- Joined
- Jul 24, 2024
- Messages
- 652 (1.97/day)
System Name | AM5_TimeKiller |
---|---|
Processor | AMD Ryzen 7 9800X3D |
Motherboard | ASUS ROG Strix B650E-F Gaming |
Cooling | Arctic Freezer II 420 rev.7 (with 6 fans in push-pull setup) |
Memory | G.Skill Trident Z5 Neo RGB, 2x16 GB DDR5, Hynix A-Die, 6400 MHz @ CL30-39-39-102-141 1T @ 1.40 V |
Video Card(s) | ASUS TUF Radeon RX 9070 XT GAMING |
Storage | Samsung 990 PRO 1 TB, Kingston KC3000 1 TB, Kingston KC3000 2 TB |
Case | Corsair 7000D Airflow |
Audio Device(s) | Creative Sound Blaster X-Fi Titanium |
Power Supply | Seasonic Prime TX-850 |
Mouse | Logitech wireless mouse for 15€, 6y old |
Keyboard | Logitech wireless keyboard, 12y old |
Also, our German Freunde have measured IPC differences between the two RDNA architectures.
While average gains in RT are wunderbar (48%), gains in standard raster load are gut (17%).
www.pcgameshardware.de
They've selected RX 9060 XT and RX 7600 XT, because these GPUs have same amount of compute units and both have 128bit/16GB VRAM.
Then they've set cores to same clocks. Memory bandwidth is also around the same for both cards (only 10 GB/s difference), so is the infinity cache (32 MB).
Comparison was made in 1080p.
(+17% perf. on average in rasterization, +49% perf. on average in RT, +130% perf. in PT based on result from CB2077 only)
Translation of final words:
That's to say, AMD has really put some efforts into improving RT performance. This was also proven by TPU reviews of RX 9070 (XT) and RX 9060 XT.
My personal opinion on RDNA4: 17% IPC improvement (in raster) is not bad, I don't care about RT/PT.
I'm happy with my RX 9070 XT's performance (had coil whiny RX 7800 XT before), but money-wise it hurts.
This generation of AMD and GPU graphics, AMD's performance improvement scales better with power than Nvidia's.
Example: RX 7900 XT has TDP of 300W and about 30% more compute units than RX 9070 XT with 304 W TDP.
RX 9070 XT has higher clocks by around 20%, but despite having significantly less units, manages to be 9% faster in raster than RX 7900 XT at the same power draw.
That's 44% more performance per compute unit for RX 9070 XT when compared to RX 7900 XT.
Similarly, I compared RTX 4090 with RTX 5090 and RTX 5080 with RTX 4080:
Compute units, TDP data are sourced from TPU's GPU specs database.
Performance numbers at 3840x2160 resolution are sourced from respective GPU reviews on TPU.
While average gains in RT are wunderbar (48%), gains in standard raster load are gut (17%).

Gigantische Zuwächse bei Radeon RX 9000: RDNA 4 vs. RDNA 3
Wie viel stärker ist RDNA 4 (Radeon RX 9000) gegenüber RDNA 3 (Radeon RX 7000)? PCGH macht den ersten Pro-Takt-Test mit 3 GHz.
They've selected RX 9060 XT and RX 7600 XT, because these GPUs have same amount of compute units and both have 128bit/16GB VRAM.
Then they've set cores to same clocks. Memory bandwidth is also around the same for both cards (only 10 GB/s difference), so is the infinity cache (32 MB).
Comparison was made in 1080p.
(+17% perf. on average in rasterization, +49% perf. on average in RT, +130% perf. in PT based on result from CB2077 only)
Translation of final words:
With the help of targeted benchmarks, it was demonstrated for the first time outside of AMD's labs that the RDNA 4 architecture has made significant progress per processing unit. The evidence in the context of ray tracing is particularly valuable. The data clearly shows that AMD's previous ray tracing weakness was not due to Nvidia-friendly code, but rather to the previously weak processing units. With traditional rasterization, the gains are smaller, sometimes less than ten percent, but still considerable in many cases. Considering the impressive scaling from Navi 44 (9060 XT) to Navi 48 (9070 XT), the lack of a high-end solution is particularly painful. A hypothetical Navi 41 with double the processing power would undoubtedly have the potential to compete with Nvidia's GB202, aka GeForce RTX 5090.
That's to say, AMD has really put some efforts into improving RT performance. This was also proven by TPU reviews of RX 9070 (XT) and RX 9060 XT.
My personal opinion on RDNA4: 17% IPC improvement (in raster) is not bad, I don't care about RT/PT.
I'm happy with my RX 9070 XT's performance (had coil whiny RX 7800 XT before), but money-wise it hurts.
This generation of AMD and GPU graphics, AMD's performance improvement scales better with power than Nvidia's.
Example: RX 7900 XT has TDP of 300W and about 30% more compute units than RX 9070 XT with 304 W TDP.
RX 9070 XT has higher clocks by around 20%, but despite having significantly less units, manages to be 9% faster in raster than RX 7900 XT at the same power draw.
That's 44% more performance per compute unit for RX 9070 XT when compared to RX 7900 XT.
Similarly, I compared RTX 4090 with RTX 5090 and RTX 5080 with RTX 4080:
Compute units, TDP data are sourced from TPU's GPU specs database.
Performance numbers at 3840x2160 resolution are sourced from respective GPU reviews on TPU.
Last edited: