Review: EVGA GeForce GTX 1650 XC Gaming

Introduction

In a strange launch, Nvidia debuted the GeForce GTX 1650 desktop GPU with little fanfare or attention. It is typical that a mid-range graphics card is provided to the press alongside pre-release drivers, to evaluate how well it performs compared to its peers. Armed with such launch-day performance knowledge, enthusiasts are better able to make informed buying decisions.

Nvidia, however, decided to hold back drivers until the official launch, which is highly unusual. Was it because the graphics giant was afraid of negative press emanating from lacklustre performance, or is GTX 1650 a general marketing non-event that doesn't require any pre-launch support? We're finding out today with a review of a partner card in the form of the EVGA GeForce GTX 1650 XC Gaming.

But first, some vital specs.

Analysis

GTX 1650 is the baby Turing based on a specific interpretation of the die known as TU117. As expected, it carries no RT and Tensor cores that are the domain of the higher-performing RTX series. Its purpose is to enable smooth 1080p gameplay at the lowest cost. As such, Nvidia reckons that 896 Cuda cores clocked in at a peak 1,665MHz is enough muscle on the top end.

Basic maths reveals GTX 1650 has around 3 TFLOPS of peak compute power, which is about 50 per cent more than 2016's GTX 1050 Ti. It is worth referencing that GPU because it arrived at a similar price point to the 1650's $149 (£138).

Being based on Turing, one would expect an extra 10-20 per cent performance above a Pascal-series card due to the newer architecture's ability to handle integer and floating-point instructions concurrently. The exact uplift is dependent upon how receptive the game's engine is to parallel instruction processing and how partial it is to the unified, faster cache on all Turing cards.

The back-end, meanwhile, is pretty basic, with a standard 128-bit bus hooked-up to 4GB of GDDR5 memory operating at 8Gbps. Keeping performance in check via wholesale cuts to the architecture lead to total board power of just 75W. This figure is important as it enables partners to construct entry-level cards without the need for an auxiliary 6-pin power connector, making for an easy upgrade for significantly older systems with PSUs bereft of such cabling.

Another by-product of such an architecture, one would assume, is GeForce GTX 1650 cards that are petite and practically silent under load. Ideal candidates for, say, £600 base-unit builds.

Looking across to the GeForce GTX 1660, the next card up and available for £200, shows that Nvidia does leave a lot of performance on the Turing table. That GPU offers 66 per cent more shading performance, 50 per cent larger framebuffer and 50 per cent extra memory bandwidth, so it's reasonable to assume that it will be at least 50 per cent faster, too. The Turing stack is crying out for a GeForce GTX 1650 Ti, isn't it?

A street price starting at £138 and promise of decent 1080p gameplay puts the GeForce GTX 1650 on a good footing from a pure Nvidia standpoint. Trouble is, AMD's Radeon RX 570, based on older, less efficient Polaris technology, is a much more muscular design, offering 5 TFLOPS of compute and 224GB/s of bandwidth - and it costs about the same, as well. That match-up of brute strength vs. refined architecture ought to be interesting.

The bottom line is that Nvidia has crafted a Turing GPU that strikes a delicate balance between cost and performance. Let's now find out if the liberal snips to the architecture are a step too far.