Nvidia could beat AMD at its own game

Nvidia may learn from AMD for the next generation of graphics cards.

The next generation of graphics cards is approaching and with it more leaks about the technical details of Nvidia’s RTX 4000. The latest rumors probably stem from data hackers stole from Nvidia’s servers.

From this we learned the first specs of Nvidia’s latest generation yesterday. More rumors are now circulating on Twitter relating to the internal memory of the RTX 4000 cards.

It should therefore be up to 1,500 percent larger than before: the top model of the current generation has 6 MB of L2 cache, while the new generation should have up to 96 MB.

All new chips should get more cache

The Twitter user @ harukaze5719 summarized the numbers in a table and compared them with the previous generation.

In the upper half you can see the AD chips, which stands for Ada Lovelace, the next generation of Nvidia’s graphics cards. Below you can see the GA chips, which describe the current generation, i.e. the RTX 3000 cards.

The new chips from Nvidia are said to have this much memory. Source: Twitter / @harukaze5719

The smaller the number, the more powerful the chip, so the AD102 chip is likely to be used in Nvidia’s upcoming top-of-the-line models like the RTX 4090 and RTX 4080 Ti.

Available graphics card: RX 6500XT

Nvidia takes a leaf out of AMD’s design

Although the numbers seem exaggerated at first, the greatly increased memory comes as no surprise. Especially if we consider the latest leaks, according to which the RTX 4000 cards should have 71 percent more cores.

Since, according to the leaks, the bandwidth does not increase from generation to generation, but the number of cores does, the additional memory could be an important component in constantly feeding the cores with data so that they can work efficiently.

This is also not a new concept in the industry. AMD is already taking a similar approach to the current graphics card generation RX 6000. They call it the infinity cache and it is similar in size at up to 128 MB.

What does this mean for the new Nvidia cards? The Infinity cache, coupled with factors such as a very high clock rate, has allowed AMD to keep up with the current Geforce generation, in which Nvidia’s cards still have far less cache. If Nvidia now uses a similar technology, it could give the RTX 4000 cards a decent performance boost.

AMD counters with a multi-chip design

But that doesn’t mean that Nvidia will be ahead of the new generation of AMD in terms of performance, because they are also developing their cards further. AMD’s new generation, RX 7000, is said to work with a multi-chip design in which up to three chips per graphics card could be used. According to AMD, the leap in performance should come this year:

If AMD manages to ensure that the chips can communicate with as little delay as possible, the technology may unleash some computing power. However, it is not yet possible to say with certainty which manufacturer will be in the lead with the next generation of graphics cards.

However, it is foreseeable that the power consumption will increase again considerably. You can read all about it here:

What do you think of the recent increased cache rumors at Nvidia? Likely and unsurprisingly, also considering AMD’s Infinity Cache? Or one of the rumors that turn out to be false in the end? Feel free to write it in the comments!