AMD Ryzen For Simracing?

[Durge Driven]

What is wrong with doing these tests at high res ?
I am told because it is pointless ....okay

Well please enlighten me how doing it at 720p isn't even more pointless as you never ever going to use that resolution ...deary me

 

[Durge Driven]

Please show me at 4K what i will gain by tweaking my 3600 C14 samsung
I am stupid for not doing it apparently

If there is no CPU gains at 4K what does tweaking ram give you at 4K ? I am going to guess ....zilch

 

[RasmusP]

Please show me at 4K what i will gain by tweaking my 3600 C14 samsung
I am stupid for not doing it apparently

If there is no CPU gains at 4K what does tweaking ram give you at 4K ? I am going to guess ....zilch

You'll gain knowledge about what cpu you should get depending on how long you wanna keep it.

It's about trying to know if your cpu will still be fine in 6 years with a RTX6070 or not. If the 50€ more for the k might be worth it or not.
Or if you go vr, if the 10700k might even be a good choice if you hate building a new system and would like to spend more now (not ridiculously much as for an i9) to just swap graphics cards for the next decade.

It's a bit like with my 2600k vs the 2500k's of 3 friends of mine.
Guess who had playable fps at 60 fps/Hz vsync in novigrad in the Witcher 3?
Right, me, because of hyperthreading and oc.

And who do you think didn't have playable fps in the Witcher 3 although he had hyperthreading?
One of my friends who wanted to save some money and bought the 2600 non-k back in 2011.

The difference in pricing was a lot different back then though. Much closer together!

And ofc it might be a "better" upgrade path to just spend less now and buy a new mobo + cpu (+ddr5) in 3-4 years instead of spending too much in relation now and hang on for the last drop in your old cpu in 8 years, like I'm doing right now

Mentioned friends are now on 8700k, 9600k, 2700x, 4790k and one poor chap still plays on his 2500k. He would be able to spend the money but he rather bought a Mac book.

 

[Locoblade]

What is wrong with doing these tests at high res ?
I am told because it is pointless ....okay

Well please enlighten me how doing it at 720p isn't even more pointless as you never ever going to use that resolution ...deary me

What you're after is a SYSTEM benchmark, NOT a CPU benchmark. Running a CPU benchmark at ultra high resolutions is like trying to find the fastest sprinter in the world by holding a 10k race. Usain Bolt is categorically the faster runner compared to Mo Farah but by putting both in a 10k race Usain is being judged using a format that doesn't allow him to use all his speed.

 

[RasmusP]

What you're after is a SYSTEM benchmark, NOT a CPU benchmark. Running a CPU benchmark at ultra high resolutions is like trying to find the fastest sprinter in the world by holding a 10k race. Usain Bolt is categorically the faster runner compared to Mo Farah but by putting both in a 10k race Usain is being judged using a format that doesn't allow him to use all his speed.

Since this is a simracing forum I'd rather say it's like doing a Gearbox test for highspeeds (400 km/h desired) with a 600 HP engine that won't go above 330 km/h.

Is more than 600 HP really realistic? Who knows but it might be interesting for guys with a Lamborghini whether or not their gearbox will be okay for extreme tuning of the engine

 

[Scblacksunshine]

I think you're right on games that are not optimize for CPU and it's intense on CPU single core usage like Assetto Corsa. Even with my 3900X and 2080Ti, while I get good performance across the board on Ultrawide resolution setting, on car and track mods that are poor optimize, I can see still framerate dipping borderline 60fps with 15-20 cars. I kind of wonder if I would gain way better performance in scenario like this with the upcoming I9 10900K processor instead.

 

[RasmusP]

I think you're right on games that are not optimize for CPU and it's intense on CPU single core usage like Assetto Corsa. Even with my 3900X and 2080Ti, while I get good performance across the board on Ultrawide resolution setting, on car and track mods that are poor optimize, I can see still framerate dipping borderline 60fps with 15-20 cars. I kind of wonder if I would gain way better performance in scenario like this with the upcoming I9 10900K processor instead.

If you want to get comparable benchmark results, the community should run 3 and 4 thread cinebench r15 (r20 uses avx, sims apart from pcars 2 don't afaik).
The problem with ryzen 3xxx is that they are on the same level on 1 thread with high boost but as soon as you load 3 threads (so 1.5 cores if you have smt/HT), the clock speeds go down, while the Intel's are still at their 5.1 or whatever GHz...

I currently have an i7 2600k but it looks like tomorrow I'll get my i5 10600K (sadly won't overclock it due to my too weak psu).

If you want we can run some 3 thread cinebenches.
Assetto corsa uses 3 big cpu threads, although 2 are maxing out the single thread performance limit, while the third thread is just chilling around.

For this you have to download cinebench r15 and then go into:
File - preferences - check custom number of render threads and set it to 3.
Then hit "run" for the CPU.

Ignore OpenGL

I'll do it again with my 2600k now and tomorrow hopefully with the 10600k.

Should he interesting!

EDIT:
2600k @ stock (3.5 GHz all-core boost): 347 points
2600k @ 4.4 GHz all core: 440 points

 

[Case_]

566 points with the 3600 on PBO and Auto OC +50 MHz (so 4250 maximum single core boost).
575 points with 4250 fixed all-core OC.

 

[Case_]

On related news about boost...

 

[RobertR1]

@RasmusP what board and mem did you end up with?

 

[RasmusP]

@RasmusP what board and mem did you end up with?

"sadly" I went for the cheapest option.. Just couldn't justify the more expensive msi & CL14 combo.

- 10600k tray, 284€
(kf isn't available yet ; we have 2 years warranty from the seller in Germany so tray is okay)

- 16gb 3200 CL16 g.skill aegis, 62€
(my air cooling has enough airflow for this naked ram I think)

- gigabyte z490 gaming X, 169€
(on top you get 10€ on steam when you write a review)

= 515€ - 10€ steam = 505€

With CL14 memory (120€ being the cheapest) and the msi tomahawk (199€ and no steam bonus) i would've paid 603€.

For 100€ I'll rather buy a 3080 instead of 3070 when they arrive or just get a new, smaller and way faster cpu in 2-3 years instead of 5 years.

Already picked up the parts from the post office so gonna assemble everything later today and come back with some comparison.

 

[karoliens]

Ryzen R5 3600@4.1Ghz here. Got 569 points.

P.S. In a NODE 202

 

[robt100]

566 points with the 3600 on PBO and Auto OC +50 MHz (so 4250 maximum single core boost).
575 points with 4250 fixed all-core OC.

605 here 4.35ghz all core on a ryzen 3600.
203 on single core, so scaling quite nicely

Interestingly, had some fun with individual core overclocking in ryzen master the other day. By using my best core and putting cinebench R20 on a fixed affinity to that core in task manager, I managed to get a 'stable' 4.6ghz out of my best core . crashed when i tired to use all of them though would lower the clocks on the other core on that ccx then crash 6 seconds later. not bad for 1.28v though.

 

[RasmusP]

2600k @ stock (3.5 GHz all-core boost): 347 points
2600k @ 4.4 GHz all core: 440 points

566 points with the 3600 on PBO and Auto OC +50 MHz (so 4250 maximum single core boost).
575 points with 4250 fixed all-core OC.

Ryzen R5 3600@4.1Ghz here. Got 569 points.

P.S. In a NODE 202

605 here 4.35ghz all core on a ryzen 3600.

For the overview I'm gonna quote your scores.

Finished a rough and quick oc with my new setup. Frequencies fluctuating between 5.1 and 4.6 GHz with the cache raised from 43 to 45.

Results for 3 threads in cinebench r15 as we said:

10600k:
595 ; stock (4.1-4.8 GHz boosting
613 ; core oc (4.5-5.0 GHz boosting)
630 ; 5.1+4.5 cache GHz (4.6-5.1 GHz boosting)

Gonna see how it performs in games when everything is running... Just was curious about cinebench first and seeing how far I can easily push it

 

[robt100]

Just to add, i notice theres seems to be a disproportionate jump to my score vs the others for a small mhz bump, this may be due to having 3600mts RAM, whereas id assume most are on 3200?

 

[Case_]

Yeah, I'm on 3200 (well, overclocked 3000). Setups with 3600 or 3800 RAM certainly perform a bit better. Though 605 seems quite impressive still, but there's probably something else limiting my performance (MB, I guess), because even with the 2600, I was getting kinda low-ish scores.
I can try what I'll get with 4350, I think I should be able to reach that for the purpose of a single benchmark safely.

 

[Case_]

598 points at 4350 all-core.

 

[RobertR1]

Daily profile. Limited to 3 threads.

 

[RasmusP]

View attachment 377616

Daily profile. Limited to 3 threads.

700... Sick score!

 

[robt100]

32 points single and 95 points multi for at least £305 extra cost.....ouch!
Though I think it'd be interesting to see the difference FPS wise (ok, the 2080ti will also help!) to see a price per frame cost of the CPU+GPU on a system like the 9900k above, vs say myself and @Martin Fiala on an R5 3600 and 1660ti on a few racing sims.