Important preliminary note
We’re doing it the same way again today with the benchmarks as we did in the previous launch review. This is important because in the sum of all games, the peculiarities of the respective architectures quickly blur. In the end, there are only 10 specially selected games, but I chose these from over 20 titles and the pre-release tests with several cards as examples because the result was almost exactly the same in the end. The weighting between the titles with pure raytracing without DXR and with DXR was done in a ratio of 6:4, with the four DXR titles coming out very differently.
Full ray tracing fun in Cyberpunk 2077, combined with more medium-weight effects like in Metro Exodus EE and the hybrid implementation of lighting to Marvel’s Guardians of the Galaxy and Shadow of the Tomb Raider, where ray tracing really only comes into play humanely. DXR is being implemented in more and more games, and the current engines almost all allow for it by now. From this point of view, it would be just as unfair to completely forego it as it would be to exclusively use such titles with DXR. Since every user has different preferences and some prefer to do without DXR completely (why actually?), I’ll accommodate all target groups a bit for once.
Sum of all games
To play reasonably performant games, even the 200 watts are easily enough, because the Palit GeForce RTX 4070 Dual 12GB even reaches a GeForce RTX 3080 10GB, of course depending on the game and resolution. That is the good side. However, the up to 15 watts more of an overclockable card would certainly still be money well spent, as the energy could be converted into more something performance in some situations, such as DLSS, when the Tensor Cores are also fully utilized. On the other hand, the advantage turns out to be much smaller in terms of pure raster performance. However, the effects of the higher power consumption are also more noticeable in the Min FPS than in the pure Average. But you won’t necessarily need it.
I normalized the FPS and percentiles and formed a geometric mean (Geomean), because this is simply more accurate statistically and is also handled this way in the industry. We compare the average OC cards with around 215 to 220 watts of power target to the non-overclocked card. This fully covers all “MSRP cards” and the so-called OC cards. I already wrote it on the first page: all cards are the same and because of the small tolerance range, two benchmark runs are enough to cover everything.
Full HD (1920 x 1080 pixels)
WQHD (2560 x 1440 pixels)
Ultra-HD (3840 x 1440 pixels)
Ultra-HD and DLSS/FSR (3840 x 1440 pixels)
Interim summary
The Palit GeForce RTX 4070 Dual 12GB is an excellent card in Full-HD when it comes to highest frame rates and also well suited for WQHD. In Ultra-HD at the latest, however, you’ll have to think about smart upscaling and that’s where DLSS comes into play. Meanwhile, games like “The Last of Us Part 1” (TLOU) subjectively look even better in Ultra HD with DLSS than native Ultra HD. NVIDIA can definitely use its advantages here, which DLSS 2.x also offers purely optically. However, if a game supports DLSS 3.0 and you would be stuck in the unplayable FPS range without Super Sampling, then this can even be the lifeline to playability. You can’t improve latencies with it, but not every genre is as latency-bound as various shooters. For TLOU, I would have desperately liked to see DLSS 3.0, but you can’t have everything.
- 1 - Introduction, technical data and technology
- 2 - Test system in igor'sLAB MIFCOM-PC
- 3 - Teardown: PCB, components and cooler
- 4 - Gaming performance
- 5 - Power consumption and load balancing
- 6 - Transients and PSU recommendation
- 7 - Temperatures, clock rates and thermal imagin
- 8 - Fan curves and noise
- 9 - Summary and conclusion
20 Antworten
Kommentar
Lade neue Kommentare
Mitglied
Mitglied
Neuling
Veteran
Urgestein
Neuling
Mitglied
1
1
1
Urgestein
Urgestein
Mitglied
Mitglied
Mitglied
Veteran
Alle Kommentare lesen unter igor´sLAB Community →