Separate names with a comma.
Discussion in 'General Discussion' started by garyjpaterson, Jun 23, 2014.
Update, With everything working correctly I get a score of 625Mbeams/s. I also did it to 60 instead of 40. The benchmark claims "realtime" at up to 58 trucks however at 58 I get 16.5fps. I Have to go down to 45 trucks to stay above 30fps. As long as I stay under 36 trucks, it can lock at 60fps on vsync.
Amd R5 3600 stock no overclock
Bananabench is a pure physics benchmark, without the overhead of input, audio, graphics, and all other game related systems. It's not necessarily representative of what the performance may be for a fully running game.
Just swapped out my 3930K for an 8 core Xeon as I found one on CL for $20 and thought, why not? Turns out, doesn't perform as well, but it sure does run a heck of a lot cooler! (left is Xeon, right is 3930K)
Here is a run with my laptop, i5-8300H turbo boosted to 4Ghz in the start but dropped to around 2.9Ghz around the middle.
And also a ryzen r5 2600 on stock settings. Boosted to 3.7Ghz.
Ryzen 2xxx or 3xxx processors will not bottleneck a gpu in 98% of cases. Usually you hit cpu bottlenecks when you have either a poorly rendered/modeled scene with too many draw calls, or you have a lot of AI or physics going on eating up all the cpu. Ex: My 3700x stock speed with stock cooler does not bottleneck my 2070 Super graphics card.
In any case you crank up the resolution or settings on gpu-specific options and keep it busy more.
The reality of it is that unless you're running a single-threaded app, you won't notice the difference between a recent mainstream desktop intel or AMD chip with the same amount of cores or threads.
The perception that AMD is somehow slower is from the FX processor days (and it's associated APU's that don't have Ryzen branding), I had one of those FX processors for three weeks in 2014 and yes it stank so badly I replaced the entire system in leu of throwing it out the 2nd floor of the house.
On-topic, 18.1.x.x (the version I edit my maps with) running on a 3700x, with 3000mhz CL-15 Micron memory (2x 16gb dual-rank sticks). Consider 3000mhz CL-15 / 3200mhz CL-16 as a base-line for memory speed / latency when assembling a Ryzen system.
Configured as I have my system, one could expect to run 15~21 vehicles in-game at any time, without dropping below 20fps for more than just a random short dip, barring any other (usually GPU) limitations. If you stick to 15 vehicles or so in play, you can run 35~40fps most of the time.
No overclocking, PBO is 'off' (default), using stock cooler with liquid metal, only XMP memory profile used, the rest of the BIOS settings I did not touch.
Old version used to get 339.x mbeams/s as I still have the old capture too. Only uploaded the new one.
Getting some pretty bad results for some reason, maybe because I have ~30 tabs open at all times.
Left is Ryzen firstname.lastname@example.orgGHz + 16GB 3000CL16, right is 3600X@PBO + 32GB 3666CL16
TWO DIFFERENT GAME VERSIONS WILL LARGELY SKEW TESTS!!! This is very important as in 0.18.x I believe the D-series pickup greatly changed, again for 0.19.x too (though not as much, just added a few things, different beds, fender flares, other hitches, so-forth and so-on).
Outside of the above very important stuff, just comparing raw numbers...
Your 2700x compared to my 3700x looks only to be 10~15% slower, which is mostly IPC-based differences due to cache (memory inside cpu) latency and setup, and right around the ball-park for generational differences between 12nm Zen plus, and 7nm Zen 2 core designs.
Definitely seeing some OS or other applications interfering on 11 and 12 vehicle tests on your 3600x. When the score drops hard before physics has to stack (e.g. more cars than threads), it's indicative of this. Notice how my scores on my 3700x and your 2700x only dropped a few % at most (3%?) on the last one before it ran out of 'free' threads on the processor? That is due to OS and driver-related sub-routines, and sometimes other applications all running in the background contending for CPU throughput and resources.
DEFINITELY close out the browser 1st, I even usually close out Steam (though I didn't this time), and close out anything else like Epic (Epic Failure) games program, or Upay (Ubisoft Uplay), or Borigin (Origin/EA games interface) and things like that. Do not need to can the Antivirus or anything similar to run the test. Just don't have media players or anything else going or it will skew the tests.
AGAIN for good measure - Please make sure all tests to compare systems run on the SAME game version. Even a minor update can skew things! Mine was run on 18.1.x which was my editing version until just yesterday.
I will update that with a new performance metric next time I restart the PC (as I have maintenance routines running atm, and Rimworld).
updated score on v0.19.
Run a test on my i5-7600K. Not overclocked but maintained full turbo at 4,19 GHz through all the test.
After a PC restart:
I have I5-7500 and this my result
Tested my laptop: i7-3537U at 2,89 GHz
This is with a very easy overclock, i used to run it at 5.1GHz but funnily enough BeamNG was the only game that managed to demonstrate its instability. Will try to overclock again for a second test.
Pretty satisfactory for a CPU I got for £100.
yep! you have 20Mbeams less than me but paid 1/4th, that's great value. Sadly i got mine when ryzen CPUs were still in their early days (1st gen)
New Banana Bench. Still I5 6600K but OCd to 4.1Ghz. Also just checked with old banana bench scores of me and it got slower not much but still. Maybe i need to reinstall Windows because its still the same installment i had when i built the PC so it has been running like this for nearly 5 years now
Ryzen 7 4700U 8C/8T Lenovo Ideapad 16GB RAM (15W CPU)
High performance profile
How much time this test take ?