Hi everyone,

While it seems the 5800X3D is pretty much the best processor for MSFS out there, with plenty of benchmarks and bar charts suggesting it trades blows or exceeds Zen 4 or Raptor Lake; I can’t seem to find much data on P3D.  I’ve only seen a small handful of posts, as I’m searching to cast a wide net for opinions and feedback.

I was wondering does anyone use a 5800X3D with P3Dv5 who could share their experience?  I have a 5900X paired with a 3080 10GB on a 1440 panel with a birthday coming up, and am on the fence if I want to ditch the 5900X for the 5800X3D as an interim upgrade before having to change platforms entirely - maybe buy another year or two before another big upgrade.  It’ll feel weird not overclocking (as I like to tinker a little bit in that realm to squeeze a few more percentage points of performance out) and dropping 4 cores for some of the light photo editing and texture work I do, but if there are significant P3D gains along with MSFS and modern titles, maybe it’s worth it despite the drop in clock speed and cores.

Thanks.

Rob Ainscough is running tests with various graphics cards and you can see his results over at the FSL forum…

https://forums.flightsimlabs.com/topic/32288-amd-announce-new-processors/

32 minutes ago, Ray Proudfoot said:

Rob Ainscough is running tests with various graphics cards and you can see his results over at the FSL forum…

But unfortunately not the 5800X3D, which outperforms all CPUs (even the newer 7000 series) across all simulators. 😉

At the same time he is benching P3D with 4K Resolution + 4xSSAA, which results in an actual resolution of 7680*4320 (8K). This is cleary a GPU limited scenario, even with a 4090 and can't tell you anything about actual CPU performance. Maybe in 10 years we will have GPUs that can handle 8K while being CPU limited 🙂 At the moment, we can only dream about this possibility..

1 hour ago, CaptKornDog said:

and am on the fence if I want to ditch the 5900X for the 5800X3D as an interim upgrade before having to change platforms entirely - maybe buy another year or two before another big upgrade

This is a actually a good decision. It's the cheapest way to get decent CPU performance without spending too much money.

MSFS, P3D and X-Plane all benefit from the large cache. The other advantage is, that RAM performance (speed and latency) is not that important anymore. The downside of the 5800X3D is, in productive (multi-core) workloads, it falls behind your current CPU due to lower clocks. But this is only important if you actually use 12 Threads at max. performance currently, a use-case I guess only 5% of home users have.

Current rumors are that the new Zen4 X3D series (ETA early next year) don't have to deal that much with downclocking. But there are no official news from AMD about these CPUs. If it's true, these CPU might be really impressive.

I for myself use a 13900K with DDR5-7200CL34. Performance is amazing (I'm CPU limited at 1440p resolution), I rarely drop below 60 FPS in MSFS (Fenix, PMDG or Leonardo planes) and P3D is running around 100FPS+ with my settings. But this platform comes with a high price and is way more expensive than just dropping a 5800X3D on your current motherboard with most likely similar (or not that much slower) performance.

One thing you should check is if you are currently GPU limited with your graphics settings (which can be easily achieved using high SSAA settings in P3D), then a CPU upgrade will not really help. Personally with my 3090 and not using 4K resolution, I'm not. But you have to check your GPU usage to be sure 🙂

Edited November 3, 20223 yr by roesti

25 minutes ago, roesti said:

But unfortunately not the 5800X3D, which outperforms all CPUs (even the newer 7000 series) across all simulators. 😉

He’s ordered a 7950X. Why not discuss it with him?

@roesti, this comparison would appear to contradict your claim...

https://cpu.userbenchmark.com/Compare/Intel-Core-i9-13900K-vs-AMD-Ryzen-7-5800X3D/4129vsm1817839

55 minutes ago, Ray Proudfoot said:

He’s ordered a 7950X. Why not discuss it with him?

There's no discussion needed because the benchmarks are available already. I just wanted to help the OP, because he specifically asked about the 5800X3D and didn't know about this thread on the FSLabs forum until you linked it.

The 7950X falls behind the 5800X3D. Notice the 3D at the end. It's only this processor, which has ~100MB of L3 cache. The normal 5800X (without 3D) is of course slower than the new released 7950X. But the 7950X is already slower in games than the 7700X from the same generation. This is due to how AMDs CPU design works (7950X has two chiplets, the communication between these two chiplets induces an increased latency which doesn't matter in productive workloads but games don't like it).

5950X, 7950X etc. are all great CPUs for productivity workloads, especially if you need all this multicore performance. Games on the other hand like low latency, fast clocks (especially for the main thread, which is still the limiting factor today), cache (and to a small extent fast RAM if the cache limits). P3D, MSFS and X-Plane all fall into the gaming category (eventhough some people don't like to call the simulators "games", but technically they are).

44 minutes ago, Ray Proudfoot said:

@roesti, this comparison would appear to contradict your claim...

Not really. I already said the 13900K is (slightly!) faster than the 5800X3D, but not much (in games! Multi-Core productive Workload, the 5800X3D can't compete with the 13900K of course). And I personally own the 13900K like I said 🙂 But for @CaptKornDog with an existing AM4 platform, the upgrade to the 5800X3D makes more sense in comparison to buy the new intel platform with DDR5. The 5800X3D costs only about 300+ bucks today and he only needs to drop it in to get almost the same performance. This is almost a no-brainer (not saying his current 5900X is "bad", it's still decent).

Second Point, in case you don't know: Userbenchmark is NOT a serious website for performance comparison. There are various articles on independent computer news portals about this "benchmarking" page. They heavily favor Intel (with "questionable" scoring algorithms, for example "hugely more recent"), even if it's not the truth. Intel denied they "co-operate" with this site, so it might be some guy who just likes Intel who runs this page. The public opinion in the hardware scene is just: Don't trust userbenchmark for serious performance comparisons.

Here's an article from computerbase (a well-known german hardware page) https://www.computerbase.de/2020-11/userbenchmark-umstrittenes-ranking-zen3/

You may need to run it through google translate, but I've already read about this topic on English news sites and hardware forums.

I'm happy to discuss about CPU and GPU performance across flight simulation (maybe in a new topic) and Rob Ainscough can chime in if he wants, but I don't think this topic here is a reasonable place for this. I just wanted to help @CaptKornDog with his decision on this specific CPU. In the end, it's all about expectation and we simulation lovers are known to chase after FPS for decades, so there are no "miracles" one should expect. But the 5800X3D is a really decent CPU for simming and I say it from the perspective of an Intel User, using current bleeding edge technology (13900K).

Edit: Here is some recent benchmark (13900K is missing, but it doesn't matter) where you see the comparison between 5800X3D, 12900K and the Ryzen 7000 series. Since no website benches P3D, I can't give you that, but I can tell you MSFS performance increases translate almost 1:1 to P3D when it comes to CPU performance. If I had a 5800X3D i'd do a P3D comparison, but I can offer benchmarking P3D on the 13900K for you (with specific defined settings).

 

Edited November 3, 20223 yr by roesti

@roesti, mine was only a passing interest so I’ll leave you to liaise with CaptKornDog.

Regarding that fps comparison chart above the one for 3840*2160 would be interesting.

24 minutes ago, Ray Proudfoot said:

Regarding that fps comparison chart above the one for 3840*2160 would be interesting.

I see your point, but benchmarking at 4K is still not "good" for comparing CPUs. The reason is, the higher the resolution, the more the GPU limits. This is my "critic" with Rob's benchmarks on the FSLabs forums, because he does CPU comparisons with an effective resolution of 8K, which isn't quite working. He even mentions there he'd expected a better performance, but if I had seen this thread before, I could've told him in advance this would be the outcome. But I don't want to talk behind people's backs. His work there is still valuable, even if it's only for people having him test this now knowing, that they might not need a new CPU when they run their simulators at this insane resolution like he does 🙂

There's no GPU which doesn't limit in 8K resolution today.. Fortunately with the 4090 released, we now might have the first card not limiting at 4K resolution anymore (in a few scenarios, but not completely).

Usually CPU benchmarks are even done in 720p. Why is that so? Because only then we know what the CPU can deliver. The CPU and GPU are always indepdendent. Meaning if my CPU1 can deliver 180FPS, CPU2 120FPS and CPU3 84FPS but my GPU only 30FPS (at 4K), I will have always 30FPS at max. So I don't know which CPU will be faster in the end, but the CPU capability remains the same and comes into play when I maybe upgrade my GPU at later time or want to drive higher frame rates at lower resolution (and get better percentile FPS / 1% Lows).

Going mentally to the future, where we might have better graphics cards (at reasonable prices, cough), the CPU still will be good enough (in this specific scenario). So this 5800X3D in my linked benchmark will still deliver 180FPS in MSFS, even at 4K - if we have a GPU that can handle 180FPS @ 4K resolution by this time. It will always give you 180FPS in this scenarios, now and also in 5 or 10 years. But if your today's GPU is too "weak", you won't see these FPS numbers now, sure.

I also found an English article about UserBenchmark: https://www.tomshardware.com/news/userbenchmark-benchmark-change-criticism-amd-intel,40032.html It's still not recommended to use. It's a shame it appears on the first place if you google about performance comparisons, but I can only say: Try to avoid it, it's heavily biased towards Intel (and I almost only buy Intel hardware. But being an enthusiast, I know better and like honest benchmarks that focus on real metrics. UB doesn't give us that)

 

Edited November 3, 20223 yr by roesti

27 minutes ago, roesti said:

I see your point, but benchmarking at 4K is still not "good" for comparing CPUs. The reason is, the higher the resolution, the more the GPU limits. This is my "critic" with Rob's benchmarks on the FSLabs forums, because he does CPU comparisons with an effective resolution of 8K, which isn't quite working. He even mentions there he'd expected a better performance, but if I had seen this thread before, I could've told him in advance this would be the outcome. But I don't want to talk behind people's backs. His work there is still valuable, even if it's only for people having him test this now knowing, that they might not need a new CPU when they run their simulators at this insane resolution like he does

I’ve had a 4K display for 4 years. I limit fps to 30 as the BenQ is 30Hz capable and I enable VSync. That keeps performance very good but I don’t have all sliders to the right of course.

If my lowly 1080Ti can handle things more powerful CPUs and GPUs will enable higher settings whilst staying at 30fps.

I really don’t want to discuss Rob’s tests here and I’m aware the OP may not be interested in my views. So time for me to bow out.

Userbenchmark is not a credible site for benchmarking and hasn't been for some time.
https://www.notebookcheck.net/UserBenchmark-gets-banned-from-major-subreddit-due-to-drama-generation.461875.0.html

A better site is cpubenchmark.net.
Alternatively, any of the popular YouTube benchmarkers like Gamers Nexus, Linus Tech Tips or JayzTwoCents.

@CaptKornDog, see below.

 

Thanks for linking this topic. It's a bit of a shame that he didn't talk about his settings. Upgrading a CPU can always be zero gains or huge ones. It all depends on settings and that's why I always recommend testing CPU / GPU limits on the users side.

It's actually that easy:

Disable any AA setting and put your resolution down to 720p. If the FPS do NOT increase, a new CPU will help. If it does increase significantly, then your GPU limits. So in the 2nd case, if the FPS you get at a low resolution is enough for you, you won't need a new CPU.

Of course there are more accurate ways to measurements with watching usage of the compoments, but this quick and simple test will give you a quick answer.

Edited November 3, 20223 yr by roesti

2 hours ago, roesti said:

I also found an English article about UserBenchmark: https://www.tomshardware.com/news/userbenchmark-benchmark-change-criticism-amd-intel,40032.html It's still not recommended to use. It's a shame it appears on the first place if you google about performance comparisons, but I can only say: Try to avoid it, it's heavily biased towards Intel (and I almost only buy Intel hardware. But being an enthusiast, I know better and like honest benchmarks that focus on real metrics. UB doesn't give us that)

But exactly what are they doing to be biased towards Intel ? And what is a benchmark that focus on real metrics ? Why is UB not such a benchmark ?

24 minutes ago, F737MAX said:

Userbenchmark is not a credible site for benchmarking and hasn't been for some time.
https://www.notebookcheck.net/UserBenchmark-gets-banned-from-major-subreddit-due-to-drama-generation.461875.0.html

A better site is cpubenchmark.net.
Alternatively, any of the popular YouTube benchmarkers like Gamers Nexus, Linus Tech Tips or JayzTwoCents.

@CaptKornDog, see below.

 

No gain in P3D5 with 5800X3D, but what was you coming from ?

 

23 minutes ago, jfri said:

But exactly what are they doing to be biased towards Intel ? And what is a benchmark that focus on real metrics ? Why is UB not such a benchmark ?

Did you read the articles I linked and also the one by F737MAX? You will also find many stuff via google.

If you don't want to read it: Basically, UB does not state which kind of "metrics" they use. On one side, they state things like "single core performance" with gains towards Intel while this doesn't reflect with real world tests or benchmarks in actual software / games tested by serious hardware news media. Then they put thing in the score like "more recent" or other stuff, like when they evaluate "slightly better latency" in favor to old Intel processors - at the same time, latency benchmarks by real world testers even contradict these "scores" of Userbenchmark while there's no way to repdoruce the claims of this site.

If you are really interested in the topic, please read the two articles I linked.

 

So what is a benchmark what focuses on real metrics? A benchmark I can run myself, which in the best case uses an application or game existing in the real world. So that I, as a user, can draw conclusions for my own situation out of it. UserBenchmark does not give one any of this while also skewing it's own scores in a weird way where some 8 year old processors seem to perform on a level like a newer one for example because of weird "evaluating" of subscores.

 

20 minutes ago, jfri said:

No gain in P3D5 with 5800X3D, but what was you coming from ?

A single user without saying anything about his settings says "no gains". That's why I wrote it's a shame no settings are known. I say there can be severe gains depending on the settings and I think I explained why there is no absolute way to say if a CPU upgrade might help or not and gave instructions how to test it. Furthermore, I explained that a CPU upgrade rarely helps when using resolution at 4K and above, because most graphics card will limit the performance (today) at this resolution. The 4090 is the first card where you can be CPU limited in 4K (in some games).

Benchmarking software or games only work if the settings are known, so that a meaningful comparisons can be made. Otherwise it might happen users spend their hard earned money on hardware they don't need. If I don't know the settings, I can't say if a specific hardware upgrade helps. If user 1 plays in 4K and user 2 plays in 1080p, then user 2 would most likely see an FPS increase with a CPU upgrade, while user 1 does not.

Edited November 3, 20223 yr by roesti

27 minutes ago, roesti said:

A single user without saying anything about his settings says "no gains". That's why I wrote it's a shame no settings are known.

I’m that user. My post obviously wasn’t super useful, but I just didn’t see the massive gains that I did in MSFS going from a 5600X to the X3D. I run at 1440p and can provide whatever settings are useful. They haven’t really changed since I made that post.