Latest Benchmarks - Intel vs AMD

Resolution: 2560x1440
Everything on minimum or off if possible, except the following:
Textures – High, Terrain Textures – High, Visib. range - high,
Res. Of cockpit Displays: 1024EF, Preload raidus: 60500,
Anisotropic filtering: 16x,[x] Full Screen, [x] Scale GUI

 

As you can see I’m probably (and hopefully) GPU bottlenecked, and difference between 6 y.o. setup is not big.

The way I measured it is I recorded a track(Caucasus map), where your location gets shot by MLRS after some time (which causes fps drop). Camera motions and plane motions in the track are the same (or close to being so I hope). All I did after is played track 3 times and averaged the values (I start playing the track by pressing pause and about half second later I press the button to record benchmark data with MSI afterburner for 3m:16s). I will attach the track I used (you need su-25A or flaming cliffs 3 to play it).

In first day of installation 3700x performance was actually worse in 1% lows by 10 fps, but today Ryzen gods had mercy and made it slightly better.

Thanks for doing this and providing the track. I'm thinking about going Ryzen, too, so I am always eager to see other people's results in DCS.

For comparison's sake, I ran your track file with the same settings on my PC:

CPU: i7 4790K @ 4.4GHz
RAM: 16GB DDR3-2400 (CL 11-13-13-31)
GPU: Gainward GTX 1080 GLH

Average FPS is obviously very different (mostly due to GPU, I guess: 1080 vs 1060). Interesting to see how your 0.1% on the Ryzen is quite a bit better despite the weaker GPU.

 

Average FPS is obviously very different (mostly due to GPU, I guess: 1080 vs 1060). Interesting to see how your 0.1% on the Ryzen is quite a bit better despite the weaker GPU.

Thing is i run tracks 3 times in a row for an average number , I think second and third tracks always get better performance in low 0.1% because probably less load stutters. I didn't get 2080Ti as i was planning and instead went for 2070S, which is probably bit closer to 1080. I only run test once (not 3 times like others, which will probably give better results) and i got these numbers: 53.6, 96.3, 169.8 (0.1%L, 1%L, Av) I also tweaked memory settings with ryzen DRAM calculator which allowed better latency.

Old setup:

CPU: i5 3570k OC (boosts up to 4.2GHz)
RAM: 24GB DDR3-1866 (CL 10-11-10)
GPU (same): Gigabyte GTX 1060 6GB

New setup:

CPU: Ryzen 3700x Stock (boosts to 4.3GHz+)
RAM: 32GB DDR4-3600 (CL 17-19-19)
GPU (same): Gigabyte GTX 1060 6GB

Game settings:

Resolution: 2560x1440
Everything on minimum or off if possible, except the following:
Textures – High, Terrain Textures – High, Visib. range - high,
Res. Of cockpit Displays: 1024EF, Preload raidus: 60500,
Anisotropic filtering: 16x,[x] Full Screen, [x] Scale GUI

Reshade: Default Lumasharpen and HDR effects on.

 

 

As you can see I’m probably (and hopefully) GPU bottlenecked, and difference between 6 y.o. setup is not big.

The way I measured it is I recorded a track(Caucasus map), where your location gets shot by MLRS after some time (which causes fps drop). Camera motions and plane motions in the track are the same (or close to being so I hope). All I did after is played track 3 times and averaged the values (I start playing the track by pressing pause and about half second later I press the button to record benchmark data with MSI afterburner for 3m:16s). I will attach the track I used (you need su-25A or flaming cliffs 3 to play it).

In first day of installation 3700x performance was actually worse in 1% lows by 10 fps, but today Ryzen gods had mercy and made it slightly better.

Thanks for doing this and providing the track. I'm thinking about going Ryzen, too, so I am always eager to see other people's results in DCS.

 

 

For comparison's sake, I ran your track file with the same settings on my PC:

 

 

CPU: i7 4790K @ 4.4GHz
RAM: 16GB DDR3-2400 (CL 11-13-13-31)
GPU: Gainward GTX 1080 GLH

 

 

Average FPS is obviously very different (mostly due to GPU, I guess: 1080 vs 1060). Interesting to see how your 0.1% on the Ryzen is quite a bit better despite the weaker GPU.

Thing is i run tracks 3 times in a row for an average number , I think second and third tracks always get better performance in low 0.1% because probably less load stutters. I didn't get 2080Ti as i was planning and instead went for 2070S, which is probably bit closer to 1080. I only run test once (not 3 times like others, which will probably give better results) and i got these numbers: 53.6, 96.3, 169.8 (0.1%L, 1%L, Av) I also tweaked memory settings with ryzen DRAM calculator which allowed better latency.

Heres my imput with same conditions:

Look what happens when DackSter94 changed from a 1060 to a 2070 Super on his 3700X (both results shown).

Conclusion: You can clearly see that the average FPS scales linearly with GPU (1080Ti is nearly 2X 1060 and 25% faster than 1080), 0.1% FPS may be linked to CPU providing no other bottleneck is introduced (like RAM shortage).

If you have a modern CPU, a good GPU is more important.

Bear in mind that when VULKAN implementation is complete it may change alot, for example by making DCS use more cores.

Benchmark.txt

Conclusion: You can clearly see that the average FPS scales linearly with GPU (1080Ti is nearly 2X 1060 and 25% faster than 1080), 0.1% FPS may be linked to CPU providing no other bottleneck is introduced (like RAM shortage).

If you have a modern CPU, a good GPU is more important.

Bear in mind that when VULKAN implementation is complete it may change alot, for example by making DCS use more cores.

Nice work putting it together.

After my first test i realised that my memory wasn't running on 17 CL. Bios took odd number and makes it even above(18). But that was for tests on gtx 1060. For my 2070s test I used ryzen dram calculator to calculate timings. So timings for RAM I used and use now with 2070s are 16-19-20. Although I doubt this change affects much.

This is my last measured results when overclock to 4.7 GHz.

- 2k resolution

- Without mirrors

DCS.exe benchmark completed, 35280 frames rendered in 198.916 s

Average framerate : 177.3 FPS

Minimum framerate : 103.3 FPS

Maximum framerate : 268.1 FPS

1% low framerate : 99.3 FPS

0.1% low framerate : 50.0 FPS

Measure your RAM latency with AIDA64. There should be 2-5ns gains from 19 to 16 primary timings.

I got down to 48ns from 57ns, 3000CL16-16-16-38 to 3200CL14-14-14-34.

Test your RAM thoroughly if you tune them.

I suggest IBTv2.6, prime95v26.6 and AIDA64.

Gesendet von iPhone mit Tapatalk

Hey
I have upgraded my system to 10700KF. Where can I download a benchmark to test it?
Sorry for bad English, I am writing via google translate)

Benchmark:

3DMark via Steam https://store.steampowered.com/

Passmark https://www.passmark.com/

Stability Tests:

Intel Burn Test IBT v2.6  https://www.majorgeeks.com/files/details/intelburntest.html

prime95  https://www.mersenne.org/download/

Aida64 https://www.aida64.com/

if you know Linux, I recommend stressapptest from Google, available via most Software Management Tools

Ultimative Test: 

14+ days no reboot and no BSOD  

09.08.2021 в 22:34, BitMaster сказал:

Benchmark:

 

3DMark via Steam https://store.steampowered.com/

 

Passmark https://www.passmark.com/

 

 

Stability Tests:

 

Intel Burn Test IBT v2.6  https://www.majorgeeks.com/files/details/intelburntest.html

 

prime95  https://www.mersenne.org/download/

 

Aida64 https://www.aida64.com/

 

if you know Linux, I recommend stressapptest from Google, available via most Software Management Tools

 

 

Ultimative Test: 

14+ days no reboot and no BSOD  

 

 

Thank you very much, comrade!
But I'm talking about the tests in DCS that people run here to get a table like this: 

Oh well, that is Hand Made with lots of sweat & time.

There is no app or modus that will let you measure DCS' performance. You have to dial the settings in by hand, try to fly the same path etc. every time, note the fps and then

rinse & repeat with a new setting, new GPU, faster RAM, etc etc...   As you can see, if you have a normal life & maybe a family on top, you dont have the time usually to do that.

Also, with that many changes happening the value of such graphs are rendered useless after a few months. New GPU drivers, DCS' updates and new modules, newer Win10 Build, Win11 soon, etc etc..  

I test it on my system with my settings and that counts for my experience. If it has too little fps I know I have to tune LOD down etc...  Graphs from others are only a rough Guide what to expect, it only takes a little overseen setting to ruin it all.