I deliberately waited a bit to not only press speculative data into a news article, but also to come up with some more detailed content. AMD and Nvidia of course continue their small-scale launch war as usual, with one leading and the other following. So good, so well known. But what further information do we have in the meantime? Of course, I can’t leak anything that manufacturer A or N may have given me in terms of information. The so-called NDAs help against this. Whereby this is not seen so narrowly in Asia rather.
So the colleagues of videocardz.com and wccftech.com have already leaked various snippets. And I am sitting here myself on a lot of own benchmarks for the cards that are already on the market, from the Radeon RX 480 to the GeForce RTX 2060 lead-free. And to say it with Nvidia’s new slogan: “And if the gap is still so small, one more card still fits between this all”. Together with what has already been leaked, the existing benchmarks and a simple calculator, the upcoming situation can be calculated quite easily.
Let’s start with AMD’s new Radeon RX 5500 XT. The chip is new and will come with 1408 shaders, which have already been reported by colleagues, in all probability a full configuration. In addition there is a 128-bit memory interface to which 4 to 8 GB of memory are supposed to be connected. With the typical 1.7 GHz clock rate already leaked, the performance can be calculated quite well. Here Nvidia has to do something against it, if they want to be further in front. Nvidia counters the RX 5500 XT with a GTX 1660 Super, almost cannibalizes their own Ti and waits relaxed with the old 2060 for a RX 5600 (XT).
The current GTX 1660 with the 6 GB GDDR5 is simply too slow for that, because the Radeon RX 5500 XT fits almost exactly between this card and the GTX 1660 Ti. The information from the board partners’ direction that Nvidia will position the new GTX 1660 Super exactly above it, i.e. just below the GTX 1660 Ti (TU116-400), but with the truncated chip of the GTX 1660 (TU116-300) is condensed. So instead of the 1536 CUDA cores there are only 1408, which is exactly the same number as the shaders of the RX 5500 XT. Which of course also offers a direct comparison.
To be really faster, one simply puts other, faster GDDR6 memories instead of slower GDDR5 on the otherwise identical board. If the GTX 1660 Ti is equipped with GDDR6 at 12 Gb/s, the GTX 1660 Super is supposed to be equipped with GDDR6 at 14 Gb/s. The boost clock of 1785 MHz would also be a bit higher than with the GTX 1660 Ti. First information should certify the GTX 1660 Super about 94 to 96% of the Ti performance, but at a much lower price point. However, this puts the Ti under pressure. By the way, talks with distributors and sales departments will in future see the Ti more in the SI business.
The gap between the RX 5700 XT and the upcoming RX 5500 XT is huge, leaving room for another card. In all likelihood, this will be the classic waste recycling of the Radeon RX 5700, although chips may still be collected at present. That’s nothing new and so you might have to wait a bit for the possible RX 5600. The following diagram shows the values of real measured cards and the new cards in a direct comparison:
Especially the position of the not even announced RX 5600 shows why Nvidia has not yet discontinued its RTX 2060 lead-free with the 6 GB memory. So the green fraction won’t even have to go crazy here, because the 2060 hasn’t been set to EOL, but lives happily on and thrives. Whereby now also the last one becomes clear, why. I still hear the sales sentence “The 2060 actually affects the sales of 2060 Super, in many countries customer only cares about the model number” in my ears, which corresponds to the sales figures of some distributors in Germany. Obviously, the Super hasn’t arrived at the last Media Markt corner yet, and you first have to look at the price anyway.
And what else? So first of all (almost) everything has been written that is known or may be written. At least I don’t write here against better knowledge. Whereby up to the half of the year 2020 Ampere should finally come (I had already written something about it from time to time). As a middle class offer, as it was said from various sources and then also with a smaller structure width. Because obviously Nvidia doesn’t dare to get close to the big guys right at the beginning, but practices with light sailors. You never know..