Upgrading my PC and running some tests (including VR)

[Eugel]

Well, the time has come for me to upgrade my system. 

And because I want to know if I actually wasted a lot of money or if it was worth it and I assume some people are interested as well if an upgrade is worth it, I thought I would run some tests before and after and share the results here.

 

So, im going from this:
Intel i7 6700K, 16 GB Ram, GTX1080Ti, and a Vive Pro VR Headset 

 

to this:
Ryzen 7 5800X, 32 GB Ram, still GTX1080Ti  and a HP Reverb G2

 

The rest of the system stays the same, DCS is installed on a NVMe M2 SSD (but I think my old mainboard doesn´t have enough PCIe lanes to utilize the full speed)

 

What I want to do:
I´ll take some FPS tests with the system as it is, both in VR with the Vive Pro and with an 21:9 2560x1080 monitor.

Then I´ll try the same with the old system and the Reverb G2.

I´ll test both Persian Gulf and Caucasus Maps.

And then do it all over on the new system.

 

What I want to find out:
Well, first of all, I´m really curious if the CPU upgrade is worth it, as DCS is supposed to be very CPU heavy. But the single clock speed of the Ryzen isn´t much higher than the 6700K.

Then I want to know if the 1080Ti can fuel a Reverb G2. I was very unhappy with the Vive Pro because of the strong screen door effect. I assume, I can afford to render at lower resolutions with the Reverb and still be able to read gauges, HUD, etc.

 

A couple of questions (any tips are welcome):
What would be a good sequence to test ingame ? I was thinking of just taking off and fliying a loop over a busy airport in Caucasus (I know that´s something my current system can´t handle well in VR)

Persian Gulf runs pretty well in VR with my current system, but I´ll test it anyway to see if the framrate might actually drop with the Reverb.

 

I don´t have the Syria map, because I think it´s hopeless to run in VR with my current system. But I might actually buy it just for the test.

 

What´s the best tool to measure ? MSI afterburner ? Does that work in VR ?

 

Well, the stuff should arrive tomorrow, so I´ll have the whole weekend to run my tests

[Terry]

Hi Eugel, I use the steam app fpsVR when tuning DCS and other games in VR. It shows FPS/and how much your CPU and GPU are being used by displaying frame times. It's handy for letting you see in game what's the weakest link using different settings, GPU or CPU. In your case it would give you an indication of how much difference the 5800X makes over the 6700K using the same settings / scenario.

[Raven (Elysian Angel)]

10 hours ago, Eugel said:

if the CPU upgrade is worth it

Yes: I went from a 6700k myself to a 5900X. It's quite the generational leap ;-)

 

10 hours ago, Eugel said:

if the 1080Ti can fuel a Reverb G2

Not properly (it's quite unlikely you'll be happy with it), but it will do with reduced resolution in Steam and reduced settings in DCS until you can get your hands on a 6800XT, 6900XT or 3090. Anything else lacks VRAM.

 

I had pretty much the same system you have before my upgrade. I only had 32GB of RAM and a CV1, but also a 1080Ti

 

10 hours ago, Eugel said:

I don´t have the Syria map, because I think it´s hopeless to run in VR with my current system.

Yes sadly it is: I bought Syria on day 1 but - except for a few instant action missions - didn't touch it until I had my new computer.

[xoxen]

Thanks Eugel for the time you are going to spend. I do not fly VR but have a triple screen setup and was playing around a lot with settings etc. When I did a comparison I took a mission (an own one or for better comparison with others a standard one), switched off the Track IR and simply started it. With MSI Afterburner I watched the FPS when entering the cockpit without doing anything else. I did so after every single change and also double checked the original FPS when putting everything back to the state I started with. As the FPS was back where it was I guess this is a good way to compare. Flying over a crowded airport gives you a non constant FPS and it is harder to compare. I also read, that going into a given mission from a module and not unpausing it is also a good way to have constant FPS.

Doing so I figured out that having 2 NVidia cards installed caps the FPS for whatever reason. I used a GT 1030 just to have additional outputs and had 15 FPS less. The same with a GTX 1050 Ti (both in combination with my GTX 1080Ti). Now I managed to run everything with the GTX 10801Ti only and do not have issues anymore.

With MSI Afterburner you can also compare the utilization of the GPU. Mine is always at 99%, this must be related to the 3 WQHD displays I´m running. But as a rule of thumb, full load at the CPU = CPU bottleneck, full load at the GPU = GPU bottleneck.

I still wonder if a CPU upgrade would bring me a benefit as the GPU is already utilized by 99%. New Hardware generation (incl. DDR4) might bring an advantage?! But to figure this out it would cost me 1000 EUR. So same situation you are in. Spending money without knowing if the outcome is worth it.

Edited March 5, 2021 by xoxen

[Eugel]

30 minutes ago, xoxen said:

Flying over a crowded airport gives you a non constant FPS and it is harder to compare. I also read, that going into a given mission from a module and not unpausing it is also a good way to have constant FPS.

 

That´s a good point. I thought I´d try something I know my "old" system struggles with. But of course, for comparison something more "standardised" would be better. I´ll keep that in mind.

 

30 minutes ago, xoxen said:

I still wonder if a CPU upgrade would bring me a benefit as the GPU is already utilized by 99%. New Hardware generation (incl. DDR4) might bring an advantage?! But to figure this out it would cost me 1000 EUR. So same situation you are in. Spending money without knowing if the outcome is worth it.

 

Well, I was going to upgrade my computer anyway, though I hoped to add a GTX3080 or 3090, but that´s not going to happen anytime soon.

And I noticed that there are a lot of questions about hardware, especially for playing in VR, around here (and Discord). And as I asked myself the same question, I would have ran the tests anyway, so I thought sharing the result might help one or two guys.

[xoxen]

I really appreciate it that you´ll share the results. This gives others a good chance to judge own ideas in updating anything. I will definately have a closer look at it, at least as you are running the same GPU and the fact that you have a better CPU already. So any benefit you will see will be even higher in my case  

Edited March 5, 2021 by xoxen

[Eugel]

Ok, after I posted, all the text was gone and it only posted the pictures 

So, here´s the short version:
I received only the G2 today, the rest of the hardware will come tomorrow. So I ran the test with my old system.

Here are the settings I use for non-VR play or screenshots etc:
 

I got somewhere between 40-56 FPS, overall pretty consistent...

 

 

 

Weird thing is that there are somtimes spikes where all CPU cores go up to 70-90%, seems like it has to do with missile launches and/or particle effects...

But as you can see, the GPU pretty much runs at 99% all the time, while the CPU cores barely go above 40% (except these spikes)

So I´m not sure if the CPU upgrade will really help.

 

I did the same with my VR settings next:

I got some more variation in FPS, between 100 and 150 FPS

The GPU usage with my VR settings went down a bit, to like 70-80%, but in densly populated areas as above, it went to 97% again.

But I didn´t get these all-core CPU spikes anymore.

 

I did the VR test next, and surprisingly, the G2 runs ok with the 1080Ti, FPS-wise, there is no difference between the Vive Pro and the G2 when I run them at similar resolutions. For the Vive, I use 1912x2124 per eye, the G2 is at about 2000x2000 and both get an average 48 FPS. But on the Caucasus map, there are some drops to around 30-35 which sucks. 

But 48 FPS is not ideal, but perfectly acceptable.

Unfortunately, fpsVR does not show up on screenshots, so no point in posting them...

 

Another thing I´d like to mention is that my 16 GB of RAM were pretty much at their limit all of the time.
So while, based on the numbers above, I won´t expect much performance increase from the Ryzen, the double amount of (faster) RAM might help somewhat.

 

Tomorrow I will hopefully run the tests with the Ryzen, but as I´ll do a clean windows installation first, it might take a while...

Edited March 5, 2021 by Eugel
text was annihilated by a GBU-12

[Eugel]

Ok, short info: I was lazy and did not do a clean windows installation. I just assembled ths stuff, plugged everything in, hit the power switch and installed all the AMD drivers...

Worked like a charm.

 

I did a quick test in DCS next. First again with my high settings on the monitor. Didn´t take any screenshots because there was nothing to see. No difference at all. I get the same 40-50 fps. Not a good sign. But, those strange spikes on all the cores I got with the i7 were gone.

 

Then, VR test with the G2, and that really came as a surprise for me. I got steady 45-50 fps, sometimes up to 60 fps on both Persian Gulf and Caucasus maps. No drops below 45 even when missiles fire, explosions, flares... 

And with motion smoothing enabled, this is a very smooth ride now, even on Caucasus. I´m tempted to buy the Syria map now ...

 

Interestingly, the GPU usage goes down to 60-70% in VR now. And what confuses me a bit is that the RAM usage barely goes over 16 GB, it´s mostly around 16.3 (according to fpsVR)

 

So was it worth it ? Considering the price tag - probably not. But on the Caucasus map those 10-15 fps elevate the game to a playable level in VR for me. And I assume, I now could also join some busy MP servers. 

But please keep in mind, I´m no expert on these things. Just plugged everything in and looked at the numbers...
Maybe it´s also better if I actually do a clean windows installation. Might do that when I got more time.

 

Unfortunately, I couldn´t run any tests from inside the cockpit as I can´t get my Warthog throttle to run anymore

[xoxen]

Did you delete the fxo and metashader2 folders in saved games DCS folder? I guess this is something you should do.

 

Edit: They will rebuild at the next start, so no worries.

Edited March 6, 2021 by xoxen

[Glide]

4 hours ago, Eugel said:

I´m tempted to buy the Syria map now ...

I also have the 1080Ti, and you will be disappointed with the Syria map.  It takes over 11GB of VRAM to run the Syria Map, and since our cards only have 11GB, we get a lot of stuttering from Windows memory manager swapping Dedicated and Shared GPU memory.  I put the Syria Map on the shelf until I get at least 16GB of VRAM because I can't bring myself to lower the textures enough to fit the map inside 11GB.  Glad to see that CPU worked for you.  First time I've seen someone just swap from Intel to AMD without reinstalling....

[Picchio]

@Eugel I'm on a Z170 chipset carrying a 6700K as well, and I'm using a 2700 Super. Running my Reverb G2 is a bit of a struggle.

I'm considering a partial upgrade (no GPU, hehe...) to a B550 with adequate RAM and a 5600X.

Have you made any more tests, how's your performance compared to your previous rig? Have you also tried other flight simulations, and measured the benefits?

[C3PO]

This week I'm due to get my new system (5800X / 32DDR 4 RAM / 1TB M.2 drive / EVGA 3090 Ultra) and am equally curious. But I have been using my current "old" i7 6700K system for creating videos ... I run at 4K resolution in all maps including Syria (even with my good old GTX 1080 8GB!!) to a level of reasonable performance - at medium to high settings, although it struggles a bit in MP in Syria.

 

I think you'll get a big improvement in performance because the Instruction Per Clock (IPC) figure is higher for the 5800X than the 6700K. That said, the i7 6700K is still very capable. I recently bought an Elgato 4K60 S+ unit to take all loads off the system - compared it with streaming video at 4K from DCS using the Elgato and nVidia Share...full details in video description with benchmarks for each attempt. Interesting results.

 

 

I'll do the same test with the new system once it arrives. 

Edited March 9, 2021 by C3PO

[Picchio]

Well, I'm primarily interested in understanding the kind of performance improvements I should expect in VR. Given the high resolutions the G2 runs at, I'm afraid the final performance gain won't be impressive. Let me know what you find.

[Eugel]

On 3/8/2021 at 9:38 AM, Picchio said:

@Eugel I'm on a Z170 chipset carrying a 6700K as well, and I'm using a 2700 Super. Running my Reverb G2 is a bit of a struggle.

I'm considering a partial upgrade (no GPU, hehe...) to a B550 with adequate RAM and a 5600X.

Have you made any more tests, how's your performance compared to your previous rig? Have you also tried other flight simulations, and measured the benefits?

I haven´t run more tests yet, because my TM warthog doesn´t like to work anymore with the new mainboards USB ports
And I don´t play any other flightsims.

 

Well, the thing with the G2 is, by default it sets the render resolution at 100% to 3096x3096 pixels, per eye ! So that´s almost 2.5 times 4K resolution. I think, the 2070 is a bit weaker than the 1080Ti, especially in VR and high resolutions due to less VRAM. What I can tell is that on the new system, with the 1080Ti I can run DCS with the G2 set to 40% render resolution in Steam VR with mostly stable 45 fps. 

It looks to me like the Ryzen CPU took a bit of the load off the GPU compared to the 6700K, according to fpsVR. I gained maybe 10-15 fps in VR, but my 11GB VRAM were used almost all of the time completely (not sure how that translates to the 2070).

 

I would like to upgrade to a 3090 because I don´t want to downgrade on VRAM, but at my local dealer, they went up to 3000 Euro now !

Edited March 9, 2021 by Eugel

[Glide]

On 3/6/2021 at 1:41 PM, glide said:

I also have the 1080Ti, and you will be disappointed with the Syria map.  It takes over 11GB of VRAM to run the Syria Map, and since our cards only have 11GB, we get a lot of stuttering from Windows memory manager swapping Dedicated and Shared GPU memory.  I put the Syria Map on the shelf until I get at least 16GB of VRAM because I can't bring myself to lower the textures enough to fit the map inside 11GB.  Glad to see that CPU worked for you.  First time I've seen someone just swap from Intel to AMD without reinstalling....

Just wanted to say the Low terrain textures are looking very nice and performance seems to have improved considerably.  There is still some funkiness going on with VRAM, but the usage seems to have come down a lot.  So, if you don't mind the odd bump in the road the Syria map is back on the rotation for my 1080 Ti.  YMMV.

[Tom Kazansky]

1 hour ago, glide said:

Just wanted to say the Low terrain textures are looking very nice and performance seems to have improved considerably.  There is still some funkiness going on with VRAM, but the usage seems to have come down a lot.  So, if you don't mind the odd bump in the road the Syria map is back on the rotation for my 1080 Ti.  YMMV.

+1

 

[Strong05]

Try updating your motherboard bios or forcing PCIE3 for your video card to fix your USB problem.  This is a known issue with AMD and a fix is due in April via a bios update.

[Eugel]

5 hours ago, Strong05 said:

Try updating your motherboard bios or forcing PCIE3 for your video card to fix your USB problem.  This is a known issue with AMD and a fix is due in April via a bios update.

 

Well, it´s rather a known Thrustmaster issue I guess. The throttle also doesn´t work anymore with the old Intel board.

 

[Strong05]

4 hours ago, Eugel said:

 

Well, it´s rather a known Thrustmaster issue I guess. The throttle also doesn´t work anymore with the old Intel board.

 

Oh, forgot about that problem.  That sucks.

[Headwarp]

6 hours ago, Eugel said:

 

Well, it´s rather a known Thrustmaster issue I guess. The throttle also doesn´t work anymore with the old Intel board.

 

 

My throttle bricked too.  I think it happened during a power outtage.    What's worse is It was my second warthog throttle.  The first one had a few mechanical failures, like one of the throttles had no resistance whatsoever.  I tried to pull the pcb out from the old one and put it into my newer one but the plastic connector snapped off the wires in the one I was putting the PCB into so I threw it in the garbage.  

 

I just got my tracking number from Virpil today for the T-50 CM3 throttle.. which I hope lasts me the rest of my simming life for the price I paid for it lol.  

Edited March 16, 2021 by Headwarp