What is the point of 9800X3D in 4k? Isn't 9950X better at the same price?

[Zazigalka]

in what way is it better at 4k

View attachment 384501

this was done on 4090, wasn't it ? you're asking about pairing it with 5090.

 

[_roman_]

What is the point of 9800X3D in 4k? Isn't 9950X better at the same price?​

Cache vs Cores vs Software - that includes the Operating System and kernel

Won't read further as there are plenty of benchmark tables already posted

 

[Jtuck9]

I've read instances of people claiming to have maxed out their cache. Is that more likely to be an "issue" if you're strictly running 4K > ?

 

[JustBenching]

I mean those are literally the settings I play at lol. I use both balanced and quality whichever gets me to around 150 fps - 150+ fps feels amazing.

Point is this - it depends on what you're trying to do... alot of these "in the real world" arguments are missing the point -- if you game at high fps get the 9800x3d, if you do production then get the 9950x.

If you want big numbers and want to overclock a 16 core chip for funsies, then get the 9950x -- totally valid reason. If you want to run as close to 230 gsynced fps as possible get the 9800x3d. There's not really a wrong answer.

The whole "My XXXX CPU is still good when I run fully GPU bound" -- that's great and valid, but not everyone likes to game like that -- my 10850k is probably still good if I run fully GPU bound as 30fps locked 8K, that doesn't change anything. The 3D cache can net you 30% in games, will you use it? Depends on your settings - if you're gonna DLAA w/ Max pathtracing at 4k then get whatever you want because it doesnt matter at that point.

Just don't gaslight yourself into thinking there's no difference between these cpus, because you're NEVER gonna turn on DLSS balanced at 4k... or more powerful cards aren't going to come out in 12-18 months that will put more stress on the CPU.

Well, in reality though with my old 12900k the prime bottleneck (a major one mind you) in games is my 4090. Even the 5090 would be one. When we talking about AAA games (KCD2 , cybeprunk, alan wake, TLOU etc.) the CPU can get over twice the framerate of what my 4090 can at 4k with DLSS Q.


Heavier games in the future will be heavier for the GPU as well.

 

[AleXXX666]

So I am building a new PC for RTX 5090. I was looking for CPU.
9950X is the same price as 9800X3D right now.
What's the point of 9800X3D if I will use my PC only in 4k+?
Isn't it better to go X and have 8 extra cores for the same price?

I'd go extra cores any day

 

[alwayssts]

I've long-argued what's the point for really anything above an overclocked 5600x/7500f, maybe an 8-core iyw, for gaming? Everything's limited by console capability.

This will change soon (I think the PS6 will be 12c @ >/~3.85ghz), which I think will use around something similar to 10ghz 128-bit mem bandwidth. (5ghz of 32gbps GDDR7, the other 27gbps for GPU, 256-bit).
Using ~7 cores like a current console, this would be similar to x3d. Using 12c, more similar to a non x3d. Still better than AMD's current memory controller is capable of, but not maxing out bw limitations.
If you look at the construct of the current 3nm zen 5c Turin CCD etc, their lane transfer is up to 32gbps. This is what makes me think this. Anyway, think ~32GB @ ~10gbps comparatively on a desktop CPU.

There are two chains of thought comparing this to available products:
1. An overclocked 8-core (to say ~5800mhz) x3d would not be dissimilar to 12c/3.85ghz. That said, PS6 could be clocked higher (and/or 9800x3d artificially constrained) so it forces an upgrade (on purpose).
2. The parallelism of greater threads is more efficient when it can be used, and 16-cores do not need to be clocked as high to compete, but they do become bandwidth-limited bc current AMD mc speed sucks.

I don't think either thought process is wrong nor right; there are pros and cons to both...but the answer is neither is future-proof. I would simply wait for Zen 6. There are many reasons, but let's go over a few:

1. AMD will almost-certainly bump up their mc speed with Zen 6 (4nm IOD?). The why is simple: 7200mhz DDR5 has gotten extremely cheap, and it's an easy win to show a (much-needed) generational performance upgrade. For all intents and purposes, this bandwidth is the difference between a vanilla or x3d 8-core CPU. It's actually less than that at current clocks. I'd say ~7000mhz to make up the difference, give or take. Current mem controller can't really do over ~6200mhz reliably. While one could argue the current design is an artificial design constraint to sell x3d CPUs, in reality it saves a ton of cost by using an older (6/7nm) process; allows them to make the MCM cheaply and for packaging of mutiple CCDs on a package an affordable reality...so I'm cool with it. Also, until recently, ~6000mhz was the common-sense option wrt affordable ram. This has changed somewhat recently with the newer chips becoming less expensive, and those now aimed at a minimum of 7200mhz (cas 34) less than $100 for 32GB.

2. BC of this likely change, for AMD to continue selling x3d CPUs (that actually make any sense) they need to increase the CCD to 12 cores, where you could see the limitation of pure ram bw shifting from ~7000mhz to >10500mhz without v-cache. Given (current) v-cache adds something not dissimilar to ~1000mhz of bw, their memory controller kinda has to clock at least 6667mhz and the limit will likely be decided by how long they plan to keep the design (and therefore scaling v-cache designs) into the future. If they plan on ramping up v-cache amount etc to accentuate the spread. My *guess* will be it would do at least 7200mhz, maybe rated up to 8000mhz, but that's just a guess. I'll be plenty happy with ~c30/6750, but if it can scale to 8000mhz+, that would be cool. Just don't expect to let them overlap.

So what is the correct answer?

Buy what makes you happy and you think will give you the most benefit right now in your current use-cases, as both have their limitations (cores/speed vs bandwidth) for the future.

As for that future? What should you BOLO for?
Probably vanilla 12c Z6 but ram clocked as high/tight as you can. An X3D iyw, but I don't think you'll 'need' it...but you never know. Some games may use more bw but less threads beyond the pure Z6 mc.
Potentially even a lower-end model (8-10?), but nobody knows how PS6 dev capabilities will be set up yet (how many threads they can use) nor the split on the PS6 vs potential desktop clock capability.
Once we know those things (how to make a PS6 cpu as cheap as possible without any limiting factors), it will be the time to buy.

Until then, go w/ your gut and you do you.

 

[bug]

Is the Ryzen 9800X3D Truly Faster for Real-World 4K Gaming? | TechSpot
^
Because it's true?

You're posting an average across all games -- that's the "bill gates in a room full of nuns" problem - everyone in that room is, on average, a billionaire - but it's not the reality.

Not actually testing 4k and not pitting 9800X3D against a similar Zen5 CPU. You really think that is an indicator of the advantage of 3D cache and not a rigged test?

 

[Frizz]

Im more interested in this extra smoothness people are talking about despite the available benchmarks available on this website. Its really hard to believe theres any difference at all at 4K for triple A titles but guess I will find out soon . Also taking this opportunity to use this comparison for content (my bet is a placebo effect)