Asenna Steam
kirjaudu sisään
|
kieli
简体中文 (yksinkertaistettu kiina)
繁體中文 (perinteinen kiina)
日本語 (japani)
한국어 (korea)
ไทย (thai)
български (bulgaria)
Čeština (tšekki)
Dansk (tanska)
Deutsch (saksa)
English (englanti)
Español – España (espanja – Espanja)
Español – Latinoamérica (espanja – Lat. Am.)
Ελληνικά (kreikka)
Français (ranska)
Italiano (italia)
Bahasa Indonesia (indonesia)
Magyar (unkari)
Nederlands (hollanti)
Norsk (norja)
Polski (puola)
Português (portugali – Portugali)
Português – Brasil (portugali – Brasilia)
Română (romania)
Русский (venäjä)
Svenska (ruotsi)
Türkçe (turkki)
Tiếng Việt (vietnam)
Українська (ukraina)
Ilmoita käännösongelmasta
In theory, the Ryzen 9 X3Ds are hybrids that trade only a bit of gaming and a bit productivity while being almost the best blend of both, but that sort of only applies for the 7950X3D. If you're doing gaming, a 7800X3D is better, and if you're doing productivity, a 7900/7900X is better.
In short, the 7800X3D, because if you have productivity needs on top of that, just go with the 7950X3D. The 7900X3D really has little reason to exist.
7700X or 7800X3D is the choice for Gaming.
Please note that if you do go with 7800X3D be sure to get a decent CPU cooler and also be mindful that the X3D CPUs are much pickier when it comes to which RAM you use.
The v-cache isn't a complete net gain. It adds cost. It perhaps adds to cooling needs. And beyond those, it can actually detract from performance. While the v-cache itself almost never hurts performance, having it means the CCD needs to be clocked lower, so when something isn't cache or I/O bound, it just loses a small bit of performance due to that lower clock speed. Not much, no, but ultimately it's a "you're paying quite a bit more yet losing slightly on performance" and that's not an attractive offering.
The v-cache chips are primarily marketed towards games as that's a prime use case that benefits from it, and the gains far outweigh the losses there so the average performance is still higher. So it's a net gain there even if there are some losses, and a big enough part of the market finds that average uplift worth the extra cost. The 5800X to 5800X3D was like a, what, 15% to 20% uplift or so on average in games (and that average is dragged backwards a lot by "doesn't gain at all so losses some due to lower clock speeds" and "only gains a bit" but a significant portion of games gain well beyond that, and the minimums are raised even more than the average). People will pay for that even when the value is worse when it represents some of the fastest gaming performance you can get at the time.
So getting back to the Ryzen 9s, you have to ask what impacts there could be if both CCDs were v-cache equipped instead of one.
Well, you'd add to cost obviously, so would it be worth it?
In the case of the 7950X3D, you now have a CPU that's an extra bit slower relative to the 7950X in stuff that isn't v-cache benefiting. So it's marginally worse again in productivity. What about games? What games are going to see enough of a gain from more than eight cores that an even more expensive 7950XD would be appealing? It's now a less appealing CPU for both tasks. As an added bonus, if you know you have a game that doesn't benefit from v-cache, you can assign it to the non-X3D CCD and gain a bit of performance. You'd lose that too.
The 7950X3D is unironically better this way, at least in the present software world.
In the case of the 7900X3D, some things a bit different, but even this one probably comes out worse. The 7900XD is unappealing because it costs more than the 7800X3D but is worse in games. A lot of thie comes down to having two CCDs, and each being 6 cores instead of 8. So not only does it have the interconnect to contend with more often, but it has less v-cache cores outright. There's definitely room in the present market for a hex core X3D... but it would need to be on its own as a lower cost option, not found in an awkward higher cost option like the 7900X3D. But even with both CCDs equipped with v-cache, it's still dual CCDs and the extra four cores are waste in games more enough than not to outweigh that (plus the extra cost). It would still be a dead CPU almost nobody wants. In the future, when 12 cores can be on the same CCD and more games need above 8 cores, maybe things change.
So ultimately in both cases you end up with costlier CPUs that are worse than the present ones. They would be less appealing and worse off for it.
There is a similar reason Intel's top SKUs would be worse if all the e-cores went away and were replaced with P-cores. There's not many things that need that many P-cores, and for the things that scale "infinitely", the e-cores provide better supplementary performance for the space they require. As it is, some gamers tend to disable those and they sometimes see increases in performance (scheduling conflicts?), which suggests the remaining 8 cores are plenty for games and more would be too niche right now.
Also, sorry for the long thoughts. I just found your comment interesting because I had the same exact "wait, why only half..." reaction when they were first announced, but over time I have come to realize AMD certainly knew what they were doing here (same as how Intel's mixed approach has its benefits).
i9 10900k 10@5.125
32GB DDR4 PC3000
3090RTX Gaming OC Gigabyte 2100@9875
2 1TB NVME M.2
1000WATT
VULKAN runs perfectly fine on NVIDIA GPUs. Even older ones like GTX 900 / 10 series.
I'm using 3080 Ti w/ Vulkan in RDR2 and it runs better then any AMD GPU out there.
I have experienced games that had issues while VULKAN was used, but i also had games that had issues with DX12.
i just got a AMD for the wife my first AMD gpu ever wont make that mistake twice it was
problematic right out of box seems they are as unstable as their cpu's
I would get 10850k or 11700k with RTX3090, E-cores can cause stuters on 12/13/14th gen Intel similar to AMDip. Even if you disable them, latency is lower on 10/11th gen and whole PC feels more responsive.
Latency and responsiveness for 7000 series was literally better than Intel's when it came out but that wasn't difficult to do because the hybrid P-core+E-core design was freaking stupid. CPUs with multiple core dies are always going to have more latency than a CPU with fewer or a single die, AMD had the same latency problems with Threadripper and Ryzen 9, still do in some instances but it's been greatly diminished as they improved the design.