iRacing Benchmarks: Does the 5800X3D beat the 5800X and 3900X?
So I have had frame drops on iRacing despite upgrading to an RTX 3070 Ti lately. Since I didn’t think that the GPU would be the bottleneck of my system I was rather looking at my CPU, the Ryzen 9 3900X at the time. I decided to upgrade that to check some results and ordered the Ryzen 7 5800X, but made a last second decision to also give the Ryzen 5800X3D a test. There are a lot of reviews of those chips out in the wild, but nearly none focus on iRacing, which was my main pain. Since I was already benchmarking all 3 for myself I thought I might share my results here — and they are remarkable.
SPECS
Before you ask me why I didn’t compare it to the new 7000 Chips: I wanted to keep my Mainboard and RAM and keep my cost of upgrade as low as possible. After all these 5000 Chips are the last generation that was made for AMDs AM4 Socket, I’m sure many of you have the same thoughts.
So lets look at the CPUs spec sheets at first, on paper these CPUs look very similar, and the 5800X looks even worse than the 3900X: Less Cores, less Level 3 Cache and just a slightly better max boost clock.
The 5800X3D looks even more confusing, since it has less base clock speed and Max boost clock, although it has a vastly higher Level 3 Cache which AMD Calls 3D V-Cache. The Level 3 Cache is necessary for the cores to speak to each other, and it is way faster than DDR Memory.
You can think of it as having Post-Its directly on your desk compared to having them hidden in your drawer which you need to open each time you want to write something.
However the CPUs are also part of a newer generation which enables them to do more calculations per cycle, therefore having more computational power at same Clock Speeds as the older 3900X. AMD claims that this equals to about 20% so I was kinda expecting this boost from the 5800X.
How Multithreading works
For pure Multithreaded Applications like Rendering 3D Stuff with Blender, exporting Photos with Lightroom or even exporting a movie both newer CPUs with less cores are in a clear disadvantage, like benchmarks from other creators show.
For games it’s a bit tougher:
While the 3900X has 50% more cores than the others, this is not a pure advantage. Only a hand full of games can scale very well to more than 8 CPUs, and since those CPUs are not that common in gaming PCs yet, developers rarely spend any time optimising for this.
You could imagine being kinda like that — although this is just an example to give you a better idea what is happening behind the scenes.
2 Cores will do the Math for the physics of your car — tyre temperatures, G-Forces, Springs and Dampeners, Force Feedback, Damage implications, Setup Values.
2 Further Cores will calculate your opponents, while I don’t think that iRacing sends all Data like Tire Temperatures to your PC, it will send the position and speed of the other cars as well as some basic information on G-Forces so that the other cars can bump on roads and have their chassis react to that.
Then you have things like time calculation for lap times, splits etc as well as communication between you and the server.
And then there is also all that background stuff like windows itself, possibly crew chief, racelabs or spotify playing in the background.
In addition to more work for developers, a bigger amount of cores also leads to less max boost in general. While the 3900X claims to have pretty much the same max boost as the 5800X this only applies if a single Core is being used.
If you need full power from 4 cores the CPU has to manage it’s power consumption and will not clock 4 of the 12 cores to its max — you gotta remember that all 3 tested CPUs have the same claimed power consumption — and therefore thermal output — which is important for the cooler to be able to deal with the heat.
12 cores at max speed would be a lot hotter than 8 cores at max speed.
Why not just take the 5800X3D anyways?
So you might have heard of the 5800X3D and its performance before, why did I even think of buying the 5800X instead, apart from cost?
It turns out the 3D is not always faster.
While the 5800X consistently beats the 3900X, the 5800X3D sometimes loses to its cheaper and older sibling. While in F1 2021 the 3D is about 13% Faster than the normal X variant as Linus Tech Tips has shown,
For CSGO the 3D is actually 2% slower.
This was the main reason why I wanted to test the 5800X as well as the 5800X3D, since I found no results online for iRacing, and there was the possibility that the 3D is slower, and why would I want to spend more money for less performance.
TESTING SCENARIO
I did all these tests in a real world scenario, so I kept all the applications that I would normally use open, although some in the background. I also fully closed iRacing between all runs. The Apps Include:
- iRacing — obviously
- Crew Chief
- TradingPaints — Although the AI wouldn’t use them anyways.
- RacelabApps with 3 Overlays open
- Relative Window
- Standings
- Fuel Calculator
- 1SimRaceLab — which is a tool by VRS to automatically download setups and ghost files
- iRacing Config 1.4.0 to jump between different graphic settings
- CapFrameX to measure the benchmarks
As well as other applications that I usually never close but could impact the performance:
- 1 Password
- Glassware
- MalwareBytes
- Riot Vanguard since I play valorant from time to time
The other system specs are as following:
- Gigabyte B550M Aorus Mainboard
- 48 GB of DDR4 Ram @ 3200Mhz
- An Nvidia RTX 3070 Ti with 528.24 Drivers
- A Fanatic CSL DD Wheel with V3 Pedals
- 3 Screens which are a bit funky in configuration
- The middle one is a 1080p Screen @ 144 Hz
- The side screens are 1080p Screens @ 60 Hz
- I know the optimal solution would be that all of them are at 144 Hz but I had them sitting around, so why spend the extra cash for just side monitors
So in total the machine is pretty good, but far from Top of the line specs. I guess this makes the test results a bit more down to earth.
For the testing itself I did try my best to make the scenario as realistic as possible. Since replays in iRacing do not stress the system in the same way as real life racing does, I tried my best to mimic online Races and went for AI Races with 30 AI Drivers, I was sitting in a Ferrari GT3 while the rest of the field were random GT3 Cars. As far as I know the AI is calculated on iRacings Servers, therefore the stress on my end should be the same as in real races — minus the voice chats.
I did 3 times 3 Laps at Spa in the first testing segment, with standing starts to make the tests as equal as possible. Then I did the same again, although only time each for higher and lower graphical settings which I will get into detail later. Then I also did 6 Minutes in each graphic setting at Long Beach to verify that it is not just one map, although the results were comparable.
RESULTS
I had one Setup that I was already using and that I was kinda happy with.
Medium Details for Sky, Cars and Pit Objects, the rest on low, high detail particles, 4x Anti Aliasing and FSR at Balanced.
The results were still quite amazing, even at high graphic settings — where I thought only the GPU would limit the output. But since the GPU always has to wait for the CPU to calculate where the Cars are, the Frames can even dip there.
So I was expecting about 20% for the 5800X and 35% for the 5800X3D when I ordered them, but boy was I wrong.
There are a lot of numbers coming up now, but I’m gonna walk you through it. Let’s first look at the v1 Settings, my default settings.
I did split the results up into 3 categories — all 3 Laps including the Start, Only the starting Lap and then Laps 2 & 3.
The reason for that is that most people will experience the biggest lags at the start since all cars are packed together, and on top of that since I have a standing start there’s also lots of tire smoke etc coming off. Also in the first round I can make sure that the results are kinda equal, since the whole pack is still very tightly together, whereas in Laps 2 & 3 the field already begins to stretch.
So I have 1080 Seconds worth of benchmarks, which are close to 3 full rounds on spa, and I always started the recording of the benchmarks when the Red Lights went on, and restarted them for Laps 2 and 3 when I crossed the Finish line.
If we now look at the average frame values between the 3900X and the 5800X you can see that the newer Chip with less cores is always at an advantage, giving be 16% more frames. I was kinda expecting more, but that’s what I got when I looked at the numbers of the 5800X3D.
It’s just no competition at this point, the 5800X3D almost doubles the frames of the 3900X, and even the 5800X from the same generation looks very dumb.
If we look at the 1% lows, so the lowest 1 Percent of Frames, also known as dips, the results are similar, although the 5800X looks even worse here. The 5800X3D proves that the lows are not only cause by my GPU having troubles with smoke and whatnot, but that the CPU can still make a huge difference.
The same story basically repeats for the .2% Lows, Which equals to 21 Seconds of the 18 Minutes of testing, so like once every minute at worst.
v2
I then turned the graphic settings up a notch to go for High Detail on all Objects, 3 Cockpit Mirrors and that’s about it.
So now we get to the better graphical settings, and I expected less of a difference here since the work should be more on the GPU side, right?
I should still mention that the total test data here is a lot less than with the v1, since I only tested one race or 6 Minutes of Benchmark, compared to the 18 Minutes for the v1, but you gotta give me some slack here, I didn’t have all the time in the world, especially since I don’t do Youtube videos in general.
But still: On average the better CPUs both performed a lot better than the old 3900X here.
Although you can see a bigger advantage on Laps 2 & 3 since the field was stretching and the GPU had lass Cars to draw.
By the way: You could basically think of this test as if you were running lower settings on a card like a GTX 1080, since the graphics are a bit more stressful for my GPU as well.
v3
For the Last setting I turned everything all the way down to make sure the GPU wasn’t in the way for the testing results.
So let’s get to the real beatdown here, the lowest graphic settings, where the CPU can really show off, and they did.
Finally the 5800X is also flexing compared to the 3900X, boosting the frame count by about 80% on average — although the difference in the 1% lows is not that big of a deal.
But the 5800X3D is just a beast in this scenario, 168% better than the 3900X is just insane, it’s also nearly 50% faster than the 5800X.
CONCLUSION
If we take an average of all my graphic settings on both Spa as well as Long Beach the 5800X provides a solid 37% increase in frames compared to my old 3900X while having the same clock speed. This was more than I was expecting, although the first round of tests were a bit underwhelming to say the least.
I am glad that I also gave the 5800X3D a test because boy… this CPU is freaking fast. I did not expect have over double the frames while using the same GPU, especially since openHardwareMonitor or the activity Manager of Windows never showed more than 70% Usage, looking at specific cores. So for me it felt like… well since the CPU isn’t running at 100% on at least one core, it can’t be the bottleneck.
Well here I am proven wrong.
So in conclusion, if you don’t want to upgrade to the AM5 Chipset yet and you are still running on AM4, the 5800X3D will boost your Frames by a lot. I think it’s even worth it upgrading from the 5800X since the values are so devastating.
I would honestly avoid buying the 5800X if you can afford to for the 5800X3D, and since you are simracing I guess you do invest in your hobby anyways, so make sure to get the best deal right away.
That’s it for me, I’m gonna go ahead and try to increase my iRating once again, since I kinda lost quite a bit amount since switching from ir04 to GT3 anyways.
