Gamers Are Wrong About 1440p vs 1080p CPU Benchmarking

[LimyG]

Huh? Are you maybe practicing English with that comment?

For someone who advocated just leaving if you struggle with a response you're given, you sure have a hard time adhering to that advice after issues in multiple of your comments were pointed out.

Just try to lighten up a bit. You take yourself far too seriously.

Watch that BP!!

 

[TurboN]

Just try to lighten up a bit. You take yourself far too seriously.

Watch that BP!!

Considering the visibly-tryhard nature of what appears to be a crash-out, is this your Alice in Wonderland impression, where you recite self-advice out loud?

I again recommend that you read my original post, which is #52 in this thread - since if you're going to base your replies on anything, it should be on the topic that you replied to, but which you oddly said you didn't bother reading despite replying to it.

 

[LimyG]

Nothing wrong with Alice in Wonderland.

I'll try and cheer you up a bit.

Everything I said is hideously ridiculous, 100% wrong. I'm crying now.

On the other hand, everything you say is golden. Perfection. 100% correct, and honestly beautifully written. It's factual, precise and everyone can benefit from your wonderul posts.

The future is bright for you, young man. Golden, and it's literally impossible for it to get any worse.

 

[TurboN]

Nothing wrong with Alice in Wonderland.

I'll try and cheer you up a bit.

Everything I said is hideously ridiculous, 100% wrong. I'm crying now.

On the other hand, everything you say is golden. Perfection. 100% correct, and honestly beautifully written. It's factual, precise and everyone can benefit from your wonderul posts.

The future is bright for you, young man. Golden, and it's literally impossible for it to get any worse.

You don't need to swing like a pendulum between extremes as you try to gaslight. There's no chance of that. But I accept that someone who isn't aware of what tryhard means, what it looks like, and when they're being it, could plausibly also be someone not keen enough to realize that.

Really good job of following your advice of just moving on when you don't like someone's response to you.

 

[poshflamingos]

Oops! That statement is true only if one assumes an infinitely powerful GPU. When you find out what vendor sells those, let me know.

That is just flat out wrong. You don't need an infinitely powerful GPU. If, as per the example that commenter gave, a CPU gives you 60 FPS at 1080p (meaning it also gives you 60 FPS and 1440p and 4K), any GPU that gives you 60 FPS or more at 1080p, 1440p and 4K respectively (which is perfectly reasonable on all 3 resolutions) already bumps against the CPU limit.

CPUs aren't infinitely powerful either. GPUs that are capable of maxing out a given CPU either already exist or will exist in the next few years. You don't need "infinitely powerful GPUs".

But they're not doing that at all. There's a certain degree of bottleneck even at 1080p, which is why these titles will all run faster even lower resolutions. If "testing only the CPU" was the sole consideration, why not test at 640p? Or better yet -- test against games that run entirely without video acceleration at all?

Steve addressed this, either in a previous article/video or in a reddit comment, I don't remember exactly. He said he did testing at 720p, like techpowerup does, but stopped doing it after he saw performance regressions compared to 1080p in some cases. According to him, flagship GPUs today are so large that they sometimes run into scheduling/saturation issues and can't keep all their SMs/CUs fed and utilized when the resolution is that low (or at least that's what he speculated to explain the performance regressions he saw). So to avoid these issues, he dropped very low resolutions below 1080p.

But yes, in theory, if those issues didn't exist, 720p testing would be even better than 1080p for CPU tests.

 

[Endymio]

That is just flat out wrong. You don't need an infinitely powerful GPU. If, as per the example that commenter gave, a CPU gives you 60 FPS at 1080p (meaning it also gives you 60 FPS and 1440p and 4K), any GPU that gives you 60 FPS or more at 1080p, 1440p and 4K respectively (which is perfectly reasonable on all 3 resolutions) already bumps against the CPU limit.

Eh? If a particular cpu/gpu combination can reach a certain frame rate at 60 fps, it in no way implies that it can reach that at "any resolution". That's true only under the assumption that the 1080p scenario wasn't even slightly GPU-limited (I.e. 100% CPU bound). Honestly, this is really basic math.

Steve addressed this, either in a previous article/video or in a reddit comment, I don't remember exactly... According to him, flagship GPUs today are so large that they sometimes run into scheduling/saturation issues

You've proven my point for me. There is no scenario that's perfectly 100% CPU or GPU bound: resolution always impacts the performance of both.

And you've also just debunked your own statement above: if a title running at 60fps at 1080p "regresses" to a slower speed at 720p, well then it can't possibly be true that this particular CPU can "maintain 60fps at all possible resolutions", now can it? Oops...

 

[LimyG]

You don't need to swing like a pendulum between extremes as you try to gaslight. There's no chance of that. But I accept that someone who isn't aware of what tryhard means, what it looks like, and when they're being it, could plausibly also be someone not keen enough to realize that.

Really good job of following your advice of just moving on when you don't like someone's response to you.

Thanks man. I appreciate it.
Just one thing. Please don't think I didn't like your responses. I loved them!!

I can't read #52 right now as I have belly ache from such great belly laughs.
I'll save it for when I feel a little down. That's pretty rare, but happens to the best of us, eh, mate!! Heh heh.

Thanks again for making my day. If you were a girl I would send you chocolates (you are a boy right?) I'm a bit old, but in good shape for a gen X.

That's all for today, and you have really made it even better than a normal day, (which on the whole are pretty good - Quite a few years back I learned not to take myself too seriously and, well, life is good - you sweetie pie make it all the better!!)

Love and Kisses xxxx

P.S. Do let me know if you are a girl, heh heh mmmm.

 

[poshflamingos]

Eh? If a particular cpu/gpu combination can reach a certain frame rate at 60 fps, it in no way implies that it can reach that at "any resolution". That's true only under the assumption that the 1080p scenario wasn't even slightly GPU-limited (I.e. 100% CPU bound). Honestly, this is really basic math.

I didn't say a CPU/GPU combo can reach that framerate at any resolution, I said CPUs specifically (not GPUs) reach the same framerate at any resolution. Because CPU performance is independent from resolution.
And yeah, testing at low resolution with the best available GPU is the reviewer's best way to get as close as possible to "not even slightly GPU-limited".

There is no scenario that's perfectly 100% CPU or GPU bound: resolution always impacts the performance of both.

First of all, that claim is categorically wrong. Resolution only impacts GPU performance, not CPU performance. CPU performance is completely unrelated to resolution, increasing resolution has zero impact on CPU performance. That's why a CPU that reaches 60 FPS at 1080p also reaches 60 FPS at 1440p and 4K.
Second, it doesn't have to be "perfectly" CPU-bound. Using the best GPU that exists at a low resolutions is the best available method to remove GPU limitations from games. Sure, an even faster GPU than the current flagship would be even better, but reviewers can't use hardware that doesn't exist yet.

And you've also just debunked your own statement above: if a title running at 60fps at 1080p "regresses" to a slower speed at 720p, well then it can't possibly be true that this particular CPU can "maintain 60fps at all possible resolutions", now can it? Oops...

You're conflating two completely different things. The performance regressions at 720p due to scheduling are exclusively a GPU issue, and only happens on very wide GPUs like today's flagships. It has no relationship whatsoever with CPU performance.
Also, buddy, you gotta stop with the "oops" thing, it's extremely cringe. Especially when you write it while still being wrong.

 

[Endymio]

I didn't say a CPU/GPU combo can reach that framerate at any resolution, I said CPUs specifically (not GPUs) reach the same framerate at any resolution. Because CPU performance is independent from resolution.

Again: learn math. Your statement is true only in the nonexistent case of a process being 100% cpu-bound, I.e. an infinitely fast GPU processing frames in exactly zero time.

Furthermore, your basic premise isn't even correct. In many engines, resolutions *does* affect CPU work per frame, in steps like distant object occlusion, for instance. Generally, a tiny factor, but factor it is.

Using the best GPU that exists at a low resolutions is the best available method to remove GPU limitations
from games.

No, the best (and only) way to do that is to test with games that don't require GPUs. Why not advocate for that?

Also, buddy, you gotta stop with the "oops" thing, it's extremely cringe. Especially when you write it while still being wrong.

Oops! The point stands: For anyone who bases purchase decisions around these reviews, the only sensible strategy are results at the resolutions you actually use.

Your premise boils down to this: "five years from now I might purchase a card faster than a 5090 Ti and pair that (by then) $5,000 videocard with my sadly aging CPU and finally see some actual benefit from the extra money I spent so long ago."

Sorry, that's not a rational approach.

 

[BallofGob]

I largely agree with the messaging of the article. But some of the benchmarks also highlights some of the really absurd comparisons that are repeatedly being presented in the the main outlets.

What is the purpose of running a super high end GPU like RTX 5090 (circa 2025) with a Ryzen 3800x? It was from 2019 and it was far from even being considered a ideal CPU for high end gaming builds at the time

I can't imagine a world someone who can afford the 5090 has a good reason for not being able to upgrade their CPU.

In order to save time the CPU comparisons with RTX 5090 should be mostly limited to CPU generations that support PCIE 5...that is an arbitrary cut-off but it makes the comparisons more relevant.

 

[amstech]

This article doesnt prove anything. It just shows that in this specific cherry picked self convincing scenario one of the CPUs isnt as good as a gamer as the other.
In 90% of situations, especially mine, a 7800X3D has a 12% or so advantage over my 265K at 1080p, but at 1440p this number drops to a virtually un-noticable 7%.
This is true for most comparisons and it bothers the AMD pundits alot because at 1440, and especially 4K their precious X3D chips don't have this big advantage anymore. In fact, in many cases Intel matches and on rare occasion, beats them.

 

[mathrawker]

"Claiming that 1080p CPU testing is useless simply because you personally game at 1440p or 4K demonstrates a misunderstanding..."
...and misunderstanding is all the kiddees do these days: Nobody cares what YOUR rig is. Go humblebrag about it in your sig. These reviews are generic for mass consumption.

 

[mathrawker]

Thanks man. I appreciate it.
Just one thing. Please don't think I didn't like your responses. I loved them!!

I can't read #52 right now as I have belly ache from such great belly laughs.
I'll save it for when I feel a little down. That's pretty rare, but happens to the best of us, eh, mate!! Heh heh.

Thanks again for making my day. If you were a girl I would send you chocolates (you are a boy right?) I'm a bit old, but in good shape for a gen X.

That's all for today, and you have really made it even better than a normal day, (which on the whole are pretty good - Quite a few years back I learned not to take myself too seriously and, well, life is good - you sweetie pie make it all the better!!)

Love and Kisses xxxx

P.S. Do let me know if you are a girl, heh heh mmmm.

OK, boomer Jesus fkn criminy. Give up.

 

[hahahanoobs]

Who wants a product tested under one condition? Use your brains, techspot.

 

[noblody]

Oops! That statement is true only if one assumes an infinitely powerful GPU. When you find out what vendor sells those, let me know.


But they're not doing that at all. There's a certain degree of bottleneck even at 1080p, which is why these titles will all run faster even lower resolutions. If "testing only the CPU" was the sole consideration, why not test at 640p? Or better yet -- test against games that run entirely without video acceleration at all?

It is a gross exaggeration to say that by running a game at 1080p you do not reduce the gpu bottleneck, surely the lowest influence is at the lowest resolution supported by the game. In the old days, anandtech ran tests at lower resolutions. I assume that for current games, 1080p is the lowest denominator across a large selection. Your suggestion of using games that run without video acceleration seems ironic, but the fact is games are tested that way internally. It is called headless testing and is used for a lot of things, including performance testing. The problem is that commercially available games do not expose this. There are synthetic tests like 3dmark time spy that returns a cpu score that is not influenced by the gpu. you can see in the list a very similar score obtained by the same cpu with 1080ti and 4090 or even 5090. the dominant cpus are multi core server cpus, so that shows pure synthetic tests are not that relevant for gaming workloads. this brings us back full circle to why they use games for testing gaming workloads and why they use low resolutions.

 

[poshflamingos]

Again: learn math. Your statement is true only in the nonexistent case of a process being 100% cpu-bound, I.e. an infinitely fast GPU processing frames in exactly zero time.

What? That is complete nonsense, where did you get this idea from? Are you somehow under the impression that the CPU and GPU do their respective workloads one at a time, taking turns? It seems you think that for frame 1, the CPU processes the CPU tasks while the GPU sits still doing nothing, then once the CPU is done the GPU starts processing the GPU portion of that frame's workload; then for frame 2 they repeat that process taking turns. Is that it?

Because that's not how it works. The CPU and GPU process their own tasks in parallel (with the CPU being at least one frame ahead). If you have a CPU that is capable of running the game at 80 FPS (when not GPU-bound), and a GPU that is capable of running the game at 60 FPS (when not CPU-bound), in reality you just get the smaller of those two numbers (60 FPS), with the CPU idling a portion of the time. Or if someone has a CPU that can do 50 FPS and a GPU that can do 70 FPS, you get 50 FPS, with the GPU idling a portion of the time. It's that simple.

That's why you don't an "infinitely powerful" GPU. They're not taking turns. The GPU does its job in parallel with the CPU. You only need the GPU to be able to produce a higher FPS "ceiling" than the CPU's "ceiling" to ensure you're not GPU-bound. In a game where the 5090 could theoretically get 400 FPS at 1080p (when not CPU-bound), that allows the 5090 to accurately test any CPU that produces less than 400 FPS in that game.

Furthermore, your basic premise isn't even correct. In many engines, resolutions *does* affect CPU work per frame, in steps like distant object occlusion, for instance. Generally, a tiny factor, but factor it is.

Sure, if you want to be pedantic, a small portion of game devs tie other aspects of the game, like LODs, to resolution. In that case resolution can indirectly affect CPU performance, not because of the resolution itself, but because in that case it forces higher LODs, which means more draw calls, which increases the load on the CPU a bit.

But that's a small minority of games, and it's not considered good practice. In the majority of games where LOD is not tied to resolution, the number of pixels on the screen has zero impact on CPU performance.

No, the best (and only) way to do that is to test with games that don't require GPUs. Why not advocate for that?

No, you can't do that, because issuing draw calls to the GPU and handling the driver overhead is part of the CPU workload of games. Even if it were feasible, without a GPU you wouldn't be measuring a real CPU workload of a modern game.

Oops! The point stands: For anyone who bases purchase decisions around these reviews

Nobody is saying that you need to buy CPU based only on those CPU tests, genius. Nobody is saying everyone must go buy a 9800X3D just because it "wins" the test. The goal of the test is to show what the uncapped performance of each CPU is, so that then you can compare it with GPU tests that are relevant to you, and see for yourself which CPUs are a good match and which are a waste for your situation.

The low resolution tests gives you the information that allows you to do what I described above. The high resolution tests don't give you any useful information, they're done with a GPU that is different from the one you have, and just shows a bunch of underutilized CPUs that are tied on the test because the test is GPU-bound. At best it's useless information, at worst it misleads people into making wrong conclusions, such as "CPUs lose performance at higher resolutions" or "these CPUs being tied in this test means that they will be tied forever".

Your premise boils down to this: "five years from now I might purchase a card faster than a 5090 Ti and pair that (by then) $5,000 videocard with my sadly aging CPU and finally see some actual benefit from the extra money I spent so long ago."

No, my premise is "I want to see the real uncapped CPU performance with zero GPU bottlenecks, because that's the only information that I need this test for and couldn't otherwise figure out myself."

 

[cfenton]

I think the problem is that people are mostly reading reviews for buying advice. They want to know if a CPU will give them enough of a performance increase to justify the cost. If you don't understand how CPU reviews work and what they are measuring, then they can be misleading. It looks like going from say a 7700X to a 9800X3D will give you a big performance upgrade, but that's only true if you pair it with a 5090. As soon as you go to even a 9070XT or 5080, the difference shrinks or disappears entirely in a lot of games. I bet there are a lot of people who have bought a 9800X3D for twice the price of a 7700X and paired it with a 4080, or worse, GPU because they thought it would give them the performance they saw in reviews.

I think, if you really only have time to do two tests, that showing how a CPU works at 1080p ultra with a mid-range GPU like the 5060 as well as the 5090 would be more helpful than showing how it works at medium and ultra on a single GPU. I understand that's not a pure CPU test, but it's useful information to people who want to know if they would get any benefit from buying a new CPU. Maybe it would be better as a separate follow-up article, but it's information people want and they clearly aren't willing to compare CPU and GPU reviews and crunch the numbers.

 

[PierreJG]

Wow, so much fighting over a damn topic. Next people going to tell me userbenchmark is real world performance. I am busy running test myself. I don't have a strong cpu, Only a Ryzen 5 5500 and an RX6600. I used a simple game. Grim Dawn. It is some of the same developers as Titan Quest and the engine does have a problem. But after I bought my 6600 in November. So start up Grim Dawn. Set to to low to high on testing as I had massive stutter in some areas. Some of it is the game engine but it was almost unplayable in some scenarios. I switched over to 1440p and only has the occasional game engine stutter and nothing else.

What this tells me is that it took pressure off the cpu to my gpu. A game like Grim Dawn should not lag on my pc, even in 1080p but it did and moving to 1440p solved the stuttering. I can understand the argument as 1440p is the new norm but it is not real world cpu performance. If you play a competitive game, hell even in dota 2 they play in 1080p where maximum frame rate is needed. You will not play a competitive game in 4k. I am happy with 60fps. I play marvel midnight suns on 30fps with raytracing on. My cpu is basically a cut down 5600 with halve the cache memory. I will rerun this test once I get my 5700x as this game is a semi gaming pc and mainly a work pc. Don't need an overpriced x3d cpu as the 5700x is at a sweet price at the moment.

Btw: My first cpu I overclocked was a 386 when I was like 5 or something. My father was strict. He bought all his old office pc's when they upgraded and told me you want a pc, build it yourself. I was freaking 5 years old lol. So put together a killer pc with 24mb memory but it was too new so it could only use 16mb. I saw on the motherboard dip switches and of course at the age of 5 I love switches. Then I notice wait my cpu mhz moved up from 33mhz to think it was 40 or more mhz can't remember. That is where cpu overclocking began but I build custom air coolers when I have spare time to see how far I can push a cpu on a custom air cooler. Sadly will buy a cooler for my 5700x. One pc that I clocked the snot out of was my old FX-4170 from 4.2 to 4.9ghz on air but guess what? It was still crap lol

 

[Endymio]

What? That is complete nonsense, where did you get this idea from? Are you somehow under the impression that the CPU and GPU do their respective workloads one at a time, taking turns? ...Because that's not how it works. The CPU and GPU process their own tasks in parallel

Are you being intentionally dense? First and most importantly, the CPU and GPU share other resources: bus, system memory, I/o calls from disk. At higher resolutions and larger textures, a GPU uses more of these, meaning there is less remaining for the CPU. Guess what happens then?

But to your direct point: while in theory the CPU and GPU run perfectly asynchronously, there is always direct synchronization overhead. Even at the lowest "cpu-bound" resolutions, the cpu spends a small portion of its time waiting on the gpu.

Sure, if you want to be pedantic, a small portion of game devs tie other aspects of the game, like LODs, to resolution. In that case resolution can indirectly affect CPU performance, not because of the resolution itself, but because in that case it forces higher LODs, which means more draw calls

Meaning I've been right all along. Glad we now finally agree.

But that's a small minority of games, and it's not considered good practice. In the majority of games where LOD is not tied to resolution

Err, what? Every major game engine uses this technique to some degree. Unreal, Valve's Source, Unity, etc. It's not "good practice" to perform it manually in the game itself because the engine handles it natively.

 

[poshflamingos]

Are you being intentionally dense? First and most importantly, the CPU and GPU share other resources: bus, system memory, I/o calls from disk.

None of this is correct.
CPU and GPUs do not share buses, they each have their own separate buses to their own memory pools. Consoles share buses because they have APUs with unified memory under a single bus. PCs, as you may have noticed, are not like that.
They do not share system memory either. While the GPU can access system RAM vie PCIe, doing that while rendering brings a huge performance penalty. In games, it keeps everything it needs in its own VRAM.
They don't share I/O calls either. I/O is handled exclusively by the CPU. Anything that goes from the SSD to the VRAM goes through the CPU/RAM first, then to the VRAM via PCIe. There are features like RTX IO that attempt to enable that "direct SSD to GPU reads" feature on PC, but no games have implemented it yet. Also, CPUs are not limited by I/O performance in games, that has never been a thing.

Meaning I've been right all along. Glad we now finally agree.
LOL, what? What? Every major game engine uses this technique to some degree. Unreal, Valve's Source, Unity, etc. It's not "good practice" to perform it manually in the game itself because the engine handles it natively.

Why do you even bother making blatantly incorrect claims like these, when it's so trivially easy to disprove them?
Hell, you can check it yourself. Open the games you own and change the resolution in them. You'll see that in 99% of games, changing resolution causes no change whatsoever to geometry/texture draw distances.

 

[Endymio]

None of this is correct.
CPU and GPUs do not share buses, they each have their own separate buses to their own memory pools.

How in the world do you think a GPU performs DMA (direct memory acccess), except by using the PCIe bus?

While the GPU can access system RAM vie PCIe, doing that while rendering brings a huge performance penalty. In games, it keeps everything it needs in its own VRAM.

Do you believe that "everything it needs" magically leaps into VRAM, without having to first trannsit the system bus?

They don't share I/O calls either. I/O is handled exclusively by the CPU. Anything that goes from the SSD to the VRAM goes through the CPU/RAM first, then to the VRAM via PCIe.

Wrong again. GPUs have had the ability to perform direct I/o access for over a decade:

"...Microsoft DirectStorage is a gaming technology that enables lightning-fast load times and improved asset streaming in PC games by allowing NVMe SSDs to transfer data directly to the graphics card (GPU), bypassing the CPU...."

Why do you even bother making blatantly incorrect claims like these, when it's so trivially easy to disprove them?

Physician, heal thyself!

 

[poshflamingos]

How in the world do you think a GPU performs DMA (direct memory acccess), except by using the PCIe bus?

PCIe devices do not impact the connection between the CPU and RAM, it's a completely separate bus.

Do you believe that "everything it needs" magically leaps into VRAM, without having to first trannsit the system bus?

"System bus" is not a thing.
You have the bus between the CPU and the RAM, a second separate bus between the GPU and its VRAM, and a third separate bus between the CPU and PCIe devices.
Also, SSD operations in current drives are limited to around 100 MB/s for Q1T1 random reads (for real-time asset streaming), and around ~1 GB/s for high QD random reads (for bulk assets during loading screens). Suggesting this usage would be able to saturate even PCIe buses (16 GB/s for 3.0, 32 GB/s for 4.0), let alone RAM buses (~50 GB/s for DDR4, ~100 GB/s for DDR5) or VRAM buses (~300 GB/s to ~1000 GB/s depending on GPU model) is hilarious.

Wrong again. GPUs have had the ability to perform direct I/o access for over a decade:

"...Microsoft DirectStorage is a gaming technology that enables lightning-fast load times and improved asset streaming in PC games by allowing NVMe SSDs to transfer data directly to the graphics card (GPU), bypassing the CPU...."

I don't know what you're quoting, but whoever wrote that is categorically wrong. DirectStorage is just a storage API, by itself it does not allow direct SSD to VRAM transfers. It does enable GPU manufacturers to build features like RTX IO on top of it, which is what allows direct SSD to VRAM transfers. But, outside of the sole exception of the Portal RTX mod (and maybe some other RTX Remix projects I'm not aware of too), no game has implemented RTX IO yet.

Physician, heal thyself!

I'll go ahead consider this point about the resolution and LODs conceded by you then.

 

[ZedRM]

Normally I think Steve's conclusions are out of whack. On this set of tests? He's come to some good and accurate conclusions.

For once, well done Steve.

 

[Endymio]

PCIe devices do not impact the connection between the CPU and RAM, it's a completely separate bus.

Now you're being intentionally obtuse. The CPU uses the PCIe bus to send data to the GPU, to perform network and disk I/o, etc ... and when the GPU is loading that bus via DMA calls, there is less bandwidth available for everyything else.

SSD operations in current drives are limited to around 100 MB/s for Q1T1 random reads (for real-time asset streaming), and around ~1 GB/s for high QD random reads (for bulk assets during loading screens). Suggesting this usage would be able to saturate even PCIe buses (16 GB/s for 3.0, 32 GB/s for 4.0), let alone RAM buses (~50 GB/s for DDR4, ~100 GB/s for DDR5) or VRAM buses (~300 GB/s to ~1000 GB/s depending on GPU model) is hilarious.

You're simply digging your hole deeper. I have no clue why you believe all game assets must be read from disc using random reads, but let's ignore that and focus on your larger error. You don't need to "saturate" a bus to slow it down -- if you're using 10% for I/o, then only 90% remains for everything else.

Still worse: you're forgetting that a GPU uses the bus *also* for direct memory access -- and this often does saturate it. Why on earth do you think we've consistently moved to higher bandwidth PCIe standards, if we weren't reaching limits on the existing ones?

I'll go ahead consider this point about the resolution and LODs conceded by you then.

Learn how game engines work. As I said earlier, every major engine performs resolution-based LOD reduction. The fact you believe you can "test this easily" by loading up and game and looking at frame rates belies an utter lack of understanding. Why not at least attempt to learn about a subject before arguing it?

"...Unreal Engine manages Level of Detail (LOD) based on screen resolution using a "Screen Size" metric, which calculates the screen-space diameter of an object's bounding sphere."

 

[Fuzzy Dunalap]

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?


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.

What cracks me up is them constantly repeating "higher resolutions don't help" when outside of a single game that they showed first, EVERY TITLE benefited from running a faster cpu at 1440p and 4k. It's like they don't want to believe their own results and just repeat the **** they always say. Im looking at the graphs and every.single.****ing.title. increased their frame rates at those high resolutions with a stronger cpu. Yea, my traffic to this place has slowed, Testing things is one things.

Knowing is another.