13 hours ago, CO2Neutral said:

Not really, more cache hits over few cores.  9800X3D 104MB total cache, 9950X3D 144MB total cache.  The 9800X3D is a Ryzen 7 part, the 9950X3D is a Ryzen 9 part … there are architectural differences.  The 9950X3D has higher base and boost clock with better single core performance.

The higher total cache capacity on the dual CCD chip is a result of the second CCD having 32MB of it's own directly-accessible L3 cache--but it's important to note that it's cache can't be directly accessed by the 3D-equipped CCD.  The 32MB of 2D L3 cache on the non-3D CCD can still be accessed by the 3D-configured CCD, but it has to be done through the IFOP "infinity fabric" interface, which makes access considerably slower, all but negating any advantage of having more total cache available.

There are architectural differences, but the CCD and cache configurations are pretty much the same as the earlier 7950X3D.

Both the 9800X3D and the 9950X3D show ~15-20% improvement in IPC performance over their respective 7000-series predecessors, so I am unsure what difference the somewhat convoluted Ryzen 9 vs 7 branding really makes.

On 5/13/2025 at 7:52 PM, supersym said:

I tried the 9800X3d and I did'nt get any better performance vs my 13900k, curiously and at the opposite of many people...

Maybe graphic card is the bottle neck? What is your graphic card?

56 minutes ago, Bob Scott said:

corner--those results are for a 9700X CPU

True, but still AM5 based and should translate to 9950X3D also … I haven’t actually tested that RAM spec with MSFS 2020 or 2024 as I can’t locate 8000+ Mts RAM modules (pair) above 32GB … for optimal performance need to run more than 32GB for current crop of flight simulators.

There is something odd about MSFS 2024 Ultra vs. MSFS 2020 Ultra (DX11) … trying to determine why MSFS 2024 is such a bad performer at Ultra with barely any visual difference … DX12 should be a benefit, not a detriment and I can’t determine … is definitely NOT CPU related but both my 5090 and 9070XT OC GPUs struggle … the CPU utilization is 20% while GPU is pegged.  Anyway, I digress … both CPUs are good, but OP indicated “money is not a problem” hence it’s an easy decision to go with 9950X3D.  Best “value” might be different.

Also rumors AMD are coming out with another gaming focus CPU later this year.

 

1 hour ago, CO2Neutral said:

True, but still AM5 based and should translate to 9950X3D also

I would be very reluctant to assume that the X3D chips have the same performance characteristics as the regular AM5, especially when it comes to memory speed.

1 hour ago, Luke said:

X3D chips have the same performance characteristics

Agree, but I didn’t say “same”, I said “translates” … it’s actually likely higher memory speed would translate better on X3D with a cache miss.

3 hours ago, CO2Neutral said:

it’s actually likely higher memory speed would translate better on X3D with a cache miss.

The point of cache is to mask latency and bandwidth issues, so while all things being equal it might be a little faster (in benchmarks), the question is - is it noticeable? I imagine the price difference is, and very few folks have unlimited budgets.

5 hours ago, Luke said:

The point of cache is to mask latency and bandwidth issues

Agree, but on fresh data (aka streaming) it should help reduce latency.

AMD UDNA GPU platform appears to be set to go head to head with nVidia which should mean the return of sane GPU prices.

AMD Q1 report was outstanding!  Glad I purchased more AMD stock while it was in the upper 80’s (117 today).

Not trying to sound like an AMD advertisement, but we desperately need them to bring competition …  $1899 for my PNY 5090 OC and $719 for my AMD 9070 XT OC.  Sure the 5090 does indeed produce more FPS under very demanding graphics settings, but it doesn’t produce 250% more FPS.

13 hours ago, CO2Neutral said:

Agree, but on fresh data (aka streaming) it should help reduce latency.

No, cache does nothing in the case of a miss - if anything it slows things down by needing to check the cache for items that aren't there.

On 5/14/2025 at 5:58 PM, yanfeng12342000 said:

Maybe graphic card is the bottle neck? What is your graphic card?

no because several raisons :

I had a RTX 4900 , I tried at very low resolution and low graphic settings.

I thing the ram at 7 Ghz with low CAS and the CPU at 6.2 Ghz max and arround 5.9 at min and the Nb at 4500 are a good point...

20 hours ago, Luke said:

No, cache does nothing in the case of a miss - if anything it slows things down by needing to check the cache for items that aren't there.

Exactly my point … there will be A LOT of misses when streaming data from a server, hence the need for faster RAM … also why MSFS implements a rolling cache on local storage (GB size not MB size).  .

When/If I can get a hold of some >32GB 2 module 8000+ Mts RAM, I’ll do some testing against my current 6400Mts 96GB 2 module and post results.

On 5/16/2025 at 10:28 AM, CO2Neutral said:

Exactly my point … there will be A LOT of misses when streaming data from a server,

If you're streaming over the network your latencies are several orders of magnitude greater than whatever L3 cache can give you on chip. You're seriously overthinking things, and if the $300 price difference is irrelevant to you then you can send me $300 🙂

https://www.guru3d.com/story/asrock-pbo-settings-linked-to-ryzen-9000-cpu-failures/

The latest from GN.

 

I've never bought a cheap motherboard (sorry ASRock, you're cheap in my book of experience) as I value the components I add to my motherboard and research what MOFSETs are used in the specific version of motherboard.  My choices are usually ASUS Extreme boards and now MSI Godlike boards ... yes, they are expensive over $1000 US, but they deliver reliability.  Typically the more layers and better MOSFET the less issues.  ASRock use cheaper Niko MOFSETs.

There  is a reason why one motherboard costs $300 an another $1200 ... it's not just about profit margins.  There is a difference.

 

4 hours ago, CO2Neutral said:

I've never bought a cheap motherboard (sorry ASRock, you're cheap in my book of experience) as I value the components I add to my motherboard and research what MOFSETs are used in the specific version of motherboard.  My choices are usually ASUS Extreme boards and now MSI Godlike boards ... yes, they are expensive over $1000 US, but they deliver reliability.  Typically the more layers and better MOSFET the less issues.  ASRock use cheaper Niko MOFSETs.

There  is a reason why one motherboard costs $300 an another $1200 ... it's not just about profit margins.  There is a difference.

 

I don`t think you get the point a cheap motherboard can give a better more stable result due the manufacturer not setting the PBO too high in the BIOS to make the motherboard look good for performance. A lot of manufacturers do that including ASUS who had to do a BIOS fix for the 14000 series CPU.