7800X3D, 7900X3D, 7950X3D..

[dureiken]

Hi,

 

does anyone did a plazma benchmark on 7800x3D here ? thanks

[EightyDuce]

Popped the 7800X3D in place of 7700X and it wouldn't post above 5800. 7700X was testmem5 stable at 6400/2100 and game stable at 6400/2167....thats a little disappointing. 

Dialed timings in and got it close to where I was with the 7700X. 

I game on MT, CPU frametime went from 3.5-5ms to 1.5-3ms. Heavily GPU limited still, honestly can't tell a difference. Going to do some more testing, depending on how it performs in Msfs2020 I may go back to the 7700X and sell the 7800X3D. 

[Fakum]

I admit i dont know alot right now about overclocking AMD but in a recent overclocking tutorial on how to, they were pretty strict about ram being set at no more than 6200. 

[EightyDuce]

37 minutes ago, Fakum said:

I admit i dont know alot right now about overclocking AMD but in a recent overclocking tutorial on how to, they were pretty strict about ram being set at no more than 6200. 

I don't see why that would be the case, other than trying to limit the troubleshooting areas.

If it's stable there's no harm in it. That being said, the gains are negligible and to effort expended trying to get it stable above 6000 is arguably not worth it unless you're benching. 

[Hoirtel]

8 hours ago, EightyDuce said:

Popped the 7800X3D in place of 7700X and it wouldn't post above 5800. 7700X was testmem5 stable at 6400/2100 and game stable at 6400/2167....thats a little disappointing. 

Dialed timings in and got it close to where I was with the 7700X. 

I game on MT, CPU frametime went from 3.5-5ms to 1.5-3ms. Heavily GPU limited still, honestly can't tell a difference. Going to do some more testing, depending on how it performs in Msfs2020 I may go back to the 7700X and sell the 7800X3D. 

Memory issues aside your testing sounds like it is an improvement. CPU frametimes are half, but yes while you are in the low single digits zone it's not going to be all that visible. Considering the 7700X is virtually the same cores, the gain must be all vcache. would be great to get some metrics of you have them. Your testing might be the most important yet! 

I should have my 7800X3D next week hopefully, Its a complete new build aside from GPU so waiting for a few other parts before I can test. Coming from a i9 9900KS I expect gains...

[EightyDuce]

9 hours ago, Hoirtel said:

Memory issues aside your testing sounds like it is an improvement. CPU frametimes are half, but yes while you are in the low single digits zone it's not going to be all that visible. Considering the 7700X is virtually the same cores, the gain must be all vcache. would be great to get some metrics of you have them. Your testing might be the most important yet! 

I should have my 7800X3D next week hopefully, Its a complete new build aside from GPU so waiting for a few other parts before I can test. Coming from a i9 9900KS I expect gains...

If you have a track you'd like me to try at whatever settings, I can try do that. We just had out first kid so free time is a bit tight.

[Th3ChosenOn3]

On 3/6/2023 at 8:48 PM, Th3ChosenOn3 said:

Nice, would be interesting to see how a 5800X3D compares.

I'm running the default High Preset, with full screen enabled:

Here is the track file that I used. I start CapFrameX capture the moment I press fly with a 1 second delay, and then I let it run for 222 seconds to have consistency between all of the runs. 

Plazma Torture Map - Level Flight.trk 623.2 kB · 22 downloads

If you get some free time, try this at 1920x1080 High Preset. Make sure to specify if mirrors and tacview are on, as it can have significant impact on the outcome. You'll also need to up the max FPS to 400.

Also, congrats on the kid.

Edited April 17, 2023 by Th3ChosenOn3

[EightyDuce]

1 hour ago, Th3ChosenOn3 said:

If you get some free time, try this at 1920x1080 High Preset. Make sure to specify if mirrors and tacview are on, as it can have significant impact on the outcome. You'll also need to up the max FPS to 400.

Also, congrats on the kid.

 

Where do you set FPS to max 400?

[Th3ChosenOn3]

I followed this tutorial, just changed it to 400 instead of 300.

 

[EightyDuce]

Think I figured it out.

Quick run at 1080 High Preset No Mirrors

 

[Fakum]

it is my understanding that the closer the 1% and 0.2% numbers are to the Average FPS numbers, the better (least amount of stuttering). if i am correct, doesnt that result look unfavorable?

[brain_delay]

You will not notice a .2% low that's still over 80fps

[EightyDuce]

1 hour ago, Fakum said:

it is my understanding that the closer the 1% and 0.2% numbers are to the Average FPS numbers, the better (least amount of stuttering). if i am correct, doesnt that result look unfavorable?

That I do not know.

 

The whole frametime/pacing/fps/1% lows and all that is a bit out of my depth. 

[peanuts0441]

43 minutes ago, EightyDuce said:

The whole frametime/pacing/fps/1% lows and all that is a bit out of my depth. 

The simplest way to think about it is like so:
Frametime is the amount of time 1 frame is shown on screen, so as an example, if you're playing at 60 frames per second, a new frame would show up every 16.5 milliseconds, thus giving a smooth 60FPS, but this is not always the case in a real scenario, due to other things, the frametime constantly goes up and down, one frame could be as short as 13 milliseconds while another lasts 40+ milliseconds, this would cause a noticeable "stutter" as its called, generally you want CONSISTENT frametime, not one that is constantly going up and down

A frametime that stays at a fixed number (e.g 16ms) is better than a frametime that is lower but jumps up periodically (e.g 14ms but spikes to 30-40ms every so often)

 

1% and 0.1% lows are a value that measure frametime (or stutter) in a more accurate manner. To make it easy to understand: The higher the 1% and 0.1% lows, the better

1% low measures the average of the slowest 1% of frames, so say your playing at 60FPS but periodically drops to 30-35 for a few hundred milliseconds, it will measure this drop and average it out, in this example; the 1% low would be 33

0.1% low does the same thing but for 0.1% of the slowest frames, so it measures an even bigger drop, lets just say periodically for a few milliseconds you drop to 25-29FPS, it will average these drops and report the 0.1% low to be 27

Edited April 17, 2023 by peanuts0441

[BitMaster]

ChatGPT says this:

 

In PC gaming, "1% low" refers to the minimum frame rate that a game experiences during the course of gameplay, expressed as a percentage of the total frames rendered. 

To calculate the 1% low FPS, the game's frame rate is recorded over a set period of time, typically one minute. The lowest 1% of those frames are then identified, and the corresponding frame rate is calculated as a percentage of the total frames. For example, if a game had an average frame rate of 100 FPS, but its 1% low FPS was 50, that would mean that 1% of the frames (or 1 out of every 100 frames) had a frame rate of 50 FPS or lower.

The 1% low FPS metric is important because it provides an indication of the game's overall smoothness and consistency. A game with a high average frame rate but a low 1% low FPS may experience frequent dips in performance, resulting in stuttering or lagging gameplay. On the other hand, a game with a lower average frame rate but a high 1% low FPS may provide a more consistent and smoother experience overall.

[Vladdo]

nullUpgraded to a 7590x3d over the weekend. Paired it with a 4090 but only running 4800MT C40 ram. Will upgrade it at some point.

That said, I can get stable/constant 72fps with a Quest 2 with the following settings. ASW is disabled. Infact, enabling ASW gives you a noticeable judder as textures go screwy from time to time in its attempt to provide smooth frame rates.

The mission I've been using is Ready on the Ramp with the Hornet on PG map. I can take off and hit downtown at low level over a lot of polys/buildings and it's rock steady at 72fps. I may get the very occasional dip to 71fps where there's a very slight pause but overall it's pretty impressive. 

The only real noticeable popup is the grass where it just appears. Is there is any way to double or triple the draw distance?

 null

[Zergburger]

my 6 year old build is on its last legs, (8700k, 1080ti, ROG maximus formula board, 32GB RAM, custom water loop).

I am totally at a loss between a 13900k build and a 7800X3D build, the delta in price is about $300-$400 higher for intel. I am unfamiliar with the intricacies of modern AMD builds so i am somewhat hesitant; while intel seems to always be pretty trouble-free and bring satisfactory performance.

Can anyone give me some advice, or show some DCS benchmark comparisons for the 13900k vs the modern X3D chips?

[brain_delay]

The x3d chips blow everything else out of the water. Get a 7800x3d, ddr5 6000 cas 30 ram, throw it together with a half decent cooler and be done.

[Vladdo]

Just go with a 7800x3d build. Intel is also unquestionably way, way hotter.. Plus, if you buy a decent motherboard now, you can use it for the next round(s) of AMD chips.

[kristod76]

Do you think that anywhere in the nearest future DCS MT will utilize more cores than in 7800x3D? I was wondering if I should get 7950x3D, disable high freq CCD and for now use it as 7800x3D. 

[brain_delay]

I wish people would read the threads even slightly.

It can already use the 7950x3d, and use it well without disabling anything.

[okletsgo]

4 hours ago, brain_delay said:

I wish people would read the threads even slightly.

It can already use the 7950x3d, and use it well without disabling anything.

yep, it works great.

[Ivanopulas]

Hello,

I bought a used 3090 and now I'm thinking of changing my processor and motherboard for DCS. The VR in its current form has disappointed me so I play at 4K on a new monitor + 2nd monitor for Helios. I'm looking at 5800x3d because an upgrade on the new 7800x3d will cost at least twice as much, and they promise a maximum increase of 30%, and it's not clear how it is in DCS. Maybe someone has already managed to feel 7800x3d in DCS? Is there a strong effect from MT at 5800? On the old overclocked i7 6700k with 3090, I intended an increase of about 5-10% at 4K, I hope after changing to 5800x3d such an assembly will be enough for flights for 5-7 years =).

[brain_delay]

2 hours ago, Ivanopulas said:

Hello,

I bought a used 3090 and now I'm thinking of changing my processor and motherboard for DCS. The VR in its current form has disappointed me so I play at 4K on a new monitor + 2nd monitor for Helios. I'm looking at 5800x3d because an upgrade on the new 7800x3d will cost at least twice as much, and they promise a maximum increase of 30%, and it's not clear how it is in DCS. Maybe someone has already managed to feel 7800x3d in DCS? Is there a strong effect from MT at 5800? On the old overclocked i7 6700k with 3090, I intended an increase of about 5-10% at 4K, I hope after changing to 5800x3d such an assembly will be enough for flights for 5-7 years =).

I'm going insane. But I will summarize a bunch of threads and posts and videos for you.

 

Using the plasma torture map, avg fps is as follows:

7kX3D ~250

13th gen ~200

7k-non3D ~170

5kX3d ~160

12th gen ~120

5k-non3d ~90

Everything else lower ~60-80

Edited April 18, 2023 by brain_delay

[some1]

People do these tests with different DCS versions, different settings and fps limit. So the numbers are not easily comparable. A 7kX3D is roughly 20% faster than 5800X3D, there's no way it could produce 55% more FPS in the same conditions.