BANANA Benchmark Results Comparison

I don't see the point in disabling the P-Cores. You'll be paying more for a product you won't use to its full potential. If all you're worried is power consumption, buy a less power hungry cpu.
According to this article which tests the 12900K, you'll end up with a cpu that's more similar to a Ryzen 3 3300X than to an i7 while paying ~$200 more.

 

S

So....you have same thinking as me
For me and I think, nearly every user will using it full potential and not using in that stupid way
And what the best CPU now on 200-250$ range?

 

Since I missed the note in the changelog of this update, I've added a new page to the Bananabench sheet for the results on version 0.28 and newer.
As always, link to the spreadsheet presentation post: https://www.beamng.com/threads/banana-benchmark-results-comparison.8595/page-49#post-1047156

 

hello, i think column "Max. Cars Before %Realtime <100" not really interesting.. cuz i play with 7 AI car's normally. even with 7 cars i feel not comfy, feel some input lag...
40 AI car absolutely unplayable.
or i don't get what mean this column

 

So finally somebody did a 7950x3d benchmark and on Jungle rock island can spawn 40 cars at around 50 fps.

So faster than 7950x, About the speed of 13900K, looses to 14900K.

But due to its upgrade path may aswell get a 7950x3d

 

That is a theoretical limit after which the simulation engine won't run at realtime making the simulation not accurate. The number of cars there has nothing to do with the number of cars you can spawn in game. As far as I know, bananabench spawns a regular d-series without the 3d mesh, textures and sounds, just the node and beam skeleton which is lighter than the vehicle you see in game.
If any dev wants to chime in, either to correct me or go more in depth please do as I myself am curious.

 

But it is 20% more expensive at my region
And cache do very much here. But for 14900K, just to climb that perf, need >350 watts

 

1: i have 4x16gb ddr4 2400mhz cl14 in quad channel which when i upgrade will be dual channel unfortunately, so the true latency is like 11.6ns, which isn't too horrible but also not that good.
2: i upgraded my gpu some time ago when i found a good on a 6700 xt so the gpu should be my bottleneck especially because i play in 4k.
3: I got lucky with a second hand Gigabyte Z690 AORUS ELITE DDR4 for a 100 euro so i couldn't say no to that. If you really want to overclock you would have to get a z690 or z790 and there aren't many m-atx boards like that, most m-atx boards are b660 and b760 which you can't overclock with although you could still undervolt when using b660 and b760 (i think you have to check on that yourself) I had a quick look and the best m-atx motherboards with good vrms are definitely MSI MAG Z690M MORTAR WIFI - ASUS PRIME Z790M-PLUS D4and ASUS ROG Strix Z690-G GAMING WIFI but the asus is the most expensive, i would buy the msi one myself if i was looking for a matx board

 

1: upgrade to the 3600 CL<20
2: yeah it is the GPU bottlenecks. 6700xt meant to play at 2K and not 4K. Probably get an 3090/6950xt which is significantly cheaper now
3: Thx for your answer

 

I AM BACK! With another maybe useful contribution to the spreadsheet?

****Here's a 14700k with a mild overclock and a 285w power limit paired with 2x2x16gb=64GB of dogshit corsair 2400mt/s cl14 ddr4 ram.

A quick test in west coast USA with 40 cars and traffic enabled with a 30fps framerate limit results in minimums around 25fps.

My overclock isnt really tuned for all core workloads like beamng but more for single threaded games, but nonetheless i get 35k+ scores in cinebench r23 so that's something.

fixed vcore of 1.3v with 4 cores going up to 5.8ghz and the rest at 5.7ghz and e cores stock****

UPDATE: Page 73, Post #1453

 

Im here to ask if a 5900x Or 3950x Would be better at beamng? Because i may be thinking of switching to a 3950x if possible

 

Thanks for posting this!

 

You should either switch to 5800x3d or 5950x. 5900x only give same performance as 3950x.
Here now I have been targeted the 7840HS laptops with integrated graphics (780M). Anyone tested it?
--- Post updated ---

You have more score than the normal 14700K. Only saw it around 900-930 Mbeams (even see 945) with 3600 cl16/6400 cl32. If you switch to 3600 cl16 and turn off power limiit, you could score up to 1000 Mbeams

 

Turns out i'm an idiot I was heavily GPU bottlenecked with those 40 cars playing with a 6700xt at 4k ultra. (Who could have known spawning cars could also affect gpu load and not only cpu?)
But when i keep the resolution at 4k but with normal graphics setting preset in west coast USA 40 cars traffic, I CAN GET +40FPS!!!!

Proof:

Also turns out that overclock was not stable when rendering a long 4k HDR video in Davinci Resolve Studio - Goes to show you that 30min of Cinebench and 3 hours of Aida 64 is not proof that your system is stable if you do "actual" work on your Computer and not just gaming. (Never Crashed in gaming)

BUT IT'S EVEN BETTER NOW! I GET 976 Mbeams/s!!!!
Changed to a fixed all core overclock of 5.7ghz P-core and 4.5ghz E-core, Fixed v-core 1.35v 300W Limit ROCK SOLID. (Funny I also used fixed 1.35v 4.5ghz with my 5930K)
Lite Load / Load Calibration at "extreme"/2nd highest setting otherwise cpu drops vcore too much like seriously i put in 1.35v but if you leave LLC at default it drops it automatically to 1.28v under load which is way too low for 5.7ghz

Maybe i won silicon lottery? Do you also have a 14700k? I never really tested banana bench at stock cpu setting. I do want the power limit because i only have 1 CPU EPS connector and i read that gives maximum 336W so i'll stay safe with 300W.
I will definitely upgrade the ram because as you know I upgraded from a 5930K which I have now given to my sister with some random 16gb ddr4 i got for free.
So I'll buy a new set of 64gb DDR4 3600 after i sell some old 16gb ddr3 1600mhz kits, I'll give my sister my 2400mhz ddr4 and I will test again! The 1000Mbeam/s WALL HAS TO BE BROKEN!!!!

P.S
I will test E-core only gaming Soon™

 

I DID IT! 1000Mbeams!!!

Screw 5.7ghz with avx offsets, 5.6ghz P-core 4.4ghz E-core 1.33v standard LLC.
I hit ~36k in CB23 and I see the cores hovering around 5.4-5.5ghz because of the power limit.

I don't need anymore, I'll probably put the power limit back to Intel spec of 255w instead of 300w and 1.35v for that extra stability because who's gonna notice a 1% difference and that'll be it for me.
This was quite fun to try breaking 1000, but it's not gonna matter when newer processors release. But such is life.

 

I havent see any proof of the 5800x3d beating the 5900x in beamng yet so i dont see it working and i haven't seen anything about the 5900x and 3950x being able to spawn the same amount of cars?

 

Nice! my best score in r23 is 30k with not everyday settings)

 

Congrats you are at 1 GigaBeam LoL
--- Post updated ---

.01 Pedabeam
.001 Exobeam
0.0001 Yottabeam

 

The integrated benchmark crashes after 20 something cars. I have a 14700K and don't think that it is because of overclocking but I will make an update if it is fixed by reducing the clock speed. Does anyone what else it could be?
--- Post updated ---

I fixed it. I reverted it back to 5.7 Ghz on core 1&2 and 5.6 Ghz for the other P-cores. The weird thing is that it is unstable at any other clock speed I tested from 5.7 Ghz for all P-cores down to 5 Ghz for all P-cores, does anyone know why that is the case?

 

When I was overclocking, I noticed something interesting. Even though Cinebench and Aida64 were running smoothly, Bananabench kept crashing after around 30 cars. It seems like Bananabench really puts a strain on your system when it uses more cars than you have threads.

Could you give my daily settings a try and see if they work for you? I've used adaptive voltage and set a negative offset below 5.6 to make sure I hit the highest frequency under all-core workload, because I'm hitting a power limitt. And above that, I've added a bit of voltage for extra stability. This way I can keep high frequencies during gaming while staying stable, and I'm getting around 980 mbeams per second in Bananabench. Oh, and I'm using a Gigabyte Z690 Aorus Elite motherboard, so your Motherboard may vary with different settings. You could probably beat my 1000 mbeam score depending on the ram you have.
Best of luck to you