Nine X3D CPUs, two platforms, and 14 games tested. We compare every Ryzen 5 and 7 X3D processor to find out how much performance has improved since the 5800X3D and where it actually matters.
We Tested Every Ryzen 5 and 7 X3D CPU: From 5800X3D to 9800X3D
Also wonder for Baldurs gate 3. That game alone has like a 50 fps difference. Would that translate to 20-30 for 7800X3D?These CPU tests give you the MAX FPS possible for each CPU.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
Well now you know what they can do at the more CPU demanding 1080p so you can extrapolate.
Resolution doesn't affect CPU load/performance. What they can do at 1440p is exactly the same as what they can do at 1080p. Except testing at 1440p would make the test worse, because more of the tests would end up GPU-limited (in which case the test says nothing about CPU performance).
That's not a thing that can be done. Each individual game stresses the CPU and GPU in different proportions, so the "perfect GPU match" for each CPU can change for each game.
Eh? This isn't true. The delta between individual CPUs narrows as resolution increases, and the size of that delta is the entire reason for such comparison tests. You might pay double for a CPU that gives you an extra 50 fps ... but if at 4K, its only adding 4-5 fps, your cost-benefit analysis changes.
This couldn't be more wrong. Resolution doesn't affect CPU performance at all. Changing resolution has no impact whatsoever on gaming performance.
You failed to read my post. Raw CPU performance isn't impacted, but overall performance deltas are.
As I intimated in my original post.
It most certainly means exactly that. While it doesn't imply the CPU has somehow become slower or less powerful, it does mean that CPU is granting a far smaller benefit. I don't know how much simpler this can be. Many people might pay an additional $300 for a CPU that gives 30% more performance, but far fewer if it results in only a 5% uplift.
Meaningless distinction at best, misleading at worst. The only reason you're measuring smaller "deltas" is because you're testing in a way that cannot show the real difference.
You only get a smaller benefit if you pair it with a GPU/resolution that gets a smaller benefit. The CPU doesn't actually get slower relative to the other CPUs at 4K, the potential is still there, it's just as fast at 4K as it is at 1080p. In the future faster GPUs and more CPU-demanding games will come out and expose that no, turns out that the 14600K was not really "just as fast as the 9800X3D at 4K" as you thought it were based on the 4K CPU tests you found, the 9800X3D was significantly faster all along, you were just testing in a way that was incapable of showing it.
That statement is still nonsense. If you want to know how your personal build would fare with those CPUs, you can cross-reference it with tests done with your GPU at your resolution to know how low you can go with a CPU. Demanding CPU tests be done in an useless way just to satisfy your ego is insane. It helps nobody to see a test where all CPUs are tied in a GPU-bound game.
No. Results from a scenario you'll never experience fail to answer the question for which benchmarks exist: "if I purchase this chip, how much benefit will I see?" The "real difference" is what you see in the real world.
You're literally admitting my point: "yes, THIS test tells me nothing. I need data from an entirely different test to know how the chip will perform."
Ouch! This is an even worse mistake. Yes the 3700X certainly DID grow faster. But there was nothing "magical" about it. The second test was done on entirely different software.
Except benchmarks prove this often wrong. Look at the link above comparing (among others), the 7800x3d vs. the 7950x3d. The latter chip was $699 when the former was $450, yet the more expensive chip actually gives you very slightly less performance in games.
In which case, they're not rendering at 4K, and thus my remarks quite obviously don't apply.
To a limited degree. A benchmark at 1080p with a 5090 tells you absolutely nothing about what performance you'll see at 4K with a 5060.
Again, this is where you're wrong. CPU tests aren't made to answer one person's quest of "how this CPU would perform on my build?". That is absurd, do you expect every reviewer to test every single combination of CPUs and GPUs in existance for your convience? Or do you expect reviewers to adopt your personal build as their test benches, to the detriment of everyone else? Either way you are delusional.
Again, you couldn't be more wrong. The actual answer is "this test tells me exactly how this chip performs in an objective manner, I need data from a different test to know if it makes sense to pair it with my particular GPU".
Except, once again, that is complete nonsense. Nothing changed in the software between 2019 and 2024. You can see that by looking at the low resolution tests, in both 2019 and 2024 you get the exact same result, the 3700X is a bit over 10% faster than the 2700X in both cases. That wouldn't have been the case if the software changed. But it was, the >10% advantage of the 3700X was always there, even in 2019.
That's completely unrelated to the topic at hand. The 3700X and 3900X are tied because past a certain point (6 to 8 cores) game performance doesn't scale with core count. You still see the exact same thing today, the 9700X is barely faster (like 5% to 10%) than the 9600X despite having 33% more cores. And the 9900X/9950X are not any faster than the 9700X no matter how many cores you add. That's just the nature of games, it has nothing to do with resolution or GPUs. You pay the extra money if you need the productivity performance, otherwise buying more than 6 to 8 cores for gaming is pointless.
It shows a 0.2% difference between them. Again sure if you play at native 4k that might be all you need but that data is just showing you how the 4090 performs because most of the top end CPUs are clearly GPU bottlenecked.
Besides the 2700x it shows a 15% difference between the slowest and fastest CPUs tested. Based on that data I should get an 11400f because the 9800x3d is only 15% faster. Is that really a significant boost? Yes people don’t need a crazy expensive CPU but data like this could lead them to cheap out too much and then they’ll wonder why their CPU is holding them back soon after.
You’re right it doesn’t. But a 5060 at 4k will be almost entirely GPU limited and so you’ll just see how the 5060 performs at 4k. Which you can already see in benchmarks for the 5060. Again there are just too many combinations. Look at CPU benchmarks. Then look at GPU benchmarks. Shouldn’t be a crazy expectation when you’re thinking about spending hundreds to thousands of dollars on PC parts.
Was this a joke? What other reason does one have for a CPU benchmark, other than to understand how that CPU can benefit them?
That's a rather nice goal-post move, but we're discussing testing at a specific resolution. Many reviewers do indeed test at both 1080p and 4K; some even test at 1440p, though the bracketing by the first two are generally sufficient. And in the the link you yourself provided, the reviewer even tests at 720p.
Again: the facts prove you wrong. The actual list of games tested are listed in both reviews; did you fail to read it?
Yes, and beyond a certain CPU speed level, game performance doesn't scale either, as it becomes entirely GPU-bound. Glad you now agree.
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. I don't understand why you struggle so much to comprehend this.
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, all within margin or error of each other, because the tests are GPU-limited and say nothing about the CPUs. That test means jack squat to anyone who has a different GPU than the one used in their test.
What facts? lmao
The point where games become GPU-bound changes per game and per GPU. Also, new, faster GPUs will continue to come out, and people will continue to upgrade their GPUs over time. Trying to measure where CPUs become GPU-limited in CPU tests is a fool's errand. The only thing that makes sense is to remove GPU limitations entirely. That's the only way CPU tests produce useful data, and allows you to make a proper decision about what CPU is a good match for whatever GPU you have.
