After graphics card prices exploded last year due to crypto mining and supply shortages, they slowly fell again in 2022. But as Nvidia boss Jensen Huang announced in September 2022, graphics cards should remain expensive in the future.
Nvidia reduced the official price recommendation of the Founders Edition of RTX 4080 and RTX 4090 by 5 percent in December 2022.
For Nvidia’s new flagship GPU, the RTX 4090, according to the comparison portal ideals, interested parties still have to pay at least 2,000 euros, for an RTX 4080 around 1,400 euros are due (as of December 22, 2022).
However, the cards are Nvidia’s top models. If you want to spend less money on a new pixel pusher, you can look forward to the RTX 4070 (Ti), RTX 4060 (Ti), and RTX 4050. COMPUTER BILD has taken a closer look at the rumor mill and ventures a look ahead to next year.
RTX 4070 Ti: An RTX 4080 at heart?
If you don’t want to put 1,400 euros on the table for an RTX 4080 but still want to pack a powerful model into your computer, you should take a closer look at the RTX 4070 Ti.
The card is said to be a renamed RTX 4080 with 12 GB (gigabytes) of memory. However, the performance difference to the RTX 4080 with 16 GB was too big, which is why the 12 GB version was canceled shortly before the market launch.
Originally, Nvidia wanted to launch two versions of the RTX 4080, one with 12 GB and one with 16 GB of memory. However, the chipmaker backtracked and announced that the 12GB RTX 4080 was a top card, it just got the wrong name.
Just like the high-end models RTX 4080 and RTX 4090, the RTX 4070 Ti could also be manufactured based on the “Ada Lovelace” architecture: Manufactured in the 4-nanometer process at TSMC, an AD104 chip with 7,680 Cuda computing units to use.
That’s about 21 percent less than the RTX 4080 (9,728 shader units), but the manufacturer raises the boost clock from 2.5 GHz (gigahertz) to 2.6 GHz and the base clock from 2.2 GHz to 2, 3 GHz on.
The graphics memory of the fast GDDR6X type is expected to be 12 GB and connected to the graphics processor with 192 bits (192-bit interface). The power consumption (TDP) is apparently 285 watts. If the RTX 4070 Ti is the new edition of the scrapped RTX 4080 12 GB, an RRP of around 1,100 euros could be expected.
RTX 4070: Still upper or already middle class?
The RTX 4070 without Ti is also manufactured at TSMC using the 4-nanometer process. According to the online technology portal TechPowerUp, Nvidia uses an AD104 chip, but in a trimmed version.
The RTX 4070 apparently gets by with 5,888 Cuda cores, but the base and boost clock frequencies remain at the RTX 4070 Ti level of 2.3 GHz and 2.6 GHz respectively. As with the big sister, the manufacturer will probably rely on the new GDDR6X memory and give the card 12 GB of memory.
Here, too, the connection is made via a 192-bit memory interface. The TDP should also be 285 watts. With a nominal computing power of 30.74 TFLOPs (teraflops), the RTX 4070 would be about 23 percent slower than the RTX 4070 Ti, but a bit faster than an RTX 3080.
It is still unclear what price Nvidia will call for the release. Nvidia is expected to launch both the RTX 4070 Ti and the RTX 4070 on January 5 at CES (Consumer Electronics Show, January 5-8, 2023) in Las Vegas.
RTX 4060 Ti: No GDDR6X memory
The RTX 4060 Ti hasn’t even been announced yet, but the TechPowerUp site is already sharing the first details. Nvidia’s upper mid-range graphics card appears to be equipped with an AD106 chip and 4,352 Cuda cores.
While the chip’s base clock is 2.1 GHz, up to 2.4 MHz is possible with a boost. The manufacturer is said to have cut back on the memory and only used conventional GDDR6 memory with 8 GB.
Data exchange between the graphics chip and memory apparently takes place over 128 lines (128-bit interface). In addition, the RTX 4060 Ti apparently also gets the new 12-pin connector despite the lower power consumption. In the recent past, this had caused adapter cables to burn through on some RTX 4090.
Good news for users with less space in the PC: According to Twitter leaker kopite7kimi, the reference board of the RTX 4060 Ti is relatively short. Compared to the almost monstrous RTX 4090, there should be fewer installation difficulties. The power consumption of the RTX 4060 Ti is probably 220 watts, the price is not yet known.
RTX 4060: An RTX 3070 for the price of an RTX 3060 Ti
The rumor mill is also bubbling about the RTX 4060. Lenovo manager Wolfstame had already made the first concrete assumptions in November 2022 on the Chinese platform Weibo.
Thanks to a performance increase of 20 percent (compared to n with an RTX 3060), the RTX 4060 presumably achieves the performance of an RTX 3070 but does not exceed the price of an RTX 3060 Ti.
According to TechPowerUp, AMD is again using an AD106 chip but reducing the cores to 3,840. As in the Ti version, 8 GB of GDDR6 memory should be at the start, connected via 128 lines.
The base clock of the RTX 4060 is said to be a strong 2.5 GHz and the boost clock at 2.6 GHz, with a power consumption of 200 watts. In terms of price, it could cost 440 US dollars (around 415 euros). According to Wolfstame, the release of the RTX 4060 is expected for June 2023.
RTX 4050: The End of the Road
It is likely that the RTX 4050 will be the entry-level model in the “Ada Lovelace” series. As a direct successor to the RTX 3050, it could be of particular interest for Full HD gaming.
As the online technology portal, TechPowerUp wants to know, Nvidia relies on an AD106 chip with 3,840 Cuda cores, as in the more powerful RTX 4060 Ti. In addition, the card is said to be equipped with 8 GB of GDDR6 memory and connected to the graphics processor with 128 bits.
The RTX 4050 also apparently clocks at 2.5 GHz and reaches 2.6 GHz in the boost. As the online mail order company Notebooksbilliger.de writes in its blog, a performance increase of 55 percent compared to the previous model can be expected. The power consumption could be 150 watts, the price is still unclear.
Nvidia Blackwell: successor to Lovelace and Hopper?
While Nvidia’s RTX 3000 graphics chips were still codenamed “Ampere”, the current RTX 4000 series is codenamed “Ada Lovelace”. Nvidia seems to have grown fond of the habit of naming graphics chips after well-known scientists.
As the site VideoCardz.com claims to have learned, the codename of the next generation has already been leaked. The next generation of Geforce cards will probably be given the name “Blackwell” – named after the American mathematician David Harold Blackwell.
It is unclear whether the code name stands for the RTX 5000 series and is intended to serve the gaming and creative market. Or whether Blackwell is intended to be the successor to “Hopper” – Nvidia’s graphics chip architecture for data centers. If the company sticks to its previous release plan, the Blackwell chips can be expected to be released in 2024.
AMD: Performance increase of up to 70 percent
Competition stimulates business. AMD and Nvidia have been courting customers for a good 20 years. With its RDNA 3 graphics card generation, officially presented on November 3, 2022, Team Rot wants to heat up the competition.
Also in the luggage: the RX 7900 XTX and the RX 7900 XT, which, in addition to gigantic graphics memory and relatively low prices, above all want to score with a better ratio of performance per watt.
Thanks to the RDNA 3 architecture, there should be a performance boost of 54 percent compared to the previous generation, with 4K resolution there is even talk of 70 percent more performance. What else is already known and what the rumor mill has served up so far, you can find out in the following paragraphs.
RX 7900 XTX: competition for the RTX 4090?
AMD’s new flagship graphics card is based on the new Navi 31 chip and is manufactured by TSMC using the 5 and 6-nanometer processes.
The 7900 XTX has 96 compute units, a base clock of 1.9 GHz, and a turbo clock of 2.5 GHz. AMD has given its top model a generous 24 GB of GDDR6 memory, which connects to the graphics chip with 384 bits.
As a competitor product for the 7900 XTX, AMD puts Nvidia’s RTX 4080 into the field – the RTX 4090 is obviously not to be taken on, despite the identical graphics memory size and connection.
The power consumption, which AMD specifies in TBP (Total Board Power), should be 355 watts. While the price for the release on December 13, 2022, was still 1,200 euros, you now have to pay at least 1,600 euros for the RX 7900 XTX (as of December 22, 2022).
RX 7900 XT: The second-best horse in the stable
There is also a Navi 31 chip in AMD’s second-strongest model. 5,376 stream processors work inside, which the manufacturer divides into 84 compute units.
The graphics chip clocks at 1.5 GHz and reaches 2.4 GHz in boost mode. The graphics memory of the RX 7900 XT is an impressive 20 GB – AMD also uses the GDDR6 type here.
The data exchange between graphics memory and graphics processor takes place via 320 lines (320-bit memory interface). With a TBP of 300 watts, consumption is below that of both its big sister and the RTX 4080. The 7900 XT has been available since December 13, 2022, for currently around 1,060 euros (as of December 22, 2022).
RX 7800 (XT): Upper middle class
If at the end of November 2022, the rumor mill still assumed that there would be no RX 7800 XT (only an RX 7800) due to a leaked code snippet, the online technology po has it
rental TechPowerUp obviously has superior knowledge.
AMD’s graphics card, which has not yet been released, is said to be based on a Navi 32 chip and equipped with 3,840 stream processors divided into 60 compute units. The base clock is said to be 1.8 GHz, with a boost at 2.5 GHz.
On the memory side, AMD is probably giving its RX 7800 XT 12 GB of the GDDR6 type, data exchange between memory and processor takes place via a 192-bit interface.
According to TechPowerUp, the TBP is 300 watts, like the 7900 XT. It remains to be seen whether AMD will decide to market the Navi 32 chip with 60 processing units as the RX 7800 XT or as the RX 7800.
RX 7700 XT: competition for the RX 6900 XT?
The RX 7700 XT is rumored to be similar in speed to the previous generation’s top model, the RX 6900 XT. That would be remarkable since the RX 7700 XT is supposed to be the new mid-range model.
According to the technology YouTuber Moore’s Law is Dead, AMD manufactures the RX 7700 XT from a Navi 33 chip, which gives 2,048 stream processors on 32 compute units.
The card’s base clock is apparently 1.9 GHz and 2.5 GHz in boost mode. With its 8 GB GDDR6 graphics memory, connected via a 128-bit interface, the RX 7700 XT should reach its limits in memory-intensive applications.
For comparison: The RX 6900 XT has 16 GB of video memory on board and thus more resources than the RX 7700 XT. According to the YouTuber, this manufacturer-side performance limitation justifies the expensive high-end cards that can deliver better 4K performance.
The power consumption of the RX 7700 XT is apparently 200 watts, the list price is 399 to 499 US dollars (376 to 470 euros).
RX 7600 XT: The entry-level class
Just like the RX 7700 XT, AMD will probably also give its entry-level graphics card RX 7600 XT a Navi 33 chip.
Compared to the RX 7700 XT, AMD will use the knife again and further reduce the number of stream processors and computing units. Unfortunately, specifications are not yet available at the time of publication of this article on December 22, 2022.
Conclusion
Whether 2023 will be the year of Nvidia or AMD is completely open. The graphics cards of both manufacturers have their strengths and weaknesses, and both are now capable of ray tracing.
While AMD will probably lose out against Nvidia’s RTX 4090, AMD’s top models should work more efficiently thanks to the RDNA 3 architecture. In the past, Nvidia’s mid-range GPUs were slightly better.
It remains to be seen whether this also applies to the 4000 series. In the budget range, experience has shown that AMD often offers cards with a better price-performance ratio.
