GTX 1660 Super has arrived-this is really weird. Rather than making a comprehensive small-scale improvement to distinguish the Super card from the original model, Nvidia made a major upgrade to replace the original last-generation GDDR5 memory with a faster current GDDR6. Elsewhere, the card is basically the same in all respects as its predecessor, except that the nominal TDP has only increased by 5W. So, how big is the memory upgrade? Is it enough to justify the new graphics card?
First, let's take a look at what this generation of changes in memory technology really means-it's actually much bigger than you think. Moving to GDDR6 can greatly increase the speed of VRAM, increasing the data rate from 8Gbps to 14Gbps and increasing the memory bandwidth from 192GB / s to 336GB / s.
Nvidia claims a 75% increase in storage speed from 1660 to 1660 Super, making the new card 20% faster than the standard GTX 1660. In terms of context, this puts the GTX 1660 Super at almost the same level as the GTX 1660 Ti, with more CUDA cores (1536 and 1408), but slower memory (14Gps and 12Gbps) and lower TDP (120W and 125W ).
So, how much does the GTX 1660 Super cost? The card goes on sale today with a suggested retail price of $ 229 / £ 209 / € 245, making it only a bit more expensive than its original GTX 1660 (price of $ 219 / £ 199 / € 229). Interestingly, the vanilla GTX 1660 is still sold at a lower price, making the card competitive with RX 590 and others. At the same time, the GTX 1660 Ti is still the most expensive product, priced at $ 279 / £ 259 / € 299. (The GTX 1650 Super with 12Gbps GDDR6 is also planned to be released on November 22, but the price and performance of this model are not yet clear.)
GTX 1060 | GTX 1650 Super | GTX 1660 | GTX 1660 Super | GTX 1660 titanium | |
---|---|---|---|---|---|
GPU core | 1280 | 1280 | 1408 | 1408 | 1536 |
Virtual RAM | 6GB GDDR5 | 4GB GDDR6 | 6GB GDDR5 | 6GB GDDR6 | 6GB GDDR6 |
Memory bus | 192 bits | 128-bit | 192 bits | 192 bits | 192 bits |
bandwidth | 192GB / s | 192GB / s | 192GB / s | 336GB / s | 288GB / s |
Boost clock | 1708 MHz | 1725MHz | 1785 MHz | 1785 MHz | 1770MHz |
processor | GP106 | TU116 | TU116 | TU116 | TU116 |
The price point of $ 229 seems reasonable if Nvidia's performance requirements are accurate, so we've tested the card in a range of classic and modern games at 1080p, 1440p, and 4K to find out what you can expect . For reference, our test bench uses an overclocked Core i7-8700K CPU, which runs at 4.7GHz full-core turbo and is equipped with 16GB of 3400MHz RAM, but any of the latest Ryzen 5 or Core i5 processors (or more (Higher versions) are paired with dual channels of memory should be sufficient to reduce bottlenecks.
There is no Nvidia reference card available, so our example for this review is the Gigabyte Windforce OC card. This dual-slot design includes two fans and an eight-pin PCIe power connector. On the port side, we are considering standard Turing GTX loading, which includes three DisplayPort and one HDMI 2.0. No DVI-D. This makes for a neat card, but this means that users without an entry-level 1080p 144Hz display with DisplayPort must rely on an adapter.
The card's boost clock is listed at 1830MHz, which is 45MHz higher than the reference design. Our original GTX 1660 evaluation unit also ran at 1830MHz, so the results between the two cards should be comparable. The retail price of this Gigabyte GTX 1660 Super card is expected to be slightly higher than RRP, but almost all cards with sufficient cooling solutions should be able to achieve this level of overclocking and can increase the frame rate by up to a few percentage points. Having said that, let's start talking about it all: our benchmark results.
Nvidia GeForce RTX 1660 Super Analysis