Intel is back! An analysis in the next few years

E-Cores are love. E-Cores are life. 


As many of us know, Intel were the first to launch with a heterogenous x86 architecture. What this means is that there's more than just one type of core. Where every single core in a processor was designed identically, Alder Lake now has two types of cores (Similar to the big.LITTLE found in ARM-based mobile processors). The purpose of this is to maximize MultiThreaded (MT) performance for its die area. Intel's large cores (P-Cores) are extremely fast. However, they're also BIG. They take up a lot of die area, and adding more of them makes the chip bigger and more expensive to manufacture. The solution is add some small cores that aren't as fast, but are much faster relative to the space they take up. 

In the above image, we can see the BIG dark blue cores. Those are the P-cores, and the small light-blue E-cores have 4 cores in about the same die area as the P-cores. In reality a cluster of 4 E-cores are slightly wider than a P-core so it ends up being about 3 E-cores per P-core, with about 60% as much performance in each E-core as in each P-core. This means it's around double the performance per die area. This lets Intel release more performance for less cost. 

Some Alder Lake vs Zen3 Benchmarks to see where we're at now:

In order to consider where we're going we have to look at where we're at. Here's some benchmarks courtesy of Hardware Unboxed that compare the 12600k and 5600X and 5800X. This will not be a complete review recap. We're not going to spend a lot of time looking at Single Thread performance mainly because that's not the focus, however Intel DID clearly regain the ST crown. We're also not going to spend any time looking at gaming performance. Most games today are GPU-bound, and gaming performance is thus not very interesting. For current gaming, anything discussed will be more than fine. 
https://www.youtube.com/watch?v=LzhwVLUVork 12600k review
https://www.youtube.com/watch?v=vlngwUuDYoc 12700kf review
https://www.youtube.com/watch?v=WWsMYHHC6j4 12900k review



(yes I used paint and did some nasty-ass chopping up of the screenshots just for more direct comparisons because HUB only puts a few in each video)

And here's a convenient image from LinusTechTips https://www.youtube.com/watch?v=-EogHCFd7w0

Current conclusion:

It's pretty clear that in terms of MultiThreaded (MT) performance, the stack has been utterly shaken up
- The 12600k as in a COMPLETELY different league than the now lagging 5600X
- The 12600k is similar in performance, though sometimes beats the 5800X
- The 12700kf comes within a breath of beating the 5900X in most cases, and consistently beats 5800X
- The 12900k consistently beats the 5900X, and comes close to the 5950X in most productivity

At this point, while AMD maintains the MT crown, it no longer has the dominating lead it once had with Intel hot on it heels. In price per performance however, Intel is posing a massive threat to AMD, most evident with the 12600k and 12700k against the 5600X, 5800X, and 5900X. 


AMD's response? 

I'm not sure AMD will do too much to respond right now. There have been some price drops on shelves, but no official price drop from AMD and I don't expect any. Micro Center had a 5800X for as low as $300, but other than that, we haven't seen massive drops yet. I think for the current generation, we'll see prices drop based on the availability of Alder Lake and how well Ryzen continues to sell. 

Moving forward however, AMD is expected to release Ryzen 6000 in early 2022 with their new V-Cache technology. 
                    price    P+E    threads
i5 12600k    $300    6+4    16
i7 12700k    $450    8+4    20
i9 12900k    $650    8+8    24

My expectation for Ryzen 6000
                                        cores/threads
Ryzen 5 6600X    $300    8c/16t
Ryzen 5 6800X    $450    12c/24t
Ryzen 9 6900X    $650    16c/32t

My justification for this is that currently, Intel once again holds the Single-Thread performance. That means, in order to be competitive, they need to control the MT value. the 6900X will be the MT king, but still lose to even the 12600k in ST. This means it can't afford to be priced too much if at all higher than the 12900k. 

Raptor Lake vs Zen4

It's expected that about this time next year, both Intel and AMD will release their following generation. The 13000 series from Intel (Raptor Lake) and the 7000 series from AMD (Zen4). We already know that Raptor Lake will double-down on the E-cores and bring 16 E-cores to pair with 8 faster P-cores. We don't know as much about Zen4 other than it's going to feature a significant Core performance improvement. Since the first release of Ryzen, each consecutive generation has increased core performance by a greater factor than its previous one. This is expected to continue with Zen4 being a larger increase than Zen2 to Zen3 was. For the sake of this discussion, I can only assume that Zen4 will have similar ST performance as Raptor Lake (RPL) does. 

This is what I expect from Raptor Lake (RPL)

i3 13300k    $185    6        12t
i5 13600k    $300    6+8    20t
i7 13700k    $450    8+12  28t
i9 13900k    $650    8+16  32t

Yes, the 13700k should be much faster than the 12900k

Given this lineup, it's my opinion that in order to be competitive, AMD needs to release a 24 core CPU on mainstream. Releasing only a 16 core would result in them losing the high-end entirely. Here's my expected lineup: 

Ryzen 3 7400       $200    8c/12t      Faster than i3, slower than i5
Ryzen 5 7600X    $330    12c/24t    Faster than i5, slower than i7
Ryzen 7 7800X    $450    16c/32t    Very comparable to i7
Ryzen 9 7900X    $700    24c/48t    Faster than i9

In fact the reason I think the 13700k will be 8+12 is because otherwise I think it will lose to the 16c. 

To be honest, at this point these numbers seem like fantasy. How could AMD release a $200 product with the same number of cores as their current $450 offering? How could the mighty 8-core slip from Ryzen 7 to Ryzen 5, then to Ryzen 3 in consecutive generations? Well, to be honest, I'm not sure it will. However it's what I believe AMD will have to do if they want to keep up with Intel. Realistically, this segmentation will not yet be decided, and depend on how it's actually going to compare to APL. 

Zen5? 

At this point we're getting into some deep speculation. These chips are nowhere near finalized, but we do have some hints as to what's going on. Tom from Moore's Law is Dead revealed in his video here that the Zen5 generation is when AMD will join Intel in creating their first heterogenous architecture by bringing in a trimmed branch of Zen4 in as little cores called Zen4D, with the D standing for 'dense'. These Z4D cores will not be as fast as the full Zen4 core, but should be very space efficient relative to Zen5. Furthermore, AMD's chiplet architecture allows AMD to mix and match manufacturing processes. Zen4 and Zen4D will both be manufactured on TSMC's 5N process while Zen5 will be on 3N. 

Contrasting with Intel, where the small cores are about 1/3 the die area of the large cores, it seems like Zen4D will be only 1/2 the die area of a Zen5, with a single die of Zen5 with 8 cores being about the same size as a die of Zen4D with 16 threads. However, given that Zen4D is made on a lower cost node, this still gives similar cost benefits to Intel's approach. 

This seems like AMD will be able to have an 8+16 configuration in late 2023, about 2 years from now. However, if I am correct about them releasing a 24c Zen4 product, this will require them to already have a 3rd core die on the package. If this is the case, AMD could release a bullish 8+32 product that would remind us of the mortality of our current products. 

As for Intel, I know a lot less about their 14th gen which is expected in 2023 as well to compete with Zen5 and would be only pure speculation on my part to discuss. As such, I'll make absolutely no commentary on which way I think it'll go. Based on Intel's recent motivation, I find it likely that they'll have something incredible at this point as well. 



Thank you for reading!

As always, come visit me in Discordland to tell me I'm wrong! 

Discord: https://discord.gg/CHfha8V
Twitter: https://twitter.com/MeyerRants

Comments

Post a Comment

Popular posts from this blog

Hardware-accelerated IO in consoles coming to PC

Image
The prevalence and performance of SSDs There's been a lot of hype around SSDs for nearly a decade now, touting anything from 10x to 100x the read and write speeds as seen on the spinning rust hard drives of old. SSDs have become nearly ubiquitous on the modern PC for several reasons including tighter boot and load times, an improvement in overall system responsiveness, and an increasing reduction in cost. However, the real-world performance improvements we see from SSDs over their Hard Drive counterparts is often nowhere near what is expected. Not only do almost all even recent AAA titles run fine off a Hard Drive, games rarely load more than twice as fast when we install them on an SSD. Even more concerning is the often imperceptible difference installing an extremely fast -- and far more expensive -- NVMe SSD makes. But why is this? Why do games take so long to load even when our storage can clearly make up the difference.  Note: If you're already well versed in HDD a...
Read more

The best competition in CPU History is NOW

Why NOW is the best time for CPU competition How did we get here?  As anybody reading this probably already knows that we're in the middle of a CPU revolution. With the rise of AMD's zen architecture in 2017 we have been faced with unparalleled innovation in the CPU space with rapidly rising core and thread counts, increasing clock speeds, architectural and node improvements, and falling prices. This is no more evident than with the release of Intel's 10th generation desktop CPUs.  I can hear many of you already shouting at your screen, and don't tear me apart. It's no secret that even now, Intel has been stuck on minor iterations of a process node introduced in 2014, and an architecture introduced in 2015.  Intel has not only abandoned the tick/tock cadence they had for years, but even their process/architecture/optimization cadence has been utterly abandoned with the 10th generation offering none of the above. This seems a far cry from the claims of innov...
Read more

PS5 vs XSX vs PC featuring storage architectrures

How do Xbox Velocity Architecture and PS5 storage compare?  Preface -- Let's support better knowledge  Anyone that's been following my blog posts and even my guest appearance on Moore's Law is Dead's podcast  Broken Silicon , you'll know that one of my recent interests is trying to inform the community about not only what the storage is in the next gen consoles and how they work, but what benefits they can bring to gaming. The need for this knowledge is apparent in the widespread misconceptions about the SSD in the consoles and the hardware that accelerate its pipeline.  Evidence of the lack of understanding among techtubers can be seen  here , with Linus at the WAN show with comments like "and if you imagine for a second that Sony's CPU in the playstation is not just an AMD chip running PCI express then you're a fool". While I don't mean to discredit Linus, as I've learned an incredible amount from watching his channel, and indeed...
Read more