3dMark looks terrible.

[derncricket]

Whenever I run any 3dMark benchmark the resolution looks terrible. It's not defined at all, very pixelated. Whether it be TimeSpy or Firestrike. It looks like I'm at best running at 720p. The raytracing one. All of them. Looks smudged and pathetic.

I'm on a Ryzen 5900x 12-core
AMD RX 6900XT
32GB of ram
And a 512GB SSD.
Running at 3840x2160

It's been this way as long as I can remember. Different drivers. Even a different graphics card. Same result.

 

[neeyik]

Without knowing exactly what you're looking at, it's hard to tell if you have a genuine issue with the software or you're simply just observing how it actually is. Here's two screenshots from the Time Spy and Port Royal standard tests, for comparison:




The default settings are at a resolution of 1440p for those tests but the 3DMark designers do have a certain fondness for going a little heavy handed with the post processing effects, resulting in some wonky DoF and smudginess.

 

[derncricket]

Techspot is erroring when I try to post my pics. I guess you're right. I just find it strange that the graphics would be so jaggy and blurry. Maybe it's a busy scene. But why do that to your benchmark?

 

[Kshipper]

The benchmark is not optimized to look pretty or run smoothly. it is meant to be punishing on your system and especially on graphics cards requiring brute force to get good results. It simulates the type of graphical loads we encounter in typical games. It has been used for many years now as a legit benchmark. Higher 3DMark scores are always going to be better for real-world game performance.

 

[neeyik]

The benchmark is not optimized to look pretty or run smoothly. it is meant to be punishing on your system and especially on graphics cards requiring brute force to get good results. It simulates the type of graphical loads we encounter in typical games. It has been used for many years now as a legit benchmark. Higher 3DMark scores are always going to be better for real-world game performance.

The best looking graphics usually induce the highest GPU loads, though. They could have easily made it look like the likes of Crysis, RDR2, Metro Exodus; unfortunately they really like upping the shader load through post-processing, rather than lighting and texturing. The latter is understandable as the benchmark needs to be a reasonable size download, not 100+ GB in size. The former is something that can be worked on and I’ve wanted 3DMark to better reflect lighting models used in actual games for years now.

 

[Kshipper]

I won't argue that it looks terrible and is not a good way to show off your system's graphics. I will only argue that the benchmark still accurately pegs a system's graphical prowess. I will often use 3DMark and Passmark's test to quickly figure out if a customer's rig has the chops to work well enough to game before they take it home. I notice that the scores totally and accurately reflect real-world performance accurately, even though this benchmark software is ancient.

Good point about the download size of the benchmark. That totally makes sense.

The Passmark software is super small and installs and runs fast but it is heavily weighted towards a good graphics card. Without a good graphics card, it will score your machine poorly.