Performance issue, help needed

Hi guys,

Some help needed.

In short, low FPS when CPU/Memory/GPU/DiskAccess are all low...

My PC was originally a Dell XPS 8910 (i7-6700, 8G RAM, Hybrid HDD+SDD, GTX XXX). Through recent upgrade, it is now:

i7-6700, 24G RAM (4Gx2, 8Gx2, 2133MHz), 512G NVMe SSD(on a PCIe card) + 1T SATA SSD + Hybrid HDD/SDD, RTX 2080 TI, 750W PSU. 3 monitors on 1920x1080 but only 1 monitor is used for DCS.

With the most recent upgrade being the RTX 2080 TI (Gigabyte Windforce, 11G vram), I was expecting the FPS to be a steady 50-70 if not more...

However, on the deck of my populated Stennis, or on my populated airbase at PG, or low pass at Dubai CBD, I get only about 25-27 FPS. When up in the air with almost nothing nearby, I can get some 70-90 FPS though.

I checked Task Manager, when I'm getting 25-27 FPS. CPU: 35%-60%, Memory: 14GB (could hit 21G sometime, but later on it would drop to 14G), GPU: 15%-25%, Disk: hard to describe in simple number, but the usage chart looks nowhere near intensive, with a few spikes though.

Checked NVIDIA Driver is the latest, used "DCS_Updater.exe repair".

Tried different missions, in some single mission like the training ones, it was quite good >100FPS, but some others as mentioned above was terrible and hopeless...

In game graphic settings are basically aligned with what Wags has shown in one of his tutorial video.

All these are in DCS OB version.

When using DCS Release version, the result is slightly better and smoother, but still doesn't feel like fully matching the expectation.

Result was a bit worse when using F-14, compared with using F/A-18.

Any idea of what could be the performance bottle neck?

The GPU utilization points to a problem with CPU bottleneck.

First try to hit ALT+Enter when you enter the mission. It will switch to Full screen. This helps me, for some reason even when you check the full screen option in options, the game does not enforce it.

If that does not help, make sure that the game is running only on physical cores and does not try to run on the hyper threaded ones. You can download Process Lasso for this, its free app.

Thanks. Interesting... Will give them a try.

I do run both two of the three monitors on graphic card while the third one on motherboard integrated graphic port. Not sure if the monitor that is on the same graphic card but is not used for DCS creates extra workload. It was normally just showing desktop and webpages or manuals.

Sent from my iPhone using Tapatalk

Your Skylake CPU is 3.4Ghz-4.0Ghz w/turbo, but you are down @20-30% clock for clock compared to Coffee Lake. Most people run 8th/9th gen Intel CPUs @5Ghz on average. That is a hefty difference in compute power (Ghz, and IPC Instructions Per Cycle).

You are basically starving your top of the line GPU (which is why it is under utilized @15-25%).

DCS is a very CPU intensive game, and until the code base changes, single threaded clock speeds (along with IPC, because raw Ghz isn't the whole story) will give you the biggest gains as long as you have a powerful video card, and you have that in spades.

Maybe ditch the onboard graphics or at least ditch it an check performance for DCS.

Your Skylake CPU is 3.4Ghz-4.0Ghz w/turbo, but you are down @20-30% clock for clock compared to Coffee Lake. Most people run 8th/9th gen Intel CPUs @5Ghz on average. That is a hefty difference in compute power (Ghz, and IPC Instructions Per Cycle).

 

You are basically starving your top of the line GPU (which is why it is under utilized @15-25%).

 

DCS is a very CPU intensive game, and until the code base changes, single threaded clock speeds (along with IPC, because raw Ghz isn't the whole story) will give you the biggest gains as long as you have a powerful video card, and you have that in spades.

Very good insight. Thanks for that!

Changing CPU might cause other issues including deactivating windows licence and might as well change motherboard and chassis... so is there a way to test/verify the above theory before I commit to purchase them?

I did check individual CPU core. Though one was noticeably higher than others, none of the cores was over 80%. Seems they were not fully stretched, yet...

Sent from my iPhone using Tapatalk

You could run Cinebench, 3dMark, Heaven benchmark, etc. and compare your results to other systems with similar specs.

If your numbers are close, then you don’t have an issue, if they are low, you know there is something to fix.

Sent from my iPhone using Tapatalk