Let us preface this section by saying that we did go into the manual overclocking testing phase with a bit of hesitation. On the one hand our experience with NVIDIA GTX 1600 series cards has been hit or miss when it comes to overclocking. For some (GTX 1650) it was simply a lack of additional power to feed the core. For others it was hard-coded in the BIOS limitations (per NIVIDA diktat) that handicap overclocking (GTX 1660). Yes, we were not expecting miracles. On the other hand… this card has been designed with overclocking in mind. So, while we had a bad case of ‘lowered expectations we were hoping for more than merely ‘good enough’.
Thankfully this card not only exceeded our highest hopes, it is a downright fun video card to play with. It really is a return to the ‘old days’ when GTX 10-series ruled the roost and before NIVIDA started to get serious with ‘big core’ desings. First and foremost, ASUS has set the maximum TDP limit to 20%… aka power slider can go up to 20 percent. While an extra 26 watts may not sound like much headroom. The truth of the matter is it makes an already very powerful card even better. Much, much better. Yes, it now sucks as much power(ish) as a GTX 1070 reference design, but in turn the ‘win some, lose some’ stock results do a one-eighty and even with fewer CUDA cores does exceed what a 150 watt TDP GTX 1070 can do.
Of course, 20 percent is still ‘only’ twenty percent, so do not expect it to walk on water, or turn water into wine. With heavy factory overclocked GTX 1070 models that can suck down 180 watts, the STRIX GTX 1660Ti O6G is still outmuscled, but it does make it a horse race to say the least.
As to specifics, we dialed in the TU116-400 core’s overclock first and yes it is impressive. Instead of an average boost of 1890 with peaks in the low 1900 range, this selfsame core pushed over the 2K range with an average in the low 2030MHz’s… with peaks routinely even higher. Even then the cards fans were not noticeable. More audible than at stock, but not loud by any stretch of the imagination.
Equally impressive is we turned 1500Mhz (12,000MHz effective) memory into 1,920Mhz / 15,360Mhz (effective) beasts. Which in turn only made the memory bus difference between it and the last generation GTX 1070 even wider. Specifically, instead of it being 256GB/s vs 288Gb/s it was now 256GB/s vs 368.6GB/s. At that width worries over 192bits vs 256bit wide memory bus are moot.
