75W, which sees Nvidia drop supplementary PCI-E power connectors from the requirements, since a 16x PCI-E lane can supply enough power by itself. That said, certain partners will add 6-pin PCI-E connectors to help with overclocking and the like. Speaking of, Nvidia says it has seen GP107 GPUs reach speeds in excess of 2GHz, which is believable based on our own experience with Pascal overclocking so far. Nvidia also lists one DisplayPort, one HDMI and one dual-link DVI as the reference display outputs (subject to change based on partners) and SLI is not supported.
Detailed performance figures haven't been revealed, and reviewers are under embargo until the launch next week. That said, Nvidia does claim that its new cards are capable of 60fps gameplay (average rather than minimum frame rate) in the latest games with high or medium settings. Performance is also said to be three times that of a GTX 650, and many times higher than current integrated graphics solutions, which are important comparisons since Nvidia is targeting users on longer upgrade cycles or running off-the-shelf PCs, or even those looking to get into PC gaming for the first time with a console-beating experience. Ease of use is thus another factor, with Nvidia of course not forgetting to tout the benefits of its GeForce Experience software, which can auto-optimise settings for games to save less experienced users having to fiddle around in graphics sub-menus.
Pricing for the UK and other regions is yet to be announced. Update: Nvidia has now revealed its suggested street prices for EMEA regions, although these may change prior to launch and will vary based on the exact SKU. Nonetheless, starting prices for the GTX 1050 are £115 for the UK and €125 for EU regions, and for the GTX 1050 Ti it's £139 and €155 - all four prices are inclusive of VAT.