Avoid expensive bottlenecks.

EFps Game Benchmarks

313
+21%
438
+16%
157
+16%
177
+18%
201
+31%
222
+38%
CSGO EFps
Average Fps
0.1% Low(Avg)
0.1% Low(Max)
1% Low(Avg)
1% Low(Max)
259
379
135
150
154
161
2060S 10900K Source
24 Jun '20 Driver:445.87
2060S 11400F Source
01 Apr '21 Driver:465.89
Stutters & Frame Drops
The graph shows lowest Fps (slowest frame) in one second buckets.

Watch Gameplay

Bottlenecks

An AMD Ryzen 3700X bottlenecks a 2070S: the 3700X costs 40% more money for 11% less performance. The lost performance is similar to downgrading from a 2070S to a 2060S. Publishing EFps data puts UserBenchmark in conflict with the marketers that represent billion dollar corporations, but it also helps users dodge marketing traps and build faster PCs. Users can verify EFps figures with Afterburner

We do

  • Focus on user verifiable facts and figures.
  • Buy all of our hardware from mainstream shops: no golden "free" samples here.
  • Provide PC details, driver versions, game settings and source video for our EFps figures.
  • Pick games that most of our users actually play. (ACO, BFV, SOTTR etc. are relatively unpopular)

We don't

So that

Intel vs AMD

Intel continues to offer better CPUs at lower prices but they also appear oblivious to social media marketing: forums, reddit, youtube etc. As a result, publishers are openly accused of bias for reporting replicable, real world performance data. Since the launch of Ryzen, AMD have carved 50 billion 100 billion dollars off Intel’s bottom line. After failing to grasp new-age marketing, and consequently losing significant market share, Intel is finally motivated to deliver material CPU improvements. The dynamic between the two companies is a blessing for well informed users who can save money without compromising on performance.

Nvidia vs AMD

AMD’s Vega (56, 64) and Navi (5000) series GPUs deliver great performance in specific games with specific hardware/software configurations. Unfortunately, following the initial surge in sales fueled by AMD's marketers, many users discovered that they exhibit widespread stability and compatibility issues (blue/black screens and driver resets). The cards are engineered to shine in benchmarks, often at the expense of real world performance and hair dryer levels of noise. For example, the reference 5700 XT has its stock clocks set so high that it overheats and stutters in popular games. Expert users, that are happy to tinker with BIOS updates and voltage curves can sometimes extract excellent price/performance but many users end up getting a refund instead. By comparison, Nvidia’s cards have better compatibility, lower return rates and rock solid stability. That said, above the 2060S-2070S tier, where there is no competition from AMD, Nvidia’s pricing remains completely unchecked.
Why does UserBenchmark have a bad reputation on reddit?
Marketers operate thousands of reddit accounts. Our benchmarks expose their spiel so they attack our reputation.
Why don’t PC brands endorse UserBenchmark?
Brands make boatloads on flagships like the 4090 and 14900KS. We help users get similar real-world performance for less money.
Why don’t youtubers promote UserBenchmark?
We don't pay youtubers, so they don't praise us. Moreover, our data obstructs youtubers who promote overpriced or inferior products.
Why does UserBenchmark have negative trustpilot reviews?
The 200+ trustpilot reviews are mostly written by virgin marketing accounts. Real users don't give a monkey's about big brands.
Why is UserBenchmark the gold standard for users?
Instead of pursuing brands for sponsorship, we've spent 13 years publishing real-world data for users.
The Best
CPUGPUSSD
Intel Core i5-12600K $159Nvidia RTX 4060 $280WD Black SN850X M.2 2TB $135
Intel Core i5-13600K $210Nvidia RTX 4060-Ti $389WD Black SN850X M.2 1TB $84
Intel Core i5-12400F $111Nvidia RTX 4070 $500Crucial T700 M.2 4TB $370
Today's hottest deals
If you buy something via a price link, UserBenchmark may earn a commission
About  •  User Guide  •  FAQs  •  Email  •  Privacy  •  Developer  •  YouTube Feedback