A notice: We are now also allowed to show the first custom designs for the RTX 4070 Super. This is what happens in the following video, have fun with it
In the past few weeks, the rumors surrounding Nvidia’s GeForce refresh for 2024 have become increasingly numerous and precise, a clear indication of an imminent release. That’s how it happens: Nvidia today officially presents the Geforce RTX 4080 Super, the Geforce RTX 4070 Ti Super and the Geforce RTX 4070 Super. PCGH listened to a press briefing in advance and summarized all the information about the eagerly awaited RTX 4000 refresh. Compared to our previous video, we now know all the relevant parameters and can therefore provide you with a refined performance forecast – along with the good news that prices are falling, not rising.
Here you will find external content from [PLATTFORM]. To protect your personal data, external integrations will only be displayed if you confirm this by clicking on “Load all external content”:
I agree to external content being displayed to me. This means that personal data is transmitted to third-party platforms. Read more about our privacy policy.
Geforce RTX 4000: Specs & Benchmarks
Geforce RTX 4000 Super: Everything is great here
With the Geforce RTX 4080 Super, RTX 4070 Ti Super and RTX 4070 Super, Nvidia is renewing its graphics card portfolio for the still young year of 2024. As you can tell from the names, the wheel is not being reinvented – it is a classic refresh . While an increase in performance was clear from the start, Nvidia surprises by implementing bold rumors: The Super Trio not only brings more performance, but also a price reduction. Anyone who has waited until now to buy a new graphics card will receive more performance from the upcoming Nvidia products, both in absolute terms and per euro. Let’s start with the spec overview:
Nvidia Geforce… | RTX 4080 Super | RTX 4080 | RTX 4070 Ti Super | RTX 4070 Ti | RTX 4070 Super | RTX 4070 |
---|---|---|---|---|---|---|
Market launch | January 31, 2024 | November 16, 2022 | 01/24/2024 | January 3, 2023 | January 17, 2024 | April 12, 2023 |
architecture | Ada Lovelace | Ada Lovelace | Ada Lovelace | Ada Lovelace | Ada Lovelace | Ada Lovelace |
GPU codename/configuration | AD103-400* | AD103-300 | AD103-275 | AD104-400* | AD104-350 | AD104-250 |
Chip size (Die) | 378.6mm² | 378.6mm² | 378.6mm² | 294.5mm² | 294.5mm² | 294.5mm² |
Transistors graphics chip (billion) | 45.9 | 45.9 | 45.9 | 35.8 | 35.8 | 35.8 |
Manufacturing process (Foundry) | 4N (TSMC) | 4N (TSMC) | 4N (TSMC) | 4N (TSMC) | 4N (TSMC) | 4N (TSMC) |
Shader clusters (SMs) | 80 | 76 | 66 | 60 | 56 | 46 |
FP32 ALUs/TMUs/ROPs | 10,240/320/112 | 9,728/304/112 | 8,448/264/88 | 7,680/240/80 | 7,168/224/80 | 5,888/184/64 |
Ray tracing units | 80 (3rd gen.) | 76 (3rd gen.) | 66 (3rd gen.) | 60 (3rd gen.) | 56 (3rd gen.) | 46 (3rd gen.) |
Matrix units | 320 (4th gen.) | 304 (4th gen.) | 264 (4th gen.) | 240 (4th gen.) | 224 (4th gen.) | 184 (4th gen.) |
Level 2 cache (MiB) | 64 | 64 | 48 | 48 | 36 | 36 |
GPU clock speed in games (MHz) | 2,700** | 2,780 | 2,700** | 2,740 | 2,700** | 2,760 |
FP32 Power ALUs (TFLOPS) | 55.3 | 54.1 | 45.6 | 42.1 | 38.7 | 32.5 |
Fill rate (Mtex/Mpix per sec.) | 864.0/302.4 | 845.1/311.4 | 712.8/237.6 | 657.6/219.2 | 604.8/216.0 | 507.8/176.6 |
Frame generation (DLSS 3) | Yes | Yes | Yes | Yes | Yes | Yes |
Memory connection (bit) | 256 | 256 | 256 | 192 | 192 | 192 |
Speed RAM (GTs/MHz) | 23.0/11,500 | 22.4/11,200 | 21.0/10,500 | 21.0/10,500 | 21.0/10,500 | 21.0/10,500 |
Storage type | GDDR6X | GDDR6X | GDDR6X | GDDR6X | GDDR6X | GDDR6X |
Memory transfer rate (GB/s) | 736 | 717 | 672 | 504 | 504 | 504 |
Storage capacity (MiB) | 16,384 | 16,384 | 16,384 | 12,288 | 12,288 | 12,288 |
PCI Express standard | 4.0×16 | 4.0×16 | 4.0×16 | 4.0×16 | 4.0×16 | 4.0×16 |
PCI Express power connectors | 1× 16-/3× 8-pin | 1× 16-/3× 8-pin | 1× 16-/2× 8-pin | 1× 16-/2× 8-pin | 1× 16-/2× 8-pin | 1× 16-/2× 8-pin |
Max. Input | 320 watts | 320 watts | 285 watts | 285 watts | 220 watts | 200 watts |
Display connectivity | DP 1.4a, HDMI 2.1a | DP 1.4a, HDMI 2.1a | DP 1.4a, HDMI 2.1a | DP 1.4a, HDMI 2.1a | DP 1.4a, HDMI 2.1a | DP 1.4a, HDMI 2.1a |
MSRP (‘US MSRP’) in dollars | 999 | 1,199 | 799 | 799 | 599 | 549 (2024) |
Launch RRP in euros | 1,109 | 1,469 | 889 | 899 | 659 | 659 |
Performance information based on the graphics cards we tested extensively (average GPU boost across all benchmarks) – the manufacturer’s information is sometimes significantly lower and therefore unrealistic. Depending on the model, deviations of up to 200 MHz are possible. *Full expansion of the GPU **Assumed, realistic average clock speed of a reference card without overclocking (real frequencies are determined in the PCGH test).
First up is the Geforce RTX 4070 Super, which goes on sale from January 17th. The new gaming graphics card takes the old price point of the Geforce RTX 4070 (599 USD), but delivers noticeably higher performance thanks to additional computing units. In fact, it is the super model with the greatest expected increase in performance compared to the respective non-super variant. Thanks to now 7,168 shader calculation units, the RTX 4070 Super (or simply RTX 4070S) is closer to the RTX 4070 Ti than to the RTX 4070. This results in further changes: The Geforce RTX 4070 Ti, which was launched just a year ago, is disappearing from the market and the Geforce RTX 4070 sees a price cut to $549 (-$50). 499 USD would have been more appropriate given the competition from AMD, but the direction is right. The Geforce RTX 4070 Super will be available from January 17th both as an Nvidia Founders Edition and in the form of custom designs from board partners.
Geforce RTX 4000 Super: Custom designs
It continues on January 24th with the Geforce RTX 4070 Ti Super, also known as the “real RTX 4070 Ti”. According to PCGH’s assessment, this model has what it takes to be a box office hit because, in contrast to the other 70s models, the RTX 4070 Ti Super (or RTX 4070 TiS) does not have a predetermined breaking point in the memory. For a decent increase in performance compared to the RTX 4070 Ti, Nvidia cannot avoid installing the larger AD103 processor instead of the AD104, so that the RTX 4070 TiS has a 256-bit interface to the memory and 16 GiByte capacity. The result is an “RTX 4080 Light”, which will cost as much as the outgoing RTX 4070 Ti (799 USD). The Geforce RTX 4070 Ti Super will be available from January 24th exclusively in the form of custom designs from the board partners, it will be a Founders Edition not give.
The Geforce RTX 4080 Super will be the preliminary conclusion of the RTX 40 portfolio on January 31st. This model replaces the Geforce RTX 4080 on the market and will take its place as the second fastest horse in the Nvidia stable in the future. In the press briefing, Nvidia was refreshingly honest about the performance of the Geforce RTX 4080 Super (or RTX 4080S): it should be two to three percent, at best five percent, faster than the RTX 4080. This matches the now complete specifications around 10,240 shader units and 16 GiByte GDDR6X with 23 gigatransfers per second. The RTX 4080S is the fastest model of the super trio, but also the one with the least increase in performance. However, Nvidia skillfully spices things up: The Geforce RTX 4080 Super starts at USD 999 and is therefore USD 200 cheaper than the original (and generally perceived as overpriced) RTX 4080. The Geforce RTX 4080 Super will be available from January 31st as both an Nvidia Founders Edition as well as custom designs from board partners.
Geforce RTX 4000 Super: performance forecast
PCGH will of course test the Geforce RTX 4080 Super, RTX 4070 Ti Super and RTX 4070 Super as soon as the time comes. Until then, we would like to provide you with a refined performance forecast. Compared to our first analysis before Christmas, all performance-relevant data and official Nvidia statements are now available, so we can estimate performance relatively accurately.
Although we assume that the following forecasts come within a few percent of the real products, they are only assumptions, not measurements!
RTX 4000 SUPER @ PCGH Performance Chart 2024
Geforce RTX 4000 Super: Performance Estimation/Estimation
|
~2.78 GHz, 22.4 GT/s, 16 GiB
0
10
20
30
40
50
60
70
80
90
100
Surname
Rasterizing Index 2024 (Relative Performance)
Ray tracing index 2024 (relative performance)
*Purely an estimate based on the official specification, no measurement (yet)!
The Geforce RTX 4070 Super makes the biggest jump; we expect a solid increase of between 15 and 17 percent compared to the RTX 4070. This means that the smallest super version would easily get past AMD’s Radeon RX 7800 XT and could instead compete with the Radeon RX 7900 GRE. The Geforce RTX 4070 Ti Super is a weight class above, it competes with the Radeon RX 7900 XT and has a good chance of almost always coming out on top in this duel. This applies despite the reduction of the level 2 cache to 48 MiByte (like the original RTX 4070 Ti), which slightly increases the gap to the RTX 4080 and RTX 4080 Super (64 MiB L2$ each). We originally placed the Geforce RTX 4080 Super ahead of the Radeon RX 7900 XTX, but the official information suggests a narrow defeat – at least when it comes to rasterizing. When it comes to ray tracing, the RTX 4080S will undoubtedly be extremely powerful and will take the position as the second fastest graphics card. You can read how the story ends in the following PCGH tests of the Geforce RTX 4080S, RTX 4070 TiS and RTX 4070S. Until then: What do you think of Nvidia’s super refresh? Join the discussion!