Gamers Are Wrong About 1440p vs 1080p CPU Benchmarking

[Steve]

Here We Go Again, or: Why GPU Bottlenecks Hide True CPU Performance.
Gamers keep demanding 1440p CPU benchmarks, but higher resolutions often hide the very performance differences CPU testing is supposed to reveal. Here's why 1080p still matters most.

Read the full article here.

Gamers Are Wrong About 1440p vs 1080p CPU Benchmarking

 

[gamerk2]

Once again guys: The point of low-resolution testing is to show which CPU is more powerful, so you get an idea how well each one will age. Testing at higher resolutions results in GPU bottlenecks that compress the results (see the first chart, where at 1440p the 3800x is equally as powerful as the 9800X3D, and therefore the superior purchase since it's several hundred dollars cheaper).

But hey, test at a high enough resolution, and I can "prove" a Pentium as just as fast as any of these newfangled multi-core CPUs.

 

[bviktor]

I never cared for 1440p. Never. Fail to see any point. It's a barely noticeable difference from 1080p.

2160p is OTOH immediately noticeable. And it's just above the limits of the human eye.

In other words, there's no point for anything below OR above 2160p. Let go of 1440p already. It was a band aid for weak GPUs, a temporary measure, like 720p. And we dumped 720p pretty fast, but 1440p somehow stuck. It's a bad excuse for poor programming. Fix 2160p performance already.

The first GPU that had a label on its box saying "4K ready" was the 980 Ti. That was 11 years ago. And we're still fooling around with temporary band aids like 1440p.

 

[Rapob]

I still see people who uses 4K/UHD claim that 1080p and 1440p is useless testing, while they use DLSS Performance mode (1080p internally) and gets the same CPU hit as if it was 1080p.

For CPU benchmarks, I only look at 1080p and lower testing. You want the CPU to be the actual bottleneck or entire test is pointless. You will see the same bottleneck when targetting high fps with a higher res anyway.

I never cared for 1440p. Never. Fail to see any point. It's a barely noticeable difference from 1080p.

2160p is OTOH immediately noticeable. And it's just above the limits of the human eye.

What? You must be blind, 1440p looks much better and much sharper than 1080p with like 80% more pixels. There is a night and day difference between 1080p and 1440p. Anyone with 20/20 vision will notice the difference instantly.

1440p with DLAA or FSR Native easily looks better than 4K/UHD with DLSS/FSR and pretty much all 4K gamers use upscaling since their hardware is too weak.

And this is also the reason 95% of gamers still use 1440p or lower. 1440p is sweet spot in terms of IQ and performance. 4K+ needs upscaling for most people, unless they accept crappy fps or use latency heavy FG which will never fix bad performance to begin with. End-result will be poor.

Even 5090 can't deliver good performance in many demanding games at native 4K.

980 Ti never was a 4K GPU by any means. Barely did 1440p well a few years after release, 6GB VRAM for 4K? Stop joking please. It did output 4K which is what they meant. PS4 Pro was also a 4K console. Did it output native 4K? No. Upscaling was used.

I had 980 Ti, I had 1080 Ti, I had 3080 and now uses a 4090. I have tried 4K both LCD and OLED many times but it never made much sense over 1440p at higher refresh rate. Smoothness is king and with features like DLAA, 1440p can look just as good as native 4K in most scenarios. DLDSR is also another option, that will make even 1080p look vastly better. At 4K you have no chance of running DLAA pretty much, you are forced to upscale, unless you accept input latency due to FG/MFG.

Foolish to claim 1440p is a temporary band aid. Has literally been the sweet spot for like 10 years and still is. Mostly casual gamers have been moving to 4K, you people playing AAA games with a goal of 60 fps. Serious (high fps) gamers stayed at 1080p and 1440p, most highly prefer 1440p over 1080p these days, by far actually.

Steam HW Survey, 95% of gamers use 1440p and lower.

Probably 95% using a 4K monitor runs 1080p internally due to upscaling as well. DLSS and FSR Performance mode is very popular here.

4K looks good but GPUs are still far too slow unless you use upscaling and frame gen, which gives other issues. 4K Upscaled don't look close to 4K Native. It is a compromise, due to weak hardware. Might not even look better than native 1440p and when you slap DLAA on 1440p, it for sure will look vastly better.

 

[Endymio]

...higher resolutions often hide the very performance differences CPU testing is supposed to reveal.

Your own article demonstrates this is false: the 'complete picture' as you say was visible only when 1080p results were contrasted against higher resolutions.

Let's put the question as starkly as possible. Is it worth paying double for a CPU that gives you 75% more performance than a cheaper model? Now rephrase it: is it worth paying double if it gives 75% more at 1080p but only 2% more at 4K ... and you're always at a 4K resolution?

Testing at higher resolutions results in GPU bottlenecks that compress the results

Sure. And if that compression was uniformly linear across all results, we could simply assume it and your conclusion would be correct. But it's nowhere near. It's linear up to the point where GPU bottlenecking occurs, and then a faster CPU ceases to matter. Where exactly is that point? Without testing at higher resolutions, you haven't a clue.

 

[jonny888]

I will take empirical proof that there's no difference at higher resolutions over "it didn't matter before so it will never matter" assumptions. But then I'm not the one putting in the work to do the benchmarks, so that's easy for me to say.

 

[llamapanda]

Great article, but it probably still won't be enough to convince the die-hards. I'm glad I understand why it's done the way it's done but a lot of the detail bypasses me: I'm still an old dog who insists on playing at native resolution, and use poor performance in worthy games to justify hardware upgrades.

BTW your paragraph referencing the good old days of Quake 3 is doubled up through a mis-edit

 

[user556]

The complains are because AMD parts are ahead. The moment Intel gets ahead those doing the complaining will vanish.

AKA, Sore loser syndrome.

Same happened with Cinebench results. Those complaints vanished when Intel introduced the mass of E-cores to pull it through on that benchmark. Although, now that the 16-core X3D2 has just pipped the lead we might start seeing more complaints about testing there too.

 

[gamerk2]

Let's put the question as starkly as possible. Is it worth paying double for a CPU that gives you 75% more performance than a cheaper model? Now rephrase it: is it worth paying double if it gives 75% more at 1080p but only 2% more at 4K ... and you're always at a 4K resolution?

Until you upgrade your GPU, and the bottleneck is reduced, and now you are held back by the CPU. That's the point.

 

[Sherwoodnt]

Seems the disconnect is that many want an at-a-glance snapshot of how a CPU will perform on their unique hardware and preferred settings, while CPU benchmarks can only really compare CPU's to one another.

Not that it would be feasible, but would trying to indicate the tier of GPU that bottlenecks each CPU at each resolution of each test game provide the info that many are looking for?

If nothing else, this topic is good for clicks!

 

[Rapob]

This article comes to mind:

Is the Ryzen 9800X3D Truly Faster for Real-World 4K Gaming?

Misconceptions in tech are endless, but myths about low-res CPU benchmarks top our list. Let's revisit this hot topic and test the new 9800X3D at 4K to...

www.techspot.com

Shows that most and almost all "4K gamers" uses upscaling and that CPU matters even in 4K when upscaling is used, as CPU is hit harder. And water is wet.

 

[redgarl]

With the "GREAT" gift of UPSCALING by Nvidia (SARCASM)...

People are not aware of rendering resolution close or below 1080p... so 1080p benchmarks actually do matter more than before, and it is sad...

 

[redgarl]

This article comes to mind:

Is the Ryzen 9800X3D Truly Faster for Real-World 4K Gaming?

Misconceptions in tech are endless, but myths about low-res CPU benchmarks top our list. Let's revisit this hot topic and test the new 9800X3D at 4K to...

www.techspot.com

Shows that most and almost all "4K gamers" uses upscaling and that CPU matters even in 4K when upscaling is used, as CPU is hit harder. And water is wet.

Exactly what I just posted. Games are so unoptimized now that using Upscaling or Dynamic Resolution is sometime the only way to get decent performances at 1440p or 2160p, but this come at a cost of rendering the games at lower resolutions which is bringing your system into a CPU bottleneck territory.

Nvidia created that mess and now devs are using upscaling as an excuse to cut corners on optimization. It is such a sad sight that I would simply recommend anyone to buy a PS5 Pro over a mid-range PC.

 

[Endymio]

Until you upgrade your GPU, and the bottleneck is reduced, and now you are held back by the CPU. That's the point.

Um, that's why CPU tests are done with the fastest available card at the time; upgrading your GPU isn't possible.

And yes, that may change in a few years time. But most buyers are far more concerned with what performance they're buying now, rather than some hypothetical future performance ... not to mention the fact that, by the time faster GPUs are available, newer games will be released, even more depending on GPU power.

 

[gamerk2]

Um, that's why CPU tests are done with the fastest available card at the time; upgrading your GPU isn't possible.

And yes, that may change in a few years time. But most buyers are far more concerned with what performance they're buying now, rather than some hypothetical future performance ... not to mention the fact that, by the time faster GPUs are available, newer games will be released, even more depending on GPU power.

But the point remains that *only* low-resolution testing shows how well CPUs will hold up years down the line.

Taking your example: Yes, GPUs get more powerful and games will continue to use them. But in your case you have a weaker CPU (because they "performed the same at 4k"), which risks becoming the bottleneck. And that's the point.

 

[Rapob]

Exactly what I just posted. Games are so unoptimized now that using Upscaling or Dynamic Resolution is sometime the only way to get decent performances at 1440p or 2160p, but this come at a cost of rendering the games at lower resolutions which is bringing your system into a CPU bottleneck territory.

Nvidia created that mess and now devs are using upscaling as an excuse to cut corners on optimization. It is such a sad sight that I would simply recommend anyone to buy a PS5 Pro over a mid-range PC.

Nvidia started but everyone literally uses upscaling now. AMD became relevant again with FSR 4. Sony uses PSSR on PS5 Pro, will soon get a watered down FSR 4 solution too (RDNA 2.x don't fully support FSR 4) and even Switch has DLSS enabled on multiple games.

FSR literally saved the handheld gaming market too. Upscaling is not going anywhere from here, this is the reason why getting a good CPU for a (good) gaming PC is more relevant than ever. High minimum 1% lows is king and CPU is responsible for this in most cases.

Barely any socalled 4K gamers actually plays at native 4K. There is simply no point and hardware is too weak in general. 99% of gamers don't own RTX 4090/5090 but even these can get in trouble at native 4K in plenty of games.

Most people like upscaling but even if you don't like it, you should not sleep on DLAA and FSR Native that will for sure make native look bad. Even DLSS/FSR can make native look bad in many cases, due to native AA and sharpening being applied. End-result is what matters for people. Much better performance and still good graphics, most people will use it.

Most people that hate upscaling, is people that can't use DLSS 4 and FSR 4. These options are very good.

RDNA 3.x and older, sadly can't use FSR 4 and FSR 3.x is far far worse. Having seen FSR 1, 2 and 3 is the reason most people "hate upscaling" these days. FSR 4 is vastly better and DLSS 4 is the king.

 

[dirtyferret]

If the ***** fan boys believed you Steve you would not have to post constant articles on how CPU work in gaming.

Much like our current state of politics, the cult wants you to prove what sounds right as opposed to what facts and testing display is right. So more alternative facts Steve and less actual results.

 

[Skully]

While I do understand the rational behind it, the problem I have with just the 1080p benchmarks is simply that it doesn't show other commonly used resolutions.

If I'm wondering if I should upgrade to this fancy new $500 CPU and I'm only seeing 1080p benchmarks, that tells me nothing. So, my CPU gives 275 fps at 1080p while this new one gives 300fps. How does that help me make an upgrade decision when I'm playing at 1440p or 4k? Sure, I can assume and maybe do some math, but I'd rather see it in black and white.

Wanting to see the full picture does not make gamers "wrong".

 

[Skully]

While I do understand the rational behind it, the problem I have with just the 1080p benchmarks is simply that it doesn't show other commonly used resolutions.

If I'm wondering if I should upgrade to this fancy new $500 CPU and I'm only seeing 1080p benchmarks, that tells me nothing. So, my CPU gives 275 fps at 1080p while this new one gives 300fps. How does that help me make an upgrade decision when I'm playing at 1440p or 4k? Sure, I can assume and maybe do some math, but I'd rather see it in black and white.

Wanting to see the full picture does not make gamers "wrong".

 

[redgarl]

Nvidia started but everyone literally uses upscaling now. AMD became relevant again with FSR 4. Sony uses PSSR on PS5 Pro, will soon get a watered down FSR 4 solution too (RDNA 2.x don't fully support FSR 4) and even Switch has DLSS enabled on multiple games.

FSR literally saved the handheld gaming market too. Upscaling is not going anywhere from here, this is the reason why getting a good CPU for a (good) gaming PC is more relevant than ever. High minimum 1% lows is king and CPU is responsible for this in most cases.

Barely any socalled 4K gamers actually plays at native 4K. There is simply no point and hardware is too weak in general. 99% of gamers don't own RTX 4090/5090 but even these can get in trouble at native 4K in plenty of games.

You talk like a real fanboy and don't get the point. You just had to lower your resolution, and most of the time, the result from the lower resolution will look better than the upscaling option. Upscaling didn't save the handheld market, Nintendo sold more Switch than Sony sold Playstations 4, so your argument is not serious. Nothing has changed, it is still pretty much accurate in 2026 and now upscaling is used as an excuse by devs to release broken products that are unoptimized mess.

The latest to the entry is Nioh 3, which look nice but nothing special, and should run at 4k 60 FPS on a 9070XT at max without any hiccup, instead it runs at 35-40 FPS, half of the performance it should run at.

 

[poshflamingos]

This article is TechSpot's version of The Onion's "'No way to prevent this', says only nation where this regularly happens". And I fully support it, I hope Steve redoes this article as many times are necessary until the problem is fixed.

What I don't understand is why the BF6 DLSS result is so weird. 1440p DLSS Q is very close in internal resolution to 1080p, and all other tests show exactly that, they're very close with 1080p being either slightly faster or slightly slower. BF6 is the only one where 1080p is significantly faster than 1440p DLSS. Why is the DLSS overhead so much bigger in that game?

 

[Endymio]

But the point remains that *only* low-resolution testing shows how well CPUs will hold up years down the line.

But it doesn't do that at all. "Performance" isn't a single monolithic value, though these tests often represent it as such. It's quite possible that a CPU which tests faster today will -- on a game released five years from now, using a larger number of cores, with a different compiler and optimized for a different instruction set mix, might actually perform worse.

Every new chip generation, both Intel and AMD optimize certain instructions, making them run in less cycles ... and very often do so by causing other instructions to run *slower* ... on the assumption that, at this particular moment in time, those instructions have become less commonly used. That's why a latest-generation mid-range chip might test slower on some particular10-year old software than a high-end chip from 2026 ... but faster on modern software. Betting on "future performance" more than a couple years ahead is risky business.

Again: if you're buying a chip **today** you're more interested on what it's giving you today ... not what it might hypothetically give you five years from now. Nor will the majority of those reading these reviews upgrade their GPU to 5090-class performance, even 4-5 years from now.

 

[poshflamingos]

Let's put the question as starkly as possible. Is it worth paying double for a CPU that gives you 75% more performance than a cheaper model? Now rephrase it: is it worth paying double if it gives 75% more at 1080p but only 2% more at 4K ... and you're always at a 4K resolution?

"Gives 75% more at 1080p but only 2% more at 4K" is a mischaracterization. Resolution doesn't affect CPU performance. If a CPU gives you 75% more performance at 1080p, then it gives 75% more performance at 4K too. The only reason you see 2% instead of 75% at 4K is because the test is GPU-limited, current GPUs aren't fast enough to expose the real performance difference. But future GPUs will be, and if in the future you upgrade your GPU you will "gain" more CPU performance (by removing that GPU limitation the you had previously). The potential to be 75% faster at 4K was always there.

If you want to see how CPU performance affects your specific GPU, that's understandable, but it's not feasible for tech reviewers to do. Testing 1440p/4K with a 5090 means jack squat to you unless you also have a 5090, because each GPU becomes GPU-limited at a different point. They'd have to test 1440p/4K with your specific GPU, but then that test is meaningless to everyone who has a different GPU than you. And they can't test every GPU there is for every CPU launch, they'd spend their whole lives doing nothing but benchmarking and still not get it done.

So if testing every CPU + GPU combination is not possible, the next best thing is to test with the best existing GPU at low resolutions (to show the CPU performance without being GPU limited) and let you figure out how it affects your particular GPU yourself (by looking at GPU tests that include your GPU on those same games, and cross-referencing the FPS values).

 

[Endymio]

"Gives 75% more at 1080p but only 2% more at 4K" is a mischaracterization. Resolution doesn't affect CPU performance. If a CPU gives you 75% more performance at 1080p, then it gives 75% more performance at 4K too.

Words fail me that someone could even express this seriously. CPU's don't operate in a vacuum -- "performance" is meaningful only as part of a complete system. My 5,000 watt three-phase power supply might have ten times the performance of a smaller PSU ... but thats meaningless if my system can't use it.

Many of these CPUs would show even larger performance deltas, if tested coupled with $25K workstation-class GPUs. Is that relevant? What's the "proper" performance increase? 200% with that GPU? 75% with a 5090? Or the 2% with a 5090, ran at the 4K resolution you typically use?

If you want to see how CPU performance affects your specific GPU, that's understandable, but it's not feasible for tech reviewers to do.

But it's certainly feasible to test at the most common resolution that people use today ... as Steve himself just proved with this article.

 

[poshflamingos]

Words fail me that someone could even express this seriously. CPU's don't operate in a vacuum -- "performance" is meaningful only as part of a complete system.

Except every system is different, systems are changeable/upgradeable, and new more powerful GPUs will exist in the future that you can upgrade your system to, genius.
The only plausible way to test CPUs that gives results that are useful for everyone is if you test at low resolutions to show the "uncapped" CPU performance.

But it's certainly feasible to test at the most common resolution that people use today ... as Steve himself just proved with this article.

The article proves the exact opposite of this. Testing at 1440p produces worse data, because you're obscuring the real CPU performance due to GPU bottlenecks.
This test done with a 5090 means absolutely nothing to everyone who doesn't also have a 5090, but also lowers the quality of the data that is useful to everyone (the uncapped CPU performance). Testing at 1440p is just a worse test overall, and testing both 1080p and 1440p doubles the workload for the reviewer while giving zero benefits to everyone who doesn't use the exact GPU the test used. Testing 1440p doesn't tell you anything that you don't already know from the 1080p test.
The only way testing CPUs makes sense is at low resolutions, to expose the real "umcapped" CPU performance. Then you can look at that data regardless of what GPU you own, and arrive at your own conclusions based on what GPU you have now or plan to upgrade to.