Interesting findings testing an overclocked i5-13600K and overclocked Ryzen 7 9850X3D's with overclocked RTX 5070Ti's and RX 9070 XT's

El_Capitan

El_Capitan

[H]ard|Gawd
2FA
Joined
Nov 11, 2012
Messages
2,026
This is kind of a spin-off of this thread. Unfortunately, due to the case sizes, I couldn't really swap all the cards. The memory speeds, timings, and amount are also different, and I may update by at least swapping the memory since I couldn't swap the cards.

If you don't want to read everything, here's the skinny:
If you're going to play at 1080P native and want the best CPU, then, yeah, get the 9850X3D and overclock it. However, an overclocked i5-13600K is -0.71% slower with an overclocked RX 9070 XT, and -1.18% slower with an overclocked RTX 5070 Ti.

If you're going to play at 2K native, and you're going to use an overclocked RX 9070 XT, then get the 950X3D and overclock it. An overclocked i5-13600K is -2.61% slower with an overclocked RX 9070 XT.

However, if you're going to play at 2K native using an overclocked RTX 5070 Ti, then get the i5-13600K and overclock it. An overclocked Ryzen 7 9850X3D is -3.35% slower with an overclocked RTX 5070 Ti.

i5-13600K at 1.4V with LLC 5 gets 5.7GHz (2 P-core usage), 5.6GHz (4 P-core usage), 5.5GHz (6 P-core usage), with E-Cores at 4.4GHz 6000MHz CL38.
Cyberpunk 2077 1080P High Native:
RTX 5070 Ti (+390 Core Clock, 3187MHz Peak) = 210.32
RX 9070 XT (+100 Mem Clock, -75MHz mV Core, 3222MHz Peak) = 201.71

Cyberpunk 2077 2K High Native:
RTX 5070 Ti (+390 Core Clock, 3180MHz Peak) = 162.21
RX 9070 XT (+100 Mem Clock, -75MHz mV Core, 3239MHz Peak) = 150.67

Ryzen 7 9850X3D PBO+200 -30CO 6000MHz CL36.
Cyberpunk 2077 1080P High Native:
RX 9070 XT (+100 Mem Clock, -75MHz mV Core, 3268MHz Peak) = 203.14

Cyberpunk 2077 2K High Native:
RX 9070 XT (+100 Mem Clock, -75MHz mV Core, 3247MHz Peak) = 154.65

Ryzen 7 9850X3D PBO+200 -30CO 4400MHz CL30.
Cyberpunk 2077 1080P High Native:
RTX 5070 Ti (+390 Core Clock, 3180MHz Peak) = 212.82

Cyberpunk 2077 2K High Native:
RTX 5070 Ti (+390 Core Clock, 3172MHz Peak) = 156.87

If you're arguing that the lows are better with the overclocked Ryzen 7 9850X3D at 2K or higher, only with the RX 9070 XT. Again, memory amount, speed, and timings may be what's making the difference.
Ryzen 7 9850X3D Min FPS: 130.07
i5-13600K Min FPS: 127.30

With the RTX 5070 Ti, the lows were better with the i5-13600K.
i5-13600K Min FPS: 140.13
Ryzen 7 9850X3D Min FPS: 136.10

As of 10/13/2026, with prices the way they are, buy a used i5-13600K for ~$150 instead of a Ryzen 7 9850X3D and save ~$300. Getting the RX 9070 XT instead of a RTX 5070 Ti will also save ~$300.

Conclusion:
When the i5-13600K was reviewed, it was with a RTX 3080. There also weren't reviewed with overclocks very well (shout out to OverClockersClub, as they used to do that a lot). With faster graphics cards like the RX 9070 XT and RTX 5070 Ti, it also changes up CPU utilization (and/or V-cache benefits). Stock i9-14900K's should still be faster than the overclocked i5-13600K, but I've not found that to be the case in regards to gaming. There's something special about an overclocked 13600K/14600K that might be worth another visit.
 

Attachments

  • 2026_10_13_Gigabyte_RX_9070_XT_i5-13600K_5.7(2)-5.5P_4.4E_AEMPII_6000MHz_CL38_Cyberpunk2077_10...png
    2026_10_13_Gigabyte_RX_9070_XT_i5-13600K_5.7(2)-5.5P_4.4E_AEMPII_6000MHz_CL38_Cyberpunk2077_10...png
    789.3 KB · Views: 0
  • 2026_10_13_Gigabyte_RX_9070_XT_i5-13600K_5.7(2)-5.5P_4.4E_AEMPII_6000MHz_CL38_Cyberpunk2077_2K...png
    2026_10_13_Gigabyte_RX_9070_XT_i5-13600K_5.7(2)-5.5P_4.4E_AEMPII_6000MHz_CL38_Cyberpunk2077_2K...png
    786.9 KB · Views: 0
  • 2026_10_13_Gigabyte_RX_9070_XT_9850X3D_PBO_200_-30CO_6000MHz_CL36_Cyberpunk2077_1080P_High_Nat...png
    2026_10_13_Gigabyte_RX_9070_XT_9850X3D_PBO_200_-30CO_6000MHz_CL36_Cyberpunk2077_1080P_High_Nat...png
    784.5 KB · Views: 0
  • 2026_10_13_Gigabyte_RX_9070_XT_9850X3D_PBO_200_-30CO_6000MHz_CL36_Cyberpunk2077_2K_High_Native.png
    2026_10_13_Gigabyte_RX_9070_XT_9850X3D_PBO_200_-30CO_6000MHz_CL36_Cyberpunk2077_2K_High_Native.png
    798.1 KB · Views: 0
  • 2026_10_13_ASUS_TUF_Gaming_RTX_5070Ti_White_OC_+390_Core_9850X3D_PBO_200_-30CO_4400MHz_CL30_Cy...png
    2026_10_13_ASUS_TUF_Gaming_RTX_5070Ti_White_OC_+390_Core_9850X3D_PBO_200_-30CO_4400MHz_CL30_Cy...png
    777.5 KB · Views: 0
  • 2026_10_13_ASUS_TUF_Gaming_RTX_5070Ti_White_OC_+390_Core_9850X3D_PBO_200_-30CO_4400MHz_CL30_Cy...png
    2026_10_13_ASUS_TUF_Gaming_RTX_5070Ti_White_OC_+390_Core_9850X3D_PBO_200_-30CO_4400MHz_CL30_Cy...png
    788.6 KB · Views: 0
  • 2026_10_13_Gigabyte_RTX_5070Ti_390_Core_i5-13600K_5.7(2)-5.5P_4.4E_AEMPII_6000MHz_CL38_Cyberpu...png
    2026_10_13_Gigabyte_RTX_5070Ti_390_Core_i5-13600K_5.7(2)-5.5P_4.4E_AEMPII_6000MHz_CL38_Cyberpu...png
    773.5 KB · Views: 0
  • 2026_10_13_Gigabyte_RTX_5070Ti_390_Core_i5-13600K_5.7(2)-5.5P_4.4E_AEMPII_6000MHz_CL38_Cyberpu...png
    2026_10_13_Gigabyte_RTX_5070Ti_390_Core_i5-13600K_5.7(2)-5.5P_4.4E_AEMPII_6000MHz_CL38_Cyberpu...png
    747.4 KB · Views: 0
Last edited:
I prefer the lowest resolution tests for CPU even though I play at 4k. I want to have realtive comparison to see how it will scale in the future with more cpu demanding games.
 
littlegreenmenfromhome said:
I keep saying 1080p only benches are useless
Click to expand...
You're stating the opposite of what is true. Without 1080p benches, we don't know how a CPU performs relative to another (in games) because otherwise with higher resolution some considerable CPU time (or lots of time in the case of 4k) is spent idling waiting on the GPU draw call to finish.

I say 1080p, when 1080p is arguably arbitrary. The resolution was chosen without committee because 1080p is easy to bench and it was silently agreed that was the ideal resolution to test CPU scaling as it puts enough distance away from the GPU bottlenecking.

It's painfully obvious from your 4k benchmark sample something just doesn't add up.
1791991730477.png

There's no linear scaling from CPUs past the 1800X or 2700X. Even without context (4k) these numbers would stick out like a sore thumb. Just look at the pattern from the lower resolution benchmarks, there's linear scaling across the CPU models. One would expect to see that here for the rest, but we're not seeing it. The answer is what I said above, the benchmarking tool hit a performance wall due to the resolution. And because of that, most of the modern CPU models show little to no delta between them. Without testing the lower resolutions, you wouldn't know that the CPUs do indeed scale linearly in a predictable pattern.

At 1080p, there is clear visible scaling across all the CPU models. No visible GPU bottleneck here.
1791993571892.png


That's my simplification of it. Bottom line is, 1080p benchmarking is still important. I'd argue all three benchmarked resolutions are important, so the reader gets some idea of where things stop scaling with the given GPU and resolution. And it goes without saying that benching with a different GPU may reshape the entire picture altogether.

One other anecdote is that if this were CPU limited, then the numbers would be nearly identical across all GPU models. That's rarely done these days because it's 1) typically not real world and 2) it's rather boring. A lot of flight sims used to be CPU limited.
 
Last edited:
Absalom said:
You're stating the opposite of what is true. Without 1080p benches, we don't know how a CPU performs relative to another (in games) because otherwise with higher resolution some considerable CPU time (or lots of time in the case of 4k) is spent idling waiting on the GPU draw call to finish.

I say 1080p, when 1080p is arguably arbitrary. The resolution was chosen without committee because 1080p is easy to bench and it was silently agreed that was the ideal resolution to test CPU scaling as it puts enough distance away from the GPU bottlenecking.

It's painfully obvious from your 4k benchmark sample something just doesn't add up.
View attachment 831182
There's no linear scaling from CPUs past the 1800X or 2700X. Even without context (4k) these numbers would stick out like a sore thumb. Just look at the pattern from the lower resolution benchmarks, there's linear scaling across the CPU models. One would expect to see that here for the rest, but we're not seeing it. The answer is what I said above, the benchmarking tool hit a performance wall due to the resolution. And because of that, most of the modern CPU models show little to no delta between them. Without testing the lower resolutions, you wouldn't know that the CPUs do indeed scale linearly in a predictable pattern.

At 1080p, there is clear visible scaling across all the CPU models. No visible GPU bottleneck here.
View attachment 831185

That's my simplification of it. Bottom line is, 1080p benchmarking is still important. I'd argue all three benchmarked resolutions are important, so the reader gets some idea of where things stop scaling with the given GPU and resolution. And it goes without saying that benching with a different GPU may reshape the entire picture altogether.

One other anecdote is that if this were CPU limited, then the numbers would be nearly identical across all GPU models. That's rarely done these days because it's 1) typically not real world and 2) it's rather boring. A lot of flight sims used to be CPU limited.
Click to expand...
That poster has spent the entire year he has been a member here telling everyone else how dumb we are because we all understand what a GPU bottleneck is.
 
Absalom said:
You're stating the opposite of what is true. Without 1080p benches, we don't know how a CPU performs relative to another (in games) because otherwise with higher resolution some considerable CPU time (or lots of time in the case of 4k) is spent idling waiting on the GPU draw call to finish.

I say 1080p, when 1080p is arguably arbitrary. The resolution was chosen without committee because 1080p is easy to bench and it was silently agreed that was the ideal resolution to test CPU scaling as it puts enough distance away from the GPU bottlenecking.

It's painfully obvious from your 4k benchmark sample something just doesn't add up.
View attachment 831182
There's no linear scaling from CPUs past the 1800X or 2700X. Even without context (4k) these numbers would stick out like a sore thumb. Just look at the pattern from the lower resolution benchmarks, there's linear scaling across the CPU models. One would expect to see that here for the rest, but we're not seeing it. The answer is what I said above, the benchmarking tool hit a performance wall due to the resolution. And because of that, most of the modern CPU models show little to no delta between them. Without testing the lower resolutions, you wouldn't know that the CPUs do indeed scale linearly in a predictable pattern.

At 1080p, there is clear visible scaling across all the CPU models. No visible GPU bottleneck here.
View attachment 831185

That's my simplification of it. Bottom line is, 1080p benchmarking is still important. I'd argue all three benchmarked resolutions are important, so the reader gets some idea of where things stop scaling with the given GPU and resolution. And it goes without saying that benching with a different GPU may reshape the entire picture altogether.

One other anecdote is that if this were CPU limited, then the numbers would be nearly identical across all GPU models. That's rarely done these days because it's 1) typically not real world and 2) it's rather boring. A lot of flight sims used to be CPU limited.
Click to expand...
What you are seeing is GPU limted gaming, they way you are meant to be playing.
IF your gaming is not GPU limited, you are doing it wrong.
 
littlegreenmenfromhome said:
What you are seeing is GPU limted gaming, they way you are meant to be playing.
IF your gaming is not GPU limited, you are doing it wrong.
Click to expand...
Why do you not understand that is not the point of cpu benchmarks? Sure look at the 4k benchmarks to see if there is an edge case on a cpu heavy game, but the whole point of those benchmarks is to show relative performance.
 
Absalom said:
You're stating the opposite of what is true. Without 1080p benches, we don't know how a CPU performs relative to another (in games) because otherwise with higher resolution some considerable CPU time (or lots of time in the case of 4k) is spent idling waiting on the GPU draw call to finish.

I say 1080p, when 1080p is arguably arbitrary. The resolution was chosen without committee because 1080p is easy to bench and it was silently agreed that was the ideal resolution to test CPU scaling as it puts enough distance away from the GPU bottlenecking.

It's painfully obvious from your 4k benchmark sample something just doesn't add up.
View attachment 831182
There's no linear scaling from CPUs past the 1800X or 2700X. Even without context (4k) these numbers would stick out like a sore thumb. Just look at the pattern from the lower resolution benchmarks, there's linear scaling across the CPU models. One would expect to see that here for the rest, but we're not seeing it. The answer is what I said above, the benchmarking tool hit a performance wall due to the resolution. And because of that, most of the modern CPU models show little to no delta between them. Without testing the lower resolutions, you wouldn't know that the CPUs do indeed scale linearly in a predictable pattern.

At 1080p, there is clear visible scaling across all the CPU models. No visible GPU bottleneck here.
View attachment 831185

That's my simplification of it. Bottom line is, 1080p benchmarking is still important. I'd argue all three benchmarked resolutions are important, so the reader gets some idea of where things stop scaling with the given GPU and resolution. And it goes without saying that benching with a different GPU may reshape the entire picture altogether.

One other anecdote is that if this were CPU limited, then the numbers would be nearly identical across all GPU models. That's rarely done these days because it's 1) typically not real world and 2) it's rather boring. A lot of flight sims used to be CPU limited.
Click to expand...
Easy to debunk ignorance these days:
Yes, the user’s handwritten notes completely misunderstand a standard GPU-limited scenario.
The chart shows Cyberpunk 2077 running at a massive 4K resolution (3840x2160) with Ray Tracing Ultra settings. At these intense settings, the graphics card (GPU) is working at 100% capacity and acts as a strict bottleneck, capping the game's maximum performance around 54 to 55 FPS.
Here is why the user's logic about "bunching up" and "linear scaling" is flawed:

1. The "Bunching Up" is Textbook GPU Bottlenecking
  • What the user thinks: The purple bracket on the top right notes, "There's a problem here. Something doesn't add up! Numbers should be following a linear path not bunching up."
  • The reality: The CPUs are bunching up because any modern CPU is fast enough to feed the GPU 54 frames per second. Whether you use an older Core i9-9900K or a newer Ryzen 9 5950X, the game cannot go any faster because the GPU cannot process the 4K Ray Tracing load any quicker. The CPU is effectively "sitting idle" waiting for the GPU to finish rendering each frame.

2. The "Linear Scaling" is actually CPU Bottlenecking
  • What the user thinks: The purple line at the bottom right notes, "Linear scaling here (roughly)," implying that this upward diagonal trend is how the entire chart should look.
  • The reality:


    This section only scales upwards because those specific processors (like the Ryzen 3 1300X, Core i3-9100, and Ryzen 5 1600X) are so old, slow, or lacking in cores/threads that they cannot even maintain 54 FPS. In this bottom bracket, you are seeing a CPU bottleneck, where the game performance scales up as you buy a slightly better processor, until it finally hits that 54 FPS ceiling dictated by the GPU.

Summary
The user expects hardware scaling to look like a continuous, straight diagonal line. In PC gaming, performance hits a hard horizontal ceiling the moment the graphics card reaches maximum utilization. The chart behaves exactly as it should.
 
littlegreenmenfromhome said:
Easy to debunk ignorance these days:
Click to expand...

Kind of ironic to say that when you have so little understanding of what is going on that you literally can't compose a coherent response without the help of AI.

Both CPU and GPU performance are important.

There is value in being able to understand the raw performance of each in isolation, without one being strictly bottlenecked by the other. Even if this means running benchmarks using settings that deviate from the settings you use when you actually play certain games.

Not all games are GPU limited, even at maximum settings. Comments like this one below are, quite frankly, retarded:

littlegreenmenfromhome said:
IF your gaming is not GPU limited, you are doing it wrong.
Click to expand...

And speaking of settings, games are chock full of adjustable settings that primarily affect GPU usage. That is to say, if you are GPU limited, that is something you can "fix" by lowering your in-game settings. Conversely, if you are CPU limited, you're mostly screwed. Knowing how much CPU headroom you have to work with is critical.

It's perfectly legitimate to look at the data and conclude that you don't need a faster CPU because you are GPU limited in the games that you play. That doesn't mean that CPU benchmarks themselves (aka 1080P benchmarks) are irrelevant. Different people can look at the data and come to their own conclusions based on their personal use-case scenarios, which may very well differ from yours.
 
Well, he was entertaining in a WTF is this fool going to say next kind of way but the inane screed and lack of reasoning got boring pretty quick and I had to put him in the bin too.
 
GotNoRice said:
Kind of ironic to say that when you have so little understanding of what is going on that you literally can't compose a coherent response without the help of AI.

Both CPU and GPU performance are important.

There is value in being able to understand the raw performance of each in isolation, without one being strictly bottlenecked by the other. Even if this means running benchmarks using settings that deviate from the settings you use when you actually play certain games.

Not all games are GPU limited, even at maximum settings. Comments like this one below are, quite frankly, retarded:



And speaking of settings, games are chock full of adjustable settings that primarily affect GPU usage. That is to say, if you are GPU limited, that is something you can "fix" by lowering your in-game settings. Conversely, if you are CPU limited, you're mostly screwed. Knowing how much CPU headroom you have to work with is critical.

It's perfectly legitimate to look at the data and conclude that you don't need a faster CPU because you are GPU limited in the games that you play. That doesn't mean that CPU benchmarks themselves (aka 1080P benchmarks) are irrelevant. Different people can look at the data and come to their own conclusions based on their personal use-case scenarios, which may very well differ from yours.
Click to expand...
They are irrelevant for 2160p in 2026, if you GPU is not peggede at +95% usage when gaming, you screwed up.
 
Absalom said:
You've already been called out as being wrong. You're in a tailspin.
Click to expand...
I bet you more people than you think overpaid for their CPU, money that would have been better diverted to a better GPU for gaming (by looking at low res charts and not understand the relation)

Just like you that went full Dunning-Kruger and tried to blame graphs for being false because you do not understand GPU limited senarios.
You forgot that part, so get off your high horse, since you were (and are) wrong.
 
littlegreenmenfromhome said:
I bet you more people than you think overpaid for their CPU, money that would have been better diverted to a better GPU for gaming (by looking at low res charts and not understand the relation)

Just like you that went full Dunning-Kruger and tried to blame graphs for being false because you do not understand GPU limited senarios.
You forgot that part, so get off your high horse, since you were (and are) wrong.
Click to expand...
Says the person who uses AI to form their responses. Do you even have an original thought? Are you even real? Keep hallucinating, bot.
 
Absalom said:
Says the person who uses AI to form their responses. Do you even have an original thought? Are you even real? Keep hallucinating, bot.
Click to expand...
Your to stupid too waste long answers on, because you will deflect and run away from being ignorantly wrong.
Shall I repost your epic brain fart?

Dunning-Kruger gets A.I.
1792243851164.png
 
littlegreenmenfromhome said:
I keep saying 1080p only benches are useless:


View attachment 831108

It like fudenmental knowlegde is being lost every singel decade...
Click to expand...

There's a phrase for this... it's called "margin of error." This means that the order would completely change if you tested it on a different day. You're severely GPU limited at these settings. That's why the CPUs bunch up. The fact that people do not understand this is alarming...
 
littlegreenmenfromhome said:
I bet you more people than you think overpaid for their CPU, money that would have been better diverted to a better GPU for gaming (by looking at low res charts and not understand the relation)

Just like you that went full Dunning-Kruger and tried to blame graphs for being false because you do not understand GPU limited senarios.
You forgot that part, so get off your high horse, since you were (and are) wrong.
Click to expand...
A higher performing CPU will last longer than a lower performing CPU. Sure you can save a few bucks and get a cheaper CPU, but the more performant CPU will have more legs. That's why nobody recommends you build a brand new system around a 5800X3D, even though it's still a fantastic gaming CPU; for not a whole lot more, you can build a 9800X3D system. You're not going to notice a difference at 4K... until you do on some new, very CPU intensive game several years down the road. Worst of all, you cannot upgrade the 5800X3D.

Seriously, do you even consider such things?
 
Upgrading the 225F on my media rig to the 250k (everything at default, ram at 6400 xmp1) led to a substantial reduction in 1% lows in gears e-day, as expected based on how cpu intensive that game is.
 
madpistol said:
There's a phrase for this... it's called "margin of error." This means that the order would completely change if you tested it on a different day. You're severely GPU limited at these settings. That's why the CPUs bunch up. The fact that people do not understand this is alarming...
Click to expand...
madpistol said:
A higher performing CPU will last longer than a lower performing CPU. Sure you can save a few bucks and get a cheaper CPU, but the more performant CPU will have more legs. That's why nobody recommends you build a brand new system around a 5800X3D, even though it's still a fantastic gaming CPU; for not a whole lot more, you can build a 9800X3D system. You're not going to notice a difference at 4K... until you do on some new, very CPU intensive game several years down the road. Worst of all, you cannot upgrade the 5800X3D.

Seriously, do you even consider such things?
Click to expand...
I build a new rig every 2. generation (from the Case to the GPU).
Old rig goes to the wife and her rig goes to the daughter.
Longevity is not a factor, so the system is always in balance.
Hence why I build to GPU as GPU requiremnts increase way faster then CPU requirements.

And I see far to many 9800X3D CPU's paried with a 5060/9060 GPU's, most likely beacuse the can afford $100 mor efor a CPU and feel good, but $3-400 for a bigger GPU they cannot and thus they misbuild their rig.
 
littlegreenmenfromhome said:
And I see far to many 9800X3D CPU's paried with a 5060/9060 GPU's, most likely beacuse the can afford $100 mor efor a CPU and feel good, but $3-400 for a bigger GPU they cannot and thus they misbuild their rig.
Click to expand...

You can't possibly draw that conclusion without knowing what games that person is playing, what resolution they are running, etc.

Steam hardware survey still shows 1080P as the most common resolution at a whopping 48%. 1440P is #2 with 27%. 4K comes in at.... less than 5%. But in your mind somehow everyone is playing games at 4K.
 
You must log in or register to reply here.
Back
Top