AMD Ryzen For Simracing?

[pattikins]

For me its looking like this may just be a time i actually do a proper upgrade and just go for the 5600x, I'm completely out of touch with the current tech so really appreciate the advice.

Which GPU would you recommend to go with the 5600x? That's another big investment lol

Is your AMD RX5700 getting maxed out?
With my current system, a AMD 3600X (missed the delivery today of my 5600x... got to wait 'til Monday now ) & 1080ti, my gpu load is less than 50% with the Samsung Odyssey+.

If you want to save some money with the 5600X system, the B450 boards are much cheaper than the B550* & can be bios updated for the 5000 series cpu.

*The VRM, lan + PCIe 4.0 are bettter than what the B450 has to offer.

 

[RasmusP]

For me its looking like this may just be a time i actually do a proper upgrade and just go for the 5600x, I'm completely out of touch with the current tech so really appreciate the advice.

Which GPU would you recommend to go with the 5600x? That's another big investment lol

That's a very difficult question!

The 5600x is a nice CPU though and if you got the money and don't want to change your CPU in a year, then spending the money now might be worth it for you!

About the graphics card:
Instead of "matching components", you have a "possible fps range depending on settings and racing situation" for both CPU and CPU, which are different for each game.

So for example:
rFactor 2:
- alone on track in the sunshine:
-- CPU = 120-170 fps depending on shadows, track details and reflection levels
-- GPU = 80-250 fps depending on all settings

- racing in the rain and sunset with 40 AI cars:
-- CPU = 60-100 fps
-- GPU = 30-180 fps

How much the settings influence the CPU and the GPU and how the racing situations differ in performance is different for each sim.

Now most people mean by "matching components" to have "medium/high settings with some ultra for eye candy on their specific monitor resolution and good amount of AI cars or online opponents".

It's important to know that settings barely make a difference for CPU performance, but make a HUGE difference for the graphics card's performance.
While the racing situation makes a big difference for the CPU performance but not necessarily a big difference for the graphics card's performance.

Anyway right now there are basically 3 graphics card available to match a 5600x depending on the used resolution:

3060 ti
3070
3080/90

And respectively the AMD counter parts of course. 6700xt, 6800 and 6800xt. Or even a 6900xt.

So please tell us your resolution or the used monitor/vr headset.
The CPU is more important for simracing anyway though since you can always lower settings and sacrifice "eye candy" to get a fluent image.
But if your CPU isn't strong enough to maintain solid fps, making things look ugly will only give you a small boost.

 

[R5GTTurbo]

Is your AMD RX5700 getting maxed out?
With my current system, a AMD 3600X (missed the delivery today of my 5600x... got to wait 'til Monday now ) & 1080ti, my gpu load is less than 50% with the Samsung Odyssey+.

If you want to save some money with the 5600X system, the B450 boards are much cheaper than the B550* & can be bios updated for the 5000 series cpu.

*The VRM, lan + PCIe 4.0 are bettter than what the B450 has to offer.

That's a very difficult question!

The 5600x is a nice CPU though and if you got the money and don't want to change your CPU in a year, then spending the money now might be worth it for you!

About the graphics card:
Instead of "matching components", you have a "possible fps range depending on settings and racing situation" for both CPU and CPU, which are different for each game.

So for example:
rFactor 2:
- alone on track in the sunshine:
-- CPU = 120-170 fps depending on shadows, track details and reflection levels
-- GPU = 80-250 fps depending on all settings

- racing in the rain and sunset with 40 AI cars:
-- CPU = 60-100 fps
-- GPU = 30-180 fps

How much the settings influence the CPU and the GPU and how the racing situations differ in performance is different for each sim.

Now most people mean by "matching components" to have "medium/high settings with some ultra for eye candy on their specific monitor resolution and good amount of AI cars or online opponents".

It's important to know that settings barely make a difference for CPU performance, but make a HUGE difference for the graphics card's performance.
While the racing situation makes a big difference for the CPU performance but not necessarily a big difference for the graphics card's performance.

Anyway right now there are basically 3 graphics card available to match a 5600x depending on the used resolution:

3060 ti
3070
3080/90

And respectively the AMD counter parts of course. 6700xt, 6800 and 6800xt. Or even a 6900xt.

So please tell us your resolution or the used monitor/vr headset.
The CPU is more important for simracing anyway though since you can always lower settings and sacrifice "eye candy" to get a fluent image.
But if your CPU isn't strong enough to maintain solid fps, making things look ugly will only give you a small boost.

On my single screen at 4k, i don't have any issues with any settings. however in VR on my Rift S i have to lower all the settings significantly, single player is not too bad i can turn up most things to medium except super sampling but add a few car's or try and play on line and i have to have everything low and i still get frame drops where the ASW thing kicks in. I am not sure how to check the GPU usage while in VR? I can do it in single screen mode but now sure how to check in VR.

My favorite track which is nordschleife tourist has frame drops in some areas even in single player with everything on low.

 

[hobsey]

Hope you don't mind me dropping in on this?

I currently have a pile of new computer components ready for my first build along with the ryzen 5 5600x and 16gb ram 3200hz and an all in one cooler for a bit of overclocking

I plan to run triple 27" 144hz 1080p monitors (only because I already have 1 and plan to buy 2 of the same model)

What gpu would you recommend for that setup? Old or new model.

I plan to stay at 1080p for the foreseeable future unless I come in to some money.

Many thanks

 

[pattikins]

On my single screen at 4k, i don't have any issues with any settings. however in VR on my Rift S i have to lower all the settings significantly, single player is not too bad i can turn up most things to medium except super sampling but add a few car's or try and play on line and i have to have everything low and i still get frame drops where the ASW thing kicks in. I am not sure how to check the GPU usage while in VR? I can do it in single screen mode but now sure how to check in VR.

My favorite track which is nordschleife tourist has frame drops in some areas even in single player with everything on low.

It might be a good idea to install msi afterburner, that way you can check both the cpu/thread & gpu usage @ the same time. After installing afterburner, go into the settings, click on the 'monitoring' tab and select: 'gpu usage' and 'cpu usage' (all... inc. 1-8). Then run AC in VR. After closing AC, look at the afterburner monitoring graph for the cpu/gpu usage.

 

[RasmusP]

Hope you don't mind me dropping in on this?

I currently have a pile of new computer components ready for my first build along with the ryzen 5 5600x and 16gb ram 3200hz and an all in one cooler for a bit of overclocking

I plan to run triple 27" 144hz 1080p monitors (only because I already have 1 and plan to buy 2 of the same model)

What gpu would you recommend for that setup? Old or new model.

I plan to stay at 1080p for the foreseeable future unless I come in to some money.

Many thanks

Depends on the game! 3x 1080p is 30% less than 4k but still 1.6x 1440p.
I'm using 3440x1440, which is 25% less than 3x 1080p.
I just upgraded from a gtx 1070 to a rtx 3070.
Before I could use AC at very high settings and about 90 fps, acc at 60 fps with medium-high settings and rf2 at 80 fps with high settings.

Now I can max out ac with the grass rendering from custom shaders patch etc, almost max out rf2 and acc, all at 90 fps. I have a 100 hz gsync monitor so 97 is the maximum I should put the limiter at. 90 fps is definitely fluid enough imo, when using gsync/freesync.

I'd say 2080/3060ti is what should be a good price/performance gpu for you! Maybe a cheap 2070 super might be a good purchase too!

If you can find a good deal, a 3070 would be nice for high settings and good eye candy.

But a 1080 or 1080ti would still do a good job at medium to high settings!

It might be a good idea to install msi afterburner, that way you can check both the cpu/thread & gpu usage @ the same time. After installing afterburner, go into the settings, click on the 'monitoring' tab and select: 'gpu usage' and 'cpu usage' (all... inc. 1-8). Then run AC in VR. After closing AC, look at the afterburner monitoring graph for the cpu/gpu usage.

The cpu usage is pretty much useless though. As long as the overall cpu usage isn't getting close to 100%, it's the hidden single thread performance of the cpu that's limiting.

Just look at the graphics card load, which is activated for monitoring at default Afaik. If that's at 95% or more, your graphics card is limiting.
If not, it's either the vr vsync (40 or 80 fps for the rift s) or your single thread performance of the cpu.

It's a bit difficult when your rift s is staying in asw mode at 40 fps but your loads are fine.
Your theoretical limit would be between 40 to 80 fps then.
I would try to lower the vr supersampling or just a few settings and see if you can get the rift s to get up to 80 fps.
Then you'll see the "headroom" in the oculus tray tool or the gpu load in afterburner.

I'm not sure if you can get the rift s to not limit via vsync so you could actually see the maximum fps, like 60-70 fluctuating?

 

[hobsey]

Legend, thank you very much Rasmus.

Sorry, I should have said which games.

Mainly Raceroom, AMS2, occasional blast on PC2 and ACC. ACC not so much now since discovering Raceroom.

 

[RasmusP]

Legend, thank you very much Rasmus.

Sorry, I should have said which games.

Mainly Raceroom, AMS2, occasional blast on PC2 and ACC. ACC not so much now since discovering Raceroom.

Raceroom would run fine even with a gtx 1070.
Pc2 would start to be difficult, ams2 is fluctuating a lot in performance so you can't really say which gpu would be good for it.
Acc is very consistent in performance but always on a high level. My 1070 wasn't really enough to make it look good and have more than 60 fps..

I would aim for a used 1080ti or 2070 super or something better.

 

[Fi22le]

I had a 2070 super. For AMS2 it can run 3x1080 / 60fps locked with decent settings for dry / daytime conditions. Dawn and dusk it will struggle on some tracks. Night racing is a no (my 3080 can just about manage large grids at night).

 

[blekenbleu]

Torvalds, Linux's creator,
finds AMD's processors deliver a much bigger bang for the buck than Intel's CPUs
"The 'modern DRAM is so reliable that it doesn't need ECC' was always
a bedtime story for children that had been dropped on their heads a bit too many times."
"You can find me complaining about this literally for decades now.
I don't want to say 'I was right.' I want this fixed, and I want ECC."
AMD, which unofficially supports ECC memory, "did it. Intel didn't."

 

[Elaphe]

I can buy now a 10700KF or a 5600X for basically the same price. I need this upgrade from my old 4790K to improve the performance of Assetto Corsa with Oculus Rift S. I have a 1080 Ti GPU, which will be replaced in the future with a 3080 or a 3070Ti. What would you recommend?

 

[RasmusP]

I can buy now a 10700KF or a 5600X for basically the same price. I need this upgrade from my old 4790K to improve the performance of Assetto Corsa with Oculus Rift S. I have a 1080 Ti GPU, which will be replaced in the future with a 3080 or a 3070Ti. What would you recommend?

Oh that's a difficult question since the 10700kf is "proven to be awesome" and has 8 cores, compared to the 6 cores of the 5600x.

But the 10700k is difficult to cool. My 10600k is at 4.9 GHz at 1.290v and with a "Le Grand Macho RT" and 4x 140mm case fans in a Fractal Meshify S2 running at about 70°c while racing.
When I do a cinebench run, it would go to 87°c with all fans at 100%...
So I limited the max power draw to 100w, which keeps it at 82°c and throttles down when all cores are under load.

No game uses all of the CPU though and especially simracing only uses 2-4 threads.

The 5600x on the other hand has a lot lower power consumption so will easily stay cool with a good cpu cooler and some case ventilation.
It has better single thread performance, which is the important thing for simracing for probably now and the next 5 years from my expectations.

But to be fair I would just wait for the 5600x to drop down to what the 3600x did cost last year.

For comparison:
Cinebench R20 single thread bench:

- 5600x = 600
- 10700k = 512
- 10600k = 483

important to know though:
- the 10600k clocks a lot lower at default but can easily be clocked to the same frequency of the 10700k.
- both Intels can be overclocked easily to 520 points for both is more realistic.
- the 5600x will probably not really be overclockable. AMD pushes them quite well.

- the AMD cpus tend to clock down a lot when using more cores. You could see this with the 3000 Ryzen CPUs: Single thread they were better or similar to the Intels but when you did a 3 thread run in Cinebench, the Intels would be better.
This is due to the Intels being easily able to be overclocked to all cores pushing to the maximum frequency, while the AMDs run better with the automatic overclock.


Sorry for the lack of structure in my post but basically what I wanna say:
- I didn't watch any videos or read reviews about the 5600x being benchmarked in a 3-4 thread run in Cinebench. That would be the important test. Not all core benches and not the pure single thread benches...
- I have no clue about overclocking behaviour of the AMDs
- In theory the 5600x would be a lot better for simracing

 

[Elaphe]

Thank you for all the information. I think I will go for the 5600x. The lack of stock here in my country is a problem, and I guess it is the reason for the prince increase. Right now it can be purchased for 350€ = $430.

 

[pattikins]

...
- the AMD cpus tend to clock down a lot when using more cores. You could see this with the 3000 Ryzen CPUs: Single thread they were better or similar to the Intels but when you did a 3 thread run in Cinebench, the Intels would be better.
This is due to the Intels being easily able to be overclocked to all cores pushing to the maximum frequency, while the AMDs run better with the automatic overclock.


Sorry for the lack of structure in my post but basically what I wanna say:
- I didn't watch any videos or read reviews about the 5600x being benchmarked in a 3-4 thread run in Cinebench. That would be the important test. Not all core benches and not the pure single thread benches...
- I have no clue about overclocking behaviour of the AMDs
- In theory the 5600x would be a lot better for simracing

I don't know if there is much difference between Cinebench r15 vs R20, but here are my 5600X cpu results for 3 & 4 threads:
3 threads - 4.6ghz, 4.6ghz, 4.3ghz
4 threads - 4.5ghz, 4.5ghz, 4.4ghz, 3.5-8ghz

It was about 3.9ghz across all cores with the 12 thread test.

 

[RasmusP]

I don't know if there is much difference between Cinebench r15 vs R20, but here are my 5600X cpu results for 3 & 4 threads:
3 threads - 4.6ghz, 4.6ghz, 4.3ghz
4 threads - 4.5ghz, 4.5ghz, 4.4ghz, 3.5-8ghz

It was about 3.9ghz across all cores with the 12 thread test.

Awesome, thanks! Did you use r15 or r20?
For simracing, r15 is better. R20 uses avx instructions but apart from pcars2 to some extend, no sim uses avx...

Also: could you give us your scores for 3 and 4 threads too?
I'll post my 10600k @ 4.9 GHz too so we'll have some good data

 

[pattikins]

Looks like I got lucky, I used R15.
Both the cpu & memory (3200mhz) haven't been tweaked so I don't know how indicative the results are in comparison to other 5600x systems.

3 threads - 743
4 threads - 961

 

[RasmusP]

The Ryzen 3xxx and 4xxx CPUs are very well tuned by amd so you probably won't gain much by tweaking.
For Intels it's quite easy: just put all cores to the speed of the single core boost and buy a good cooler. Sometimes there's room for going a bit higher.

For AMD it's a different story. You can't overclock all cores to clock as high as the standard single core boost without a lot of tweaking.
And if you go lower, you'll lose performance when only loading 1-3 threads.

So the only viable route for the Ryzen 3000 CPUs was "pbo", precision boost overdrive, which dynamically overclocks all cores a little bit. But mostly it's 100 mhz to maybe 200 mhz... Not really worth it.
But I never did that myself and I don't know how the 4xxx series behave.
All I've read mentions the 4xxx series being even better optimized by Amd.

A 10600k on the other hand will drop to 4.5 GHz when running a 4 thread cinebench. Raising all cores to 4.9 GHz gives a noticeable boost!

I'll do the cinebench runs tomorrow!

 

[robt100]

Looks like I got lucky, I used R15.
Both the cpu & memory (3200mhz) haven't been tweaked so I don't know how indicative the results are in comparison to other 5600x systems.

3 threads - 743
4 threads - 961

Shows the jump between generations, 3 threads with my 3600 (non-x) hit a score of 606.....

For AMD it's a different story. You can't overclock all cores to clock as high as the standard single core boost without a lot of tweaking.

So the only viable route for the Ryzen 3000 CPUs was "pbo", precision boost overdrive, which dynamically overclocks all cores a little bit. But mostly it's 100 mhz to maybe 200 mhz... Not really worth it.

I think later 3000 series CPUs were a lot better as the process refined. My 3600 (non-x) quite happily went to 4.35GHz all core at 1.25v(150mhz above normal peak single core boost and about 0.2v less too), with about 5-10mins of adjusting. Tuning the ram on the other hand is very time consuming...and a pain when one move too far means a BIOS reset to get the system to boot again

But yeah, in general you could say you cant get much more out of an AMD processor, but from what I see the intel overclocks are mainly the boost clock plus maybe 100mhz, but put on all cores rather than just the one and removing the power/time limit. I would explain what I mean a bit better, but I'm rather tired

 

[RasmusP]

Looks like I got lucky, I used R15.
Both the cpu & memory (3200mhz) haven't been tweaked so I don't know how indicative the results are in comparison to other 5600x systems.

3 threads - 743
4 threads - 961

Shows the jump between generations, 3 threads with my 3600 (non-x) hit a score of 606.....

EDIT: my scores are for some reason really bad.. Seems like I'm having something eating up performance!
Single core is 211, which is okay... but 3 threads were only 515...
I've found Screenshots from June 2020 with 624 points, which makes a lot more sense...

My 3600 (non-x) quite happily went to 4.35GHz all core at 1.25v(150mhz above normal peak single core boost and about 0.2v less too), with about 5-10mins of adjusting.

As I said I never did it myself so thanks for "correcting" me! That sounds very nice and will definitely give a boost in simracing performance!

 

[pattikins]

EDIT: my scores are for some reason really bad.. Seems like I'm having something eating up performance!
Single core is 211, which is okay... but 3 threads were only 515...
I've found Screenshots from June 2020 with 624 points, which makes a lot more sense...

As I said I never did it myself so thanks for "correcting" me! That sounds very nice and will definitely give a boost in simracing performance!

I hope you find & solve the problem that is holding back the cpu performance. Here is a screenshot of my Cinebench r15 3 thread test result:

I've technically also got some issues with the 5600x performance across all cores. It seems to be held back by a 76W PPT limit (should be 88W) and it seems to have a definite impact on the overall score of all 12 threads (maybe even 4 threads!) - 1727 (probably less than what @robt100 is getting with his Ryzen 3600 ).