How Accurate Is GPUBoss?

| | , ,

GPUBoss lets you quickly compare graphics cards, but how accurate is it? Let’s find out.

GPUboss lets you compare the stats and benchmarks of graphics cards and then scoring them from 0.0 to 10.0 which are based on the graphics card specifications and real-world game benchmarks.

It’s a reasonably popular tool, and it was even more popular some time ago. Unfortunately, its audience is slowly declining.

This tool lets you compare gaming, graphics, computing, performance per watt, value, noise, and power scores. Based on that, the overall rating is calculated, which is represented in the review section.

Gaming Scores

Gaming scores are determined running benchmarks on Battlefield 3, Battlefield 4, Bioshock Infinite and 21 more. At least this is what this site claims because I couldn’t find what other 21 benchmarks were. I tried clicking underlined “21 more” which looks like it’s linking somewhere, but it’s not. Maybe this was a working link in the past so you could see more stats in-depth.

Anway Gaming score is supposed to be the average of all those benchmarks, but I can’t check how those ratings are calculated because this webpage doesn’t show all the criteria. I think they are just arbitrary or subjectively picking these scores, and that’s it.

Graphic Score

It’s the same story with the graphics scores. The difference is they used synthetic benchmarks like T-Rex, Manhattan, Cloud Gate Factor, Sky Diver Factor, and 1 more. This test is mostly based on scores of well-known 3DMark and GFXBench 3.0. So they took an average of those scores and somehow transformed it 0.0 – 10.0 scale.

What is interesting to me that on this test GTX 1070 managed to get the graphic score of a 9.6 while GTX 980 Ti got 6.6? Ok, I have to admit this looks really fishy to me let’s investigate further.

In 3DMark Fire Strike and Time Spy synthetic benchmarks GTX 1070 just barely defeats GTX 980 Ti. Although GTX 1070 has 8GB GDDR5 while GTX 980 Ti has 6 GB GDDR5, rating it for 3.1 more it quite a stretch and I also don’t believe that GFXBench showed the difference that big. Furthermore, real-time gaming performance on GTX 980 Ti is by 0.3 higher than GTX 1070, it should be the opposite.

Computing Power

They are testing computing power based on Face Detection, Ocean Surface Simulation, and 3 more. I assume that with the computing power, GPUBoss is referring to GPU acceleration of a CPU? So basically GPU can enhance or help CPU to accelerate specific task or applications for those who don’t know.

Site is not giving any information which software is using for face detection and ocean surface simulation comparison, so I can’t say much about that other than they got their data from Compubench which is a credible source.

Performance Per Watt/ noise and power

Nothing exciting to point out in this topic. The less power-hungry your graphics card is, the better. It’s as simple as that.

Value Score, Empty?

GPUBoss shows us empty graphs about graphics cards value. So my guess is they tried to implement value rating, but they scraped it or just didn’t finish it. Also, the value of a graphics card is hard to determine when prices are constantly changing, but they could calculate a value based on released date prices.

Problem with Graphics Cards Comparison

There are numerous problems compering graphics cards from scale to 10. Some graphics cards had a perfect 10.0 scores in some stats. How on earth will you compare better graphics cards in the future when you leave no headroom for better scores. By this logic, in the future, all cards would have perfect scores. Based on that, I would guess that this tool would be best suitable for comparing GPUs in the same generation but even comparing graphics cards in the same generation with this tool, stats are just all over the place.

Way too many mistakes or wrong data for anyone to rely on this tool for even a bit. For instance, comparing GTX 1060 6GB and GTX 1050 Ti yields better graphics scores and that should not happen since GTX 1060 6GB is a much better card? This tool is just super bad and inaccurate.

IMO best would be just to compare raw data of benchmarks, for instance, this X card has this x amount of FPS in this x game while another has a certain amount more, etc.

GPUBoss GPU Difference Section or Actual GPU Stats Comparison

In this section, GPUBoss is comparing actual graphics cards stats. Sometimes GPU performance stats translate poorly into real-time benchmarks, but people love to see those numbers and comparing them with each other. Anyway, this section is actually helpful and straightforward.

Stats GPUBoss is comparing in this section are: rendered output processors, shading units, mapping units, clock speed, manhattan score, memory, effective memory clock speed, turbo clock speed, and TDP or thermal design power.

Gpuboss Real-World Test Benchmarks AKA Most Helpful Section on the Gpuboss Website

This is the most helpful section that can actually tell the user a performance difference between graphics cards and educate the potential buyer in the right direction. This section should be first you see when comparing graphics cards, but unfortunately, it is the last.

This is the data user relly most on and contains actual facts and not some silly scoring system that is suitable for movie ratings or something. On the other hand, this section is unfortunately covered mostly with synthetic tests.

The Conclusion Is GPUBoss Is a Horrible User Experience Overall.

The whole website looks outdated, with too many ads and with so many issues, It just can’t be credible.

At the time reviewing this site, there were six ads on only one page. Furthermore, the whole thing looks like it’s one big ad for Amazon, and even Amazon links don’t work correctly. Links are linking to some laptops instead of graphics cards.

This site feels really neglected because there is no data for the new Nvidia RTX series and some information is missing ads are all over the place. Too bad because the website is still somewhat trendy and could be useful with some small corrections.

The Irony in Decimal Rating

They are using some decimal rating system from 00.0 to 10.0 to appear more credible witch is such an irony since they have wrong data for so many comparisons. I get it they don’t update this site anymore.

Conclusion

This site exists because someone is still milking passive income from Adsense and Amazon affiliate links, which makes this site poor user experience. Just don’t trust data on GPUBoss because it’s inaccurate, and it may lead to wrong decisions.

Photo by Fritzchens Fritz

Previous

If You Are Not Sure What to Look for in Gaming Mouse, Read This

Everything You Need to Know About Monitor UFO Test

Next

Leave a Comment