known issue Cluster bomb detonation rendering - still drags system into a crawl

[einarabelc5]

So in average from inside the Cockpit I got a 10 FPS increase. Considering my average was 30 to 40 that is ok.

 

3900X with 4*8 at 3600 16-16-16-16 CAS.
2070 Super.

 

The test I wanted to make was to count the FPS for cluster bomb external view. Using the Russian 500U Frag bomb from a Su-27 still lowers the engine to a crawl, first 17 FPS but still running. The real problem/bug is when the game goes to almost 0 (3 fps) and then just hangs/locks in that thread, as soon as the cluster munitions detonate.

 

I'd thought I'd bring it up since multi-threading can be leveraged to help with this type of simulation.

[einarabelc5]

Update, simply hitting f7 and looking at vehicles with explosions on the background slows the FPS to a crawl. The more objects (tree, houses, fences, etc) that are between the vehicle and the explosion the less fps drop. It has serious issues rendering the smoke particles. What could be the cause and is it GPU bound? Attached is dcs.log of latest test where multiple bombs were dropped and the game froze one more time, ultimately releasing the CPU threads and then locking. (I did not wait to return to desktop)

dcs.log dcs_frame_mt.log

Edited May 30, 2023 by einarabelc5

[BIGNEWY]

Hi, 

this is a known issue and the team are looking for solutions. 

thank you

[einarabelc5]

18 hours ago, BIGNEWY said:

Hi, 

this is a known issue and the team are looking for solutions. 

thank you

Oh, thanks be to God. Thanks sir!

[einarabelc5]

On 6/1/2023 at 3:35 AM, BIGNEWY said:

Hi, 

this is a known issue and the team are looking for solutions. 

thank you

Hey Newey,

 

More than likely your team is aware but I wanted to let you know that while the load on the CPU spikes from the 30 % to the 80 % when I render cluster bombs, now the FPS only drops to the 27 FPS (being down to 2 or 0 and freezing before with no significant increase in CPU usage).

The ONLY difference is the CPU, so it is related to how the memory for these calculations is managed. Before I was using a Ryzen 3900X with multiple CCDs sharing cache that's away from the multiple chiplets. Now I am using a 5800X3d with a single 8 core CCD with access to the cache on top of the chiplet, so that significant reduction on latency to the L2 cache helps. The number of core reduction came with it, but the back and forth in data transfer was reduced as well as the IPC count.

Edited June 4, 2023 by einarabelc5