AMD's Ryzen 9000 series launch is characterized by its promise of "the ultimate performance meets incredible efficiency." When we subjected the Ryzen 5 9600X and Ryzen 7 9700X to our testing process for the LTT video for this launch, Labs identified a noteworthy observation: the Ryzen 7 9700X appears to be better suited for a higher TDP classification. As highlighted in the video, once the PPT was set to unlimited, the 9700X readily consumed 140W of power while operating well within its TJMax constraints.

Definitions

Before delving into our observations, it is essential to understand a few key terms related to CPU performance: 

PPT: Package Power Tracking 

PPT quantifies the power delivered to the CPU socket on the motherboard, rather than the CPU itself. The PPT threshold defines the maximum permissible power consumption across the voltage rails supplying the socket. Applications with high thread counts or intensive threads may encounter PPT limitations, which can be mitigated by increasing the PPT limit.

TDP: Thermal Design Power 

TDP indicates the maximum heat a CPU can dissipate under standard operating conditions, expressed in watts. The TDP is influenced by the component's architecture and design, with higher-performance CPUs typically exhibiting higher TDP values, thereby dictating their cooling requirements and power consumption.

TJMax: Thermal Junction Maximum

TJMax represents the highest temperature a CPU can safely reach before triggering thermal protection mechanisms. The "junction" refers to the specific point on the CPU die where this temperature is measured, and manufacturers determine TJMax values to ensure safe operation.    

Overclocking

Overclocking entails running a CPU at speeds beyond its default clock settings. While this can enhance performance, it also introduces risks, including potential damage to the processor and a reduction in its lifespan. 

PBO: Precision Boost Overdrive

PBO allows the CPU to operate beyond its standard voltage thresholds, enabling higher performance by extending the power budget automatically. This feature serves as an automated alternative to manual overclocking, pushing the CPU to higher voltages and clock speeds within safe limits. 

Out of the box defaults on the Ryzen 7 9700X

AMD Eco Mode

AMD's Eco Mode allows users to configure the processor to operate at a lower TDP, such as reducing a 105W TDP CPU to 65W or a 170W TDP CPU to 105W. This adjustment is made possible through the Ryzen Master software and BIOS, which modify the socket current and power settings accordingly. The power savings achievable through Eco Mode are substantial. As the 9000 series evolves, Eco Mode is being integrated into motherboard BIOS options, although initial support may vary across different models.

The prior Ryzen 7000 series introduced the Curve Optimizer, a feature that allowed for PBO-aware undervolting by dynamically adjusting voltage across the curve. The new Curve Shaper builds upon this, enabling users to fine-tune voltage curves across 15 distinct frequency and temperature bands, thereby optimizing undervolting potential.

(Photo from AMD Ryzen 9000 Series Tech Day, July 9-10, 2024)

Labs Testing

AMD suggests PBO is still the best option for gaming and mixed-use cases. Manual overclocking can enhance performance in specific scenarios, such as all-core use cases with high uptime. AMD provided a general template for optimizing PBO overclocking, which we followed in our testing to determine the optimal wattage for the 9700X.

During our core tests for the LTT video, we found that the 9700X seemed constrained by the default 88W PPT limit, which is beneficial for keeping temperatures low. To test if this limit was holding back performance, we adjusted the PPT limit in the BIOS by setting PBO to advanced mode, and further set the PBO limits to manual to have full access to the PPT wattage.

With the PPT limit effectively set to unlimited (the default maximum of 810W is rarely, if ever, reached due to other constraints like temperature and current limits), the CPU drew an average of 149W, peaking at 155W during Cinebench 2024 benchmarks. This power consumption is comparable to the Ryzen 7 7700X, which drew an average of 144W with a peak of 146W under similar conditions.

Removing the wattage restriction yielded immediate performance gains. In Cinebench 2024, the multi-core score increased from 1187 to 1305—a nearly 10% improvement over the 7700X’s score of 1157. In Blender 4.2.0, rendering times dropped from 7 minutes 31 seconds to 6 minutes 38 seconds, a marked improvement over the 7700X’s 7 minutes 41 seconds. Similar performance gains were observed across tests like Handbrake, 7-Zip, Godot Compile, and PugetBench.

120W Testing

Our next objective was to identify the point of diminishing returns by testing the 9700X at a 120W PPT limit. Historically, this is where a base TDP of 90W would land under PPT boost load. Although there were rumors[1][2] that AMD might increase the base TDP to 120W, this did not come to fruition. Nevertheless, performance at 120W PPT was impressive, with a Cinebench multi-core score of 1274, compared to 1187 at the default PPT and 1305 at unlimited PPT.

In our F1 23 tests, the default 9700X maintained an average clock speed of 5478 MHz with a maximum clock speed of 5525 MHz. At the unlimited PPT setting, the CPU held steady at around 5512 MHz while drawing 92-95W, slightly above the 85-87W drawn at the default PPT. This suggests that a 120W PPT might be an optimal setting for gaming. We’ll get back to this hypothesis later.

In Cinebench, the efficiency, measured as score-per-watt, demonstrated expected diminishing returns as wattage increased: 13.49 at the default 88W PPT, 10.62 at 120W PPT, and 8.76 at unlimited PPT. This trend is characteristic of overclocking, where increased wattage often results in reduced efficiency.

Though of course Cinebench is only one piece of the whole testing puzzle, let’s take a look at the performance in some of the other tests.

7-Zip 

In 7-Zip, the 120W PPT closely matches the performance of the unlimited PPT, achieving 112.33 MB/s compression and 1611 MB/s decompression compared to 112.47 MB/s compression and 1619 MB/s decompression for the unlimited setting. The difference between the two is minimal, making it a tight race in this benchmark. However, the 120W PPT still delivers about a 3% improvement over the stock 88W PPT. 

Blender 4.2.0 

In Blender, the 120W PPT is approximately 14 seconds slower than the unlimited PPT - roughly a 4% decrease in speed. Despite this, the 120W setting offers an 8% improvement over the stock 88W PPT, which recorded a time of 7 minutes and 31 seconds.

Handbrake 

Handbrake tests showed an average encoding speed of 103.62 FPS in x264 and 91.70 FPS in AV1 with the 120W PPT, which is about 3% slower than the unlimited PPT. However, when compared to the stock 88W PPT, the 120W setting resulted in a 9-10% performance boost. 

Godot Compile

In the Godot compile test, the 120W PPT and unlimited PPT delivered nearly identical results, with times of 5 minutes and 50 seconds and 5 minutes and 47 seconds, respectively. The 120W PPT still offered a 5% improvement over the stock PPT, which took 6 minutes and 11 seconds to complete the task. 

PugetBench - Photoshop

In PugetBench Photoshop, the performance gains from increasing the PPT were negligible. The score only slightly improved from 12049 on the stock PPT to 12098 on the 120W PPT, identical to the score under the unlimited PPT. This suggests that the test might be limited by factors other than wattage.

PugetBench - Premiere Pro

In contrast, PugetBench Premiere showed a respectable gain, with the score increasing from 13426 on the stock PPT to 13832 on the 120W PPT—about a 3% improvement. Interestingly, this score is nearly on par with the unlimited PPTs 13880, indicating that 120W might be an optimal setting for Adobe applications on this CPU. However, further testing at 100W is needed to draw a definitive conclusion.

100W Testing

Further testing at 100W is necessary to determine if it strikes the right balance, particularly for gaming. If our previous assumption holds true - that 120W is optimal for gaming - then the 100W setting should yield similar results in F1 23.

Indeed, the data supports this, showing behavior similar to that of the 120W PPT. Testing the 100W PPT in F1 23 showed the chip consuming between 91-95W with an average clock speed of 5504 MHz. So for gaming, setting the PPT to 100W should allow the CPU to maintain max turbo for an extended duration.

In Cinebench 2024, the 100W PPT achieved a multi-core score of 1235, a 4% improvement over the default setting. However it falls short by 3% compared to the 120W PPT and by 5% compared to the unlimited PPT. In terms of efficiency, the 100W PPT yields a score of 12.35 per watt, which is decent but represents an 9% decline in efficiency compared to the default PPT’s score per watt of 13.49, but 14% better than the 120W PPT’s score per watt of 10.62.

7-Zip

In 7-Zip, the 100W PPT landed between the default and 120W settings, with a compression rate of 111.10 MB/s and a decompression score of 1560 MB/s. This is a modest 2-3% improvement over the stock PPT, indicating that the slight increase in wattage doesn’t significantly impact performance in this benchmark.

Blender 4.2.0

Blender results tell a similar story. The 100W PPT clocked in at 7 minutes and 12 seconds - 19 seconds faster than the stock PPT but 20 seconds slower than the 120W setting. This outcome supports our hypothesis that the 100W setting is a middle ground.

Handbrake

In Handbrake, the 100W PPT delivered an average encoding speed of 87.90 FPS for AV1, marking a 5% improvement over the default. However, in x264, it barely underperforms compared to the 7700X, with an average encoding speed of 99.62 FPS versus the 7700X’s 99.98 FPS. Notably, the 9700X matches the 7700X’s x264 performance while drawing less power, which is a benefit. Still, for noticeable gains over the previous generation in x264, at least a 120W PPT is recommended.

Other Benchmarks

Godot Compile, PugetBench Photoshop, and PugetBench Premiere Pro follow the same pattern. The Godot Compile test shows the 100W PPT completing the 5 minutes and 59 seconds, which is 12 seconds faster than the stock PPT but 9 seconds slower than the 120W setting. In PugetBench Premiere Pro, the 100W PPT scored 13599, a slight 1% improvement over the stock PPT. Meanwhile, in Photoshop, the score of 12000 shows no significant change in the stock setting, as the variance in Photoshop scores can swing by about 50 points in some runs.

When it comes to performance, if you’re aiming to maximize the capabilities of the 9700X, it is better to stick with a higher PPT setting (120W or above). The 100W settings does not provide enough gains to justify the potential risk of voiding your warranty, as the performance improvements are minimal across most benchmarks.

Cinebench MHz Performance Across Different PPT Settings

Now that we’ve seen how these wattage settings affect application performance, let’s quickly examine how they influence clock speeds in Cinebench.

Stock 88W PPT: The CPU manages an average clock speed of 4706 MHz, with a peak of 5233 MHz. In CPU-intensive tasks, it is clear that the lower wattage is limiting performance, preventing the processor from reaching its full potential. 

100W PPT: There is a modest improvement here, with an average clock speed of 4887 MHz. The maximum clock speed recorded was 5459 MHz, just shy of hitting full turbo speeds. This suggests that while the CPU is nearing its full turbo capacity, it only sustains this peak for brief moments before dropping back to the average. 

120W PPT: The clock speed sees a significant boost, averaging 5137 MHz and finally surpassing the 5 GHz threshold. The maximum recorded speed is 5501 MHz, supporting the idea that 120W might be the optimal PPT setting for this chip.

Unlimited PPT: With an average power draw of 149W, the CPU reaches an impressive average clock speed of 5309 MHz. Although the maximum speed is only slightly higher at 5468 MHz, the clock speed remains the most stable across the board, thanks to the unrestricted power profile. 

These results suggest that while increasing the PPT does push the clock speeds higher, there’s a point of diminishing returns, especially when moving beyond the 120W PPT. The unlimited setting provides the most consistent performance, but the gains over 120W are relatively modest. 

Conclusion

Across the productivity tests, the 120W PPT generally resulted in a 3-4% performance decrease compared to the unlimited PPT, except in PugetBench and 7-Zip, where the difference was less pronounced. This reduction in performance is offset by improved efficiency, given the roughly 20% lower power consumption under full load. 

Depending on your workload, simply increasing the PPT from the stock 88W to 120W can yield a 3-10% performance boost in productivity tasks, making it an attractive option for users looking to extend the CPU’s lifespan before their next upgrade.

When we tested the 100W PPT, it became clear that this setting occupies a middle ground. While it offers moderate gains over the stock PPT, it does not quite match the performance of the 120W or unlimited PPT settings. In Cinebench, for example, the 100W PPT scored 1235 in multi-core performance—better than the stock but lagging behind the 120W and unlimited settings.

When it comes to gaming, the extra wattage has little impact on FPS. While the higher wattage allows the clock speed to remain near maximum turbo, the difference in average clock speed is minimal—about 5.48 GHz with the stock PPT versus 5.51 GHz with the increased PPT. Consequently, the FPS improvement is negligible and within the margin of error.

With the increase in wattage, there is some rise in temperatures, though it varies depending on the load. For our testing we used an Arctic Liquid Freezer III 360, therefore temperatures listed here may be lower than you might experience with an air cooler. In gaming, the temperature increase is minor; for instance, in our F1 23 test, the average temperature only rose from 52°C at the stock 88W PPT to 54°C with a 92W average draw. In Cinebench, the impact on thermals is more pronounced. The stock 88W PPT resulted in an average temperature of 53°C, with a maximum of 59°C. When compared to the 100W PPT, we see an average temperature of 57°C, with a maximum of 63°C. And the 120W PPT’s average temperature of 66°C, with a maximum of 68°C, slightly higher but still keeping under 70°C. And finally the unlimited PPT, which drew an average of 149W with an average temperature of 82°C, with a maximum of 85°C.

Hence, unless you have the cooling to tame the unlimited PPT, our recommendation is to stick to the default PPT if you are looking for the coolest temperatures, or if you need some extra oomph set the PPT to 120W.

Overall, our findings indicate that while overclocking the Ryzen 7 9700X results in diminishing returns, it still offers significant generational improvements over the Ryzen 7 7700X when operated at similar wattage levels. As our Senior Test Technician aptly noted, "overclocking won’t suddenly reveal the universe’s secrets, but it’s reassuring to see that there’s more potential beyond the 88W limit”.

---