Initial comparison 7800X3D on MSFS, XP12 and P3D

[GCBraun 2,392]

30 minutes ago, IanHarrison said:

Which ones? All of them?

Yes. All the pictures you posted are illegible.

Well, sorry about that. Besides the HDR issue on the P3D ones, here I can see all the numbers on my 1440p and 4K monitors. If I click on them twice, I get a full copy of the image and am able to zoom even more.

 

 

Edited April 15, 2023 by GCBraun

[Ray Proudfoot 9,009]

6 hours ago, Mace said:

GCBraun clocked the *vcache* cores to 5.4.  But the NON-vcache cores on a 7950X3D can clock to around 6 ghz.   

Even default a 7950X3D's non-vcache cores clock to 5.7 ghz straight out of the box.   A 7950X3D and 7900X3D are split personality cpu's -- each CCD -- each "half" so-to-speak -- operates at a different clock speed.   Asymmetric architecture.

This is getting very complicated. It feels like you have to have a degree in computer science to get the best out of these AMD CPUs. Still sitting on the fence but leaning towards Intel just because it’s less complex to setup.

[Ray Proudfoot 9,009]

2 hours ago, GCBraun said:

5.4Ghz is what top overclockers managed to achieve using a series of different and non-straightforward tweaks.

That’s 11% slower than the 6.0 of the i7-13900KS out of the box. Maybe it can go higher. Obviously serious cooling is required but that’s down to Chillblast to recommend suitable cooling and install it.

[GCBraun 2,392]

Coming back to MSFS, here is another comparison between my previous 13700K and the 7800X3D. Not seeing any difference using my previous test scenario (4K @ Ultra, TAA + FG, KJFK, 12:00, Overcast, AI Traffic with the Fenix).

13700K - 55fps:

 

7800X3D - 55fps

 

Here it is possible to see that MSFS makes much better usage of all cores/threads in comparison to P3D. All of them are similarly used and none is reaching 100%:

As for the 7800X3D, even though I am seeing some gains here and there, I must say that I am bit underwhelmed with it. It is more efficient, for sure. On the scenario above the 13700K was using around 150W, whereas the X3D can accomplish the same using 84W. 

Nevertheless, if I could back in time, I would probably have kept my 13700K. At least I should be able to upgrade this AM5 chip keeping the rest of the system for the foreseeable future...

Edited April 15, 2023 by GCBraun

[IanHarrison 522]

9 hours ago, GCBraun said:

Well, sorry about that. Besides the HDR issue on the P3D ones, here I can see all the numbers on my 1440p and 4K monitors. If I click on them twice, I get a full copy of the image and am able to zoom even more.

 

 

Don't worry, not the end of the world.

[Mace 1,936]

8 hours ago, Ray Proudfoot said:

This is getting very complicated. It feels like you have to have a degree in computer science to get the best out of these AMD CPUs. Still sitting on the fence but leaning towards Intel just because it’s less complex to setup.

Not for the 7800X3D.  It is very simple.  All of its cores run at the same speed, all have the big L3 cache.  You install it and forget it.

Now, for the 7950X3D and 7900X3D, yes, they are not as simple to set up and maintain.  Some may say they are install and forget also, but, as I said...experience tells me that won't happen.  They'll be fighting with xbox game bar, every Windows Update, the provisioning software, and Process Lasso.  My opinion.

But this should mean very little to you.  For your use (P3D) you should be looking at 7950X or 13700K or 13900K.

Edited April 15, 2023 by Mace

[Mace 1,936]

8 hours ago, Ray Proudfoot said:

That’s 11% slower than the 6.0 of the i7-13900KS out of the box. Maybe it can go higher. Obviously serious cooling is required but that’s down to Chillblast to recommend suitable cooling and install it.

I'll point this out again -- the 7950X3D out of the box clocks to 5.7 ghz.  And that would be the CCD you'd be running P3D on anyway.

The 5.4 ghz would be the top-end clock speed of the 7950X3D's *vcache* CCD -- the one you would likely NOT run P3D on.

And yes, making sure P3D ran on the clock speed cores would be priority #1.  Unless it gets built into some provisioning driver, you'd likely have to set that up using Process Lasso.

You mention that being complicated.  It kind of is.  Or from my perspective, more of a pain and leaving your rig open to potential problems down the line.  It's a niche cpu for niche use cases and niche users who don't mind the setup and continued vigilant maintenance.  Not ideal for your use at all, in my opinion, based on what you've said and of course not ideal for P3D.  I'd take a regular 7950X over the 7950X3D for P3D.  Or a 13700K or 13900K for Intel.

[GCBraun 2,392]

Just for the record and for other potential 7800X3D buyers: the expected Cinebench R23 results for this CPU is around 18000 points with an all-clock speed of 4.8Ghz. I was getting just 16.5k with my chip with only 4.450Ghz on all cores, something that caught my attention.

After debugging for some hours, I have decided to just drive to a nearby store and get a new CPU. After switching the processors, I was already seeing the correct frequencies and score on Cinebench.

Moreover, after some quick curve adjustments on the Bios (set  to -30), I can now see a stable 5Ghz speed on all cores, both on benchmarks and on MSFS/XP12. Even though I've managed to gain almost 1000 points on Cinebench and scored the 7th. best result in the world in 3DMark's CPU profiler, the fps increase on MSFS and XP12 were negligible (a couple fps, at most), showing that both Sims did not benefit much from the clock increase, even on an extremely CPU-limited test case...

Unfortunately I am not able to test P3D anymore, as I have restored Windows, and the installation process for that sim is very cumbersome when compared to the others. I assume the stable 5Ghz clock would make a bit of a difference, but not enough to surpass the best Intel processors.

Edited April 15, 2023 by GCBraun

[Ray Proudfoot 9,009]

1 hour ago, Mace said:

I'll point this out again -- the 7950X3D out of the box clocks to 5.7 ghz.  And that would be the CCD you'd be running P3D on anyway.

Not according to the site below who posted this… “Ryzen 9 7950X3D has 128MB of L3 cache and operates at 4.2 GHz by default, but can boost up to 5.7 GHz, depending on the workload.”

https://www.techpowerup.com/cpu-specs/ryzen-9-7950x3d.c3024
 

Quite frankly I don’t want all the apparent hassle the AMD brings so the Intel looks the more likely choice.

I haven’t a clue what CCD means.

[Mace 1,936]

43 minutes ago, Ray Proudfoot said:

Not according to the site below who posted this… “Ryzen 9 7950X3D has 128MB of L3 cache and operates at 4.2 GHz by default, but can boost up to 5.7 GHz, depending on the workload.”

https://www.techpowerup.com/cpu-specs/ryzen-9-7950x3d.c3024
 

That link is exactly what I said.  Think of the 7950X3D as two cpu's -- one cpu that runs at 4.2-5.2 depending on settings, and the other cpu that runs up to 5.7 ghz by default.

[Ray Proudfoot 9,009]

2 hours ago, Mace said:

That link is exactly what I said.  Think of the 7950X3D as two cpu's -- one cpu that runs at 4.2-5.2 depending on settings, and the other cpu that runs up to 5.7 ghz by default.

I don’t see the point of that. Intel for me.

[GCBraun 2,392]

Today I had a bit more time to use MSFS instead of benchmarking and managed to complete my first flight with my second 7800X3D chip.

Flew from EDDM to EBBR with the Fenix A320 and performance was nothing but stellar. Below is a screenshot showcasing over 100fps on a full IFR approach into Aerosoft's EBBR with Live Traffic. 

Not sure this would have been possible with my previous CPU's, especially not with just 72W average power consumption! The 7800X3D is slowly growing on me...

[spitzer45 99]

On 4/15/2023 at 9:08 AM, GCBraun said:

Coming back to MSFS, here is another comparison between my previous 13700K and the 7800X3D. Not seeing any difference using my previous test scenario (4K @ Ultra, TAA + FG, KJFK, 12:00, Overcast, AI Traffic with the Fenix).

55 fps with 7800x3d at jfk seems a bit lower than I expect. I have just tested inibuilds jfk with 7950x3d and 4090; taa, overcast, 4k, tlod 200, just static aircrafts, all ultra without fg gives apprx 50 fps on runway 33, with frame generation about 100 fps, Tested with pmdg 738.

Edited April 16, 2023 by Ray Proudfoot
Images and most text removed from large quoted post.

[orchestra_nl 947]

On 4/14/2023 at 5:27 PM, GCBraun said:

Which ones? All of them?

On the flight-deck the difference is 10%, but on the spot image, about 20%. I have the latest BIOS (1410) and Chipset drivers (5.0.3.24.2328), just confirmed. 

After updating to the latest HWINFO Beta, I am doing a test flight on P3D from KMCO to PANC and I am now seeing 100% C0/T0 usage, both on the airport and on cruise. 

Nevertheless, performance is still underwhelming @ about 4.4Ghz, with some spikes to 4.7Ghz. Around 20fps on the airport with AI and now 30-35fps on cruise. Similar, but definitely a bit lower, when compared to my 7600X (with base speed of 4.7Ghz and boost up to 5.3Ghz).

I will still conduct additional tests but, as of now, it seems that P3D prioritizes clock speed in detriment to cache. If that is the case, a 13900KS would be a better option for that sim...

A 9750x3D might be the best option if one is using both MSFS / XP and P3D. MSFS and XP can take advantage of the CCD0 cores with the extra cache and P3D could be pushed to CCD1 with the higher clockspeed.
If one is only using P3D then the Intel or non x3D AMD's seem to be the better choice at the moment.

[GCBraun 2,392]

8 hours ago, spitzer45 said:

55 fps with 7800x3d at jfk seems a bit lower than I expect. I have just tested inibuilds jfk with 7950x3d and 4090; taa, overcast, 4k, tlod 200, just static aircrafts, all ultra without fg gives apprx 50 fps on runway 33, with frame generation about 100 fps, Tested with pmdg 738.

I had about 70 live traffic using PSXT + Real Traffic. on that test. That makes a big difference. Just with static traffic, my fps basically doubles.