• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.
  • The forums have been upgraded with support for dark mode. By default it will follow the setting on your system/browser. You may override it by scrolling to the end of the page and clicking the gears icon.

Questions about 9900x performance for purely gaming and the dual-CCD

Francis.g

New Member
Joined
Dec 9, 2024
Messages
3 (0.02/day)
I bought the Ryzen 9 9900x instead of the 7800x3D because it was cheaper, and the 7800x3D was out of stock everywhere or ridiculously priced. I've done some research, and from what I understand, the dual-CCD nature affects it in gaming, and it becomes equivalent to the 9600X. But after looking even more, I saw the driver updates on my new motherboard say that there was an update reducing latency a substantial amount.

Does this mean core parking is disabled in gaming and it will take advantage of the whole CPU or at least get better gaming performance than the 9700X, which from what I’ve seen can have been better performance than the 9900X?

I just can't find any recent gaming tests on benchmarks with these new updates in effect and I purely intend to do gaming on this CPU but I'm sort of worried that I wasted my money instead of getting this productivity CPU for gaming.

If it adds any context, I normally play BeamNG. Drive and Assets, Corsa, and other car simulation games.
 
You can test it yourself.

Lock your benchmark program into 6 cores on one CCD, time it.

Lock it into 6 cores, 3 on one CCD, 3 on the other, and time it.

For Assetto Corsa I can't imagine it matters.
 
All of those games play just fine on my 5900X.
 
I bought the Ryzen 9 9900x instead of the 7800x3D because it was cheaper, and the 7800x3D was out of stock everywhere or ridiculously priced. I've done some research, and from what I understand, the dual-CCD nature affects it in gaming, and it becomes equivalent to the 9600X. But after looking even more, I saw the driver updates on my new motherboard say that there was an update reducing latency a substantial amount.

Does this mean core parking is disabled in gaming and it will take advantage of the whole CPU or at least get better gaming performance than the 9700X, which from what I’ve seen can have been better performance than the 9900X?

I just can't find any recent gaming tests on benchmarks with these new updates in effect and I purely intend to do gaming on this CPU but I'm sort of worried that I wasted my money instead of getting this productivity CPU for gaming.

If it adds any context, I normally play BeamNG. Drive and Assets, Corsa, and other car simulation games.
How you getting on with your 9900x? I hear the Agesa 1.2.0.2b BIOS update improved performance. Interested to see what your performance is like on something like that new Indiana Jones IDTECH 5 game, or that Matrix Awakens demo that uses Unreal Engine 5.4 and should take advantage of more cores. Spread over the two CCD's?! I've also read people mention how the 9800X3D generally has better memory compatibility (better memory controller?!) for higher speeds, not sure how accurate that is?
 
I bought the Ryzen 9 9900x instead of the 7800x3D because it was cheaper, and the 7800x3D was out of stock everywhere or ridiculously priced. I've done some research, and from what I understand, the dual-CCD nature affects it in gaming, and it becomes equivalent to the 9600X. But after looking even more, I saw the driver updates on my new motherboard say that there was an update reducing latency a substantial amount.

Does this mean core parking is disabled in gaming and it will take advantage of the whole CPU or at least get better gaming performance than the 9700X, which from what I’ve seen can have been better performance than the 9900X?

I just can't find any recent gaming tests on benchmarks with these new updates in effect and I purely intend to do gaming on this CPU but I'm sort of worried that I wasted my money instead of getting this productivity CPU for gaming.

If it adds any context, I normally play BeamNG. Drive and Assets, Corsa, and other car simulation games.
Initially core parking was introduced for 7000X3D CPUs for the reasons most of us know. 1 CCD better at gaming and the other CCD better in all most anything else.
Those CPUs in order to run properly the driver needed to recognize the task and assign it to the right CCD.

There is a windows service that these dual CCD X3D CPUs should run for proper CCD selection.

Untitled_227.png

When 9000series first came about, AMD recommended that all dual CCD 9000 CPUs should be treated as X3Ds in terms of core parking.
That I believe was due to the increased latency between the CCDs that was almost doubled from 7000.
I am not sure if at that point the nonX3D dual CCD 9000 CPUs was indeed run the above service.

After the AGESA update that brought down the CCD latency to 7000 levels I believe the nonX3D dual CCD 9000 CPUs are treated just like any previous dual CCD CPU.
Like any 7000 and 5000 nonX3D dual CCD.
So in that content the core assignment should be dictated by core frequency capability and placement of the cores. And that is depending on the CCD/core binning that each CPU has.
Its literally a lottery. Some CPUs have all their best (as high freq) cores at one CCD and others have them scattered in both.

The cores are numbered in a certain order
1. The CPPC order (Collaborative Processor Performance Control) that windows choose to load first
2. The hardware-fused preferred core order assigned by manufacturer evaluation.

The exact order of the above 2 can be seen in HWiNFO64 (sensors mode) like the following screenshot

1735823565346.png


Perf #n/n is the 2 separate orders.
1. First n is the CPPC (win assignment)
2. Second n is the hardware-fused by AMD

As you can see, windows suppose to load the 0-5 cores first and then the 6-11 even though the AMD frequency/voltage evaluation tells otherwise. Cores 7 and 8 are better than core 5 but they are 7th and 8th in CPPC order. Just be aware that those cores are numbered 6 and 7 in CPPC because cores 1 and 3 are both marked as 1st (in CPPC).

So we can say that on this specific CPU cores and thread loading are not distributed ideally because core 5 (8th in evaluation) gets loaded before the cores 7 and 8 even though they are better (6th and 7th in evaluation).
I've seen other CPUs from screenshots around that have all best cores in first CCD and yet others that have more messed up order than my 5900X.

I hope this makes some sense. It may take some time to digest.
 
Initially core parking was introduced for 7000X3D CPUs for the reasons most of us know. 1 CCD better at gaming and the other CCD better in all most anything else.
Those CPUs in order to run properly the driver needed to recognize the task and assign it to the right CCD.

There is a windows service that these dual CCD X3D CPUs should run for proper CCD selection.

View attachment 378059

When 9000series first came about, AMD recommended that all dual CCD 9000 CPUs should be treated as X3Ds in terms of core parking.
That I believe was due to the increased latency between the CCDs that was almost doubled from 7000.
I am not sure if at that point the nonX3D dual CCD 9000 CPUs was indeed run the above service.

After the AGESA update that brought down the CCD latency to 7000 levels I believe the nonX3D dual CCD 9000 CPUs are treated just like any previous dual CCD CPU.
Like any 7000 and 5000 nonX3D dual CCD.
So in that content the core assignment should be dictated by core frequency capability and placement of the cores. And that is depending on the CCD/core binning that each CPU has.
Its literally a lottery. Some CPUs have all their best (as high freq) cores at one CCD and others have them scattered in both.

The cores are numbered in a certain order
1. The CPPC order (Collaborative Processor Performance Control) that windows choose to load first
2. The hardware-fused preferred core order assigned by manufacturer evaluation.

The exact order of the above 2 can be seen in HWiNFO64 (sensors mode) like the following screenshot

View attachment 378060

Perf #n/n is the 2 separate orders.
1. First n is the CPPC (win assignment)
2. Second n is the hardware-fused by AMD

As you can see, windows suppose to load the 0-5 cores first and then the 6-11 even though the AMD frequency/voltage evaluation tells otherwise. Cores 7 and 8 are better than core 5 but they are 7th and 8th in CPPC order. Just be aware that those cores are numbered 6 and 7 in CPPC because cores 1 and 3 are both marked as 1st (in CPPC).

So we can say that on this specific CPU cores and thread loading are not distributed ideally because core 5 (8th in evaluation) gets loaded before the cores 7 and 8 even though they are better (6th and 7th in evaluation).
I've seen other CPUs from screenshots around that have all best cores in first CCD and yet others that have more messed up order than my 5900X.

I hope this makes some sense. It may take some time to digest.
omg thank you, you've just reminded me about my cpu since I couldn't find much about it. So from what I understand in general I should just leave both ccds active since the latency isn't to bad now but if I want the best performance Id have to go through the process of measuring which cores are good on which ccd and use the one with the best cores if most of them are on that ccd.
 
Just leave it alone, don't mess with it as there is no point, differences between 9900X and 9700X are within margin of error.
 
omg thank you, you've just reminded me about my cpu since I couldn't find much about it. So from what I understand in general I should just leave both ccds active since the latency isn't to bad now but if I want the best performance Id have to go through the process of measuring which cores are good on which ccd and use the one with the best cores if most of them are on that ccd.
Just leave it alone, don't mess with it as there is no point, differences between 9900X and 9700X are within margin of error.
Agree...
@Francis.g, there is no point of doing anything. AGESA, Chipset drivers and windows are suppose to manipulate the CPU the best way.

In your case I would only maybe try to run benchmarks (gaming and nongaming) with CPPC on/off from BIOS to see if anything is changing.
And that just for the sake of information and curiosity.

Do you have HWiNFO64?
I am curious to see what it tells you on those orders.

1735828311578.png


On the other hand if you like spending time by disabling CCDs/cores and have some fun, by all means...
Just dont expect miracles.
Same applies to DRAM timings and frequency tuning.

I would 100% tinker with CurveOptimizer though as I did with 5900X. That and custom PBO limits (advanced settings).
 
I bought the Ryzen 9 9900x instead of the 7800x3D because it was cheaper, and the 7800x3D was out of stock everywhere or ridiculously priced. I've done some research, and from what I understand, the dual-CCD nature affects it in gaming, and it becomes equivalent to the 9600X. But after looking even more, I saw the driver updates on my new motherboard say that there was an update reducing latency a substantial amount.

Does this mean core parking is disabled in gaming and it will take advantage of the whole CPU or at least get better gaming performance than the 9700X, which from what I’ve seen can have been better performance than the 9900X?

I just can't find any recent gaming tests on benchmarks with these new updates in effect and I purely intend to do gaming on this CPU but I'm sort of worried that I wasted my money instead of getting this productivity CPU for gaming.

If it adds any context, I normally play BeamNG. Drive and Assets, Corsa, and other car simulation games.
If you're only a gamer, no need for the 9900X, more cores are usefull for productivity tasks, but games are more affected by single core performance. You should have gone for the 9600/9700x, performance in games are the same between the 3 processors.
 
but games are more affected by single core performance.
Once he gets comfy and learns how to tune, the big CPUs are good for games too. The boost clocks are hard to ignore.
 
If they come down in price enough I might just end up pulling the trigger myself.
 
Once he gets comfy and learns how to tune, the big CPUs are good for games too. The boost clocks are hard to ignore.
Based on most reviews, performance are on par between the 9900x/9950x and 9600/9700x. So for a pure gamer it makes no sense to buy the 9900x/9950x.

Of course there is the marketing/placebo effect of buying the "higher" end. So if it makes anyone feels better and that same person has money to waste, sure go for the "big" cpus.
 
Looking at the sales data from Mindfactory in Germany a drop in price would be wunderbar!

Also saw mentioned cache being saturated with raytracing / path tracing and more cores being beneficial. Perhaps the reason for the supposed increase to 12cores on Zen 6?!
 
I trust my tuning over most reviewers.

No offence meant jic some actually come here.
Any CPU can be tuned, and when it comes to OC, more cores aren't an advantage. Also, when you buy a CPU, its always a lottery when it comes to tuning/OC, may it be the more expansive or the cheapest.
 
No issues gaming on my 5900x. I average around 55fps with everything maxed (no upscaling used and motion blur turned down) in STALKER 2 with my 3080Ti at 1440p. When I was playing the game, outside of testing overall performance, I played by turning down a couple of stupid graphic options (such as hair....majority of NPCs wear a mask or hood, why the hell do you need an option for hair?).

I make use of it gaming and it also kills it when using handbrake. Not sure why anyone would care if someone has a 9900x that they utilize for gaming.
 
No issues gaming on my 5900x. I average around 55fps with everything maxed (no upscaling used and motion blur turned down) in STALKER 2 with my 3080Ti at 1440p. When I was playing the game, outside of testing overall performance, I played by turning down a couple of stupid graphic options (such as hair....majority of NPCs wear a mask or hood, why the hell do you need an option for hair?).

I make use of it gaming and it also kills it when using handbrake. Not sure why anyone would care if someone has a 9900x that they utilize for gaming.
Can't say I care, just trying to spare some people some money, buying a 9900x for gaming only is a waste of money, the 6/8 core version is always as good, even if the 12/16 core vesrions have higher base frequencies, the difference is so little anyway and boosting higher does not mean anything, its how long a processor can maintain the boost clock, may the max be lower, and in that regard, the 6/8 core version has an advantage in the fact they overheat less and thus can boost longer.
 
Can't say I care, just trying to spare some people some money, buying a 9900x for gaming only is a waste of money, the 6/8 core version is always as good, even if the 12/16 core vesrions have higher base frequencies, the difference is so little anyway and boosting higher does not mean anything, its how long a processor can maintain the boost clock, may the max be lower, and in that regard, the 6/8 core version has an advantage in the fact they overheat less and thus can boost longer.
I am sure you mean well. But I dont get how this repeated statement is helping the OP and his thread.

...that already got the CPU
 
Most people cant even get a 9800X3D anyways. It will probably be that way for some time..
 
I am sure you mean well. But I dont get how this repeated statement is helping the OP and his thread.

...that already got the CPU
He can sell the 9900x and buy a 9600/9700x, it can also help people reading the thread who hesitate between those processors.
 
I wouldn't drop a 9900X for a 9700X. With PBO, CO and the curve you can extract a lot more performance in the way of higher clock speeds all around.

To me it sounds like you are saying these big CPUs cant play games..
 
I wouldn't drop a 9900X for a 9700X. With PBO, CO and the curve you can extract a lot more performance in the way of higher clock speeds all around.

To me it sounds like you are saying these big CPUs cant play games..

Seriously. The cross-CCD delay is not that bad and in any case you can lock an application into one CCD.
 
Seriously. The cross-CCD delay is not that bad and in any case you can lock an application into one CCD.
I don't even bother with that. I just tuned it. Runs mint.. though X3D is a little better sometimes.. dammit.

Also doesn't run at 200w playing a game like my 5900X lol..
 
I wouldn't drop a 9900X for a 9700X. With PBO, CO and the curve you can extract a lot more performance in the way of higher clock speeds all around.

To me it sounds like you are saying these big CPUs cant play games..
I'm saying 9900x aren't better than 9600/9700x for gaming, basically same performance, even with PBO, CO and Curve, which you can use with 9600/9700x as well.

I've read shitloads of official and non official reviews and the difference varies, but is always within margin of errors, so basically you have no practical benefit from using a 9900x over a 9600/9700x for gaming.

I'm not saying the 9900x is bad for gaming, i'm saying its not worth the difference in price because the performances in gaming are basically identical.

Now if you do games and productivity, I'd pick the 9900x of course.
 
@Shakallia
lol, guess you need to tell that to my 5950, as it improved perf by a lot, even with my old 2080S, no stutter in any game that did, with my old 5800X,
same for win (10), as its snappier/runs better doing more than one (low load) thing at same time, all on savings plan with clocks limited to 1.7, even some older games.
so where the 5800 had to run at "unlimited" (stock clocks) for "smooth" use, the 5950 i have now, does it with lower power/temps on power savings plan.

where exactly is that making less sense, to have 16C?
completely ignoring what happens in 5y running an 8C vs 12/16, as not everyone upgrades every year...
 
Last edited:
Back
Top