Intel Core i9-14900K Delivers Up to 23% Higher Performance than AMD Ryzen 9 7950X3D in Starfield and Other Titles, Official Gaming Benchmarks Show

Tsing

The FPS Review
Staff member
Joined
May 6, 2019
Messages
13,407
Points
113
HXL (@9550pro) has shared a new slide from Intel, revealing how the company's new flagship CPU, the Core i9-14900K, should fare against the AMD Ryzen 7950X3D in some of today's most popular games. According to Intel's benchmarks, the new Raptor Lake-S chip is able to outperform red team's option by up to 23% in 14 games under the 1080p/high setting, including Metro Exodus and Starfield, although it doesn't win every battle, showing worse performance by up to a similar percentage in titles that include DOTA 2, Cyberpunk 2077, and Red Dead Redemption 2. VideoCardz, which translated the titles, has estimated that the Core i9-14900K will only be 2% faster than the Ryzen 7950X3D on average in gaming based on Intel's benchmarks.

See full article...
 
Same cores, different options - unlocked, w/o GPU, locked, low power...

Is something for everyone a bad thing?
For us no. For someone uninformed expecting a 14900 to be the same as any other... yes.
 
For us no. For someone uninformed expecting a 14900 to be the same as any other... yes.
Well the 14900 will be an OEM product that you'll normally only find in big box OEMs. Sure, these will be available for DIY use but I feel like people who can build their own computer should be able to figure out the differences.
 
This is a terrible performance comparison.

Two things:

1.) If you want to compare against the best AMD has for gaming, use the 7800X3D, not the 7950.

2.) Starfield is a weird outlier with some very strange performance and is not at all reflective of typical CPU performance in games.

Other than that, yes, AMD needs to get ready to leapfrog Intel again.

I tend to think the infinity fabric is what is holding them back in this gen. It doesn't allow for sufficient RAM bandwidth. They need to really improve the infinity fabric in the next gen.
 
This is a terrible performance comparison.

Two things:

1.) If you want to compare against the best AMD has for gaming, use the 7800X3D, not the 7950.

2.) Starfield is a weird outlier with some very strange performance and is not at all reflective of typical CPU performance in games.

Other than that, yes, AMD needs to get ready to leapfrog Intel again.

I tend to think the infinity fabric is what is holding them back in this gen. It doesn't allow for sufficient RAM bandwidth. They need to really improve the infinity fabric in the next gen.
The issue with AMD's Ryzen and all its successors is their horrid internal latency compared to their Intel counterparts. They've mitigated this over time with improvements to the CCX / CCD layout, Infinity Fabric and of course, the biggest improvement has been throwing more L3 cache at the problem. But that stuff only goes so far.
 
The issue with AMD's Ryzen and all its successors is their horrid internal latency compared to their Intel counterparts. They've mitigated this over time with improvements to the CCX / CCD layout, Infinity Fabric and of course, the biggest improvement has been throwing more L3 cache at the problem. But that stuff only goes so far.

Yeah, latency is an inherent downside to chiplet designs.

The interconnects are never going to be able to be as fast as something that is on the same physical die.

It worked well when the chips and memory bandwidth were slower, but infinity fabric performance has not kept up with chip speed and memory bandwidth. If they can improve infinity fabric enough to keep up, I don't think they have a problem, but if they can't, they may be forced back to monolithic designs.
 
Yeah, latency is an inherent downside to chiplet designs.

The interconnects are never going to be able to be as fast as something that is on the same physical die.

It worked well when the chips and memory bandwidth were slower, but infinity fabric performance has not kept up with chip speed and memory bandwidth. If they can improve infinity fabric enough to keep up, I don't think they have a problem, but if they can't, they may be forced back to monolithic designs.
I don't think AMD will go back to monolithic designs anytime soon if ever. The chiplet really gives them their primary advantages right now. Sure, there is a cost but they have largely been able to mitigate it with more L3 cache. Predominately. the issue comes from gaming which is one of the more latency sensitive applications. For general productivity work, its not necessary. They've done well on that front. Plus, don't forget that the X3D cache versions closed that gap a bit more.
 
I don't think AMD will go back to monolithic designs anytime soon if ever. The chiplet really gives them their primary advantages right now. Sure, there is a cost but they have largely been able to mitigate it with more L3 cache. Predominately. the issue comes from gaming which is one of the more latency sensitive applications. For general productivity work, its not necessary. They've done well on that front. Plus, don't forget that the X3D cache versions closed that gap a bit more.

True, but then again, general productivity work doesn't require a high end CPU. You could do most modern general productivity work on a C2D from 2005. :p

But I'm with you, chiplets helps AMD's bottom line a lot. It VASTLY improves overall yields compared to large monolithic designs. For this reason they are not going to want to move away from them unless they are absolutely forced to. At least not unless FAB capacity gets a lot cheaper again, and yields go way up, neither of which seem likely any time soon (or ever)

Intel is pushing some pretty high performance chips right now, but they must have some razor thin margins due to the largish dies.

This is why I think AMD has to focus on doing whatever they can to bring down infinity fabric latency and bring up infinity fabric bandwidth. This problem is only going to get worse as chip and memory speeds improve if they don't.
 
For us no. For someone uninformed expecting a 14900 to be the same as any other... yes.
For someone uninformed it might as well be.

Wait is that the 14900, the 14900K, the 14900KF, 14900T, or the 14900F?

They said the K.... buuut who's double checking with so many options!?
You forgot the KS, or is there no KS this time?
 
This is a terrible performance comparison.

2.) Starfield is a weird outlier with some very strange performance and is not at all reflective of typical CPU performance in games.

But it is representative of Starfield performance, if you play Starfield, will matter. I always like to say that things are "game-dependent" cause they really are. But what this means, is some games are better in some regard on Part A vs Part B, but that is relevant information still, and worth testing because it does tell you how that game performs, and if it's a popular game, if it's a game that millions are playing, and if it's a game that matters to you, then it's important and relevant.

To me, it's all about telling the reader how games perform, and what's better or worse, or the experience in that game. I'd love to cover every game, naturally, but if you stand back, and just look at it from the perspective of telling a story for that game, then that's the angle.

The fact that it is a game, means it is worth exploring, IMO.
 
Same cores, different options - unlocked, w/o GPU, locked, low power...

Is something for everyone a bad thing?
For us no. For someone uninformed expecting a 14900 to be the same as any other... yes.
I can see both sides of this and unfortunately, the average Joe is "sool" unless they do some research or are able to look at something explaining it as simply and to the point as @LazyGamer did.
 
Become a Patron!
Back
Top