GPU Review Test System 2025.2 Games & Planning

[W1zzard]

Hardware

No changes in hardware planned, people seem happy with my choices

Just newest OS patches, new BIOS, new chipset drivers, and of course newest GPU drivers

Games

- See screenshot above for changes/new titles
- Couldn't find a game that should get kicked out for Civ7
- DOOM Eternal will be replaced by DOOM Dark Ages next round
- Like a Dragon Pirate kinda flopped, so giving KCD2 its spot, also CryEngine, which makes it an interesting addition
- Rise of the Ronin only 2.3k reviews, mixed
- No plans for mixing in runs with DLSS/FSR just yet

 

[adilazimdegilx]

Last of Us 2 to replace 1?
And maybe let go of Elden Ring too. Pretty much everyone plays it on locked 60FPS and even midrange cards can do that easily now so modding to unlock FPS just to test kinda pointless I think. I know Nightreign is coming but network tests kinda showed that it's same as ER in terms of looks and consoles still runs it perfect (and its coop anyway). I dont think missing Souls Engine in the list is that important until they release something next gen, and demanding (or at least not locked to 60FPS). Maybe you can put CIV 7 to its place. Although that is also a game where nobody really cares about FPS so I dont know. Maybe GTA 5 Enhanced?

 

[Dragokar]

KCD2 is welcome.

 

[Assimilator]

Dump Last of Us, Monster Hunter Wilds and Spiderman 2. They're all horribly optimised console port rubbish that aren't in any way representative of average performance of PC games, and are thus going to unfairly bias results. Yeah they're "AAA" titles and people will whine they're not present, but those people can be ignored.

 

[Hyderz]

F1 2025 i think is being revealed soon and set to release around may if you want to replace the 2024 ver

 

[Macro Device]

- Couldn't find a game that should get kicked out for Civ7

Maybe Witcher? CP77 is essentially Witcher on steroids and anyone who can afford 200 dollar GPUs can play Witcher 3 on stupid settings at high framerates anyway. That's why I'm also not sure one should include DOOM games because they run at 100+ FPS even on ancient hardware.

 

[Chrispy_]

Can I please request that CP2077 testing on the new set of results moves to the RT:Overdrive preset?

It's an old game at this point and there are so many other games that have hybrid raster/raytracing mixes that CP2077's inclusion for hybrid RT performance is becoming just another data point in a sea of similar data points.

On the other hand, its pure path-tracing mode is still quite rare among games and far more interesting to me, maybe to others as well?

 

[Event Horizon]

Looks good to me. I'm fine with the list as long as there's at least one game representing the UE4, UE5, Source, Creation, and Unity engines. The id Tech and Divinity engines are appreciated as well.

 

[TumbleGeorge]

When ZEN 6 arrives next year, the hardware change will occur naturally.

 

[alwayssts]

I would ditch Witcher and Elden Ring. Logged on to say that; saw others already did. GJ, guys.

Also, thank you for getting rid of Silent Hill. Suite looks pretty good.
I still kind of miss Control, Plague Tale, and Collisto Protocol; those are really great benches imo and I don't know if they've outlived usefulness.

Also not a fan of counterstrike being in the averages...I've said that already. It throws everything off badly. If you keep it, please make it own section for esports or something. Perhaps same w/ Doom.
Ofc course I don't know how you'll test the new Doom. If it'll be in absurd framerate mode or all-bells/whistles RT etc mode. If the later, it'll probably be okay in the regular list imo.
Maybe add Fortnite and such; maybe Hunt:Showdown and whatever else that scene enjoys. I bet that would make a lot of people happy that are FPS chasers in those games...but don't include them in averages.

Please ignore the guy whom said to remove literally all the pertinent games, because it sucks when that actually happens (because I like your layout but they need to be relevant games), and it sometimes does.
I can't quote him because I have him on ignore bc...well...he says ridiculous things like that.

Glad to see you add SM2, so I can stop looking back to that page and try to figure everything out (given it scales very well in raster and rt).
I will very much appreciate when raster/RT mins are given for new cards in that game. I also think the others are all GREAT additions (Indy, AC. MHW). Every one of those I will massively appreciate being there.

Also, if you're not going to add up-scaling w/ mins across resolutions/methods (which I truly know is a lot of work...but I don't know what to say other than we need it), please add minimums for ray-tracing.

Not having those feels a lot like obfuscation and borderline playing into a certain company's narrative (of high averages and bad lows that people often don't see). I also hate those average up/down graphs.
I can tell you (after talking to many) the 100% positive truth that is people do not understand those in the scope of everything else about cards comparatively, and those graphs give the averages too much weight.

I honestly would prefer the up-scaling/rt 1% and .1% (or minimum) lows over testing more games.
Maybe you could even lose a couple more older titles that most-anything can play at higher than their intended resolution in newer titles.
This is what many other reviewers have done, and I think it's for the best.

I know you pride yourself on having lots of info, and you should take pride in it (I very much appreciate it, even when I sound like I'm complaining), but we need more quality information, not just more information.

If I could ask one more thing, any way you could have a page where which model is used shown for stereotyped testing on the chart?

EX: What does 5080 on the chart mean? A Founders Edition? What about a 5070Ti where there is no FE? I ask so can check clocks of that review to understand real performance (I am in-fact that particular).
Perhaps the model could be added to the TEST SYSTEM page with a hyperlink to each review of that particular card. EX: I want to know if it's a 7900xtx if it's running ~2631 in raster, ~2397 in RT, or WAYY higher.

I think some of these changes would help a great deal, and I appreciate it. As I say, I think 1080->1440p/4k and 1440p->4k (with 1%/.1% and/or minimums) are worth having even if testing 1/2 the amount games. I guess I should refer to them as 1440 'quality' and 4k 'quality' and "performance" up-scaling, if not also 'balanced'. Many cards are, and will be, targeted towards those markets over the next year or so. This will depend upon if using raster, RT, or PT...and people need to know if they can hit/maintain acceptable frames while using those technologies (which often require up-scaling of that exact type).

As long as those games are somewhat current and/or popular and/or particular examples (like Snow Drop/UE5/etc) and don't already hit absurd FR (and/or their cap).
(except in the case of esports titles, ofc).

Thanks for the changes, thanks for hearing me out, and thanks for taking other suggestions from the community. I appreciate all your work (and I know you do a lot).

 

[80-watt Hamster]

Maybe Witcher? CP77 is essentially Witcher on steroids and anyone who can afford 200 dollar GPUs can play Witcher 3 on stupid settings at high framerates anyway. That's why I'm also not sure one should include DOOM games because they run at 100+ FPS even on ancient hardware.

DOOM is valuable, I feel, because it scales so well. It's kind of like: "Here's what your shaders can really do when they're not being held back by the engine."

 

[alwayssts]

DOOM is valuable, I feel, because it scales so well. It's kind of like: "Here's what your shaders can really do when they're not being held back by the engine."

That's fair. I think it's interesting on that level as well, and ofc some people use the same arguement for CS.
The thing is...is it really a better representation of important information where instead in another game two competing cards are measured and one achieves 50fps and the other over 60? I don't think it is.
I think people want to know if they can play that game and/or use that engine at their intended resolution/setting and be happy, which is not a question for Doom. It almost never was, and certainly not anymore.

 

[W1zzard]

is it really a better representation of

Just to provide a different perspective: more people are playing CS right now than all the other 24 games in my bench combined, by a pretty huge margin

EX: What does 5080 on the chart mean? A Founders Edition? What about a 5070Ti where there is no FE? I ask so can check clocks of that review to understand real performance (I am in-fact that particular).

FE/Reference design where it exists, otherwise the closest possible representation

Just checked, this has been written in every GPU review since 2022: "All cards used for comparison are reference designs. When a reference design does not exist, we go the extra mile and buy the closest possible match, using reference clocks and default power limit."

 

[Denver]

KCD2 is very welcome.

- CS2, for being light and producing abnormally high framerates compared to modern AAA games, should be removed or not computed when calculating the average. In fact, it could have a section dedicated to popular E-Sports.
- Indiana Jones is too buggy to be part of this, despite the positive reviews the game isn't really a success... well short of that.

 

[_roman_]

Please dump counterstrike 2. That high 600 FPS value is nonsense in comparison with usual frame rates of 60-120 fps.

 

[nomdeplume]

Yakuza LaD IW → Yakuza LaD Pirate in Hawaii?

 

[80-watt Hamster]

That's fair. I think it's interesting on that level as well, and ofc some people use the same arguement for CS.
The thing is...is it really a better representation of important information where instead in another game two competing cards are measured and one achieves 50fps and the other over 60? I don't think it is.
I think people want to know if they can play that game and/or use that engine at their intended resolution/setting and be happy, which is not a question for Doom. It almost never was, and certainly not anymore.

A good test battery presents the specimen with a wide variety of conditions. Focusing too heavily on the card-killers may deny us data on what happens at high-load+high-refresh. This is admittedly not a very good parallel, but how many vehicles are subjected to temperatures of -40 or below in service? Not that many, but I'm sure glad they're tested that cold.

 

[W1zzard]

Last of Us 2

F1 2025

the new Doom

Those are too far out, maybe I'll be able to sneak in Atomfall next week, but can't even wait for TLOU2, because 5060 Series reviews are not that far away I hear. Those titles will be included in the next round of retesting, after Computex

 

[Assimilator]

Yakuza LaD IW → Yakuza LaD Pirate in Hawaii?

- Like a Dragon Pirate kinda flopped, so giving KCD2 its spot, also CryEngine, which makes it an interesting addition

Reading. Still preventing you from seeming like an idiot in 2025.

 

[Onasi]

Leaving Civ 7 off is probably fine. It’s not a game that’s reliant on high framerate. I suppose late-game turn times can be interesting in the context of the CPU bench, though.
Not sure how I feel about Wilds replacing RE4. I understand the logic (newest game on the engine), but it’s extremely poorly performing and, just like DD2, kind of confirms that the REngine just isn’t suited to open-world games and completely shits the bed.
Would echo the sentiment that Elden Ring can go - just isn’t much point in keeping it. Though, I suppose, it did provide some weird outliers in the past (like with the 7600XT), so YMMV on that one.

Edit: Oh, just as an aside - are the tweaks you provide in the GPU Bench thread for Windows still current to this version, no notable changes?

 

[nomdeplume]

Reading. Still preventing you from seeming like an idiot in 2025.

Pretty weak troll even within a vastly marginal repertoire of insulting predispositions.

Since a question was raised. This and the previous thread had a directly stated purpose of choosing unique game engines. If one felt so inclined they might've taken into consideration what context the post was exploring. Why the Dragon Engine was no longer interesting. Thank you for allowing a chance to fully expand on the intended meaning.

 

[W1zzard]

are the tweaks you provide in the GPU Bench thread for Windows still current to this version, no notable changes?

You mean my OS install script? I use it on every fresh install as it gives me a workable baseline without all the Microsoft fluff and junk

 

[blinnbanir]

City Skylines 2 is the most PC intensive Game a lot of people play. It will use every single CPU core and big populations will lead to up to 85% CPU usage. That sounds like a great CPU test. I am sure someone can send you a late Game save to use as well.

 

[bajito93]

You mean my OS install script? I use it on every fresh install as it gives me a workable baseline without all the Microsoft fluff and junk

Hi, where can I find this script? Thank you

 

[W1zzard]

Hi, where can I find this script? Thank you

Thread 'Windows 11 Tweaks for GPU Benchmark'

Oct 5, 2021

Updated for 24H2

• Install without Internet
• When it prompts you to go online, press Shift+F10 and type "OOBE\BYPASSNRO" (that's an o not a zero at the end). After the automatic reboot you can install without network
• Install on systems without TPM, UEFI or other requirements: https://www.techpowerup.com/287584/windows-11-tpm-requirement-bypass-it-in-5-minutes
• Name the user "TPU" (this specific user will be deleted at the end of the scripts and you'll be using "Administrator")
• Install VGA driver, from USB, still without network, so Windows Update won't install a...

• W1zzard
• Replies: 62
• Forum: Benchmarking

• 
• 

Google: Windows 11 tweaks gpu
will give it as one of the first results, so no need to bring the link to a new os install