AMD 5800X3D, the new King for flight simulators?

[some1]

6 hours ago, peachmonkey said:

so ..

5900x + 3090+3600mhz16cal 
vs 
5800x3D + 3090+3600mhz16cal
 

who wins at VR in DCS?

There's no scenario where 5900 would win in DCS, but there are scenarios where 5800x3d won't make a difference if you're limited by GPU. I posted a comparison a few messages above. With 3080, so with 3090 you'll be less GPU constrained that I am.

Edited May 26, 2022 by some1

[M1Combat]

8 hours ago, some1 said:

There's no scenario where 5900 would win in DCS, but there are scenarios where 5800x3d won't make a difference if you're limited by GPU. I posted a comparison a few messages above. With 3080, so with 3090 you'll be less GPU constrained that I am.

 

Your comparison showed that IN THAT EXACT SCENARIO the 3D was 7% faster.

I've already pointed that out to you.

[some1]

Huh? I own 5800X3D myself and compared it to my own 5900X here. I've posted a screenshots from a mission where there's no difference between the two in FPS. And from another mission where the new CPU yields 30% gain. It really depends on the scenario used for testing. 

 

 

 

[M1Combat]

I mean your graphic you posted a while back showing a 7% gain when the system was fully GPU bound.  7% at 45FPS is about 3FPS.  That's nothing to sneeze at coming directly from a CPU when you're GPU bound.  That's actually pretty dang phenomenal.

[some1]

Yes, you've said that three times already, ignoring everything else posted in this thread since then, apparently. 

 

2 hours ago, M1Combat said:

I mean your graphic you posted a while back showing a 7% gain when the system was fully GPU bound. 

Not really, it's just a status that MSFS puts in one of the debug windows. Does not say "fully" anywhere, and doesn't have to be 100% accurate, it's just what the game sees as it measures the time spent calling GPU-related functions.

Edited May 27, 2022 by some1

[M1Combat]

OK...  I'll go read the rest of the thread :)...

[M1Combat]

OK...  I read it.  Every word.  I hadn't missed much :).

Looks like my original assessment is still very correct.  No matter what you're running the 5800x3D will be an improvement and not an insignificant one.  Your post showed 13% when GPU bound and I just ran with 7% because it was a conservative estimate.  Sorry man but that's a HUGE performance increase coming from a CPU upgrade when the system is GPU bound already.  Would you get more from a GPU when you're gpu bound...  well yeah...  of course..  as I said before.  BUT...  Normally you get damn near nothing from a CPU upgrade when you're GPU bound.

Please keep in mind I'm not suggesting that everyone just go run out and buy an x3D...  There may be more gains in a GPU depending on your system for sure.  I mean I'm sure there's somebody out there running with a 12900K and a GTX970 in VR or something...  But even then I'd bet they'll see an improvement.

 

Anyway...

I don't mean to pick on you Some1 :)...  I promise :).  It's just that this statement "but there are scenarios where 5800x3d won't make a difference if you're limited by GPU." doesn't seem true.  

I will admit...  I am an AMD fan and have been since the K6-2.  I run whatever is best/close to best bang for the buck at the time though...  I'm running a 3800X right now...  But I had been running Intel exclusively since not long after the K6-2 days because any time AMD was the right choice for me...  I didn't need an upgrade.  I like AMD though.  they're a good company and when they're doing good I like to point it out :).

They are currently doing VERY good ;).

[edmuss]

Once mine lands I'll hopefully be able to offer some insight on the GPU bound front and as to whether the X3D helps in such situations.

I'm thinking along the lines of the A10 easy instant access, with the 3600/3070 I'm typically running 10-12ms CPU and 13-15ms GPU. I'll do my best to do a couple of tests with my normal settings and some with higher resolution to further bind the GPU and repeat the process with the new CPU dropped in.

[some1]

2 hours ago, M1Combat said:

OK...  I read it.  Every word.  I hadn't missed much :).

I don't mean to pick on you Some1 :)...  I promise :).  It's just that this statement "but there are scenarios where 5800x3d won't make a difference if you're limited by GPU." doesn't seem true.  

Still missed a couple things. Like these shots I took on my PC and posted here before:

5900X: https://imgur.com/B5VQrip

5800X3D: https://imgur.com/vBnQQ3H

See the FPS gain? Me neither.

That does not change the fact that 5800X3D is probably the best processor for flightsims right now, and a worthy upgrade under certain conditions. But it won't help much if on your system the bottleneck is elsewhere.

[M1Combat]

2 hours ago, some1 said:

Still missed a couple things. Like these shots I took on my PC and posted here before:

5900X: https://imgur.com/B5VQrip

5800X3D: https://imgur.com/vBnQQ3H

See the FPS gain? Me neither.

That does not change the fact that 5800X3D is probably the best processor for flightsims right now, and a worthy upgrade under certain conditions. But it won't help much if on your system the bottleneck is elsewhere.

Well...  It literally says 97 and 99 FPS...  That's >2 percent...  but thank you for changing what you're saying from "but there are scenarios where 5800x3d won't make a difference if you're limited by GPU."

I know...  splitting hairs :)...  but numbers are numbers and that 2% is the absolute smallest delta I've seen.

Edited May 27, 2022 by M1Combat

[some1]

Except its the other way round, 5800X3D is the slower one here.

That's with my limited DCS testing, as I wasn't seeking GPU limited scenarios in DCS - on the contrary, I picked two places where GPU does not show 100% use. But there are other games I checked that showed literally no difference at the resolution I play (Asassin's Creed, Tomb Raider, Forza Horizon).

[skypickle]

The next gen 7800x is around the corner. It will prob have 3D cache since am5 is already having trouble differentiating itself from am4 based on published specs- yes there is pcie5 but for the real world where pcie5 devices are still rare- the biggie will be 5.5ghz clock speed. What is the beneficial effect of more cache  as a function of clock speed?

[some1]

The next generation of AMD CPUs won't have 3D cache, at least not initially. 

[Hiob]

Has anybody done a proper testing with a defined (replay) and repeatable benchmark scene and average frame counter?

Because - no offense - but a still doesn't prove anything at all. Even less when the screenshot is not from the exact (and I mean exact!) same spot and perspective.

I would absolutely love to do proper testing with a 5800X3D, but sadly I don't have one - and I'm hesitating to sacrifice my 5900X (being reasonable for once.... )

[joca133]

Just dropped a 5800X3D to replace 5600X and holy moly it made a difference! Not a technical benchmark, just my amateur thinkerer observation, but I did ended up turning on msaa x2, mfaa, changing fsr from 70ish% to 90%, reprojection lock from 30 to 45 and openvr set at 110%. Running smooth enough that I increased tree coverage and set clouds to high.

Not enough to push my g2 anywhere close to 90fps in this dated game, but good enough. It does run a bit hotter though, close to 80° peak (62° avg) on water, but my home AC is busted and its summer.

Im kind of new to OpenVR but the improvements were very noticeable.

5800X3D, evga xc3 ultra 3080, 32Gb 3200MHz, Reverb g2.

[edmuss]

Still waiting for mine, ordered 12th May expected end of May, now expect middle of June earliest!

[dureiken]

Hi, could someone planning to change from XXX to 5800X3D, make a real benchmark with high load CPU scene just before changing it ?

 

As @Hiobsaid, to have some real numbers to compare ?

 

Thanks

[Hiob]

I will get a 5800x3d next week for testing. And will thoroughly test it against my 5900x. I‘m just collecting base data with the 5900x right now.

I will only focus on DCS and heaven benchmark (since I feel that this is the benchmark that comes closest to DCS in terms of load).

Can someone give me a hint on how to record and replay (and benchmark) properly in VR?

Also I would need a CPU-heavy (in terms of many units and lots of AI-stuff going on) replay scene for testing, because I normally focus on the graphics settings only.

A scene about 2-3 minutes long would be great. 

Edit:

I will test only in 4K, because that’s the only res I‘m interested in. (except anyone really really wants me to test a different resolution)

If anyone wants me to test a very certain setting or feature, I need to know now, to make a baseline, before switching CPUs!

Edited June 11, 2022 by Hiob

[edmuss]

Likewise if someone has a CPU heavy test scenario that can be run through I'll do a comparison with 3600 vs 5800x3d with a 3070

[Hiob]

Ok, Results: (short preview, I will post detailed Results and procedures tomorrow later, when I find the time)

In a nutshell: - for Pancake-Players "Nay", for VR-Players "Yay"!

There were no fps-gains in 2D whatsoever. I tested DCS-4K-high-details, DCD-1440p-medium-details, Heaven in 4K, 1440p and 1920*1080.

In VR I tested Marianas in the Apache (-25% cpu-frametime average) and Huey in Incirlik(Syria) (-20% cpu-frametime average).

Good night!

Edited June 15, 2022 by Hiob

[edmuss]

Mine is arriving tomorrow

[edmuss]

Mine is installed and running with the 3070, like for like settings in still seeing fps uplift across the board, especially on the lows.

A10c easy instant action peak fps is up about 10-15% but the average is consistently higher.

KA50 meggido to haifa tour gains 25-30% more fps on the ramp and consistently 20% higher in general flight.

Both of these are lightweight missions with regards to the CPU and it's still giving a very healthy improvement on a saturated mid range GPU. I set up a quick mission with 475 units and CPU times according to openxr performance overlay have dropped from 120ms to 80ms and it's almost flyable!

Coming from the 3600@4.3ghz to the 5800X3D (@4.45 consistent boost clock) is a significant improvement

edit: further testing, KA50 flight low overland on Marianas is giving 50fps, was lucky to get 40 before!

Edited June 18, 2022 by edmuss

[7Coolhand]

I’m thinking of an upgrade of my CPU, I have i5-9600K and a 2080ti.

Do You think I will see any big improvement with a 5800X3D?

 

[edmuss]

Yes you will but I wouldn't bother unless you're already on the am4 platform (which you're not). I would look at getting new motherboard and processor once the new AMD processors have come out and there are comparisons between the Intel 12th gen vs AMD 7000

Obviously the X3D is the only CPU currently with the monster L3 cache that DCS loves so much and whilst I suspect it will make an appearance in the 7000 lineup I don't think it will be on initial launch.

I don't know how much performance delta there would be going up to a 12th gen Intel with your 2080ti but I've certainly seen a healthy increase with my 3070 (which is marginally slower) simply by switching to the X3D.

[BusPilot]

Guys, I’m about to building a new rig, I got a credit to expend exclusively on a PC, veeery generous budget, so I’m not worried with price, because if I don’t spent the credit, the money is gone. I read the whole topic but I’m still can’t decide about the 5800X3D and i9 12900K, even with the AM5 about to be released, it worth go with 5800X3D? Or should I go with 12900K? Paired with a RTX 3090Ti. Any thoughts on specs?