Cyberpunk 2077 Bug Bulldozes AMD Ryzen: Here's a Fix, Tested (Updated)

AMD Ryzen getting bulldozed
(Image credit: Shutterstock)

Regardless of your processor or GPU, Cyberpunk 2077 can push your system to the limits. However, resourceful redditors have discovered that the game code significantly harms performance on some AMD processors, because it apparently checks to see if your CPU uses Team Red's 2011-era Bulldozer platform and, if not, limits the number of available threads. 

Fortunately, the same folks have also crafted a simple fix that can lead to big performance gains with some AMD Ryzen processors under specific conditions. We put the fix to the test with several Ryzen CPUs, including some of the best CPUs,  to determine which processors benefit and under what conditions. We logged up to a 23% gain in some Cyberpunk 2077 benchmarks, so the fix is definitely worthwhile if you're limited by the performance of your Ryzen CPU. 

Our tests show that the fix works well for AMD chips with lighter core counts, so any Ryzen chip with eight or fewer cores will benefit. The performance gains apply to all four generations of Ryzen processors, but the improvements will vary based upon how close you are to a GPU bottleneck. We'll also outline how you can apply the fix yourself.

Cyberpunk 2077 is one of the most hotly-anticipated games of all time, but it's absolutely punishing on your PC, especially if you want to play at heightened resolutions and fidelity (see our evaluation of Cyberpunk system requirements). As usual, the community has begun investigating the best settings for the game and searching for useful hacks. 

The biggest hack to emerge centers around Cyberpunk's apparent inability to use AMD Ryzen processors to the fullest. Instead of using all of the threads available to the processor, the game code only runs on physical cores in some circumstances. As a result, some threads are left unused, leaving potential performance on the table. These same problems don't appear to apply to Intel's processors.

We're working on getting a better description of the problem from the game developers. For now, AMD's problem appears to stem from the GPUOpen software suite that Cyberpunk 2077 uses to detect the processor installed in the system. Apparently, this check determines whether or not the AMD chip is a Bulldozer model, and if it isn't, the software assigns the chip half the number of scheduler threads. That leads to the underutilization of some of AMD's processors.    

The fix is relatively simple, but as with any unsanctioned and unofficial fix, proceed at your own risk. Also, the fix doesn't work under all conditions. As we'll show below, you should be aware that some hardware configurations (typically processors with more than eight cores) can actually lose performance as a result. Also, be aware that the fix tends to result in higher CPU utilization on Ryzen processors, even if the particular model doesn't benefit from the patch.      

How to Apply the Cyberpunk 2077 AMD 'Fix'

You can find detailed instructions on applying the fix in several locations; there's a PCGaminWiki listing, a video, and a reddit post with more info on applying the fix manually. New software that automatically applies the 'fix' is already in circulation, but always exercise caution with unverified software - there could be malware lurking.

Here's a simple bullet point list of instructions on how to make the change, but head to the linked resources above if you're not experienced with making these kinds of modifications:

  • Open a hex editor to open the Cyberpunk 2077 .exe file
  • Search the executable for the following hexadecimal string: 75 30 33 C9 B8 01 00 00 00 0F A2 8B C8 C1 F9 08
  • Change to: EB 30 33 C9 B8 01 00 00 00 0F A2 8B C8 C1 F9 08 (yes, you only modify one entry). (address 02A816B0)
  • Save and exit.

And with that fix, the game code treats all CPUs equally and will spread the workload across all available cores. Now you're ready to test to see if your particular configuration responds well to the patch. We did just that with a limited selection of AMD processors to see which models benefit.

Cyberpunk 2077 Test Setup

We tested with the highest number of NPC's available and followed the same benchmark methodology used for our Cyberpunk 2077 PC Benchmarks, Settings, and Performance Analysis article. However, we tested with the latest 1.04 Cyberpunk patch.

We begin our test sequence at the entrance to V's apartment building and walk a path down the stairs, across the street to the left, hang a right at the vending machine, then head across the street towards the two policemen on the other side. From there, we proceed to the right across the pavilion, down another set of stairs, and to the building on the far side. 

Be aware that performance can, and will, vary in other scenes and locations (up to a 20% performance range), but this scene gives us a good-enough sense of the type of performance improvement, or degradation, that you'll see with the 'fix' applied.

Our test systems easily outweigh the recommended minimum system requirements for Cyberpunk 2077. We outfitted our test systems with the Gigabyte GeForce RTX 3090 Eagle to reduce the GPU bottleneck as much as possible. Be aware that you'll see less of a performance delta with lesser graphics cards. Also, if you're already shackled by a GPU bottleneck, you might not see appreciable gains with the same processors we've tested below. You can see a further breakdown of our test system at the end of the article.

Cyberpunk 2077 CPU Scaling RT Ultra Settings Benchmarks

Processors marked in the charts with 'Modified' have the fix applied. We tested with three resolutions at various settings, but our first album of results focuses on the 1920x1080 resolution with the RT Ultra preset applied. This test dials DLSS up to the Quality preset and also engages ray tracing, making it the most demanding preset. After we applied the patch, it also yielded big advances for a few of the AMD Ryzen processors.

The six-core 12-thread Ryzen 5 5600X trails the Core i9-10900K with the native game code, but the fix speeds up the game by a whopping 23%. We can also see that this improves the chips' performance across the full sweep of fps percentiles, indicating that the game runs much smoother as a result. In fact, the faster gameplay and improved smoothness were very noticeable during our test runs. We also noticed much faster game loads, particularly at higher resolutions. 

Those gains also apply to the eight-core 16-thread Ryzen 7 5800X, which takes a solid 16% jump forward. We also tested the six-core 12-thread Ryzen 5 3600XT with the fix and recorded an 18% improvement to average fps. This chip is basically the same as the second-gen Ryzen 5 3600X, so we expect to see similar gains when that chip lands on our testbed. 

We also tested the first-gen Ryzen 7 1800X, but while the fix helps with these older chips, the gains aren't quite as pronounced - we're simply running against the limits of the processor. However, we did see a respectable 9.5% improvement in this benchmark sequence. 

It's important to note that from our testing, and the testing we've seen shared by enthusiasts, AMD Ryzen processors with more than eight cores don't seem to benefit from the fix. In fact, in many cases, the fix can actually result in less performance. That's illustrated here with the beastly 16-core 32-thread Ryzen 9 5950X. This chart-topping chip leads the pack with 78.2 fps with unmodified game code, but performance drops by 8% after we've applied the fix.   

This is likely due to the changes forcing an even distribution of threads across the cores. That isn't always ideal for Ryzen processors because the operating system targets threads at the fastest cores first. We also see the reduced performance across our full sweep of the 5950X's fps percentiles. As you'll see time and again in the expanded testing below, the Ryzen 9 5950X doesn't benefit from the fixes - not that it matters much; as we can see, the 5950X leaves the Core i9-10900K in the rearview mirror. 

With ray tracing and DLSS performance accounted for, we'll take a look at multiple resolutions with standard rasterization after a quick look at why we've seen performance improve on the smaller Ryzen models.

Here we've plotted CPU utilization for each processor core during our test sequence, though it is notable that, to avoid possible overhead, these measurements weren't taken during our performance measurements. 

The first image shows the Ryzen 5 5600X running Cyberpunk 2077 with the native code. As you can see, several cores are under-utilized and fall below the 20% range. Flipping to the second chart shows the magnitude of the change with the fix applied: Now all cores are more fully utilized, which, as we've seen above, leads to increased performance. 

The third slide outlines core utilization on the Intel Core i9-10900K with the native code, and it's clear that the game uses the cores evenly with the standard game code.

The final two slides measure the same metrics, but with the 16-core, 32-thread Ryzen 9 5950X. Here we can see that several cores languish at the bottom of the chart with sub-10% utilization, but applying the fix (final slide) results in a more even distribution. However, we theorize that the fix doesn't result in increased performance due to incorrect thread targeting.

Cyberpunk 2077 CPU Scaling Medium Preset Benchmarks

This series of tests steps through 1080p, 1440p, and 4K resolutions at the medium preset. This preset features lower fidelity settings than the Ultra preset below, so we see more substantial gains. 

It's readily apparent that the magnitude of the performance gains relies heavily upon how close the system is to a GPU bottleneck – performance gains steadily decline as we increase the resolution. 

Again, the Ryzen 9 5950X runs slower with the fix applied - and that occurs at all resolutions except 4K, where the results fall within the margin of error due to a graphics bottleneck. Regardless, it's the fastest chip in our test pool by a large margin, so it doesn't need much help. 

At 1080p, the Ryzen 5 5600X delivers 11% more performance after the fix, climbing past the Core i9-10900K in the performance rankings, while the Ryzen 5 3600XT gets a 14% bump. At 1440p, we see the 5600X again jump out ahead of the 10900K due to a 5% increase in performance, and the 3600XT improves by 7%. These gains all translate over to smoother gameplay as well, as can be seen in our fps percentile charts. 

We see very limited gains with the Ryzen 7 5800X at 1080p, which might indicate the processor is giving us near-peak performance with the number of threads allotted at default settings. However, we do see a 2.5% gain at 1440p, but that isn't very significant.   

Similar trends carry over to the Ryzen 7 1800X - this chip doesn't get a meaningful boost at 1080p or 1440p and moves forward a rather insignificant 2.5% at 4K. It's clear that gains with the first-gen Ryzen processors will be far less pronounced. 

You'll notice that, with the exception of the vanilla Ryzen 5 3600XT configuration and the 1800X, we're almost completely GPU-bottlenecked at 4K. The fix allows the 3600XT to push the RTX 3090 to the saturation point. We also see the unmodified Ryzen 5 5600X configuration squeeze past the fixed version of the code by 1%, but that's within the margin of error.  

Cyberpunk 2077 CPU Scaling Ultra Preset Benchmarks

Here we turn the dial up to the Ultra preset, and again, this is without ray tracing active. As expected, the gains again shrink as we increase resolution and push closer to a graphics bottleneck that makes CPU performance less of a factor.

At 1080p, the Ryzen 5 5600X improves by 10%, while the Ryzen 5 3600XT jumps 9.5%. At 1440p, those gains shrink to the negligible range for the 5600X because it has already pushed the RTX 3090 to a bottleneck. The Ryzen 5 3600XT still has some room to maneuver beneath the GPU-imposed ceiling, and it does by 13.7%. 

The Ryzen 7 5800X scrapes out a 3% gain at 1080p and falls within the margin of error at 1440p. Basically, this chip is fast enough with the standard code to not benefit much from the fix. 

Once again, the 4K resolution pushes us to the land of the margin of error, and it's clear that here the CPUs don't gain any meaningful performance from the fix. Conversely, we see small gains for the 1800X in these tests, but that's likely due to the chip already operating near its limits.  

Overall, our first round of benchmarks with the Cyberpunk 2077 fix indicates that it works well for AMD chips with lighter core counts, meaning chips with eight or fewer cores, and the performance delta will vary based upon how close you are to a GPU bottleneck. As such, those with lesser GPUs will see varying levels of success. You'll see more noticeable performance improvements with newer generations of Ryzen processors, but even the first-gen models benefit in some scenarios. In many cases, the best approach to determining if your particular configuration will benefit simply consists of making the relatively simple change and giving it a spin. Proceed with caution, though, as the 'fix' isn't officially acknowledged by CD Projekt Red. 

Update from Jarred: I tested the Ryzen 9 5900X with and without the SMT modification. As expected, it generally doesn't help. The unmodified code was faster in most test runs, though it was often within 1-2 percent (which is basically the margin of error for OCAT testing of the game). I did measure slightly higher performance at 1080p with the RT Ultra preset (4 percent faster) and 2 percent faster at 4K medium, but neither one is a particularly noticeable change.

Swipe to scroll horizontally
Cyberpunk 2077 CPU Benchmark Test Configurations
Intel Socket 1200 (Z490)Core i9-10900K
Row 1 - Cell 0 MSI MEG Z490 Godlike
Row 2 - Cell 0 2x 8GB Trident Z Royal DDR4-3600 - Stock: DDR4-2933
AMD Socket AM4 (X570)AMD Ryzen 9 5950X, Ryzen 5 5600X, 3600XT

MSI MEG X570 Godlike
Row 5 - Cell 0 2x 8GB Trident Z Royal DDR4-3600 - Stock: DDR4-3200
All SystemsGigabyte GeForce RTX 3090 Eagle

2TB Intel DC4510 SSD

EVGA Supernova 1600 T2, 1600W
Row 9 - Cell 0 Open Benchtable

Windows 10 Pro version 2004 (build 19041.450)
CoolingCorsair H115i
Paul Alcorn
Managing Editor: News and Emerging Tech

Paul Alcorn is the Managing Editor: News and Emerging Tech for Tom's Hardware US. He also writes news and reviews on CPUs, storage, and enterprise hardware.

  • Flayed
    Can't help but think Intel had a hand in this
    Reply
  • TCA_ChinChin
    Flayed said:
    Can't help but think Intel had a hand in this
    I don't think it's Intel, but its disappointing that CDPR didn't catch something so easy that could have improved performance for many AMD CPUs.
    Reply
  • Makaveli
    Flayed said:
    Can't help but think Intel had a hand in this
    Based on how easy it is to fix just looks like lazy programming more so than something done on purpose.
    Reply
  • woot
    Flayed said:
    Can't help but think Intel had a hand in this
    nonsense, it would be really petty and ultimately pointless for Intel to do something like that...
    Reply
  • JarredWaltonGPU
    Makaveli said:
    Based on how easy it is to fix just looks like lazy programming more so than something done on purpose.
    Yup. Looks like CDPR just used a compiler or compiler flag that had some code in to try and help certain older AMD CPUs that didn't know what to do with Ryzen.
    Reply
  • hotaru.hino
    Missing a simple spot check happens all the time in programming.

    Speaking from experience. :tearsofjoy:
    Reply
  • clsmithj
    You guys test graphs gave me a headache. I wish you went with simple blue bars to show the performance difference, and left the Core i9 CPU off since it's Intel and not relevant to this patch.
    Reply
  • randomizer
    I find it odd that AMD didn't pick this up. Surely for such a heavily anticipated title there would have been development support from Intel, NVIDIA, and AMD.
    Reply
  • Annorax_
    I made an account just to say that with a Ryzen 5 3600X my game used to run at 50-70 fps but now it runs at 80-100 frames after applying the fix.
    Reply
  • Russell78
    Crippled by devs legends apart, I have to say this time great useful job done here by Tom's, a few more Intel CPUs wouldn't have hurt but thanks anyway ^^
    Reply