We Tested Every Ryzen 5 and 7 X3D CPU: From 5800X3D to 9800X3D

[Endymio]

The reason for CPU benchmarks is to give everyone objetive information about how CPUs perform in games without any GPU bottlenecks. Not to give each individual reader personalized information about how each CPU performs in each resolution with their individual GPUs

Oops! Reviewers test CPUs at multiple resolutions and GPUs, because people DO wish to see results close to their own environment. Even when a reviewer tests against a single GPU, they generally do so at multiple settings. Take this Techspot review of the 250K. If the goal was to "eliminate GPU bottlenecks entirely", why test at 1080p medium settings AND 1080p Ultra plus?

https://www.techspot.com/review/3106-intel-core-ultra-5-250k-plus/

If you have a RTX 3080, and force every review to do their tests with a RTX 3080 for your convenience, then those tests are useless to everyone who has any other GPU.

Nonsense. If a review tests with a low-end, mid-range, and high-end card, then those three datapoints alone allow everyone to interpolate to their own setup.

Yeah, and techpowerup's 4K tests show exactly how stupid 4K tests are. You just end up with most of the CPUs tested tied at the top

Most -- but not all. That's the point. Those tests show at what point purchasing more CPU power gives little or no benefit.

the 3700X beats the 2700X by 12% in the 2024 test....new, faster GPUs will continue to come out, and people will continue to upgrade their GPUs over time.

This utterly demolishes your argument. Anyone who upgraded their 2700x to a 3700x in 2019 for 4K gaming made the worst possible purchasing decision. It took 5.5 years to open a 12% performance gap ... had a buyer instead deferred that upgrade until it actually benefitted them, they could have either saved a few hundred dollars, or spent the same money for a far faster, more modern CPU.

Also, you've again contradicted yourself. The only reason we know that 12% performance gap exists is because a reviewer tested the CPUS at 4K resolution. You're quoting a test that you yourself called "stupid". Oops again!

 

[Luminousbeing99]

Nonsense. If a review tests with a low-end, mid-range, and high-end card, then those three datapoints alone allow everyone to interpolate to their own setup.

So you’re willing to interpolate to get a rough idea of what your performance might be like. When you could just look at benchmarks for your GPU and get a much more precise idea of the performance you can except. Mostly just sounds like laziness.

 

[poshflamingos]

Oops! Reviewers test CPUs at multiple resolutions and GPUs

No, they don't. Out of the high profile reviewers, techpowerup and guru3d are the only ones that test higher resolutions. Every other reviewer (TechSpot/HUB, GamerNexus, Tom's Hardware, Eurogamer/DF, TweakTown, overclock3d.net, PCMag, Linus Tech Tips, etc) tests at 1080p or lower.

Even when a reviewer tests against a single GPU, they generally do so at multiple settings. Take this Techspot review of the 250K. If the goal was to "eliminate GPU bottlenecks entirely", why test at 1080p medium settings AND 1080p Ultra plus?

Why do you think, genius?
It's because, unlike resolution, graphics settings DO affect CPU load. Higher settings generally mean more API draw calls, which directly impacts CPU performance.

Nonsense. If a review tests with a low-end, mid-range, and high-end card, then those three datapoints alone allow everyone to interpolate to their own setup.

So you want to triple the workload of reviewers to have them test 3 GPUs, just to generate redundant data that you can figure out yourself by cross-referencing GPU tests?
In reality, because reviewers don't have infinite time, your short-sighted suggestion would just make CPU reviews worse across the board. If reviewers test 15 games with 1 GPU, they wouldn't spend triple the time (that they don't have) to test 15 games with 3 GPUs, they would just use the same amlunt of time to test 5 games with 3 GPUs, which just makes the game selection less representative across the board.

Most -- but not all. That's the point. Those tests show at what point purchasing more CPU power gives little or no benefit.

No, they don't. They only show at what point purchasing better CPUs provide no benefit for the specific GPU used on the test. If you have any other GPU, that test means jack squat for you, because for your GPU that point will be different.
You know what does give you exactly what you want? Looking at low resolution CPU tests with no GPU bottleneck, and then cross-referencing with GPU tests of your specific GPU.

This utterly demolishes your argument.

lmao
You know someone is desperate to save face when they start trying to use dramatic/bombastic language like this.

It took 5.5 years to open a 12% performance gap

No, you dim. It didn't take 5 years to "open the performance gap". The performance gap was always there. Any time you're not GPU limited, the performance gap shows up. That includes any time you play a game that is more CPU-demanding than GPU-demanding (like Baldur's Gate for example), any time you do a GPU upgrade (you don't need to wait 5 years for the 4090, any GPU faster than the 2080 Ti will create more scenarios where you're no longer GPU-limited), any time you use DLSS, and so on. That's why knowing the real performance difference between CPUs is important, rather than just being tricked by a 4K test graph that every CPU is the same.

Also, you've again contradicted yourself. The only reason we know that 12% performance gap exists is because a reviewer tested the CPUS at 4K resolution. Oops again!

WTF are you smoking?
No, the reason we know that is not the 4K tests. We knew the 3700X was 12% faster already in 2019, because of the low resolution tests. The 4K tests in 2024 just reiterated what we already knew.
I just used those tests as an example to illustrate to you why your claim that "CPU performance deltas decrease at 4K" is nonsense. No, CPU performance deltas do not decrease at 4K or any other resolution. They only decrease when you're GPU limited, but once new GPUs come out they stop being GPU limited even at 4K, so the real performance difference comes out even at 4K.
This isn't a thing only for those two CPUs only on those two years, that was just the first example I found on google. That's true for every CPU on every year. Every CPU will invariably stop being GPU-limited at 4K at some point once faster GPUs come out. Which is why, again, it's important to see the real difference between CPUs (using low resolution tests), instead of misleading yourself into thhing they perform the same by looking at GPU-limited 4K graphs.

 

[Endymio]

No, they don't. Out of the high profile reviewers, techpowerup and guru3d are the only ones that test higher resolutions. Every other reviewer (TechSpot/HUB, GamerNexus, Tom's Hardware, Eurogamer/DF, TweakTown, overclock3d.net, PCMag, Linus Tech Tips, etc) tests at 1080p or lower.

Ouch! You're 0 for 4 now. This Toms Hardware review of 40+ CPUs has scores at 1080p and 1440p (see pg 2), and the PCMag review of the 250K below it tests the CPU at 1080p, 1440p AND 4K.

https://www.tomshardware.com/reviews/cpu-hierarchy,4312-2.html

I Tested Intel's New $199 Core Ultra 5 250K Plus Desktop Chip. It's the New Lord of Budget CPUs

Intel's Core Ultra 5 250K Plus boosts gaming performance over its "Arrow Lake" predecessor at a price that's hard to pass up. It's not the fastest gaming CPU around, but no other chip goes as hard for so little money.

www.pcmag.com

No, you dim. It didn't take 5 years to "open the performance gap". The performance gap was always there.

There is no fixed "gap" between the CPUs. In 2019, there was no gap at all. In 2024 (if you wish to pair a $3000 video card with a $150 CPU), there was a 10% gap, and a few years from now, that "gap" will widen further when even faster cards are introduced.

Again: anyone who paid a huge premium for a CPU that they saw zero benefit from for 6 years wasted their money. If you disagree, I have some nice ocean-front property in Kansas for sale. Of course, you'll need to wait on future plate tectonic shifts to actually SEE that benefit ... but it's still there, I promise!

We knew the 3700X was 12% faster already in 2019

12% faster in 1080p. Tied at 4K. And probably 25% or more faster if you scaled down your resolution to 640x480.

For someone who exclusively uses 4K, however, only one of those three results has meaning.

 

[poshflamingos]

This Toms Hardware review of 40+ CPUs has scores at 1080p and 1440p (see pg 2)

Maybe on this individual, special feature they made they did. But if you look at their actual reviews, there's only 1080p tests.

and the PCMag review of the 250K below it tests the CPU at 1080p, 1440p AND 4K.

Funny, because on this other PCMag review there's only 1080p tests.

There is no fixed "gap" between the CPUs.

Yes, there is. Once you remove GPU bottlenecks, you get a fixed gap between CPUs. That's why the 3700X was 11% faster in 2019 and was still 12% faster in 2024. That's the fixed gap between those two chips, and you see it whatener you're not GPU-limited.

In 2019, there was no gap at all.

Yes, there was, it was 11%.
The only reason you couldn't also see those 11% at 4K was because the 2080 Ti was too slow. Once we got faster GPUs, what a surprise, it was there at 4K as well! Who could have seen this coming?

In 2024 (if you wish to pair a $3000 video card with a $150 CPU)

You didn't need to wait until 2024 or spend $3000 to see that performance advantage, genius. Any GPU faster than the 2080 Ti would have shown it. If you spent $700 on the RTX 3080 in 2020, you already saw it. If you spent $600 on the RTX 4070 in 2022, you saw it. If at any point you played a game that is more CPU-demanding than average, you saw it.
Just because the review I linked is from 2024 and uses a $1600 GPU (not "$3000"), doesn't mean that was the first time since 2019 that this was observable.

Again: anyone who paid a huge premium for a CPU that they saw zero benefit from for 6 years wasted their money. If you disagree, I have some nice ocean-front property in Kansas for sale.

Where did you get this moronic idea from? Nobody here is talking about wasting money on CPUs you don't need. The point is that low resolution tests produce useful information, and 4K tests don't. That doesn't mean you should go out and overspend on a CPU you don't need. You still look at the low resolution tests and buy what is appropriate for your GPU.

12% faster in 1080p.

No, it was 12% faster in every resolution.

Tied at 4K.

No, it was not tied at 4K, it was 12% faster at 4K. The 2080 Ti was just too slow to show it. Now, with faster GPUs, we know it is indeed 12% faster at 4K as well.

For someone who exclusively uses 4K, however, only one of those three results has meaning.

Yes, the low resolution one.
The 4K test has absolutely no meaning whatsoever, unless you have the exact same GPU used on the test.

Seriously, you're too unintelligent to participate in this conversation. I'm just repeating myself now. It will be abundantly clear to everyone reading this who between the two of us knows what they're talking about, and who doesn't. At this point I feel satisfied to just leave it at that, there's nothing I need to add here.

 

[Endymio]

Maybe on this individual, special feature they made they did....Funny, because on this other PCMag review there's only 1080p tests.

You fail to understand how logic works. When you claim something doesn't exist, a single counter-example suffices to disprove it. I've given you three, from three different sites. A quick Google search reveals dozens more. Dozens!

You're also missing the point. When reviewers limit their tests to a specific GPU or resolution, it's not because testing over a wider range is "stupid" or "useless", but simply because their resources are limited.

The only reason you couldn't also see those 11% at 4K was because the 2080 Ti was too slow.

Sigh. I'll make this as simple as possible. If the only way to "see" this mysterious invisible performance boost is to wait 6 years and buy a new $3000 video card -- you're wasting money upgrading the CPU now, for zero benefit. Far better to wait until you CAN see the benefit, when either the chip is much cheaper, or for the same money you can purchase a faster, more modern CPU.

You didn't need to wait until 2024 or spend $3000 to see that performance advantage, genius. Any GPU faster than the 2080 Ti would have shown it. If you spent $700 on the RTX 3080 in 2020, you already saw it.

Oops! You're 0 for 5 now and sliding fast.

"...Test System: A high-end RTX 3080 Ti is used to test these processors, demonstrating that the GPU is the primary factor at 4K.

• Performance Result: The 2700X and 3700X deliver very similar FPS at maximum 4K settings in Far Cry 5....."

 

[ddferrari]

Ermmm...what about the 9950X3D??

The 9800X3D is ~ 5% faster than the 9950X3D in gaming overall. So ermmm use your maths

 

[amghwk]

The 9800X3D is ~ 5% faster than the 9950X3D in gaming overall. So ermmm use your maths

I see, thank you friend. I was just wondering why the graph was not included. But thank you, anyway.

So, I guess, it's better to buy a 9800X3D compared to a 9950X3D... Since I was shopping a for a 9950X3D.

 

[dankbot420]

1080P benchmarks are a great way to see raw performance differences with big irrelevant numbers but at 4K the difference between a 5800X3D and 9800X3D is around 5% nobody except maybe niche e-sport gamers are buying a 9850X3D and 5090 to play at 1080P

 

[Roboyt0]

Great test. My 7800X3D is still great and up there. Some games seems to run a lot better on the newer ones tho. I wish we can test these CPUs with a more normal card... like 5080. I know why we test with the top of the top but still.

Makes me wonder how much of a difference is there on a lower tier card. 5090 has never been in the stores here even lol. I got a 5080 and id love to see how much id gain with a 9800X3D. Is it 5? 10 fps? Cant be as much as 5090 obviously Also wonder for Baldurs gate 3. That game alone has like a 50 fps difference. Would that translate to 20-30 for 7800X3D?

Techspot did this previously using 4 GPUs and 4 CPUs. With a lower tier card, the performance doesn't scale nearly as much.

As an example, RX 9070 at 1440P there's only a ~14% difference in 1% lows between a standard R5 5600 and the 9800X3D.

https://www.techspot.com/review/3021-ryzen-5800x3d-cpu-gpu-scaling/

Seeing things like this are why I have lost the itch to upgrade my hardware so frequently. Add in this ridiculous 'shortage' due to the AI boom and I'm more than content with a 13700K and R7 7700. Both are paired with standard RX 9070's and are driving 3440*1440 and 4K, so, upgrading makes little sense for me.

 

[Vanderlinde]

It would have been nice to see some basic productivity benchmarks even if that was limited to the main x3D chips and omitted the binned variants.

I know HUB is primarily gaming but it would have been interesting to see more.

Usually they are a small percentage slower (esp the 5x series) then the counterpart without X3D.

Only a few apps can use the additional cache, but it mostly shows in gaming.

Whole X3D was originally a datacenter based chip (Epyc) and marketing team of AMD was asked, "Make a use case for this chip" and so the X3D was born.

 

[urielejh]

Buying a 5800x3d at launch has been a smart choice. At 4k with a 7900xtx cpu-bound scenarios are very rare.

 

[hahahanoobs]

I think the more important thing here is to know what it can do at 1440p since most people are starting to move in that direction. Also it would be nice to know what graphics cards would be paired nicely with each of these but it seems according to Google that every iteration of amd's new X3D generation. I gives more life to older graphics cards. As an example my 3080 scales exponentially with each generation of x3d compared to my 11400f. If I go to 9th generation I'm leaving about 80 to 100% fps on the table which is a lot and my minimums I suspect are going to be even higher. The question at this point remains is it better to just put in 11700 and upgrade the video card to a similar price point when Nvidia/AMD's next generation arrives or upgrade the system?

I agree, but Steve refuses because "the CPU, an RTX 5090 and 1080p is all gamers need to see for a CPU review." I strongly disagree so I've limited my time here recently. I'm sure that makes Steve happy.

Try Hardware Canucks YT Channel for testing that gamers that have no intention of owning an RTX 5090 or 9800X3D can benefit from.

 

[hahahanoobs]

Maybe on this individual, special feature they made they did. But if you look at their actual reviews, there's only 1080p tests.


Funny, because on this other PCMag review there's only 1080p tests.

Gamers Nexus
Techpowerup
Hardware Canucks

All the above include at least 1440p in CPU reviews to name a few...

 

[ashkjd404]

More than

I think the more important thing here is to know what it can do at 1440p since most people are starting to move in that direction.

According to Steam's recent Hardware survey from March 2026.
More than 50%(51.93%) of all users are still playing in 1080p, and the users who are playing at 1440p is actually decreasing and only makes up for about 20.7% of all users on Steam.
I think you need to get your facts checked and wake up.

Many of the most popular games on Steam are highly competitive,
and that's where 1080p and 1440p are still the kings, 1080p makes the most sense
as it is the most affordable resolution for most average gamers.

 

[longjohn119]

I think the more important thing here is to know what it can do at 1440p since most people are starting to move in that direction. Also it would be nice to know what graphics cards would be paired nicely with each of these but it seems according to Google that every iteration of amd's new X3D generation. I gives more life to older graphics cards. As an example my 3080 scales exponentially with each generation of x3d compared to my 11400f. If I go to 9th generation I'm leaving about 80 to 100% fps on the table which is a lot and my minimums I suspect are going to be even higher. The question at this point remains is it better to just put in 11700 and upgrade the video card to a similar price point when Nvidia/AMD's next generation arrives or upgrade the system?

I can tell you my experience although it's with Ryzen only. Using the same 4070 Ti and 1440p Ultrawide I upgraded from a 5800X and 3200 DDR4 to a 7800X3D and 6000 DDR5. I play mainly with the highest graphics settings I can that will still get me around 70 FPS. I benchmarked Shadow of the Tomb Raider, AC Odyssey, AC Valhalla, HZD Remastered, and Cyberpunk 2077 all using the exact same graphics settings. With the way I play the difference between the two was only 3% to 4% with Cyberpunk getting the biggest uplift of about 5%

I was a bit disappointed at the results but it did confirm that buying a 9800X3D that was $120 more would have been a waste of money because it wouldn't have gained me anything but another 1 FPS at the graphics settings I use at the resolution I use. Unless you are playing competitive online games with low graphics settings at 1080p the X3D chips aren't really worth the extra money in my opinion based on the tests I made because there is little to be gained on games where you are GPU bound.

 

[Athlonite69]

Ermmm...what about the 9950X3D??

That's a R9 series CPU not an R5 or R7 series and this is what the Tests were aimed at low to mid range X3D cpu's not the high end

 

[OortCloud]

While I understand that all these tests have to be done using a 5090 so the underlying performance of the CPU can be the only bottleneck, readers need to understand that in reality most these CPU's are more than capable of rendering the scenes at resolutions and frame rates they are likely never to achieve (and are plenty good enough for their needs) because they will be limited by their 5070's, 3060's and 4K output etc etc.

 

[ScottSoapbox]

Usually they are a small percentage slower (esp the 5x series) then the counterpart without X3D.

Only a few apps can use the additional cache, but it mostly shows in gaming.

Whole X3D was originally a datacenter based chip (Epyc) and marketing team of AMD was asked, "Make a use case for this chip" and so the X3D was born.

Yes, but that non-x3D isn't nicely at hand. Thus, it would have been a nice inclusion in this article.

 

[HC1Gunner]

You left the Ryzen 9 9950X3D out of the test because?

 

[Loadedaxe]

5500X3D is pretty much the only X3D I can get for my AM4 build at the moment, and it's only $160, would say it's quite fair..

where is it $160?

 

[Loadedaxe]

Ermmm...what about the 9950X3D??

ermmmm...where is there a R5 or R7 9950X3D?

 

[Loadedaxe]

You left the Ryzen 9 9950X3D out of the test because?

Its not an R5 or 7 maybe ....

 

[Athlonite69]

You left the Ryzen 9 9950X3D out of the test because?

As previously stated the tests were for R5 and R7 X3D CPU's not the R9 high end of which the R9 9950X3D is

 

[koblongata]

where is it $160?

Taiwan