3DMark Will Now Estimate Your GPU’s Performance in Popular Titles

[Tsing]

Image: 3DMark



3DMark’s tests are used by gamers all around the world to determine how their graphics cards might fare in their favorite titles, but wouldn’t it be neat if those Fire Strike and Time Spy scores could immediately be translated to real-time game performance? UL, the team behind 3DMark, has brought that brilliant idea to life with its latest update, which introduces an estimated game performance panel to benchmark results.



The new panel comprises two drop-down categories that allow users to quickly determine what kind of frame rates their GPUs are capable of delivering in five popular games at either 1080p Ultra or 1440p Ultra settings. These titles are Apex Legends, Battlefield V, Fortnite, GTA V, and Red Dead Redemption 2.



“This new game performance...

Continue reading...

 

[Zarathustra]

Interesting.

Will be interesting to see how accurate it is compared to real world benchmarking.

 

[Grimlakin]

Would be a good comparison article to see how accurate it is.

 

[AntiQuark]

Dangerous unless they keep up with OS, game and gfx card updates, and reflect correctly users on different versions of any.
ie it could see the reputation of this feature drop like a stone if it doesnt reflect different versions of all the above correctly.
If they can instead give a % or fps error expected, that could work.
Will be really cool if they pull it off.

 

[Brian_B]

How is it anything other than just rehashing what every other review site does already? They are just putting the interface in their application.

 

[Dan_D]

This is actually a pretty cool feature if it works. I've always said that the problem with 3D Mark is that it's completely arbitrary. It doesn't use a game engine for its own testing and the numbers it generates aren't necessarily influenced by the same variables as games are. We have never been able to say: X system gets 11,420 3DMarks and therefore will get 90FPS in Ghost Recon Breakpoint at 4K or whatever. Often CPU changes or other system variables have impacted the 3D Mark score, but that effect has been disproportionate to what you get in actual games making such a change. This has made the software even less useful.

Even arguments for it being a good stability test really don't hold up as there are games or other software packages out there that do a far better job of stability testing. ASUS' Realbench being one example of that.

The only thing 3D Mark has been good for up until now has been comparing the score with other, similar systems. But again, it's almost meaningless beyond bragging rights. One could use it to determine whether their system was running right or not compared to others, but even that was hard to do given how vastly different scores can be even among similar configurations.

How is it anything other than just rehashing what every other review site does already? They are just putting the interface in their application.

I fail to see the connection. Reviewers test and benchmark games and provide the results. They do not correlate with 3DMark 10 as its entirely separate from the rest of the testing software. You cannot look at a 3DMark score and translate that into anything yourself. 3DMark doesn't respond to CPU clocks, RAM increases, or GPU overclocking necessarily the same way games do. It's numbers are arbitrary. How they'll translate it into something and how accurate it will be remains to be seen, but it is a cool feature. Especially for people who have no idea whether or not they'll be able to play a new game on older hardware. They'll be able to download this and estimate the game's performance. If it works.

 

[Brian_B]

We have never been able to say: X system gets 11,420 3DMarks and therefore will get 90FPS in Ghost Recon Breakpoint at 4K or whatever

I fail to see the connection

I guess my point is I don't see how it will be anything other than what your first quote implies. And that makes me wonder how it will be a valid tool at all, which was where I was coming from.

Maybe your right, and it will end up being something more than just a good guess based on 3DMark score, but somehow I doubt it will be any more sophisticated than that.

 

[Dan_D]

I guess my point is I don't see how it will be anything other than what your first quote implies. And that makes me wonder how it will be a valid tool at all, which was where I was coming from.

Maybe your right, and it will end up being something more than just a good guess based on 3DMark score, but somehow I doubt it will be any more sophisticated than that.

Well, what I mean is that right now 3D Mark isn't a predictor of anything because its totally isolated. The numbers are arbitrary and don't coincide with anything else. As I said, variables that impact 3DMark scores do not have the same impact on games. For example: I've gathered data for Ghost Recon Breakpoint at 4K for the 3900X, 9900K, and the 10900K. These scores are all within about 3FPS from worst to best. This is due to being GPU limited. However, these systems score quite a bit further apart in 3DMark 10. Thus, you can't draw a real world conclusion off 3DMark 10.

At best, you could see what a given system with similar 3D Mark scores gets in a given game. However, that means finding a review or posted score for that system. The only way that would work would be to try and manually collect data from different reviews and make estimates yourself. However, this isn't necessarily all that useful given the configurations used. That is, you'd need identical hardware and software configurations for this to be meaningful.

It's too much work and still not directly comparable. 3DMark being able to do this for you is potentially useful. We'll also be able to figure out how accurate 3DMark is at this very easily. If it works, the tool will be able to tell someone if their machine is fast enough for XYZ game at X resolution. That can quickly answer the question: "Do I need to upgrade for XYZ game?"