NVIDIA GeForce RTX 3080 Performance Benchmark on iRacing

We continue with the races on iRacing, this time, we have new collaboration by Davo. An interesting sample of the potential of these graphics on iRacing. Let’s start.

There are many articles and forum posts that I have read about this, but I have decided to perform my own tests (with the help of CapFrameX software) and show you the results.

My goal is to help us to identify which graphic options are the most resource consuming and, analyzing the results, to obtain the best possible graphic configuration according to our needs. In my case, we start from the following hardware configuration:

• AMD Ryzen 5 3600x
• 32Gb RAM DDR4 3200Mhz
• GPU Asus TUF RTX3080 OC
• Monitor: Gygabyte g34wqc ultrawide (3440×1440) 144hz

Therefore, in my case the goal is to run the game at the highest possible resolution on a single monitor at 144fps constant. We start from the assumption that we will not always have those 144 fps because, let’s not fool ourselves, the graphics engine of iracing has many “secrets” and, depending on the circuit, number of cars in the race, etc. fps oscillate wildly. That’s why we will look for a configuration that allows us to keep those 144 fps constant most of the time.

Without further ado, let it be clear that my intention is not to confuse you with technicalities or a low level detail by changing options one by one, let’s get to the point and see which options are the ones that take away the most fps. This guide should be valid, at least, for those who have nvidia graphics, regardless of whether you play on a single monitor (as in my case), with triple monitor or vr.

For this we are going to start from the minimum configuration of iRacing, and we will add graphic elements progressively to see the performance penalty. On the other hand, we have 3 other elements active by default and, later, we will see if we lose or gain performance by disabling them:

• Resizable BAR (equivalent to the SAM in AMD)
• Game Mode (Windows 10 Settings)

• Windows 10 Hardware Accelerated GPU Scheduling

The measurements will be performed during the first lap of a replay of a race at the Nurburgring GP circuit, during the daytime.

The graphic options that we are going to comment, you will have to change them in the “Graphics” tab in Options. In my case, to make the tests, I have changed them in the tab “Replay”. Although there are some more options in the Graphics tab, the most important ones are also found in Replay.

Before we start, let’s clarify 3 points: the first one is that, in the Graphics tab, you must have the sliders of both graphics memory and system memory at maximum. The second is that, in this study, we have disabled the LOD, so that it does not affect the result of the tests. The advisable by iRacing regarding the LOD is to set the fps value to the minimum desired (60, for example) and leave the default values. Personally I have the LOD disabled because I notice micro-cuts and stuttering when I have it enabled. In the concussions we will comment a little more on this. The third and last point has to do with the option to configure the number of simultaneous cars we see on screen. Since the replay we have used to make the measurements involves 22 cars, we have left the configuration as follows:

The value in parentheses represents the number of simultaneous cars that are drawn in the rear-view mirrors. In our case we do not care about this value, since the camera we use in the replay is TV1. Also note that, for obvious reasons (we will lose a lot of fps), we will not test activating / deactivating the rear-view mirrors.

Let’s get down to business:

TEST 1

---

We start from the lowest values that we can configure:

Not bad, about 293 fps on average. That’s plenty. For those who have a 3080 and miss the 1845Mhz, comment that I have it with undervolt from day one. Later we will comment on the consumption as well.

Accumulated performance loss: 0%

TEST 2

---

Let’s play with the values on the left side. These values are CPU dependent so, depending on the CPU you have (the more single-core performance the better), it will affect us to a greater or lesser extent.

We put all the options in Low Detail:

 

As we can see, performance has hardly been affected, the CPU can handle it.

Cumulative performance loss: 1%.

TEST 3

---

We are going to give a little more power and we are going to give more quality to important elements of the simulator, keeping the part of the stands low:

We obtain a performance loss of 10%. It seems that the CPU power starts to be not enough. Curious that the values of % of use of both gpu and cpu come out practically nailed to the previous test. We begin to see the “secrets” of the iRacing graphics engine.

Accumulated performance loss: 10.4%.

TEST 4

---

Come on, everything at full throttle and let’s see how this affects performance:

We add 8% more performance loss compared to the previous test.
Curiously we see that the % of average CPU usage has dropped to 73%, to see if it is going to be that the iRacing engine does not take advantage of the CPU….hahaha (ironic mode)

Accumulated performance loss: 18.1%.

TEST 5

---

We move on to the central part of the options. From here it is supposed to be already GPU dependent options. We are going to activate only the shadow maps.

The “Night Shadow Map” option will not be activated in this case since our tests are performed during the day, not at night. This option is advisable to have it activated by default since it even helps to improve the performance in night races as it is based on predefined shadows.

First hack… Yes, I’m afraid we’ve hit the most resource-intensive option, and added 11.2% more performance loss. And yes, the average GPU usage has dropped.

Cumulative performance loss: 29.3%.

TEST 6

---

Activate the dynamic drawing of shadows:

We get a 19.5% performance loss compared to the previous test. And yes, we continue with the decrease of the average CPU and GPU usage…ains those “secrets” of the iRacing engine.

Cumulative performance loss: 48.8%.

TEST 7

---

We can’t miss the opportunity to test this famous graphic option, those big trees. In nurbur gp there are few, but well, there are some.

In this case there is hardly any performance loss, 0.4%. I understand that in circuits with more forestation it is possible that this % may increase, but it should not be too much either.

Accumulated performance loss: 49%.

TEST 8

---

Let’s go with the options that are supposed to affect the memory consumption of the GPU.
As I mentioned at the beginning, it is important that in the Graphics tab we have the memory usage bar at maximum.

The Motion Blur, in these tests I have not used it, in fact I have never activated it in iRacing…moreover, do any of you have it?

Come on, let’s activate all the options:

We have a performance loss of 4.6% with respect to the previous test.

Cumulative performance loss: 51.3%.

External iRacing Graphics

---

Taking as reference this last test we have done, with most of the graphic options at maximum, we are going to play with several graphic options that we find outside iRacing.
2 of them we have in Windows 10:

• Game Mode
• Hardware accelerated GPU programming

I remember that the tests have been performed with these options active. So, we start by disabling the game mode:

It seems that the game mode does not affect iRacing too much, we even gain some performance by disabling it, but nothing remarkable.

Let’s disable the GPU theme by HW:

This time we do have a gain of 0.8%. Something is something, let’s leave it enabled by default. Finally let’s disable the Resizable BAR that we have in the RTX 3000 series (Ampere), also known as SAM in AMD:

Although the average fps performance is the same, there is an important point. A higher percentage in the 0.1% and 1% percentiles of the test tells us that having BAR enabled helps the minimum fps values to be higher. To make it clearer, let’s put the same graph only with the fps values:

We have gained 2fps in the worst frame drop we have ever had. Although it is still a mere 1%, this is important in iRacing, something is something.
So, if you have the option to activate it, let’s leave it activated.

Conclusions

---

This is the graphical summary of the tests we have been carrying out:

The graphic options that consume more resources are those that have to do with the management of shadows, and we should start here. My advice is to remove them directly or, at most, leave only the “Shadow map…” options enabled. Regarding the GPU memory consuming options, I advise to have all of them enabled (except Motion Blur) if you have a GPU with at least 6Gb of VRAM.

As for the CPU dependent options, I would sacrifice elements that only give us a more beautiful environment, such as the sky, the stands, the people…and the pits, well man, it’s always nice to see them in detail, let’s leave it at least medium.

About the LOD…I’ll sum it up quickly: if you have a monitor with freesync premium don’t activate it, why? Because freesync premium already has LFC (low frame compensation) which helps us not to notice a drop in frames below the freesync range (48-144 in my case). With LOD activated I have micro-cuts and stuttering in a more pronounced way.
And finally, the number of cars shown on screen, although we have not tested it here, obviously affects. In the replays we can put as many as we want because it looks nice. In the race with 20 I think it is more than enough, even we can lower it even more if we have a tight pc. And the mirrors… I’m used to the central mirror, I always prefer to be more aware of what happens in front of me than what comes behind me. For multiclass put a good spotter to warn you and that’s it.

Let’s make a final test with the recommended options, they would be these:

We see that we have a good balance with this configuration. The minimum fps we had in the test was 58.4fps. That said, if we remove all the shadows we will get a plus more performance, and if we go tight with the CPU we will reduce the options on the left.

FPS Limit and Perfomance

---

As we can see in this last test, the average fps is above the capacity of our monitor (144). So, let’s limit the fps and see what happens. Whenever we limit the fps we do not do it to the maximum fps that the monitor supports, let’s leave a margin to ensure that we are always in the freesync range, so let’s put as a limit 141 fps and see what happens:

Obviously, the average fps has remained at 141 but, as you can see,
we get less sudden fps drops, and we get a 68.2 fps minimum.
minimum. Let’s look at the sensors:

On the left we have the values without limiting fps and, on the right, with the fps limited to 141.

We can see that both the % of GPU and CPU usage (to a lesser extent) have been reduced and, therefore, so has the power consumption (and, of course, the temperature as well, which in summer is appreciated). In the GPU we have a reduction of 50W and, in the case of the CPU, 3.5W less or, what is the same, by limiting the fps we reduce by 20% the light consumption.

Without further ado, I hope this personal study will be useful to the rest of the community. Any doubt or suggestion here we are.

See you on the track!