AMD’s and Nvidia’s latest GPU generational leaps compared in native benchmarks, with one obvious winner
Table of Contents
Now that both Nvidia and AMD have released the majority of their new-generation graphics cards into the world, you can genuinely see the performance gains from third-party testing that these provide over the previous generation. That’s what German publication ComputerBase has decided to do in its latest tests.
With some help from Google Translate in Chrome, we looked over these results, and the conclusion is rather one-sided when it comes to the native performance improvements between generations. It’s clear that AMD’s cards see the biggest boost with the RDNA 4 cards offering significant intergenerational improvements over RDNA 3, while Nvidia’s RTX 50 series is more incremental.
Prime Day may have closed its doors, but that hasn't stopped great deals from landing on the web's biggest online retailer. Here are all the best last chance savings from this year's Prime event.
- Sapphire Pulse AMD Radeon™ RX 9070 XT Was $779 Now $719
- AMD Ryzen 7 7800X3D Processor Was $449 Now $341
- Skytech King 95 Ryzen 7 9800X3D gaming PC Was $2,899 Now $2,599
- LG 77-Inch Class OLED C5 TV Was $3,696 Now $2,996
- AOC Laptop Computer 16GB RAM 512GB SSD Was $360.99 Now $306.84
- Lexar 2TB NM1090 w/HeatSink SSD Was $281.97 Now $214.98
- Apple Watch Series 10 GPS+ Smartwatch Was $499.99 Now $379.99
- AMD Ryzen 9 5950X processor Was $3199.99 Now $279.99
- Garmin vívoactive 5 Smartwatch Was $299.99 Now $190
*Prices and savings subject to change. Click through to get the current prices.
RDNA 4 vs RDNA 3 & Blackwell vs Lovelace
For AMD, ComputerBase compared the RX 9060 XT versus the RX 7600 XT at 1440p in three areas. The first is pure rasterization, where RDNA 4 improves by 20% over RDNA 3, as it averages 60.3fps compared to 50.3fps. Next up was the enabling of ray tracing, which RDNA 4 improved upon, and it shows, as it is 31% better than RDNA 3, with 49.6 vs. 37.9 fps. Lastly, the punishing path tracing is tough on both, as the new gen is only at 23.4 fps compared to 12.8 fps, which is 83% better.
As for Nvidia, the RTX 5070 Ti is compared to the RTX 4070 Ti Super at 4K. There, the Blackwell card averages just one percent, or 0.4fps, higher than its predecessor. That trend follows in ray tracing as the improvement is just 0.3 fps and a 1% boost. Even path tracing is only 0.2fps higher in Blackwell, which suggests that Nvidia’s focus is on other technologies beyond pure performance, with the use of DLSS 4 and its frame generation being a key factor and major selling point of the new Blackwell-era cards.
After all this, ComputerBase concludes that Nvidia’s Blackwell isn’t faster than Lovelace by default. Instead, it wins out in some games but generally utilizes more execution units, higher clock speeds, power, and features instead. In contrast, AMD offers a big leap in generations, alongside utilizing higher speeds, too, to gain a massive improvement in native performance.
RDNA 4 vs RDNA 3
RDNA 4 vs RDNA 3 1440p raster performance, Source ComputerBase
RDNA 4 vs RDNA 3 1440p pathtracing performance, Source ComputerBase
RDNA 4 vs RDNA 3 1440p raytracing performance, Source ComputerBase
Blackwell vs Lovelace 4k raytracing performance, Source ComputerBase
Blackwell vs Lovelace
Blackwell vs Lovelace 4k pathtracing performance, Source ComputerBase
Blackwell vs Lovelace 4k raster performance, Source ComputerBase
Blackwell vs Lovelace 4k raytracing performance, Source ComputerBase
The setup and limitations
ComputerBase attempts to compare intergenerationally by using similarly specced-out graphics cards and seeing how they perform across the same games on the same platform. For example, the RTX 5070 Ti and RTX 4070 Ti Super are the two closest options available. However, since the 5070 Ti has 512 more CUDA cores than the 4070 Ti Super (8,960 versus 8,448), this has been compensated for by running the last-gen card at a higher clock speed.
The power limits of the two were maximized, and fortunately, they use the same L2 cache. The 3D accelerators aren’t comparable to the VRAM of the 50 series, which uses the GDDR7, which has a faster total bandwidth. Although the impact on gaming performance is not significant, the overclocked 4070 Ti Super memory reaches 784GB/s compared to 864GB/s of the 5070 Ti.
As for the AMD GPUs, both the RX 9060 XT and RX 7600 XT are configured identically with cores, VRAM, and some cache. But the L2 and clock speeds have increased. Even at the lowest clock rate, the 9060 XT drops to 2,860 MHz and doesn’t quite reach the 2,800 MHz achieved by the 7600 XT.
However, this is accounted for in the benchmarks, and the memory doesn’t require much adjustment or consideration. Since the two are within 1GHz of each other, they still use the same memory setup.
Sapphire 11350-04-20G Pulse AMD Radeon™ RX 9060 XT Gaming OC Graphics Card
ASRock Ultimate Gaming Bundle
XFX Swift AMD Radeon RX 9060 XT OC
MSI Gaming RTX 5070 12G Ventus 3X OC Graphics Card
ASRock Steel Legend PC Build Bundle
About the Author