Full HD resolution (1920 x 1080 pixels) remains relevant in gaming, as it has lower hardware requirements and enables high refresh rates, which is particularly advantageous for competitive gaming. Despite the optimization of new graphics cards for higher resolutions, Full HD remains useful for some users. Combining the FPS values from 10 games and calculating an average value is usually used to provide a general assessment of a system’s gaming performance. The frame rates of the games are added together and divided by the number of titles. However, this average value is only of very limited significance as it does not take into account the different requirements and performance profiles of the individual games. Games often have considerable differences in their processor, graphics card and memory requirements, which leads to strong fluctuations in performance. A hardware-intensive game can push the average down, while some less demanding titles can raise the value unrealistically. But it is already a first indication.
So, of course, I have another, much more precise consideration. Normalizing FPS values and creating a percentage rating for each individual game is much more useful here than cumulative FPS values. Combining all the indices of the various games into a common average serves to present the performance of a hardware objectively and comparably. Such deviations can be recognized very quickly by a slightly different bar sequence, in which individual games can contribute less to distorting the overall impression. I use this metric both for the average FPS and for the P1 Low, i.e. the percentile with the slowest rendered frames.
Important: I would like to explicitly point out once again that all Blackwell cards correspond to the reference data and are not overclocked, while the RDNA4 cards have both already been provided with a lot of overclocking headroom ex works and, depending on the card and setting, come to the customer with at least 5 or a maximum of 10 percent performance increase. This also applies to the RX 7900XTX, as my MBA card now also has a defective vapor chamber and is therefore unusable. The Blackwell cards (all in the reference design) can also be increased by around 5 percent through simple overclocking and without extensive RAM OC. Depending on the cooling solution and RAM quality, even a little more.
Although this puts the results into perspective a little, the RX 9070 is around 10 percentage points ahead of an RTX 5070, so you might be inclined to see a tie here if you adjust the entire preset to the actual reference values. The performance index in the actual state puts the RX 9070XT on a par with the RX 7900XTX and the GeForce RTX 5070 Ti, although the note about factory overclocking also applies here. Without this boost, the card would be behind, albeit only slightly. But let’s copy the approximately 10 percentage points advantage of the RX 9070XT over the RX 9070 to the human clipboard, because that will be very interesting in a moment and also explains the title of today’s article.
The term P1 Low describes a value used in benchmarks that indicates the lowest 1% of the measured FPS. In contrast to the average value, it shows drops in performance that can disrupt the gaming experience. The same elaborate method of calculating the cumulative individual score of all games is used as I did for the performance index. However, the values for P1 and the general frame time progression, including the variants for RDNA4, are significantly better than for RDNA3 and even Blackwell. You have to be fair and I refer you to the individual metrics of the games further down in the text.
Now we look at our clipboard and find the 10 percentage points advantage of the RX 9070XT over the RX 9070. However, the RX 9070XT requires a whopping 31.5 percent more electrical energy to achieve this advantage! Compared to an RX 7900XTX, this is even 60.8 percent, which shows us how broken AMD’s crowbar was with RDA3 and what a waste of energy the AIB cards celebrated in order to keep up to some extent. At least that’s over now.
The measurement in Full HD enables a targeted analysis of the performance of the Ryzen 7 9800X3D, as the GPU is only slightly utilized in this resolution. This highlights potential CPU bottlenecks more clearly, which is particularly useful for assessing whether the CPU has a limiting effect in modern games. In addition, the following values for GPU and CPU power consumption provide information about the scenarios in which each component works under full load.
The efficiency is then the counter calculation, so to speak, where the amount of watts used (average over the respective game) and the average FPS of the game in question are compared. The lead is very clear, because the RX 9070 is now in the midfield of the Ada cards and the RX 9070XT is at least still significantly more efficient than all the RDNA3 cards tested. I’m also sure that a card like the Nitro can still be properly optimized, although time and AMD’s Adrenalin software are currently getting in the way. But there will certainly be follow-up tests.
Individual metrics and details
Of course, as always, there are also the individual metrics for all the games tested:
- 1 - Introduction and details - RX 9070 and RX 9070XT
- 2 - Test system and equipment
- 3 - Gaming: Full-HD 1920x1080 Pixels (Rasterization Only)
- 4 - Gaming: WQHD 2560x1440 Pixels (Rasterization Only)
- 5 - Gaming: Ultra-HD 3840x2160 Pixels (Rasterization Only)
- 6 - Gaming: WQHD 2560x1440 Pixels, Supersampling, RT & FG
- 7 - Gaming: Ultra-HD 3840x2160 Pixels, Supersampling, RT & FG
- 8 - Power consumption, transients and PSU recommendation
- 9 - Clock rate, temperatures and noise
- 10 - Summary and conclusion
303 Antworten
Kommentar
Lade neue Kommentare
Veteran
1
Veteran
Veteran
Mitglied
Urgestein
Urgestein
Urgestein
Urgestein
Neuling
Urgestein
Urgestein
Urgestein
Veteran
Urgestein
Urgestein
Veteran
Urgestein
Urgestein
Alle Kommentare lesen unter igor´sLAB Community →