Sehatnews – In the test, the Nvidia GeForce RTX 4080 Founders Edition turns out to be a very fast graphics card that is perfect for UHD gaming. With ray tracing enabled, you may need to use DLSS, which is available with Ada Lovelace v3 and delivers a massive FPS boost in select games thanks to DLSS frame generation.
Another positive aspect is the energy efficiency during gaming. However, the gap to the RTX 4090 is huge and the price is not only set too high compared to the current flagship, but also to the predecessor RTX 3080.
Technical data in a quick check: The AD103 graphics processor of the RTX 4080
A quick look at the technical data already reveals two things: First, that the jump in performance from the RTX 3080 to the RTX 4080 will not be nearly as big as from the RTX 3090 to the RTX 4090, and second, that the RTX 4080 will be noticeably slower than the 4090 is.
Specifically, the RTX 4080 does not even rely on the same graphics chip as the RTX 4090, but on the smaller AD103 with 45.9 billion transistors and a chip area of 379 mm 2 . When fully configured, the GPU consists of six graphics processing clusters (GPC) with 40 texture processing clusters (TPC) and 80 streaming multiprocessors (SMs).
However, the AD103 of the RTX 4080 is trimmed by four streaming multiprocessors, which ultimately results in 9,728 Cuda cores, 304 fourth-generation Tensor cores and 76 third-generation RT cores. There are also 112 render output units (ROPs). That means the RTX 4080 has just under 60 percent of the processing units of the RTX 4090.
The GPU has 16 GB of GDDR6X video memory at its disposal, operating at 22.4 Gbps and connected to a 256-bit memory interface. This results in a memory bandwidth of 717 GB/s, which is even less on paper than the RTX 3080. To avoid the problem of the memory bottleneck, Nvidia has increased the L2 cache from 5 MB to 64 MB.
The base clock of the RTX 4080 is 2,205 MHz and the boost clock is 2,505 MHz, both of which are only marginally less than the RTX 4090. Nvidia specifies the total graphics power (TGP) of the graphics card as 320 watts, which is exactly the value which corresponds to the RTX 3080.
Since the dimensions of the heat sink of the RTX 4080 with 304 x 137 x 61 millimeters correspond to those of the 4090, quiet operation and low temperatures are possible. Power is supplied via a 16-pin connector, with a corresponding adapter to 3x 8-pin included. In the following table we have summarized the most important technical specifications of the “Ada Lovelace” graphics processors compared to the previous generation “Ampere” for you:
RTX 4080 lands between 3090 Ti and 4090 in gaming performance
Nvidia itself states that the RTX 4080 should be up to twice as fast compared to the RTX 3080 Ti , and even up to three times faster by using DLSS 3. We couldn’t measure such a big jump in performance in our test in any of the games in the image display via rasterization. Nevertheless, the increase in performance, especially compared to the RTX 3090 Ti, is greater than might have been expected based on the technical specifications.
On average over 13 games in 2160p resolution, the RTX 4080 is 49 percent ahead of the RTX 3080 (130 vs 86 FPS) at the end of the day , which corresponds to a fairly normal generational leap. As already indicated above, the increase in performance is significantly lower than from the RTX 3090 to the RTX 4090, which averages 75 percent, but that’s to be expected, the RTX 4080 lacks too many computing units for that.
Both AMD’s current flagship Radeon RX 6950 XT (93 FPS) and Nvidia’s RTX 3080 Ti (94 FPS) are beaten by almost 38 percent. Even the former top model, the RTX 3090 Ti (102 FPS), can outperform the RTX 4080 by around 26 percent. This places the RTX 4080 in the middle between the RTX 3090 Ti and the RTX 4090, because the latter performs 32 percent better. The RTX 2080 Super (52 FPS), so to speak the predecessor, is slower by a factor of 2.4.
In our test, the RTX 4080 is therefore able to play all games smoothly with the highest game details in UHD resolution, although in Cyberpunk 2077 (59 FPS) it is just enough for a spot landing at the magical 60 FPS limit. In this game, however, the 4080 tends to be one of the worst performers, so the advantage over the 3080 is only 41 percent and over the 3080 Ti only 34 percent, both values are well below the average. A very similar picture can be seen in the two games Horizon Zero Dawn (116 FPS) and Metro Exodus (102 FPS), whereby the gap to the RTX 4090 is also slightly below average in all three games, these seem to be the “Ada Lovelace” -Architecture not so good to lie in general.
The absolute prime example of this is Doom Eternal (273 FPS), in which the 4080 can extend its lead over the 3080 (166 FPS) to 65 percent and over the 4080 to 51 percent. However, we are miles away from doubling the performance, as Nvidia states. A game that Ada also likes is Assassin’s Creed Valhalla (110 FPS), although the P1 values for the RTX 4080 are relatively poor here, the small memory interface and the even compared to the 3080 (57 FPS ) to make lower memory bandwidth somewhat noticeable.
A few interesting facts are revealed in QHD resolution. Already during the tests of the RTX 3090and RTX 3090 Ti, Ampere has shown itself to have trouble bringing the processing units to the monitor at lower resolutions. As a result, the RX 6950 XT (163 FPS), which in UHD still competes with the RTX 3080 Ti (157 FPS), is in QHD between the RTX 3090 (162 FPS) and the RTX 3090 Ti (170 FPS) can push. This same fact now also has an impact on the performance ratio to the RTX 4080 (222 FPS), because interestingly it is now 29 percent ahead of the RTX 3090 Ti, which has even increased the lead somewhat, which is extremely unusual. But when you look at the number of computing units, that seems plausible again. However, the distance to the RTX 3080 has not changed at all, it remains at a lead of 49 percent. The RTX 3080 Ti is again beaten by 40 percent, here, too, there is a slight plus. The lead over the RX 6950 XT is smaller at just 33 percent.
The tests also show that the RTX 4080 in 1440p resolution is significantly less affected by the CPU limit than the RTX 4090. We were only able to determine a limitation in two of the 13 games tested: Cyberpunk 2077 (110 FPS) and Horizon Zero Dawn (199 FPS), which may be the reason why the FPS values here are below expectations even in UHD resolution. Ultimately, this means that the 4090 is no longer 32 percent ahead of the 4080, but only 14 percent, so the gap has more than halved. So the 4080 cuts a fine figure in WQHD – although it is actually designed for UHD resolution.
The RTX 4080 (263 FPS) is definitely oversized for Full HD gaming, but we still provide the measurement results, of course. In seven of the 13 games tested, and thus a little more than half, the RTX 4080 runs into the CPU limit in 1080p and can therefore not develop its full potential. It’s even worse for the 4090 (277 FPS), which is just four percent ahead of the 4080. Nevertheless, the 4080 is still a good 18 percent ahead of the third-placed, the RX 6950 XT (216 FPS). The RTX 3080 Ti (202 FPS) is still beaten by 27 percent and the RTX 3080 (192 FPS) by 34 percent. Still, we wouldn’t recommend the 4080 for Full HD resolution: it’s like locking a wild animal in a cage.
Ray tracing is the main discipline of the RTX 4080 – thanks to DLSS 3
We continue with the tests with activated ray tracing. While we called the RTX 4090 the first graphics card with activated ray tracing that is also suitable for UHD gaming, the RTX 4080 has to struggle a lot more here. The missing computing units compared to the flagship are definitely noticeable, but the performance is still convincing. Over seven games in UHD resolution, the RTX 4080 (69 FPS) is again 25 percent ahead of the RTX 3090 Ti (54 FPS), so absolutely nothing has changed compared to the image display via rasterization. The lead over the RTX 3080 Ti (48 FPS) is marginally larger at an average of 40 percent, and the gap to the RTX 3080 (44 FPS) has grown even more, which is now 54 percent. AMD’s still fastest GPU, the RX 6950 XT (39 FPS), is beaten by a whopping 89 percent, although a successor is already in the starting blocks. The RTX 4090, however, can use its computing units a little better with ray tracing and is now 40 percent ahead of the 4080.
It can be observed that the RTX 4080 can make up ground on the RTX 3090 Ti, especially in games in which relatively few or simple ray tracing effects are used. The best example of this is Shadow of the Tomb Raider, which relies exclusively on raytraced shadows. Here, for example, the 4080 (87 FPS) is 52 percent ahead of its direct predecessor (53 FPS) and 34 percent ahead of the RTX 3090 Ti (65 FPS). In Cyberpunk 2077 (28 FPS) and Control (55 FPS), both flagship titles for ray tracing due to countless effects, the 4080 then performs well below average and is only 16 and 18 percent respectively ahead of the RTX 3090 Ti (24 and 47 FPS). Several points are noticeable here: the comparatively small number of computing units, the smaller video memory compared to the 90s cards and above all the only 256-bit wide memory interface. In such demanding titles, the 4080 then no longer makes it to 60 frames per second, here you either have to reduce the graphics settings a little or activate DLSS, we will come to that later.
The phenomenon, which has already revealed itself in the image display via rasterization in WQHD, has become even stronger through the activation of ray tracing. The 4080 (121FPS) in 1440p can extend its lead over the 3090 Ti (91FPS) to 31 percent. However, the distance to the RTX 3080 Ti (83FPS) and RTX 3080 (76FPS) does not change, it remains at 40 and 54 percent respectively. On the other hand, the gap to the 4090 (158FPS) has shrunk a bit, which only performs 31 percent better. This means that the RTX 4080 is able to enable smooth frame rates with the highest graphics settings in almost all games. Only in Cyberpunk 2077 (58 FPS) does the AD103 GPU just barely miss 60 frames per second.
In a 1080p resolution with activated ray tracing, we hit the CPU limit in just one game, which is why the 4080 (159 FPS) can still develop its performance relatively well. As a result, it’s still 26 percent ahead of the RTX 3090 Ti (125 FPS), 36 percent ahead of the RTX 3080 Ti (116 FPS), and 50 percent ahead of the RTX 3080 (104 FPS). With the RTX 4090 (186 FPS), on the other hand, the advantage is halved from 31 percent in QHD to just 15 percent in Full HD, which is mainly due to the CPU limitation.
DLSS 3 including frame generation for the ultimate FPS boost
If you want even more FPS, you can activate DLSS and, for example, in Cyberpunk 2077 even in the new Version 3 including DLSS Frame Generation. DLSS 3 analyzes the current frame as well as the previous frame to determine how the scene has changed. The Optical Flow Accelerator provides the neural network with all information about the change in direction and speed of the pixels in the image comparison. This neural network then generates the intermediate images. So DLSS Frame Generation can generate completely new images without taxing the graphics pipeline. This means that both GPU and CPU limited games will benefit from DLSS 3. A technical explanation video for DLSS 3 is available here. We looked at DLSS 3 in the three games F1 2022, Cyberpunk 2077 and A Plague Tale Requiem and the performance gain is significant.
DLSS 3 can significantly increase the FPS values even in Quality Mode, i.e. the highest quality level. In F1 2022, activating DLSS including frame generation doubles the FPS. In performance mode, we can even measure an increase by a factor of 2.6. Of course, we also determined how big the additional boost from frame generation is. In concrete terms, this is 16 percent for the average FPS in F1 2022. The P1 value is much more significant here, because we can measure an increase of 45 percent here. Unfortunately, this also increases the latency by 47 percent despite the activation of Nvidia Reflex, although at 25 milliseconds it is still very low and significantly better than with the native resolution.
Cyberpunk 2077 benefits even more from DLSS 3 and enables absolutely smooth game rates even in quality mode. Specifically, the FPS increase compared to the native resolution by a factor of 2.8 in quality mode and by a factor of 3.8 in performance mode. The activation of frame generation also plays a much larger role here, because it not only increases the average FPS by 46 percent, but also the P1 value by 59 percent. As expected, the latency unfortunately increases by 42 percent, but is again below the value of the native resolution.
After all, the last game A Plague Tale Requiem delivers very similar values to F1 2022. Such large increases in performance as in Cyberpunk 2077 are more likely to be the exception, although this ultimately always depends on the implementation of the game manufacturer. In quality mode, the FPS can be doubled compared to the native resolution, in performance mode we can measure an increase by a factor of 2.4. The performance gain through frame generation is 34 percent for the average FPS and 32 percent for the P1 value. However, the latency has to be listed negatively here, because despite the Nvidia Reflex, it is marginally worse than in the native resolution due to the use of frame generation.
According to Nvidia, the following ten games support DLSS 3 when the RTX 4080 is released on November 16:
- A Plague Tale: Requiem
- Bright Memory: Infinite
- Destroy All Humans! 2 – Reprobed
- F.I.S.T.: Forged in Shadow Torch
- F1 ® 22
- Marvel’s Spider-Man Remastered
- Microsoft Flight Simulator
- SUPER PEOPLE
Fluctuating performance of the RTX 4080 in the application area
With the RTX 4080 you can of course not only gamble, but also work creatively. However, you have to accept a few significant trade-offs compared to the RTX 4090. The most obvious is the video storage that has shrunk from 24GB to 16GB. For very large video or render projects, this can sometimes be an exclusion criterion. In addition, there is the memory interface, which is only 256-bit wide, which means that the memory bandwidth is even lower at 717 GB/s than with the RTX 3080 with 760 GB/s, despite the higher clocked video memory.
In Blender, this does not seem to play a major role in the selected project, here the RTX 4080 is 89 percent ahead of the RTX 3080 and even 50 percent ahead of the RTX 3090 Ti. The lead of the RTX 4090 is comparatively manageable at 26 percent. In the Octane Bench or in V-Ray, however, the difference quickly increases to 40 to 50 percent. The missing computing units, the smaller video memory and the trimmed memory interface quickly become noticeable here. This can also be seen in the Indigo benchmark or in LuxMark. In both tests, the RTX 4080 is only a single-digit percentage behind the RTX 3090 Ti and well behind the RTX 4090.
Top: RTX 4080 shines with low power consumption, high energy efficiency & cool GPU temperature
Nvidia specifies the total graphics power (TGP) of the RTX 4080 as 320 watts, which is just as high as that of the RTX 3080. The power is supplied via the new 16-pin power connector, with a corresponding adapter to 3x 8-pin included, to be compatible with older power supplies. However, our test of the RTX 4090 showed that it only rarely reaches its power limit of 450 watts when gaming, and the power consumption is usually significantly lower. The same picture emerges in the tests of the RTX 4080. In none of the 20 gaming tests – regardless of whether with ray tracing or without – can we measure a higher power consumption than 305 watts.
On average, the power consumption is “only” 283 watts in UHD, 261 watts in QHD and 221 watts in Full HD. The 4080 needs almost exactly 100 watts less than the 4090, a good 40 watts less than the 3080 and almost 70 watts less than the 3080 Ti.
In terms of efficiency, represented by the Watts/FPS ratio, this means that the RTX 4080 can secure the top position just ahead of the RTX 4090. The positive impression of the Ada Lovelace architecture also continues with the smaller AD103 GPU. Compared to both the RTX 3080 and the RTX 3080 Ti, the efficiency has improved by a factor of 1.7. AMD specifies an improvement in efficiency of 54 percent for its RDNA 3 architecture compared to its predecessor, so it could be a very close race here.
We can only determine higher consumption rates when the GPU is fully loaded with the FurMark stress test for 30 minutes. Specifically, 316 watts are recorded in the Power Capture Analysis Tool (PCAT) after half an hour. The maximum value in the 10 millisecond range was 343 watts. This makes it clear that Nvidia has the voltage peaks under control with Ada Lovelace much better than with Ampere.
Since the dimensions of the GPU cooler of the RTX 4080 at 304 x 137 x 61 millimeters are identical to the cooler of the RTX 4090, it was already to be expected that the temperatures would be very good with a 130 watt lower TGP. This is also evident in the worst-case scenario, namely a stress test using FurMark for half an hour. The chip temperature rose to a maximum of 64 degrees Celsius and the hot spot temperature to up to 75 degrees Celsius, putting the 4080 at the top of the charts. The fans of the graphics card turned at a maximum of 1380 RPM, which means that the GPU is virtually inaudible from the closed housing.
Due to the low temperatures and the fact that the RTX 4080 almost never reaches its power limit when gaming, the GPU can work almost continuously at the clock limit. According to our measurements, this is at 2790 MHz. The gaming average is 2737 MHz in 2160p, 2759 MHz in 1440p and 2774 MHz in 1080p. Compared to the RTX 3080, the clock speeds have increased by almost 900 MHz, which corresponds to an increase of 48 percent. These values reveal that there is still some potential for manual optimization, especially for undervolting .
RTX 4080: Price classification and availability
The Nvidia GeForce RTX 4080 Founders Edition is said to be available from November 16 at 3:00 p.m. for 1,469 euros on Nvidia’s official website . The board partners should all also offer at least one model for the specified RRP, with the prices of the factory overclocked models increasing significantly in some cases. We were given the prices of the three Zotac models in advance. The Trinity is said to start at 1,489 euros, the Trinity OC is said to cost 1,559 euros and the AMP Extreme 1,629 euros. Ultimately, however, the retailer’s prices should be a little higher, because even if the RTX 4090 is officially priced at 1,949 euros, there is currently no model for less than 2,199 euros in the price comparison .
But even the official MSRP is outrageously high for several reasons: First, let’s compare the price of the RTX 4080 within the “Ada Lovelace” portfolio. According to the RRP, the RTX 4090 should cost almost 33 percent more than the RTX 4080. But in UHD resolution, it also delivers a good 32 percent more performance in image display via rasterization and 40 percent more performance with activated ray tracing. In addition, there is the VRAM, which is eight gigabytes larger at 24 GB, from which creative users in particular benefit.
So the top model offers a slightly better price-performance ratio than the second-fastest Ada GPU to date. This is extremely unusual, because normally a disproportionately high surcharge is due for the top model, but it also makes it clear that the RTX 4080 is far too expensive.
Nvidia itself compares the RTX 4080 to the RTX 3080 Ti, which is a nice move since both GPUs have an official MSRP of $1,199. At this point, however, it’s important to remember that the RTX 3080 Ti came out in the middle of the great chip shortage, when graphics cards were sometimes costing three to four times their MSRP. The RTX 3080, on the other hand, was priced before the crisis and has an official MSRP of $699. Accordingly, the RTX 3080 Ti officially costs 72 percent more, for just ten percent more performance and two gigabytes more video memory. For this reason, we rather use the RTX 3080 for comparison:
For around 50 percent more performance in UHD resolution and six gigabytes of additional video memory, Nvidia requires a surcharge of 72 percent. Of course, one must not forget the efficiency improved by a factor of 1.7 and the support of DLSS 3, as well as global inflation, but this price increase cannot be justified in any way.
This is also revealed when we look at the price-performance ratio for gaming, which we calculate from the RRP divided by the number of frames/s. For the RTX 4080 we have assumed the official RRP of 1,469 euros, although we strongly assume that the market price for the cheapest model is more likely to be in the direction of 1550 to 1600 euros. The RTX 3080 performs significantly better than the RTX 4080, even though the 3080 is still sold above the MSRP. AMD’s entire RX 6000 portfolio is also well ahead of Ada Lovelace. Nvidia even benefits from the fact that the RTX 3000 top models have increased in price again in recent weeks. A 3090 Ti was still available for 1199 euros a few weeks ago, now at least 1407 euros are due.
Test-Fazit: Nvidia Geforce RTX 4080 Founders Edition
The Nvidia GeForce RTX 4080 Founders Edition is a good graphics card, there’s no question about that, but it still stands in the shadow of the RTX 4090. While the top model can increase the image display by rasterization in UHD resolution by 75 percent compared to the predecessor and with activated Raytracing is even 87 percent ahead of the 3090, so the generational leap in the 4080 with a performance gain of 50 percent compared to the 3080 is significantly less spectacular. The increase in efficiency by a factor of 1.7 compared to Ampere and the ray tracing turbo DLSS 3 are particularly noteworthy – which is currently limited to a very manageable number of games and from our point of view should only play a minor role in the purchase decision.
Since the RTX 4080 also only has 16 GB of video memory to offer, it ultimately lacks the unique selling point, because for creative users the card is no longer as attractive as the 4090. The biggest sticking point – the “key issue” – is on End of Nvidia’s completely exaggerated asking price: For the performance and range of functions offered, the RTX 4080 with an RRP of 1,469 euros is simply far too expensive. We can only advise you that the release of the AMD Radeon RX 7900 XTX and 7900 XTstill to be seen on December 13th. If AMD’s performance specifications are correct, the two GPUs should definitely be able to compete with the RTX 4080 or even beat it – and that with an official RRP of 999 and 899 US dollars. The cards are not cheap either, but they are still significantly cheaper than the RTX 4080.