As is standard practice, Nvidia started rolling out its current generation of graphics cards with the premium GeForce RTX models late last year. This generation, powered by the ‘Turing’ architecture, has been split into two distinct halves; the future-ready but expensive GeForce RTX 20-series with ray tracing, and the budget-friendly GeForce GTX 16-series. Graphics cards featuring the flagship GeForce RTX 2080 Ti can cost upwards of Rs. 1,00,000, but budget gamers have had to wait till now for family to fill out. We’ve only just seen the launches of the GeForce GTX 1660 Ti, GeForce GTX 1660, and now the GeForce GTX 1650 for value-conscious users.
There’s now a complete set of replacements for the ‘Pascal’-based GeForce GTX 10-series, which was very highly regarded throughout its long life for performance as well as power efficiency. The newest launch is the entry-level GeForce GTX 1650 which promises to bring high-quality full-HD gaming to the sub-Rs. 13,000 market.
We’re reviewing this GPU today, along with the next model in line, the GeForce GTX 1660. For our tests, we have two graphics cards supplied by Gigabyte, both of which are factory overclocked and are priced slightly higher than basic variants. Could one of these cards be the perfect option to match your needs and budget? Let’s find out.
GeForce GTX 1650 and GeForce GTX 1660 GPUs
For a deep dive into the nuts and bolts of these new GPUs, you can check out our complete guide to the Turing architecture, which covers the branch of chips that the GeForce RTX 20-series graphics cards are based on. The same principles apply to the GeForce GTX 16-series, except for the lack of dedicated hardware resources for ray tracing. By stripping this out, Nvidia has been able to reach lower prices. We don’t think any gamer on a tight budget will mind, since there are still very few games that support it.
Even without ray tracing, the GeForce GTX 16-series has all the benefits of the new Turing architecture, including concurrent execution of integer and floating-point instructions, improved scheduling, a unified cache, and new shading algorithms. These GPUs should handily outperform their GeForce 10-series predecessors. For more about how the RTX and GTX versions of Turing differ, you can read our Asus ROG Strix GeForce GTX 1660 Ti OC 6GB Review.
The GeForce GTX 1650 GPU is codenamed TU117, and it features 896 cores called CUDA units (the chip physically has more, but these are selectively disabled to improve manufacturing yields). These are arranged into 14 clusters called Streaming Multiprocessors.
Nvidia has set the base and boost clock speeds of the TU117 at 1485MHz and 1665MHz respectively, and all GeForce GTX 1650 graphics cards will have 4GB of GDDR5 RAM on a 128-bit bus, for a 128GBps of memory bandwidth. The TDP is an impressively low 75W and this GPU can run off motherboard power alone with no need for a PCIe power connector.
The GeForce GTX 1660 uses the same TU116 GPU as the GeForce GT 1660 Ti, but with some of its resources disabled. There are 1,408 CUDA cores instead of 1,536 which means there are 22 SM clusters rather than 24. Interestingly, the at 1530MHz and 1785MHz base and boost clock speeds are very slightly higher than those of the GeForce GTX 1660 Ti, though it’s still obviously the less capable variant.
You get 6GB of GDDR5 RAM instead of GDDR6. That takes memory bandwidth down from 288GBps to 192GBps. This is still a whole lot better than the 3GB of RAM that this GPU’s predecessor, the GeForce GTX 1060, had at around the same price level. The TDP is 120W, which is still very low.
One little twist is that TU117 lacks the new Turing video encoder. which can accelerate HEVC, H.264, and H.265 video encoding. On the other hand, TU116 does have this feature. Game streamers will be the most affected by this.
Just like with the GeForce GTX 1660 Ti, there won’t be any Founders Edition graphics cards for either of these GPUs designed and sold by Nvidia itself, but thankfully there’s a robust ecosystem of partner companies with multiple models in the market since launch day. In fact, some are selling for slightly less than Nvidia’s recommended launch price.
Gigabyte GeForce GTX 1650 Windforce OC 4G features and specifications
Gigabyte offers four graphics card models with the GeForce GTX 1650 GPU and four with the GeForce GTX 1660. One in each series is a Mini-ITX card for small-form-factor PCs with a single-fan cooler and less aggressive factory overclocking. Of the other three in each line, there are dual- and triple-fan coolers, varying degrees of factory overclocking, and slightly different cosmetic touches. The two cards that we’re reviewing today fall in the middle of their respective product families.
Starting with our GeForce GTX 1650 Windforce OC 4G, this is a relatively compact two-slot card, though the cooler is a little taller and longer than the PCB itself. Given that this is a low-end GPU, it has modest cooling needs and the two 90mm fans are probably overkill. The heatsink itself is quite small. Gigabyte has designed its fan blades with a ridged surface that is supposed to improve airflow. The two fans spin in opposite directions, which Gigabyte says is more efficient because it reduces turbulence.
Interestingly, although the GeForce GTX 1650 GPU does not require power over and above the 75W that the motherboard itself can provide through the PCIe slot, this model still has a single 6-pin PCIe power connector. This is most likely for stability with the factory overclock. We tried running this card unplugged to see if the extra power was optional, but all we got was a warning on screen telling us our test bench would not boot without it.
This particular GeForce GTX 1650 graphics card has been factory overclocked to run at 1785MHz boost speed clock speed, but all other specifications are stock. Interestingly, Gigabyte has gone with three HDMI 2.0b ports and a single DisplayPort 1.4 for video output. Low-end cards usually have legacy DVI connectors for those upgrading older PCs. You get dustcaps for all the ports, which is always a nice touch.
GeForce GTX 1660 OC 6G features and specifications
Moving on, we also have the Gigabyte GeForce GTX 1660 OC 6GB with us for review. This model doesn’t use the Windforce name but the cooler shroud and fans look identical to the ones on its more affordable sibling. Both cards have very nearly the same dimensions, but on closer inspection the GeForce GTX 1660 OC 6GB is actually slightly shorter. Despite that, it has a much beefier block of aluminium under the shroud with two copper heatpipes winding around it. There’s also a black backplate which is most likely more for looks than rigidity.
We have the same two 90mm fans with ridged blades, and they also spin in opposite directions. There’s an 8-pin PCIe power connector on the top, and a more conventional arrangement of three DisplayPort 1.4 and one HDMI 2.0b video outputs on the rear, also with dust caps.
The OC in this model’s name tells us that there is some factory overclocking, and what we have is a boost speed of 1830MHz as opposed to the stock 1785MHz speed. Everything else is the same as Nvidia’s stock specifications.
One thing that’s missing completely on both of these graphics cards is RGB LEDs. We’re used to having at least illuminated fan blades or a brand logo, but there’s nothing of the sort here. Some buyers might actually like this, and we at least don’t mind at all.
GeForce GTX 1650 and GeForce GTX 1660 performance
Our test bench consists of an AMD Ryzen 7 2700X CPU, Gigabyte Aorus X470 Gaming 7 Wifi motherboard, 2x8GB of G.skill DDR4 RAM, a 1TB Samsung SSD 860 Evo boot drive, Corsair RM650 power supply, and Asus PB287Q 4K monitor. We used Windows 10 (1803) and Nvidia’s GeForce Game Ready Driver version 430.39 for both cards.
Starting with the GeForce GTX 1650, what we see is that performance across the board is almost exactly half of what we got with the GeForce RTX 2060. Considering that GeForce GTX 1650-based graphics cards cost much less than half of the GeForce RTX 2060’s recommended price, this is a strong showing. Gigabyte’s default factory overclocks help this GPU look good in our tests, and there isn’t a significant price premium over stock-clocked cards.
There isn’t a huge performance difference between the GeForce GTX 1660 Ti and the GeForce GTX 1660, which was perhaps to be expected. However, stepping down to the GeForce GTX 1650, the numbers take a much bigger hit. The scaling is consistent with how graphics cards based on these three GPUs are priced.
We ran through 3DMark’s DirectX 12 Time Spy and DirectX 11 Fire Strike test suites. It’s clear that the GeForce GTX 1650 is not suited for the higher-end versions of these tests, which render at 4K. As expected, we get reasonable results at the lower end of the scale. Its beefier sibling, the GeForce GTX 1660, held up much better.