7800X3D, 7900X3D, 7950X3D..

[LucShep]

 

1 hour ago, shazam253 said:

At the very beginning, you'll see that two different DCS version were used. The AMD run is on a slightly newer version than the Intel run.

There's an updated video, however, it arrives at the same results.

I would guess that DCS is actually benefiting from the more cores on the Intel 12900k. Would be interesting to see, how it fares against the 7950x3d

 

 

55 minutes ago, Hoirtel said:

Interesting... and a bit of a blow along with the current X3D issues... To be really sure what's happening it would be nice to see a proper benchmark that has additional load on it. I don't think flying round an empty Caucasus map is the best test, however I wouldn't expect big leads either way. I would prefer to see it over a big city on Syria or over Marianas with some added units.

What is a bit more unknown for me is that all the tests are really at the frame lock of the headset so not really unconstrained, it does look like the 12900 has more to give and the 7800 doesn't, but the constraining behaviour could just be different. There is terrible aliasing so I would assume the resolution is low which is how they test 2D games, but for a headset with an FPS lock this doesn't really work. 

Other thing to consider is the recording itself, intel may be better at multitasking like this.

Having said the above, can't see there would be a be massive difference rather than trading blows at best. Anyone wanna buy a unused x3d...

Agree on all accounts.
I'm a bit surprised with the results on that video but, yes, it could be the recording interfering and giving advantage to the 12900K(?).

Edited May 5, 2023 by LucShep

[some1]

1 hour ago, shazam253 said:

 

I would guess that DCS is actually benefiting from the more cores on the Intel 12900k. Would be interesting to see, how it fares against the 7950x3d

I find it very unlikely. DCS MT is not a heavily multithreaded application. We've only got rendering and main thread separated. Both CPUs show limited utilization on these videos and have some idle cores that could do more work. Also while 13rth gen Intel paired with fast memory (7000MHz+, tweaked timings) can trade blows with 7800X3D and often come on top, that would be a very unusual case if a 12th gen Intel with very slow memory is faster too (IIRC he said Intel had 4800MHz sticks). 

One very important thing that he is not showing on these videos are CPU and GPU frametimes as measured by VR software. All other tools and metrics are simply not designed to work with VR and may show garbage instead. The DCS "CPU bound" indicator is straight up broken in VR. The Nvidia overlay is also designed for 2D games, on my system it doesn't even show anything in FPS/frametime section when DCS is launched in VR.

For example here you have a similar situation I've just snapped on my PC: GPU usage above 90%, but not 100% yet, DCS indicator screams "CPU BOUND", yet the VR software tells the truth. My GPU can't maintain frametime low enough for 90 FPS (orange spikes) which causes FPS drops. The CPU on the other hand sits at very low numbers in the green, there's so much headroom I could swap it for something slower and not see any difference, because here the GPU is the limiting factor. For a steady 90 FPS both CPU and GPU frametime should be below 11ms. GPU is at the limit, CPU is not.

And that's what is probably happening on these videos. He's testing the game near the GPU limit. For some reason on the AMD system GPU usage is higher and that's what we can see on the video. Why is that, I don't know. But this means he hits the GPU limit more often, which causes FPS to drop. If your GPU is at the limit, then you are not testing CPUs any more. 

TLDR: Do not look at FPS when comparing hardware in VR, look at CPU/GPU frametimes. With FPS you simply don't know if the limiting factor is the CPU, GPU or the VR software itself. 

Edited May 5, 2023 by some1

[Cavemanhead]

On 5/2/2023 at 9:16 AM, Fakum said:

I realize this is not really relevant to this thread at this point, but since I mentioned it here, I figured I would share this as my issue wasnt the BIOS at all and I resolved it via phone call to ASUS. Here is what I posted in the ASUS & ANANDTECH forums when I was trying to get some help:

"Dis-regard. After 3 days, I called ASUS. Bottom line was. I had my Arctic CPU AIO cooler plugged into the CPU AIO Header, instead of the CPU Fan Header (As the manual states to do) and when plugged into the CPU Fan header now, all boots up fine. WTF with all the hours of research! At least im good now, was ready to order a new MOBO and send this one back via prime,,,,"

Fakum, thanks so much for the post.  I'm the other guy from the Anandtech forum with the same sort of questions.  I've been debating about whether to jump into the 1414 beta bios or wait for the non-beta, "stable" version...  Wondering if a Kraken Z73 AIO will have the same stupid issue you had or whether it's different.  Do you think this might be a case of the CPU fan header is actually physically reversed with the CPU AIO header and both the silkscreen label on the MoBo and the instructions got reversed?  Weirder things have happened...  Or maybe just a peculiarity of the Arctic AIO?  I loved the simplicity of that cooler but the Z73 was trading blows in a few reviews and I really wanted the CPU & GPU temps easily readable in real-time so I went for the Z73.  I'll let you know how it goes once I dive in.

 

Just checked Asus website for the new bios and there is a new 1602 version released...  And it's still a beta!  Wow, this is so ridiculous.   

Edited May 5, 2023 by Cavemanhead

[Fakum]

I don think there is a reversal of anything. for me, it was a simple case of following the manuals instructions and plugging my AIO into the AIO header as the manual states. I had no idea that I had to set the CPU FAN in the BIOS to ignore. And that was going to cause the problem I had, and that was being able to boot up right off of the bootmanager or a bootable thumb drive.

Thanks for the BIOS update tip, I have been checking daily, just havnt checked yet today

Edited May 5, 2023 by Fakum

[Hoirtel]

16 hours ago, Cavemanhead said:

Just checked Asus website for the new bios and there is a new 1602 version released...  And it's still a beta!  Wow, this is so ridiculous.  

Yeah it's crazy. For the x670 there is two betas in a row and last "stable" is long enough ago at the moment to not be correct. They were taking the old ones down but now I have a choice of three betas and and out of date stable. Nice. Build on hold..  still.  

[dugite57]

I'm running my ASRock Tachi 670E 7800X3D in their 1.21 Bios AGESA 1.0.0.6 and everything seems fine.  But I'm not game enough to try going into bios and start tweeking, not that it needs tweeking, it plays DCS and stuff briliantly...

https://www.thefpsreview.com/2023/05/05/amd-has-advised-motherboard-vendors-to-hold-off-on-rolling-out-bios-based-on-agesa-1-0-0-7-bios-firmware-after-its-discovered-to-be-a-buggy-mess/#:~:text=YouTube Channel-,AMD Has Advised Motherboard Vendors to Hold Off on Rolling,to Be a Buggy Mess&text=AMD has advised motherboard vendors to wait until its next,inclusion in their BIOS updates.

[chev255]

Well some of the benchmarks above made me hesitate but on "the other sim" the 7800X3D and especially the 7950X3D are way better than the 13900K so I just pulled the trigger last night and will get the 7950X3D!

[kerlcat]

7800X3D on DCS with lastest ASUS stable bios, works great with DCS MT.

Smooth like butter without any stuttering issue which I had in ST version. And using air cooling only as this CPU is reletively cool during gaming. No trouble with AIO stuff. 

Wish for 7950X3D... while I mainly use this PC for gaming so it's a bit overkill. 

Some people was comparing this against Intel gen12/13 . The most of major medias reported it's faster than 13900K in "gaming". And much more advantage in simulation type of games. Maybe it's not the case for DCS. For me, the mother board of AMD could enable me to upgrade CPU in 3+ yrs away while intel will abandon me next year. 

Before 7800X3D, I was using 4790K since 2014 with multiple upgrades... motherboard is the last thing I would bother to upgrade or I would consider to replace the whole rig. 

That's only my personal humble opinion. Everyone has his/her own criteria to pull the trigger.

 

[Fakum]

@kerlcat Did you make any changes in the BIOS for EXPO or AMD overclocking? I built my rig last week and I havnt had time to even start messing with the Bios to set EXPO or anything else yet. I did run a quick mission for DCS with all setting to MAX just for an observation and I was amazed at the frame rates over 120 and no stuttering, My intent was to install some free benchmark software and test at these default BIOS settings to get a baseline, but havnt done so yet.

[kerlcat]

2 minutes ago, Fakum said:

@kerlcat Did you make any changes in the BIOS for EXPO or AMD overclocking? I built my rig last week and I havnt had time to even start messing with the Bios to set EXPO or anything else yet. I did run a quick mission for DCS with all setting to MAX just for an observation and I was amazed at the frame rates over 120 and no stuttering, My intent was to install some free benchmark software and test at these default BIOS settings to get a baseline, but havnt done so yet.

I just enable both EXPO and PBO. Simple for AMD CPU.

4090 to reach 120fps with max out everything sounds incrediably good. My 3090 only could get me to 70-90fps w/ ASW on and selective settings to high. Without ASW on, would be ~45. 

[Jive_68]

Hi everyone,

I recently switched over as well. Until a few weeks ago, I was running a 10700k CPU and a1080GTX. I use DCS exclusively in VR and had 32 FPS in the past with extremely low settings. Except that I have EXPO enabled for the RAM in the bios, my system is not further optimized.

Here are some of my performance tests using the plazma torture map:

Greetings

Dirk

[Fakum]

@Jive_68 I am trying to figure out what this " plazma torture map" is? I presumed it to be some sort of benchmarking software so I searched for info. Basically the only thing I have come across are some uTube vids that show guys running some tests in some different games. I have seen 2 that show them using DCS. Can you point me in the direction that I can get some info about this so i can educate myself? I would rather use DCS as a benchmarking tool than Heaven or Cinebench exclusively. Thanks

[Hoirtel]

1 minute ago, Fakum said:

@Jive_68 I am trying to figure out what this " plazma torture map" is? I presumed it to be some sort of benchmarking software so I searched for info. Basically the only thing I have come across are some uTube vids that show guys running some tests in some different games. I have seen 2 that show them using DCS. Can you point me in the direction that I can get some info about this so i can educate myself? I would rather use DCS as a benchmarking tool than Heaven or Cinebench exclusively. Thanks

It's a DCS simple and heavy track file that allows us all to run the same gameplay to compare hardware. The track file is linked in this thread in the earlier posts. Scroll back through and you will find it. 

[Fakum]

Ahh,, ok,,, so just find it and run it? I presume I can run it at my resolution of (3840 x 1600)?

 

[SilentSierra]

7 minutes ago, Fakum said:

Ahh,, ok,,, so just find it and run it? I presume I can run it at my resolution of (3840 x 1600)?

 

In the video description have the track for replay.

[Fakum]

@SilentSierraThank you sir! I have been scrolling

through all pages, so far I found 2 track files on page 5. I will check out what you posted. Thanks

 

EDIT: Ah, ok. I watched that vid earlier before I posted. Thats when I was searching for how these files were made to be called "Plazma Torture Test". I still dont know but I will download that Track and see what I can learn. Thanks

Edited May 10, 2023 by Fakum

[Th3ChosenOn3]

1 hour ago, Fakum said:

Ahh,, ok,,, so just find it and run it? I presume I can run it at my resolution of (3840 x 1600)?

 

For testing CPU performance, we ran Plazma's torture map at 1920x1080 resolution, with the high preset to reduce the GPU bottleneck. You will also have to update the frame cap to 400 if you can run the benchmark close to the 180 FPS cap. There's a video earlier in the thread showing how to update it.

You can also test GPU performance on Plazma's, like we did in the 4080 vs 7900 XTX thread, but the benchmark is mostly to test and compare CPU performance.

You should also look at CapFrameX to gather the statistics, there's a post I made earlier in the thread explaining how to benchmark with it. While benchmarking, having mirrors enabled along with Tacview can reduce average FPS by about 30%, so it's important to check if it's on and keeping it in mind when comparing numbers here.

[Fakum]

@Th3ChosenOn3Ok, I really and truly appreciate the feedback, but some of this stuff I simply have no familiarity with. And I dont want to derail this thread by going down the rabbit hole further.  As I said earlier, I had/have no idea what "Plazma Torture Test" was/is? Again, I assumed it was Benchmark software, then I came to understand that it seemed not to be, and it was basically someone making a DCS Trackfile and sharing it so others can play it back as and use it as a tool to Benchmark performance. 

Now, reading what you stated above "For testing CPU performance, we ran Plazma's torture map" I read that as something other than a DCS Track File?

Then you state "You can also test GPU performance on Plazma's" I will ask the dumb question, what does that mean? Plazmas Web Site, Plazmas Software, Plazma TV's?

I also presume when you say "You will also have to update the frame cap to 400". Are you referring to the "MAX_FPS" settings needs to be set to 400?(If so, I am familiar with doing that.

• 
   

 

 

[Jive_68]

vor 27 Minuten schrieb Fakum:

@Th3ChosenOn3Ok, I really and truly appreciate the feedback, but some of this stuff I simply have no familiarity with. And I dont want to derail this thread by going down the rabbit hole further.  As I said earlier, I had/have no idea what "Plazma Torture Test" was/is? Again, I assumed it was Benchmark software, then I came to understand that it seemed not to be, and it was basically someone making a DCS Trackfile and sharing it so others can play it back as and use it as a tool to Benchmark performance. 

Now, reading what you stated above "For testing CPU performance, we ran Plazma's torture map" I read that as something other than a DCS Track File?

Then you state "You can also test GPU performance on Plazma's" I will ask the dumb question, what does that mean? Plazmas Web Site, Plazmas Software, Plazma TV's?

I also presume when you say "You will also have to update the frame cap to 400". Are you referring to the "MAX_FPS" settings needs to be set to 400?(If so, I am familiar with doing that.

• 
   

 

 

Hi Fakum,

I'll try to break it down. The Youtuber Plazma has created a mission that makes high demands on the system due to the many different units and actions. For the mission, he has recorded matching track data that you can play on yourself. You can get both the track data and the mission data via the links in his video. The mission data is needed if you don't own the aircraft from the track recordings. Then you perform the mission with your own airplane and record a track yourself. Since the tracks always do the same thing you can change the settings and then see which settings affect the performance. So you have a cleanly defined lab environment that always runs the same way.

The mission or the track files only provide the basic conditions and the measurements are done with another software. Many here in the forum, like me also use the already mentioned tool CapFrameX. This tool is started and you can relatively easily give the program, which should be measured. So for my measurement I made sure that DCS is measured. In addition, you can specify a time for the measurement and a key combination that starts the measurement.

For example, I entered 180 seconds for the measurement.

To perform the performance measurement you first start CapFrameX, then DCS. Then you load the track file, start the playback of the track and activate the measurement by key combination. After the specified time the measurement stops, even if the track is still running.

You can repeat this process as often as you like and adjust resolutions or settings like shadows or clouds and see the effects of the settings in the measurement results.

CapFrameX looks a bit confusing when you first start it, but basically it is easy to use for basic tests. It took me about 15 minutes and a Youtube tutorial video to get my first measurement running.

[Fakum]

@Jive_68LOL! So Plazma is a person, I had no indication of that when I searched today, all I got was basically Videos with different files attached for testing. Thank you so much for the explanations. I know have a clearer understanding sir. The only thing I would like to clarify is the "You will also have to update the frame cap to 400"  Thats is done within CapFrameX then correct?

Many thanks again.

[Jive_68]

vor einer Stunde schrieb Fakum:

@Jive_68LOL! So Plazma is a person, I had no indication of that when I searched today, all I got was basically Videos with different files attached for testing. Thank you so much for the explanations. I know have a clearer understanding sir. The only thing I would like to clarify is the "You will also have to update the frame cap to 400"  Thats is done within CapFrameX then correct?

Many thanks again.

To increase the frame cap you have to create/modify a config file in the saved games section of DCS.

You can find instructions here:

Youtube Tutorial increase FPS Cap

[Fakum]

Thanks, got the CapFrameX software installed and gave it a test run, many thanks

[SilentSierra]

"ASUS' actions relating to the Exploding Ryzen CPU debacle are disgraceful and abrasive to the trust that the brand has earned. ASUS has demonstrated clearly it wishes to not only avoid supporting users, but actively engineers ways to abandon them. ASUS' updates haven't even fixed the problems, yet they posture as if they have while simultaneously suggesting that users 'just run defaults' on their $700 motherboards, as if that makes any sense whatsoever. So, to accommodate ASUS' request, we ran defaults and re-benchmarked the Ryzen 7000 series. It sucks. Big surprise. They also don't support their own BIOSes for the ASUS ROG boards."

Edited May 11, 2023 by SilentSierra

[shazam253]

Not sure what to make of ASUS right now. I just got my 7950x3D yesterday with an ASUS TUF board and it runs stable so far. I haven't tried too much yet, but I also refrained from overclocking or XMPing the board/RAM yet.

In DCS, I saw an improvement of roughly 20 FPS (80) on average over an i9900K(60) while using a 4090 RTX card. However, it seems the bottleneck in VR (G2) is still the graphics card. Also the 1% percentile didn't change too much, I tested it with the hot start of the AH64 in Syria.

Just on a side note, in 3DMark, it not only improved the CPU score, but also somewhat the GPU score, which I assume comes somewhat from the faster memory.

Sorry for not giving you an exact science, but in my opinion, for VR, we are still GPU bound. The current processors and DCS in MT mode seem to do just fine for now.

Edited May 11, 2023 by shazam253

[Fakum]

Yeah, Im not to thrilled with current events, I did alot of research for a couple of months and was initially committed to Intel, but after alot of discussion, the swaying factor was price and performance and HEAT! I havnt done any overclocking at all, nor have I set Expo. I just started doing some benchmarking last night with Cinebench, Heaven and Plazma Torture Test to get a baseline before I decide to make any adjustments. I have run DCS a few times and like I said in an earlier statement, i just cranked up all of the DCS settings just to see what this thing would do stock, and I am just so happy with the difference at stock. I was struggling with my older rig with MT. I dont know where this is all going, but I really dont want to have to think about trying to send back the CPU & MOBO to switch to Intel, way to much effort has gone into this build. I am very satissfied with what I have BUT, I want to dabble in VR by end of year, i dont know if these problems will hinder that?

1st Torture test I did last night.

null