- Performance Baseline and Test-Setup
- Overclocking the Ryzen 5 7600X
- Delidding & Direct-Die cooling
The Ryzen 5 7600X is a 6-Core, 12-Thread processor for AMDs AM5 platform launched at the end of September 2022. The 7600X is built on the "Zen 4" architecture using a single I/O-Die paired with a single CCD hosting the six CPU-Cores.
Built for the AM5 platform also means the 7600X only supports the newer DDR5 standard as well as PCI-E 5.0. Built on TSMCs 5nm FinFET process technology, the CPU runs at a 4.7 GHz base clock, boosting up to 5.3 GHz according to AMDs specifications.
What's also interesting is the fact that the Ryzen 5 7600X and all other 7000-Series processors launched so far come with an integrated GPU (IGPU), a feature that up until now was only found in the "G" variant of previous Ryzen processors.
Today I want to explore the overclocking capabilities of the Ryzen 5 7600X and find out how to get the best performance from the Zen 4 CPU.
Most Important Specifications
|# of Cores||6|
|# of Threads||12|
|Max. Boost Clock||Up to 5.3GHz|
|Memory Support||DDR5-5200 (1R1D)|
|Max. Operating Temperature (Tjmax)||95°C|
|CPU Boost Technology||Precision Boost 2|
For the full specification, visit amd.com.
2. Performance Baseline and Test-Setup
For this test I'm using a Gigabyte B650 Aorus Elite motherboard paired with 2x16GB DDR5 dimms from G.Skill alongside the 7600X, all running on an open-air test-bench. The memory is a 5600-CL36 kit with Samsung ICs, manually tuned to run DDR5-6200 34-36-36-89 as well as tightened sub-timings.
|Motherboard||Gigabyte B650 Aorus Elite|
|Memory||G.Skill Flare X5 DDR5-5600 CL36|
|Power Supply||Seasonic Prime PX-1300|
Memory tuning will not be part of this article, so here is the memory configuration I used across all tests.
|Memory Speed||6200 Mbps|
|Memory Controller Clock||3100 MHz|
|Infinty Fabric Clock||2067 MHz|
|Memory ICs||Samsung B-Die|
Stock Performance and Boosting Behavior
For Ryzen 7000-Series CPUs, it's not so easy to define an actual stock performance due to the aggressive boosting behavior these chips are designed to run out of the box. To put it simply, the 7600X will try to run the highest possible clock speed that's still within specification as long as certain requirements are met, one of them being the maximum operating temperature of 95°C. This means that depending on the cooling solution used, one might end up with different boost clocks and performance numbers. This is very important to take into consideration when evaluating the performance of the Ryzen 5 7600X or other Ryzen 7000 Series CPUs.
Because of that, I have decided to use multiple different cooling solutions to highlight the impact it makes on the overall performance and the overclocking experience of the CPU. The coolers I'm using are an AMD Wraith Prism stock cooler, an Arctic Liquid Freezer II 420 AIO and a custom water cooling loop equipped with a Heatkiller IV Pro water block and a MO-RA3 420 radiator.
To get a first impression of the stock performance of all three coolers, I started with a five minute Cinebench R23 multi run and observed the clock speed, temperature and power draw. All coolers always run at 100% fan and pump speed in all tests.
|Cooler||CB23 multi (5 min)||avg. Clock||Power Consumption||max. Temperature||avg. VCORE|
|Wraith Prism||15305||5247 MHz||95 W||88.9 °C||1.24 V|
|420mm AIO||15810||5410 MHz||94 W||77.6 °C||1.26 V|
|Custom Loop||15853||5445 MHz||88 W||67.3 °C||1.23 V|
From the numbers above it's already obvious that the Wraith Prism cooler isn't sufficient to unlock the full potential of the Ryzen 5 7600X. Both the AIO and the custom loop yielded roughly 3-4% better performance than the stock cooler and maintained about 150-200 MHz higher clock speed. The custom loop also allows the CPU to run 6-7 W more efficiently due to the lower operating temperatures and reduced VCORE.
I moved on to run a few more benchmarks to confirm the findings from the first test. The water cooling solutions generally performed up to 3% better than the stock Wraith Prism air cooler.
|Wraith Prism||420 AIO||Custom Loop|
|Cinebench R23 multi||15305||15810||15853|
|Cinebench R23 single||1970||1970||1976|
|Geekbench 5 multi||12501||12504||12544|
|Geekbench 5 single||2246||2234||2249|
|HWBOT x265 4K||21.807 fps||22.466 fps||22.611 fps|
|SuperPI32M||352.187 s||347.496 s||339.908 s|
|y-cruncher 1B||30.324 s||29.908 s||29.666 s|
3. Overclocking the Ryzen 5 7600X
With our stock performance baseline established, we can now move forward and start overclocking our CPU. The two most common options are using a static overclock across all CPU cores, or makeing use of Precision Boost Overdrive 2 (PBO2) to tune the boosting behavior of the processor.
Precision Boost Overdrive 2
The Ryzen 5 7600X uses Precision Boost 2 which automatically adjusts the CPUs clock speed and core voltage depending on various factors in order to deliver optimal performance. The most important factors that influence Prevision Boost 2 are:
- CPU temperature
- Type of workload
- Number of active cores
- Power draw from the CPU socket
- Current draw from the motherboards VRM
- Maximum boost frequency limit
The CPU monitors these factors constantly and adjusts the frequency up to 1000 times per second to adapt to changes. We can use Precision Boost Overdrive 2 to manually change some of these factors which gives the CPU more headroom for boosting in certain scenarios.
Overclocking with Precision Boost Overdrive 2 (PBO2)
PBO2 allows the user to tune some of these parameters to get potentially higher boosting frequencies and ultimately increase the overall performance of the Ryzen CPU. It can either be configured in the BIOS or via Ryzen Master, AMDs own overclocking and monitoring utility.
PBO2 allows us to configure several parameters to change the Precision Boost 2 algorithm.
Package Power Tracking (PPT): PPT is the amount of power the CPU can draw from the socket before the boost levels off. Higher PPT is especially relevant when running multithreaded workloads.
Thermal Throttle Limit: This is the maximum temperature the CPU is allowed to reach before throttling.
Electrical Design Current (EDC): The peak current that the motherboard's VRMs can supply under transient conditions. Higher EDC can improve short-duration, high-intensity workloads.
Thermal Design Current (TDC): The amount of current the VRM can supply over a sustained period. Higher TDC can improve heavily multithreaded workloads.
Precision Boost Overdrive Scalar: This tells the boost algorithm to use higher voltages for longer periods to sustain boost activity that may otherwise be throttled back. Higher scalar values can improve boost performance in scenarios that might be voltage limited.
Boost Override: The boost overdrive is used to set a higher frequency than the stock boost frequency.
Curve Optimizer (CO): Allows adaptive over- or undervolting that can free up additional voltage and thermal headroom for higher frequencies and longer boost periods. You can either apply a positive or negative voltage offset for all or individual CPU cores. This effectively tells the boosting algorithm that the CPU does need less or more voltage (depending on the offset) for a given frequency. It is usually beneficial to use a negative offset to lower the power draw and temperature of the CPU but at the cost of decreased stability.
Overclocking with PBO2
When overclocking with PBO2, I usually max out the motherboard-specific settings EDC, TDC and PPT as I am aiming for maximum performance. Also, the 7600X won't be able to draw power anywhere close to the limit of what my motherboard is capable of delivering, so I'm not worried about causing any damage with too high settings. The Thermal Throttle Limit is set to 95°C and I don't recommend setting it any degree higher than that. Boost Override is always set to the maximum offset of +200MHz to enable the highest possible frequencies the algorithm allows.
I have tested a few different Curve Optimizer settings to showcase the impact of different values on the performance of the CPU and again used my three different cooling solutions for each scenario.
Let's have a look at the results of the 420 AIO cooler for different Curve Optimizier settings at -20, -25 and -30 respectivley.
|No CO||CO -20||CO -25||CO -30|
|HWBOT X265 4K||21.782 fps||22.324 fps||22.496 fps||22.568 fps|
What's interesting here is that enabling PBO2 without Curve Optimizer can hinder heavily multi-threaded workloads while increasing single-threaded performance. The best result is achieved with a -25 Curve Optimizer offset, but the difference to the -30 offset is only marginal. In general, multi-threaded performance didn't change compared to the stock results while less-threaded workloads achieved slightly better results.
Now let's have a look at how the stock air cooler and the custom loop perform with the -25 CO offset compared to stock levels. For the y-cruncher and SuperPI results, the relative performance is always interpreted as "XY percentage faster relative to stock", because a lower value is better for duration-based results.
|Wraith Prism||420 AIO||Custom loop|
|CB23 multi||15350 (+0.3%)||15861 (+0.3%)||15917 (+0.4%)|
|CB23 single||2021 (+2.6%)||2034 (+3.2%)||2027 (+2.6%)|
|GB5 multi||12514 (+0.1%)||12612 (+0.9%)||12656 (+0.9%)|
|GB5 single||2309 (+2.8%)||2314 (+3.6%)||2313 (+2.8%)|
|HWBOT X265 4K||21.842 fps (+0.16%)||22.496 fps (+0.13%)||22.675 fps (+3.26%)|
|y-Cruncher 1B||30.274s (+0.16%)||29.823s (+0.28%)||29.705s (+0.13%)|
|SuperPI32M||306.5s (+12.97%)||305.211s (+12.17%)||305.747s (+10.05%)|
The results of the two other coolers are inline with what we have observed from the AIO cooler. Single-threaded performance is slightly improved while multi-threaded workloads stayed within the run to run variance. It's also interesting that all three coolers deliver almost identical single-threaded performance which makes sense given the much lower power draw in such scenarios.
Overall though, the performance gains compared to the stock results are disappointing. If we have a look at the relative performance gain it becomes pretty obvious.
Most benchmarks improved by less than 3%, SuperPI32M being the only exception here. Maybe there are other benefits of PBO2 like reduced temperature and power draw, so I've rerun the five minute Cinebench R23 multi loop.
|Cooler||avg. Clock||Power Consumption||max. Temperature||avg. VCORE|
|Wraith Prism||5241 MHz (-6)||93 W (-2)||89.9 °C (+1.0)||1.24 V|
|420mm AIO||5419 MHz (+9)||95 W (+1)||77.1 °C (-0.5)||1.27 V (+0.1)|
|Custom Loop||5463 MHz (+18)||95 W (+7)||73.8 °C (+5)||1.28 V (+0.5)|
Again, there aren't any noticeable improvements regarding clock speed, power consumption or temperature and I also didn't expect anything to improve significantly after seeing the results for the multi-threaded workloads.
When searching for the optimal Curve Optimizer offset, it's important to keep in mind that the stability of your system will decrease with a larger CO offset. For example, using the Wraith Prism air cooler, I started to see stability issues with a -30 offset in some benchmarks. While the other two coolers were able to pass all tests with the -30 offset without issues, it's good to know at which point stability issues might occur and your system starts crashing. As I found no measurable benefit between an offset of -30 and -25, using the -25 offset is a better choice in my opinion.
If you are testing your PBO2 settings and Curve Optimizer offset for high stability, you will find unstable settings easier when running heavily multi-threaded benchmarks like y-Cruncher or prime95 as they will produce a higher power output and put more stress on your motherboards VRM and your cooling solution.
Now that we have explored different settings for PBO2 and CO, let's summarize our findings when applying it to the Ryzen 5 7600X:
- A negative Curve Optimizer offset will yield the best performance when using PBO2 and CO. I've found an offset of -25 to produce good results while maintaining stability in all performed tests.
- Multi-threaded performance will only change very slightly or not at all in most cases I've tested.
- Single-threaded performance will improve slightly depending on the workload executed. I've found that in the few tests I've run, the improvements were between 0-3% but can be higher in specific workloads.
- A better cooling solution will not lead to significantly higher single-threaded performance due to the relatively low power output.
Static All-Core Overclocking
Of course, there is always the option to apply a "traditional" static all-core overclock to your CPU to maximize multi-threaded performance. This approach however comes with a few drawbacks. First, you probably won't be able to run the same frequencies across all cores as what the CPU achieves using Precision Boost 2 during less threaded workloads. Therefore the single-threaded performance of the CPU will most likely decrease. Secondly, the CPU temperature can exceed the maximum safe operating temperature of 95°C.
For this test, I am running my benchmark suite again with the three different coolers, but this time applying a static all-core overclock. Due to the different performance levels of the three cooling solutions, the overclock will be different on each cooler. Better coolers will be able to dissipate more heat and therefore enable us to use higher voltage levels while maintaining safe operating temperatures.
For dialing in the overclock you can use the BIOS or the Ryzen Master utility. To improve stability, I've also set the LLC setting to HIGH in the BIOS of my Gigabyte Aorus B650 Elite that I am using. LLC is important to control the amount of voltage drop you get when running the CPU under load. I recommend setting LLC to at least medium or high to increase the stability of your overclock.
Overclocking the Ryzen 5 7600X
Finding good overclock settings always requires a bit of trial and error as you try to find a balance between voltage, temperature and performance.
- Start by setting the core voltage to a reasonable high voltage and a moderate core clock. You can usually just take some values the CPU uses in stock conditions under load as a starting point.
- Increase the core multiplier and run a stability test or benchmark until you experience instability.
- Raise the voltage level by a small amount, I like to use 25mv steps for this.
- Repeat this process until you don't see any scaling with higher voltage levels anymore, or your temperature gets too high.
In the table below you can see the settings I have used for the different cooling solutions.
|Wraith Prism 5.40GHz @1.2V SET||420 AIO 5.50GHz @1.25V SET||Custom loop 5.60GHz @1.275V SET|
|CB23 multi||15755 (+2.9%)||16050 (+1.5%)||16264 (+2.6%)|
|CB23 single||1948 (-0.11%)||1991 (+1.1%)||2017 (+2.1%)|
|GB5 multi||12504 (+0.0%)||12733 (+1.8%)||12605 (+0.5%)|
|GB5 single||2229 (-0.08%)||2286 (+2.3%)||2304 (+2.4%)|
|HWBOT X265 4K||22.499 (+3.17%)||22.963 (+2.21%)||23.173 (+5.25%)|
|SuperPI32M||332.237 (+5.66%)||318.56 (+8.33%)||307.883 (+9.42%)|
The results confirm my initial expectations for the most part. All three coolers achieve higher multi-threaded performance across all tests while sacrificing some single-threaded performance. The only exception to this is the custom loop that can maintain about the same single-threaded performance as with PBO2. This is due to the maximum possible boost speed of 5.65GHz when using Precision Boost 2, which is just 50 MHz higher than the 5.60 GHz overclock achieved with the custom loop.
The actual improvement however is again below 3% in most cases which is barely relevant.
Using an all-core overclock enables the CPU to run higher clock speeds across all cores as one can achieve using PBO2. It's no surprise therefore to see higher temperatures and increased power consumption across all coolers compared to the stock run.
|Cooler||avg. Clock||Power Consumption||max. Temperature||avg. VCORE|
|Wraith Prism||5400 MHz (+153)||99 W (+4)||91.5 °C (+2.6)||1.25 V (+0.01)|
|420mm AIO||5500 MHz (+90)||109 W (+15)||88.0 °C (+10.4)||1.30 V (+0.04)|
|Custom Loop||5600 MHz (+155)||109 W (+21)||81.6 °C (+14.3)||1.32 V (+0.09)|
Running the Ryzen 5 7600X with an all-core overclock can easily result in seriously high temperatures if you are not careful. Under stock conditions or while using PBO2, the CPU will lower its core clock and voltage automatically to stay below the 95°C maximum operating temperature. Once you manually change the core voltage of the CPU however, this 95°C limit does not apply anymore. When I was playing around finding an all-core overclock, at one point when running y-Cruncher, I noticed that the temperature of the CPU was reaching over 115°C which honestly left me kind of surprised. I wouldn't expect that to happen because I haven't changed any setting regarding the maximum allowed operating temperature of the CPU. Keep that in mind when running an all-core overclock and make sure your voltage is appropriate for the workload you are running.
You probably noticed that no cooler was able to finish the y-Cruncher 1B test with the applied overclock. You could of course lower the frequency or try to increase the core voltage to improve stability at the cost of lower performance. However, you have to decide for yourself what level of stability you want or need to achieve. Y-Cruncher is a very heavy workload that puts a lot of stress on the CPU and not everyone will encounter this type of workload on a daily basis.
My goal here is to give an impression of what's possible with the CPU and a given cooling solution when running the CPU pretty much on the edge. Nobody should be disappointed if he cannot run 5.60 GHz stable, but rather use my results as an orientation on what is reasonable to achieve and what is not.
- Using a static all-core overclock can increase multi-threaded performance in a range of 0-5% while sacrificing a very small amount of single-threaded performance.
- Power consumption and temperature will be higher compared to stock settings or using PBO2.
- The CPU doesn't automatically throttle down to stay below the default 95°C maximum operating temperature and can reach temperatures well above 95°C and up to 115°C before shutting down.
4. Delidding & Direct-Die cooling
If you're not satisfied with the level of overclocking you can do with the 7600X, there are some more advanced things you can do to your CPU to improve cooling efficiency and maximize the overclocking potential of your CPU. However, these techniques require some amount of skill, patience and it can be quite costly.
Removing the IHS and using direct-die cooling can potentially result in a huge temperature reduction. However, removing the soldered IHS from a CPU involves a lot of risk of accidentally damaging or destroying the processor in the process. Additionally, there are a lot of factors that influence how well the direct-die cooling will work with your particular cooling solution.
Before your decide to delid your Ryzen 5 7600X, let me tell you just a few problems you might encounter:
- Your CPU might get physically damaged or destroyed during the delidding process.
- If you are using liquid metal, it could spill over some parts of the CPU or the motherboard during the cooler installation, potentially shorting out and damaging multiple components of your system.
- Your cooler may be incompatible with direct-die cooling due to the reduced height of the CPU without the IHS.
- Your cooler might perform poorly because the cold plate may not be flat enough to make proper contact with the CCD of the CPU.
- If the cold plate of your water block isn't nickel-plated, the liquid metal will react with the copper, forming an alloy over time which also impacts the performance of your cooler.
How to delid your Ryzen 7000 CPU
There are a few different techniques to remove the IHS. The most common way is to use a specialized delidding tool like the Thermal Grizzly Ryzen 7000 Delid-Die-Mate. While using the Delid-Die-Mate, it's very important to take your time and follow the instructions of the tool precisely to ensure you have the highest possible chance of success and minimize the risk of destroying or damaging your CPU.
If you were able to remove the IHS successfully, any remaining solder must be removed from the CPU dies. The easiest way to do this is to spread a small amount of liquid metal on top of the CPU dies and let it sit there for about five minutes. The liquid metal will react with the indium solder and you should be able to wipe it off with some cleaning alcohol afterwards. Repeat the process if there is still some solder left after the first round of cleaning. You can also use a razor blade to scrape away the solder residues, but you have to be extra careful not to cut off an edge or corner of one of the CPU dies.
Installing the delidded CPU
Once your CPU is cleaned, you need something to secure it in the socket of your motherboard, because, without a heatspreader, you can't use the standard AM5 ILM anymore. The easiest way to mount the CPU in the socket is with the Thermal Grizzly Ryzen 7000 Direct Die Frame that screws directly into the existing AM5 backplate of your motherboard. Screw in the frame fully tight and then loosen the screws up just a tiny bit again. You might experience memory detection issues if your CPU sits too loose or tight in the socket, but with this technique, I found it relatively easy to socket the CPU properly.
With the CPU in the socket it's time to apply the liquid metal on all the CPU dies as well as on the cold plate of your cooler. If you are lucky, your cooler won't collide with any components surrounding the socket and can make proper contact with the CPU. It's a common issue that the included mounting solution of your cooler won't work because of the reduced height of the CPU without the heatspreader.
With the mounted cooler it's time to test if everything worked out correctly and see how the temperatures are. If your system doesn't start, it's probably due to improper mounting pressure and you'll most likely have to tighten or loosen the direct-die frame a little bit and try again.
Dealing with cooler mounting issues
For my testing, I went on to delid my Ryzen 5 7600X as described above and mounted my Heatkiller IV waterblock that I used in the previous tests. I set the clock and voltage back to default and booted up the system to perform some tests. When running the first Cinebench R23 benchmark to verify the setup, the temperatures immediately jumped to 95°C and the CPU throttled down to stay at the maximum allowed operating temperature.
This is a clear indicator that something isn't right with the contact between the water block and the CPU, but even after multiple re-mounts, the temperatures were horrible. After investigating the cold plate of the water block, I could see clear markings of the CPU dies on it, which means the block was making contact with the dies. Maybe I didn't clean the CPU properly and some residues messed up the contact with the cooler. I cleaned off the liquid metal from the CPU dies and then used a very fine piece of sandpaper to wet sand the CPU very gently. After that the CPU looked absolutely spotless and I was hoping that this would make a difference, but sadly it didn't change anything. Sometimes the system wouldn't even make it into windows because it just crashed while booting.
At this point, I tried replacing the liquid metal with the new Thermal Grizzly KryoSheet. If not anything, it would at least make it a lot easier for me to remount the water block when necessary. Unfortunately however, the temperatures weren't anywhere near where they should be when using the KryoSheet. I also tried using a different water block with a flattened copper cold plate, but didn't change anything about the current situation. I even tried screwing the water block directly into the AM5 back plate with the included screws that are used for the standard AM5 mounting hardware in order to eliminate any mounting pressure restriction, but it didn't work either.
An alternative to direct-die cooling
My last hope to rescue the situation and recover the CPU was to try the Thermal Grizzly AM5 High Performance Heatspreader instead of direct-die cooling the CPU with the direct-die frame. This heatspreader is manufactured with high-precision tools which means the surface flatness of it is as good as it gets, which lets me hope it will solve the issues I had with direct-die cooling. The surface area of the heatspreader is also a lot larger compared to the stock IHS, which should improve cooling efficiency and lead to better temperatures.
I applied liquid metal again on the CPU and the inside of the Thermal Grizzly Heatspreader and then mounted everything in a similar way as with the direct die frame. With the new heatspreader the system finally booted up fine with reasonable temperatures. The CPUs boosting behavior was as expected and the system didn't overheat or crash. To compare the actual improvement over the stock IHS, I dialed in the same overclock settings I've used previously for the static all-core overclock running 5.6GHz @1.275V core voltage with HIGH LLC setting. After running Cinebench R23 for five minutes, I noticed an actual improvement of just about 2-3°C. Unfortunately, such a small temperature improvement won't have any impact on the overclocking capabilities of the Ryzen 5 7600X.
Overclocking the Ryzen 5 7600X doesn't have a significant impact on the overall performance in the tests I have done. Using a static all-core overclock can lead to temperatures exceeding the max. safe operating temperature of 95°C when using poor settings while negatively impacting performance in single-threaded tasks. You can use PBO2 with Curve Optimizer, there is nothing wrong with that, but again the performance impact isn't that significant.
Advanced techniques like direct-die cooling simply didn't work for me on the 7600X and even with an alternative solution like the high-performance heatspreader, the risk of damaging the CPU and the additional cost for all the tools and parts needed are not worth the potential temperature and performance improvements in my opinion.
The best way to get the most out of your 7600X is to use a high-performance cooling solution in the first place.
If you have any feedback or want to share your experience overclocking the Ryzen 5 7600X, I would be happy to hear it from you in the comments.