Triple Screen Discussion

damn I really want to upgrade my i5-8400 to a 5800X3D now but it's so expensive. Why they don't make a 5600x3D

 

Not sure what got you all excited about the 5800X3D, but it's not worth it for simracing. I don't think even a 5800X is worth upgrading too tbh. It doesn't make that much difference. The 5600X for sure has the best value for your money.

 

Well it depends. With low resolutions you can quite easily be cpu bound and benefit from the upgrade if super high fps is what you're seeking. Not really worth it with 7560x1440 res triples.

 

What I'm saying is, it won't get you those super high fps compared to a 5600X. Especially not in ACC.

Just a friendly tip for @DCN.

 

You may have missed this: https://www.assettocorsa.net/forum/...-framerate-decrease.67725/page-3#post-1216172

 

Definitely missed that. Those are some big numbers for the X3D.

Not sure how usefull these are tho, since the lowest graphic settings possible were used. Not really a good use case if you ask me.

But I guess there will always be those who are willing to spend big bucks for pointless gains.

 

A 1440p low detail replay benchmark does not tell much about triple screen 7560x1440 performance in a race. Unfortunately almost all benchmarks are done at low settings (or a replay, skipping all the physics calculations). I'd like to have some benchmarks with high resolution to see if there is any improvement left.

 

As a VR user exclusively I tested according to my use case scenario, and in VR, that's low settings or 45fps with 90Hz reprojection which is a big no no.

The test were done to assess the CPU/RAM bottleneck, so although I know replay isn't perfect by any mean as a lot of the CPU load isn't done in replay, still, it is apple to apple comparison, in both CPU upgrades with same GPU, and the uplift came almost only from the CPU upgrade each time (at constant core count).

And I did the benchmark on my 1440p monitor avoid any change in the view point in VR (as I never kept the HW to do a back to back test, I just kept the methodology for each upgrade)

What I can tel you as well is that the HP Reverb has 2 panels of 2160x2160 ran at 90Hz/fps and I run them with 150% super sampling, my output resolution is 2704 x 2644 per eye.

That's 14 millions pixels driven at 90HZ when when a triple 1440p is 11 Millions times your refresh rate of choice.



The TLDR is that ACC is very CPU bound, and some graphics settings are pulling on the CPU too. So increasing the eye candies isn't only stressing your GPU, it also picks quite a bit of CPU power, it does so also with replays, so while playing with AI, or simply while driving when the physics engine is working full tilt, it's even worse performance for sure.

And with the pretty close amount of pixel triples 1440p drives in comparison to a good VR headset, I see no reason to have so much higher eye candies available for a smooth experience.
We run the same physics so our CPU constraints are the same, you have just a little more room to play with your GPU if you aim for 90Hz, but not so much if you expect 120Hz as the pixel throughput would then be the same than mine.

 

Wrong....the 5800X3D it's so much better than the 5800X for sim racing and literally any other game that is heavy on CPU single thread workloads (think nearly all DX11 games which are limited by pseudo-multithreaded congestion caused by the DX11 API). It's even stronger than the 12900K - I have friends who have FPS dips trying the exact same AI race I setup to do my testing (see my videos below).

The 5800X3D resolved all of my massive FPS drops at race starts, especailly in VR using the Reverb G2 - I was able to hold solid 90FPS no matter the situation, where as the 5800X was dropping to low 60's at race starts.
Before changing CPU's from 5800X to X3D (all other PC parts exactly the same - 3080ti, 32Gb DDR4), I created videos of the exact same 30 AI field at race start starting the same position, same weather, same time of day etc.
It's also much better for triples (I run 1440p), again for the same reasons as it was in VR. I can max out pretty much all the graphics in iRacing without issues, and of course any drops in FPS are not noticable with a good G-Sync certified monitor setup, unlike VR where it's a very unpleasant experience!

Clip of 5800X using G2:
https://drive.google.com/file/d/1FMETKnqyhp6TgHgtiHAI9idmCdGv48Rt/view?usp=sharing

Clip of 5800X3D using G2:
https://drive.google.com/file/d/1Pk91XCoamBNsk1cPpD7t2WGZjcKZW_I_/view?usp=sharing

@Snoeipaard - you are correct that the 5800X would be no better than the 5600X for gaming/sim racing at least in the majority of today's games.

@DCN - at this stage I would think about what you planned to do when the next generations of AMD/Intel RaptorLake CPU's launch in the coming months. If you planned to upgrade everything such as board, RAM (DDR5), CPU then it would make sense to wait for the launch. Since you are on 8th gen Intel you are kind of stuck in your choices, and its probably time for a total overhaul if you can afford to.
I'd think there will be a lot of used current gen parts for sale towards the year end if you can wait, but I would still wait and save a little more to get into the latest gen products to protect your upgrade path - on that note, I would likely go with AMD as their new socket for their upcoming Ryzen 7000 should last a lot longer than Intel, who tend to require board upgrades nearly every/every other generation - the current AM4 socket was good for 5 years, during that time Intel would have had you purchase quite a few new boards if you were upgrading to the latest CPU.

 

there are many factors such as
-Multiplayer vs single player with AI
-Resolution
-video settings (e.g Epic vs high)
to determine whether the CPU or GPU will reach the limit first.

I have a triple 1440 setup, my old cpu is the bottleneck. I use a custom video setting which gives me over 100fps during the hot lap mode but only 6x fps with a few AI

It’s a complicated topic with a lot of variables. You may be right that a 5600x is enough. It’s really hard to tell the minimum cpu upgrade I need to obtain a GPU bottleneck.

 

There's definitely a lot of factors that come in to play with this. The outcome will be different for every build.

Just wanted to warn you not to get too excited about the X3D. The test referred to in this thread isn't of any use for the average simrace setup. I've seen a lot of tests in different more common configurations and ofcourse there are gains, but none of them come close to what is advertised here.

 

Epic settings at high fps on tripples or in VR is like dreaming of a peaceful world... Not gonna happen anytime soon.

gains from 5800X to 5800X3D are real and the CPU bottleneck is real too HU tested with a 3090ti and DDR4 3800c16 you can't do better than this atm, and even then here are the results at MEDIUM settings





so you can see the CPU's are the bottlenecks at 1080p and 1440p MEDIUM, but the GPU is already (a 3090ti excuse me) a bottleneck at 4K MEDIUM, nevertheless the 5800x3D is performing way better (the % lows influences a lot more the smoothness perception imo)

So obviously you need the best performing CPU to get the AI, the physics and everything working full tilt while still being able to feed the GPU.

And by the way, no one is "advertising" anything in here, no one has any other interests than helping others.

 

Haven't watched the video, but going from what you're saying this is a test with a 3090ti? Again, hardly information that is of any use for the average person in here. The X3D won't generate these kind of gains on most builds.

 

Yes the specs are described in a specific chapter, and they announce the GPU at 3:25 ish

My point is why everyone is so eager to ask "we should have EPIC or HIGH" benchmarks when even a 3090ti with top of the line CPU isn't capable of achieving the dream of high res high refresh rate.

I agree with you, 3090ti isn't of any use for average person, so are EPIC and HIGH presets when someone seeks for ultimate performance, so my LOW / MEDIUM are a good and practical use case.

What I said as well, is that at my low settings (that aren't stressing the GPU, so that would fit in average Joe GPU capabilities), it's only when I upgraded from 5800X to 5800X3D that I was able to run 35 AI at 90 fps locked in VR at pretty high resolution.

If a 5800X wasn't able to do it, a 5600X won't either.

In my post linked by Andy70 you can see my test from 3700X to 5800X to 5800X3D, every CPU upgrade gave me the best uplift.

 

I understand what you're saying, I just don't agree .

I don't think testing with a 3090ti or even a 3080ti on low to medium settings will give you a general idea of what to expect when upgrading (to a X3D). Upgrading a system without such a high end GPU, won't have anywhere near the gains you have gotten. No matter the settings.

So that is why I commented. Just trying to manage expectations .

 

I agree we disagree, but I don't know why would you as facts are clear in my opinion. I do understand you may not share it but let's try one last time

If you haven't followed the posts I did (which is totally understandable ^_^) please go back to this post where I compare strictly 3700X to 5800X both on a 1080ti which by today's standards is an average GPU (2070/3060 like) and you have a more powerful GPU with your 3070 (although you have only 8Gb of VRAM and you would benefit from lowering the texture quality a bit, something to monitor at least)
https://www.assettocorsa.net/forum/...ed-but-framerate-decrease.67725/#post-1177908

This post linked above is the first iteration, and you'll see that the uplift of the CPU upgrade was massive, still not enough to run as much AI as I wanted, nor prevent from the occasional stutters even in MP.

5800X3D fixed that last part, so regardless of the resolution you want to drive, regardless of the level of details you want to use, the CPU limitation was always there and the 5800X3D fixed it.

The CPU limitation would be even more prevalent as you increase eye candies (some eye candies use CPU as well and pretty heavily) and resolution.

Now, I think I tried everything to convince you, if not, keep the 3600x, we're all free right?

 

I can agree with you on most games and sim racing titles, ACC is somewhat different. the X3D makes a huge difference, especially in 1% lows where you notice the stutters.

 

Haha, well I'm not staying with my 3600x thats for sure. After my switch to tripples, I'm now heavely CPU bound. A cheap upgrade(about €100) to the 5600x should solve that just fine. Paying an additional €300 for an extra 15 fps give or take (which I don't need btw) doesn't make sense to me.

Here's a video of a test with the X3D on a more average system and settings. Here the differences are more what you'd expect. Not anywhere near your 50% increase. Which however is awesome ofcourse. And logically you can expect the gains to be even less with systems with even lesser GPUs.

 

Benchmarking 1920x800

 

Is that weird? Seems to me that is just what you would need to do in order to show the actual difference the CPU upgrade makes, is it not?