Friday , February 28 2020
Home / india / NVIDIA GeForce GTX 1660 Super review. EVGA SC Ultra: Recalibration of the mainstream market

NVIDIA GeForce GTX 1660 Super review. EVGA SC Ultra: Recalibration of the mainstream market

After a break after the launch of the Super GeForce RTX 20 cards from the NVIDIA series, as well as the new Radeon RX 5000 cards based on the AMD RDNA architecture, the graphics card market is again rising in time for the most important holiday shopping of the season. To fill the mainstream market, AMD is already sending its Radeon RX 5500 series cards to OEMs – soon also for retail sales – while NVIDIA is doing its own machinations on the market.

NVIDIA announces the first GeForce GTX 1660 Super. This is a relatively small but interesting version of the GTX 1660 family that adds SKU 1660 (vanilla) with faster GDDR6 memory to improve performance. Together with the GeForce GTX 1650 Super (delivery at the end of November), these two cards will form the basis of NVIDIA's mainstream efforts to close the year. And while other NVIDIA GTX 1660 cards are not going anywhere, as we will see today, at a price of USD 229, equipped with GDDR6 GTX 1660 Super will make the remaining 1660 cards superfluous.

The fourth Super card that will be introduced by NVIDIA, the GTX 1660 Super is slightly more difficult to read than the high-end cards that appeared last summer. In terms of performance, the card is just shy of the GTX 1660 Ti and priced north of the GTX 1660 (vanilla). But unlike other Super Cards, this premiere is not the same type of transparent price draw as the previous premiere; This time, NVIDIA is not giving up any cards, and store shelves will be common to all three GTX 1660 cards in the coming months. Nevertheless, this latest card performs at least the same basic role that serves as a kind of kicker in the middle generation – even if the original GTX 1660 only appeared 7 months ago.

NVIDIA GeForce GTX 1660 Series
CardConfigurationPrice (10/29)
GTX 1660 TiFull TU116
with GDDR6 12 Gb / s
$ 279
GTX 1660 Super (new)Cutting TU116
with GDDR6 14 Gb / s
$ 229
GTX 1660Cutting TU116
with GDDR5 8 Gb / s
219 USD

The ambiguous factor here is, as is often the case, AMD. The company has already announced part of its RX 5500 family of products. However, it is unclear what to expect from the cards in terms of performance, especially since AMD frames them against the lower GTX 1650 NVIDIA. So, although at first glance, the GTX 1660 Super seems to be NVIDIA, unlike the upcoming AMD launch, outside of AMD and NVIDIA testing labs, it remains to be seen where this will end.

One thing is clear, however, that it could be a de facto price reduction of NVIDIA's GeForce GTX 1660 Ti. Released for $ 50 less than the fastest NVIDIA GTX 1660 card, the GTX 1660 Super is only slightly slower than the GTX 1660 Ti, providing about 96% of its performance – barely beyond our regular version. When NVIDIA began to reduce the GTX 1660 Ti to create the original GTX 1660, one of the biggest changes was the replacement of GDDR6 with the cheaper GDDR5. Now that they've successfully undone it – and then some – most of the performance that NVIDIA has shaved for the GTX 1660 has returned in a cheaper Super card. As a result, although the GTX 1660 Super does not offer anything in terms of performance that we have not seen before, it lowers this level of Turing performance to a lower price, and NVIDIA hopes that something more tempting for GTX Series 10 and 9 owners, who were discouraged by relatively high prices latest cards.

Comparison of NVIDIA GeForce specifications
GTX 1660 TiGTX 1660 SuperGTX 1660GTX 1060 6 GB
CUDA cores1536140814081280
Core clock1500 MHz1530 MHz1530 MHz1506 MHz
Increasing the clock1770 MHz1785 MHz1785 MHz1708 MHz
Memory clockGDDR6 12 Gb / sGDDR6 14 Gb / sGDDR5 8 Gb / s8 Gb / s GDDR5 (X)
Memory bus width192-bit192-bit192-bit192-bit
Single Precision Perf.5.5 TFLOPS5 TFLOPS5 TFLOPS4.4 TFLOP
TGP120 W.125 W.120 W.120 W.
(284 mm2)
(284 mm2)
(284 mm2)
(200 mm2)
The number of transistors6,6B6,6B6,6B4.4b
Production processTSMC 12nm "FFN"TSMC 12nm "FFN"TSMC 12nm "FFN"TSMC 16nm
Start date02/22/201910/29/201903/14/201907/19/2016
Run price$ 279$ 229219 USDMSRP: $ 249
FE: $ 299

What exactly is GeForce GTX 1660 Super? In short, take the GTX 1660 (vanilla), replace the GDDR5 8 Gb / s GDDR6 14 Gb / s, raise the TDP a few watts to meet the increased power demand, and you have the GTX 1660 Super.

On the GPU front nothing has changed compared to the GTX 1660. It's the same GPU TU116 in the same partially turned off configuration with the same clock speeds. To tell you the truth, I'm not even sure if the NVIDIA GPU binning operations have changed for this product, or if they only had partners who clapped TU116-300 graphics cards on the circuit board connected to GDDR6. Either way, the net result is that the performance on paper has not changed since the GTX 1660, because the GPU itself can process as many pixels, textures and data elements as before.

Instead, memory support has changed and that is where the vast majority of the GTX 1660 Super performance increases. While the original GTX 1660 shipped with GDDR5 8 Gb / s as a means to reduce costs and distinguish / segment the card from the full-fledged GTX 1660 Ti, the GTX 1660 Super receives all the memory bandwidth it can handle. NVIDIA gave him 6 GB GDDR6 14 Gb / s, the fastest memory type available today. This gives the GTX 1660 Super 336 GB / s memory bandwidth, 75% more bandwidth than the original GTX 1660 or as much as the RTX 2060. And such a large bandwidth is almost completely responsible for the 12% performance increase that we see above the original card.

In fact, the GTX 1660 Super has more memory bandwidth than the higher class GTX 1660 Ti. This card comes with a slightly slower (and cheaper) GDDR6 12 Gb / s, which gives the GTX 1660 a super bandwidth advantage, which further destroys the small advantage that GTX 1660 Ti holds. At the current stage, the TU116 is almost overloaded by GDDR6, which is why the GTX 1660 Super does not overtake the GTX 1660 Ti, despite its advantage; at this level of performance, the additional memory bandwidth is not as important as the additional two SMs that the GTX 1660 Ti receives.

Finally, as a member of the GTX 1660, the latest card receives a very similar, but not exactly identical TGP with a power of 125 W. This is a small increase in power by 5 W compared to the GTX 1660 and 1660 Ti (as well as GTX 1060/960 families) and it seems it is here mainly to meet the greater energy demand resulting from shipping with fast GDDR6 memory – otherwise the GTX 1660 Super will lose some clock speed compared to the original GTX 1660. This is also a small change that suppliers boards should not have a problem with meeting the increased power demand in their existing GTX 1660 Ti designs, which means that the GTX 1660 Super cards can easily be built from existing parts. The only drawback in this case is that the higher TDP and lower performance compared to the GTX 1660 Ti means that the GTX 1660 Super is not as energy efficient as the card, which it otherwise makes redundant.

The disordered state of the NVIDIA stack

If NVIDIA seems to have released many GeForce graphics cards over the past year, that's because it is. Within 13 months of the first launch of the Turing architecture and GeForce RTX 20 cards, NVIDIA launched 12 different GeForce SKUs. And even if we throw away RTX 2080 and RTX 2070, which have been formally discontinued, but are still sitting on some store shelves, the current stack of NVIDIA products still has 10 depth cards.

NVIDIA GeForce 20/16 product series (Turing)
RTX 20 seriesGTX 16 series
RTX 2080 TiGTX 1660 Ti
RTX 2080 SuperGTX 1660 Super
RTX 2080 (deprecated)GTX 1660
RTX 2070 SuperGTX 1650 Super
RTX 2070 (deprecated)GTX 1650
RTX 2060 Super
RTX 2060

For comparison, the stack of GeForce GTX 10 series products was only 8 depth cards at the sloppy point, and this was after launching the GTX 1070 Ti about 16 months later.

To be fair to NVIDIA, some of them stem from the fact that the GeForce stack is more expensive than before, especially with the 1100 RTX 2080 Ti. Nevertheless, NVIDIA released more cards at a faster pace than at any point in the last decade. In particular, Super cards have grown significantly because NVIDIA offers refreshment in the middle generation, even within a year of launch. Compared to the GTX 9 and GTX 10 series, in which NVIDIA had clear two-year product cycles and introduced very few fillers between them, the RTX 20 series and GTX 16 series basically brought us back to the annual premieres of cards with an alias for two years GPU cycles .

Ultimately, this means that a very large number of NVIDA graphics cards are available on the market, with NVIDIA offering 5 SKUs under $ 300. Include card providers who overclock their factory premium cards, and we check GeForce at any price. As a sales strategy, market saturation is definitely a viable way to maximize sales – making sure there is always a card at a price someone is willing to spend – but it does mess up the NVIDIA stack, especially now with three different GTX 1660 cards.

Product positioning and competition

Changing gear, let's talk about the equipment behind today's premiere. Since the new GTX 1660 Super cards are the first generation mainstream cards, there are no reference cards here. Instead, it is a pure virtual premiere, and all the cards available on the market are partner projects.

At this point, partners have been sending cards based on TU116 for over 6 months, so they could improve their family GTX 1660 models. This means that they hit the ground using their existing GTX 1660 Super cards. At the moment I haven't seen any new projects especially for the Super card and I don't expect any. These cards will therefore look and work similar to the GTX 1660 cards already available on the market.

It is no wonder that the GTX 1660 Super takes the same general place in the NVIDIA stacks as the easier GTX 1660 cards. Which means that for both constructors and system updaters, NVIDIA puts the card in front of the same audience as before. The fact that the GTX 1660 Super offers better performance at this price is definitely an improvement here and is likely to ensure NVIDIA overtakes upcoming AMD cards.

Moderators, however, are likely to be more demanding; Turing is not a generation ahead of performance in terms of performance GTX 10 based on Pascal, and the GTX 1660 Super does not change that. So, although it is about 35-40% faster than the GTX 1060 6 GB, the GTX 1660 Super will be much more tempting for owners of the GTX 960 and the like, where the new card provides more than double the performance.

Meanwhile, as I mentioned at the beginning of this article, while NVIDIA is releasing a new GTX 1660 Super today, they are not going to withdraw any other cards. So on GTX 1660 Super paper it will be set very narrowly between GTX 1660 Ti and GTX 1660 (vanilla).

At current prices of the GT9 1660 Super 229 USD is 50 USD cheaper than the GTX 1660 Ti. And since the card ends only on the best NVIDIA 1660 card by about 4%, I would argue that the GTX 1660 Ti was released at this point; The GTX 1660 Ti is not big enough to provide significant value. Instead, the next real step after the GTX 1660 Super is the GTX 2060, which is 22% faster but costs around $ 100.

Otherwise, the GTX 1660 vanilla, from the other side, is still regularly available for around $ 219. With a price difference of just $ 10 between the GTX 1660 Super, the Super card is now the obvious choice, because 12% higher performance is more than worth $ 10. Despite all this, NVIDIA insists that they and their partners continue to sell vanilla GTX 1660. It will be interesting to see what NVIDIA is actually doing here, both in terms of card delivery and prices. GTX 1660 must lower the price to have a future; and if it did, it could be a good spoiler under $ 200.

It's also worth noting that NVIDIA doesn't offer any packages or other promotional offers for any of the GTX 16 series cards, so you can see what you get with the new cards.

Finally, the competitive situation for the GTX 1660 Super is mixed. AMD is still doing what it can, and in this case puts Radeon RX 590 on its closest competitor GTX 1660 Super. It is possible that the late cycle Polaris card was originally created to beat the GTX 1060 6 GB in terms of performance (while throwing energy through the window), which it certainly did. However, this means that the GTX 1660 Super is ahead of it by about 20% (and much less power), which is why AMD uses the card instead to place its claim at $ 199. For buyers who absolutely cannot exceed USD 200, this is an option; but at the moment the GTX 1660 Super does not face any serious competition.

Source link