So this one is gonna be a bit to read, sorry to all lazy people! Tests were done on my i7 6950X Extreme Edition, it was overclocked to 4Ghz (dont trust the banana bench, it says 3Ghz there). First thing i noticed was, that it was waaaaay over 100% realtime when just testing 20 vehicles, so i changed max vehicles to 60 and bam i overloaded the engine, not the hardware: At that point I still had 52GBs of RAM free, so, because I didnt want to modify the engine, I just set max vehicles to 48 and got a complete run: Impressive results if you ask me... keeping 110% realtime at 40 vehicles and a max of 355.285Mbeams/s And it seems like this CPU is unkillable with dynamic collisions off, I would love to see how many vehicles it takes to kill this CPU (i want it to atleast go under 25% realtime), but I would need help with getting the engine to not crash. If anybody has an idea MSG me.
This is from 0.7.0.2: This from 0.9.0.0: 3,5 means just the frequency. So 3.5ghz. Interestingly I seem to have Windows 6.2
You are hitting the limits of dynamic collision subsystem. The dynamic collision subsystem is constrained to 65536 nodes in total (over the whole map). I've never imagined that the core would be able to run 40+ vehicles in real time so the limit seemed reasonable at the time. Your beast of a machine proved me wrong I'll add it in my todo list to remove the limit, but it has low priority. Most of the machines are a lot less powerful, and the limit actually provides a little bit of performance. Edit: On second thought you might be hitting a different limit that is Lua's reserved memory (all the vehicles have a common Lua memory pool right now). This is more tricky to remove and it might take some time...
On my i7-3770 a change from 84.574 MBeams/s on 0.5.10 (8 vehicles real time) to 106.332 MBeams/s on 0.9.0.1 (12 vehicles real time) It's nice to see how well Ryzen is doing. I definitely know my next upgrade. PS: The UI cannot handle the vehicles of data over 14.
Amd......., what was that? I can kill that CPU with 4 cars, will take 6 at the most and won't need much of GPU for that either, just start spawning off road roamers at ECA town and have high beams on in each one of them, of course it will be quite unfair, but that is probably hardest for any CPU after these great improvements to physics performance.
Anyone with a Ryzen 5 (1400/1600) able to upload their BananaBenches. My i5-4288u MacBook pro is in dire need for an upgrade.
For those interested here is how Ryzen preforms at different core counts all with SMT enabled I tested 8-6-4-2 core which interestingly my idle temps never changed but load definitely did. sorry about the formating :/
Your earlier post was SMT disabled? I see it had bit higher scores at least with 1 and 2 cars which numbers I did look from every post with curiosity.
No my last post was also smt enabled which is why it said 16 core. It was probably higher as I just restarted.
Also if you run benchmark from folder without running BeamNG it gives bit higher results, so that adds bit to variability too. Have you tested without SMT? Has it much effect like I have read some claiming that SMT would slow down actual performance, which is bit odd to me? Either way, we would need a funnel to Cliff map so that we could send hordes of Cement mixers down and see that beautiful chaos
Yes, stock speed and voltages. RAM is only having slight adjustment, 1.25V instead of manufacturer's 1.2V, because I couldn't get them running 2666 speed with stock voltage. But other than that, stock settings. Amazing isn't it ?
Upped my vehicle limit to 48, it crashes at anything above that. Managed to get 264 Mbeam/s with 48+ cars in realtime, 28 with collision. Really good results though!
How do you have a higher score than me, my 6700k is at 4.8ghz. --- Post updated --- How do you change the max cars from 20 to more?