Question Intel Raptor Lake vs AMD Zen 4 vs Apple M2

[Carfax83]

These CPUs are all going to square off against each other at some point this year assuming nothing catastrophic occurs to delay any of the product launches. So going by what we know from official sources and informed rumor mongers (many of which were very accurate before Alder Lake and the M1 launched), which CPU do you think will win out in these categories?

1) Single threaded performance
2) Multithreaded performance
3) Gaming performance
4) Performance per watt
5) Overall performance (who wins the majority of applications)

While I've been keeping a close eye on rumors and leaks for Zen 4 and Raptor Lake, I have not admittedly been doing so for the M2; as I'm unrepentant Apple hater  At least I'm honest about it... That said, this is my ranking based on what I've seen and heard:

I think the single threaded crown will go to Raptor Lake, and I say this based on informed rumors that Raptor Lake will have up to 10% more IPC from microarchitectural updates, cache upgrades and higher clock speeds than Alder Lake. From what I've seen, gauging IPC performance isn't easy as it varies so much based on application, but I'd say Alder Lake already has at least a 15% across the board IPC advantage over Zen 3, so Raptor Lake could conceivably have 25% better IPC than Zen 3, which is similar to what Zen 4 will reportedly possess. But I doubt Zen 4 will match Raptor Lake in clock speeds and memory latency performance, which is why I'm predicting Raptor Lake will take the single threaded performance crown.

For multithreaded performance, Zen 4 should easily take it due to having more big cores than its Intel counterpart and similar IPC.

Gaming performance is more complicated because while some games are inherently more reliant on single core performance (strategy games for instance), more and more 3D engines are becoming increasingly parallel due to the adoption of Vulkan and DX12 in addition to modernized programming methods. Still, very few 3D engines can scale beyond 8 threads and 6 to 8 cores remains the sweet spot for gaming and will be for some time. So overall, I feel more comfortable going with Raptor Lake for the gaming crown. Also if rumors are correct, Raptor Lake will officially support DDR5-5600 off the bat while Zen 4 will reportedly use DDR5-5200. The raw memory speed won't likely be a significant factor, but Intel's memory controller will be right next to the CPU cores while Zen 4's will be in the I/O die which while still on the same package will definitely incur a significant latency penalty; which I'm sure will be offset by a massive L3 cache.

On performance per watt, one would think the M2 should take this category easily......but from the small amount of research that I've collected on it, it seems that there won't be much of a performance increase with the M2, if any. Some rumors are even suggesting there may be a bit of a regression in that aspect. Also since Zen 4 will be on TSMC's 5nm node, it will undoubtedly have excellent performance per watt and I believe it will also easily crush Apple's best in single core. So for performance per watt, I'm going to go with Zen 4.

When it comes to overall performance, I'm leaning towards Zen 4 but it will be close. Raptor Lake will supposedly double the amount of Gracemont efficiency cores which will certainly help in multithreaded performance per watt, but ultimately they won't be a match for Zen 4's 16 big cores with SMT. AMD will have the core count advantage and when that's combined with IPC parity with Raptor Lake, Zen 4 will win the majority of the benchmarks.

 

[igor_kavinski]

Intel is welcome to put more than 8 Golden Cove or Raptor Cove cores on their consumer CPUs using tiles, and charge a premium for the product. But they won't do it.

16 GC cores even with tiles might exceed 400W. It's not that they don't want to. They can't.

 

[Carfax83]

The benchmarks from a few pages back show 12600K being 50% faster in LLVM while using 27% more power.

I've always sucked at math. Isn't the 12600K 34% faster than the 5600x, while the 5600x is 75% slower than the 12600K?

With that out of the way @tamz_msc probably had an issue with you extracting conclusions about the strength of Golden Cove from a product that wins against the competition using a larger mix of P+E cores. You compare 6+0 from AMD with 6+4 from Intel and conclude Intel Cove architecture is much better because.... 12600K is priced equally with 5600X. It doesn't get more apples and oranges than this.

I wouldn't mind seeing the performance breakdown of the P vs E cores in that benchmark, but I wouldn't be surprised if they are a small factor in the 12600K's performance since its just 4 of them at relatively much lower clock speeds and IPC. Also, the advantages of having a hybrid architecture is one of the "ace up their sleeve" strategies Intel hopes to demolish AMD with in the coming years.

As @JoeRambo said, Intel is just going to keep piling on the efficiency cores.

 

[igor_kavinski]

16 GC cores even with tiles might exceed 400W. It's not that they don't want to. They can't.

Well, maybe they can, with 3 GHz all core boost?

 

[Zucker2k]

16 GC cores even with tiles might exceed 400W. It's not that they don't want to. They can't.

The concern is not power, it's size. As for @DrMrLordX suggestion, the reverse is also true, and AMD needs to be worried because the 12600k is trading blows a tier above it's category, with the 5800x. That to me means, a 6+0 6600x is going to be barely faster than the 12600k, much alone the 13600k with potentially more cores. They're going to have to add more cores or Zen 4 will be DOA in multithreading.

 

[Abwx]

Let's match process nodes, cache sizes, and latencies as well, while we're at it. 5950x v 12900k is 32 threads vs 24 threads all day, not to mention the 5950x is more expensive but you don't call out people for comparing to two, not to mention the 5600x is more expensive on Newegg.

12900K vs 5950X is one of the most flawed comparison, let s cap ADL to 125W and set the 5950X to 220W, would you say that the latter is a greatly performing CPU with such conditions.?.

 

[epsilon84]

12900K vs 5950X is one of the most flawed comparison, let s cap ADL to 125W and set the 5950X to 220W, would you say that the latter is a greatly performing CPU with such conditions.?.

Considering the law of diminishing returns WR to overclocking/overvolting, an argument can be made that such a scenario might actually put Intel in a better light, save for the benchmark chart warriors. AMD's main remaining advantage against Intel right now is performance/watt (when comparing the 5950X to the 12900K) and that would just throw that last advantage out the window.

The 5950X would of course then dominate in MT workloads whilst chugging a lot more power, while a 125W 12900K would be hailed as an efficiency champion - still retaining its ST/gaming crown albeit losing ~15% of its MT performance: https://www.club386.com/intel-core-i9-12900k-at-125w/3/

As it stands: 5950X wins in performance/watt, competitive in MT, loses in ST/gaming

Your scenario: 5950X loses in performance/watt, dominates MT, loses in ST/gaming.

12900K becomes the efficiency king whilst retaining its lead in ST and gaming, whereas the 5950X would be the 'brute force' MT king. I honestly think the current status quo suits AMD better.

 

[Abwx]

Considering the law of diminishing returns WR to overclocking/overvolting, an argument can be made that such a scenario might actually put Intel in a better light, save for the benchmark chart warriors. AMD's main remaining advantage against Intel right now is performance/watt (when comparing the 5950X to the 12900K) and that would just throw that last advantage out the window.

The 5950X would of course then dominate in MT workloads whilst chugging a lot more power, while a 125W 12900K would be hailed as an efficiency champion - still retaining its ST/gaming crown albeit losing ~15% of its MT performance: https://www.club386.com/intel-core-i9-12900k-at-125w/3/

As it stands: 5950X wins in performance/watt, competitive in MT, loses in ST/gaming

Your scenario: 5950X loses in performance/watt, dominates MT, loses in ST/gaming.

12900K becomes the efficiency king whilst retaining its lead in ST and gaming, whereas the 5950X would be the 'brute force' MT king. I honestly think the current status quo suits AMD better.

My point is that we remove TDP from the equation then all comparisons are meaningless, at same TDP and MT 12900K would compete against a 3950X or a 5900X but certainly not against a 5950X.

Intel Core i9-12900K, i7-12700K & i5-12600K im Test: Benchmarks in Anwendungen

Intel Alder Lake im Test: Benchmarks in Anwendungen / Ein rekordverdächtiger Testumfang / Windows 10 vs. Windows 11

www.computerbase.de

 

[mikk]

Plus gracemont cores will have more l2 and raptor cove will have more ipc than golden.. amd will be massacred in laptops

The big deal for laptop will be Meteor Lake and not Raptor Lake, at least when it comes to Intel. Meteor Lake is full of innovations because it's a chiplet design, it's a shrink from Intel 7 to 4, it comes with a new Cove and a new Mont (Redwood Cove+Crestmont) apparently and also comes with a new iGPU architecture from 12.2LP to 12.7HPG. If there is 1 year between ADL mobile and RPT mobile it comes during Q1-Q2 2023 which is likely very close to Meteor, even though I expect that Raptor mobile is required because Meteor won't fulfill the volume Intel needs for the entire mobile business.

 

[epsilon84]

My point is that we remove TDP from the equation then all comparisons are meaningless, at same TDP and MT 12900K would compete against a 3950X or a 5900X but certainly not against a 5950X.

Intel Core i9-12900K, i7-12700K & i5-12600K im Test: Benchmarks in Anwendungen

Intel Alder Lake im Test: Benchmarks in Anwendungen / Ein rekordverdächtiger Testumfang / Windows 10 vs. Windows 11

www.computerbase.de

Its plain and obvious that a 241W 12900K exists purely for overall performance crown reasons, the 'geomean champion' if you will. The 125W version is what Intel would have released, had the 5950X not existed. However, we don't live in a competitive vacuum, and the stock 241W 12900K is the result, for better or worse.

I agree with you that at the same power, a 12900K would not compete with a 5950X in MT. It is at a distinct thread disadvantage, and there is no possible way to overcome that except for what Intel is currently doing - overclocking the P cores to compensate. OR... to add more E cores, ala Raptor Lake...

 

[epsilon84]

Getting back to the thread topic though... I'm curious as to how so many are already declaring a victory for Zen 4 based on IPC and clockspeed gains alone? Whilst RPL is getting an additional 8 E cores that should add 25-30% more MT throughput vs the current 8+8 design. Zen 4 may well bridge the gap in ST and clockspeed, but Intel is also bridging the gap in thread count.

Looking back, Zen 1 -> Zen 2 - > Zen 3 all yielded less than 20% improvement (core for core) per iteration on average. That 20% factors in generational increases in clockspeed as well, if we are talking IPC gains it is probably closer to 10-15% gains per iteration.

Yet I'm hearing people are expecting a 40% improvement from Zen 4 over Zen 3?! Wishful thinking or am I not aware of some magic sauce that AMD is cooking up?

 

[tamz_msc]

This sentence to me is nonsensical. The reason why I drew the comparison between the 5600x and the 12600K is because they are priced similarly, and the big core count is the same. Actually, going by Newegg prices, the 5600x is more expensive than the 12600K.



It's very rare to see Intel or AMD have such a large gap between them in any workload not factoring core count these days. Cypress Cove looks paltry to me compared with Golden Cove.

Again, you're making statements about the core itself, while supporting those statements by using a comparison between a 6+4 core SKU and a 6+0 core SKU. If you want to speak about the strengths of Golden Cove, do so with examples of Zen 3 core vs Golden Cove core.

An apples to apples comparison would be something like the 5600G vs 12400, both 65 W, both having same boost clock/turbo. Fortunately PTS has the comparison that you would want i.e. compilation benchmarks.

Intel Core I5 12400 Linux Benchmarks Performance - OpenBenchmarking.org

openbenchmarking.org

You can see that the performance characteristics depends heavily on the code base that is being compiled. Now you can indeed say that Golden Cove is impressive against Zen 3 in compilation benchmarks.

 

[tamz_msc]

Looking back, Zen 1 -> Zen 2 - > Zen 3 all yielded less than 20% improvement (core for core) per iteration on average. That 20% factors in generational increases in clockspeed as well, if we are talking IPC gains it is probably closer to 10-15% gains per iteration.

That's simply not true. Zen 1 > Zen 2 SC perf was theoretically 1.15*(4.7/4.1) = ~1.32x, this is comparing the ST clocks of the top SKUs(1800X and 3950X).

 

[Abwx]

Getting back to the thread topic though... I'm curious as to how so many are already declaring a victory for Zen 4 based on IPC and clockspeed gains alone? Whilst RPL is getting an additional 8 E cores that should add 25-30% more MT throughput vs the current 8+8 design.

8 E cores add 35% perf to 8 P cores, so 8 extra E cores wont add 25-30% to 8 + 8, not counting that they require 50W to be added on top of 241W.

 

[epsilon84]

That's simply not true. Zen 1 > Zen 2 SC perf was theoretically 1.15*(4.7/4.1) = ~1.32x, this is comparing the ST clocks of the top SKUs(1800X and 3950X).

Indeed, my mistake, I mistook Zen refresh for Zen 2 when I was looking at the differences between various Zen CPUs (via TPU charts)

It looks like the biggest jump was from Zen 1 to Zen 2, which was over 30% as you said, though Zen 2 to Zen 3 was indeed under 20% (3900XT vs 5900X)

 

[epsilon84]

8 E cores add 35% perf to 8 P cores, so 8 extra E cores wont add 25-30% to 8 + 8, not counting that they require 50W to be added on top of 241W.

That 50W budget can easily be achieved by simple voltage optimisations with very little performance loss: https://www.techpowerup.com/review/...er-lake-tested-at-various-power-limits/2.html

Basically Intel has to avoid overvolting the P cores to stupid levels to eek out that last 100-200MHz like they do with the 12900K.

As for potential 8+8 vs 8+16 gains, I was looking at this and did some rough estimates:

https://www.techpowerup.com/review/intel-core-i9-12900k-alder-lake-12th-gen/6.html shows that in CB 8+0 gets 20105 while 8+8 gets 27780. the 8E cores nets an extra ~7.7K points. Adding another 8E cluster will add ~7.7K on top of that, which takes a hypothetical 8+16 12900K to ~35.5K. 35.5/27.8 = 1.28x the performance

This is assuming perfect core scaling of course, which CB seems to exhibit.

 

[Zucker2k]

8 E cores add 35% perf to 8 P cores, so 8 extra E cores wont add 25-30% to 8 + 8, not counting that they require 50W to be added on top of 241W.

This is far from what would happen. 8 gracemont cores running at 3.9GHz consume 48w (Anandtech review). This means 16 cores running at 3.3GHz could be consuming 70w or less, leaving 150w+ for golden cove cores alone, caches, etc. The 12900k (with full 16 cores) at 150w is already a potent unit. 150w is more than enough for the 8 p cores in raptor lake.

 

[DrMrLordX]

16 GC cores even with tiles might exceed 400W. It's not that they don't want to. They can't.

They probably can't reliably fab anything that large as a monolithic die, but with tiles - assuming there isn't something inherently broken with their tile approach that's also holding back Sapphire Rapids - they could do it. And if they dialed back all-core turbo to somewhere in the 4GHz ballpark then power wouldn't be that out-of-control either.

Stubbornly, Intel refuses to bring tiles to consumer products until Meteor Lake. Kinda makes you wonder why?

AMD needs to be worried because the 12600k is trading blows a tier above it's category, with the 5800x.

We don't even know if AMD will bother selling 6c products anymore. They don't need to be worried, though.

 

[eek2121]

Getting back to the thread topic though... I'm curious as to how so many are already declaring a victory for Zen 4 based on IPC and clockspeed gains alone? Whilst RPL is getting an additional 8 E cores that should add 25-30% more MT throughput vs the current 8+8 design. Zen 4 may well bridge the gap in ST and clockspeed, but Intel is also bridging the gap in thread count.

Looking back, Zen 1 -> Zen 2 - > Zen 3 all yielded less than 20% improvement (core for core) per iteration on average. That 20% factors in generational increases in clockspeed as well, if we are talking IPC gains it is probably closer to 10-15% gains per iteration.

Yet I'm hearing people are expecting a 40% improvement from Zen 4 over Zen 3?! Wishful thinking or am I not aware of some magic sauce that AMD is cooking up?

AMD is moving from N7 to N5. A 20-25% performance improvement is actually a very conservative estimate. If AMD correctly utilizes N5, the gains could easily be in the 40% range. Shoot, a straight up Zen 3 port would add 25% MT performance…

At the end of the day, even with optimizations for power, Intel still has a hard power limit of 240W. Adding more cores is going to make the chip even less efficient.

Also, for the record, the 5950X is currently $685 on Amazon. A 12900k is $614. Those two chips are priced right where they should be.

A little monkey tells me AMD has nothing to worry about until Intel transitions to Intel 4.

 

[Abwx]

This is far from what would happen. 8 gracemont cores running at 3.9GHz consume 48w (Anandtech review). This means 16 cores running at 3.3GHz could be consuming 70w or less, leaving 150w+ for golden cove cores alone, caches, etc. The 12900k (with full 16 cores) at 150w is already a potent unit. 150w is more than enough for the 8 p cores in raptor lake.

I assumed that they would be clocked at same frequency than ADL, there is more benefit in doing so than to lower the whole E cluster clocks at 3.3.

With such an implementation they could shave 50W out of the P cores and getting them running in a more favourable part of the V/F curve.

 

[Henry swagger]

That 50W budget can easily be achieved by simple voltage optimisations with very little performance loss: https://www.techpowerup.com/review/...er-lake-tested-at-various-power-limits/2.html

Basically Intel has to avoid overvolting the P cores to stupid levels to eek out that last 100-200MHz like they do with the 12900K.

As for potential 8+8 vs 8+16 gains, I was looking at this and did some rough estimates:

https://www.techpowerup.com/review/intel-core-i9-12900k-alder-lake-12th-gen/6.html shows that in CB 8+0 gets 20105 while 8+8 gets 27780. the 8E cores nets an extra ~7.7K points. Adding another 8E cluster will add ~7.7K on top of that, which takes a hypothetical 8+16 12900K to ~35.5K. 35.5/27.8 = 1.28x the performance

This is assuming perfect core scaling of course, which CB seems to exhibit.

A souce told me intel is aiming fir 38k to 40k multi-core on cimebench with raptor lake they.ll overclock to the heavens to reach it 😀

 

[epsilon84]

AMD is moving from N7 to N5. A 20-25% performance improvement is actually a very conservative estimate. If AMD correctly utilizes N5, the gains could easily be in the 40% range. Shoot, a straight up Zen 3 port would add 25% MT performance…

At the end of the day, even with optimizations for power, Intel still has a hard power limit of 240W. Adding more cores is going to make the chip even less efficient.

Also, for the record, the 5950X is currently $685 on Amazon. A 12900k is $614. Those two chips are priced right where they should be.

A little monkey tells me AMD has nothing to worry about until Intel transitions to Intel 4.

What portion of those gains are IPC and what portion are clockspeed increases? I'm assuming you think a 6950X will hold an all core turbo speed of 5.0GHz then? Which then leaves 15% IPC gains?

FWIW I'm aware of the 5GHz demo of Zen 4 running Halo. Running games at 5GHz is a far cry (excuse the pun) to a 5GHz all core workload like rendering. Even the most heavily multi-threaded of titles don't come close to utilising all cores on a 16 core CPU - and Halo Infinite can only utliise 4 cores/threads: https://www.dsogaming.com/pc-perfor...nfinite-pc-performance-analysis/#&gid=1&pid=6

It was an impressive tech demo nonetheless, but there is a ways to go before we can declare ~5GHz all core is a reality.

Unless AMD is willing to 'do an Intel' and have the 6950X running at 200W+, I would bet that the sustained all core boost will be under 5.0GHz. Happy to be proven wrong by the wonders of the N5 process though. If AMD can really hit those clocks without insane power levels then I think my next upgrade would be a no brainer

 

[repoman27]

The big deal for laptop will be Meteor Lake and not Raptor Lake, at least when it comes to Intel. Meteor Lake is full of innovations because it's a chiplet design, it's a shrink from Intel 7 to 4, it comes with a new Cove and a new Mont (Redwood Cove+Crestmont) apparently and also comes with a new iGPU architecture from 12.2LP to 12.7HPG. If there is 1 year between ADL mobile and RPT mobile it comes during Q1-Q2 2023 which is likely very close to Meteor, even though I expect that Raptor mobile is required because Meteor won't fulfill the volume Intel needs for the entire mobile business.

Minor correction, Meteor Lake will have updated IGP but it's still Xe LP (low power, integrated), not Xe HPG (high performance, discrete). I believe the specific flavor for Meteor Lake is Gen12.722, vs. Gen12.71 for Alchemist (Xe HPG) or Gen12.721 for Ponte Vecchio (Xe HPC).

 

[epsilon84]

A souce told me intel is aiming fir 38k to 40k multi-core on cimebench with raptor lake they.ll overclock to the heavens to reach it 😀

Some people might like it, but Intel 'pre overclocking' everything kinda takes the fun out of well... overclocking. Maybe I'm just old school.

400W for 40k?

 

[coercitiv]

This is far from what would happen. 8 gracemont cores running at 3.9GHz consume 48w (Anandtech review). This means 16 cores running at 3.3GHz could be consuming 70w or less, leaving 150w+ for golden cove cores alone, caches, etc. The 12900k (with full 16 cores) at 150w is already a potent unit. 150w is more than enough for the 8 p cores in raptor lake.

You're trying to disagree with Abwx while partially reinforcing his point. RPL is bound to drop clocks versus ADL (in fulll MT loads), though I expect less than 10% as you mention, maybe 5%. With that in mind, take the 28% estimate for MT improvement from @epsilon84 , and deduct 5%. The resulting gain is 22%.

Obviously this 22% is more of a baseline, as we need to factor in any IPC gains for Raptor Cove, any optimization on the voltage rails and also probable refinement of Intel 7. On the downside Cinebench scaling is the best case scenario, let's not forget that every increase in core count sees diminishing returns in other consumer oriented benchmarks.

It looks like the biggest jump was from Zen 1 to Zen 2, which was over 30% as you said, though Zen 2 to Zen 3 was indeed under 20% (3900XT vs 5900X)

Zen2 to Zen3 was done on the same process mind you (well obviously a tweaked version). Zen4 will benefit from both the jump to 5nm and the move to the new socket (new electrical specs). The info we have so far is AMD may increase the TDP on their flagship consumer desktop. To summarize the list of probable factors adding towards higher performance:

• Zen4 core will be considerably bigger in terms of transistor count
• AM5 max PPT will likely increase
• 5nm will offer better overall efficiency
• we already know the new core on 5nm can sustain 5Ghz all-core in moderate workloads (gaming), which means dynamic range is excellent (they're bound by power and/or thermals)

The point of all this is Zen4 will likely have the headroom to increase all-core clocks in heavy workloads vs Zen3, unlike RPL which will see stagnation at best. Last time AMD changed nodes and architecture (Zen1>Zen2) they claimed 15% better IPC and 25% better performance at ISO power. Add more power this time around, since I'm pretty sure they'll join the stupid power limit game, and gains of over 25% are very likely.

As you can see, the napkin math says the gains are quite evenly matched (as far as we can extrapolate obviously), in fact it pretty much requires Raptor Cove to be better than Golden.

 

[gruffi]

AMD needs to be worried because the 12600k is trading blows a tier above it's category, with the 5800x.

No. It's not above its category.

5950X <-> 12900K
5900X <-> 12700K
5800X <-> 12600K
5600X <-> 12400/12500

When Intel still needs more die area with every core configuration, ranging from ~15-75%. 5800X and 12600K perform similar at multicore and gaming. 12600K is only faster single core. But we know Zen 4 will increase single core performance significantly. So, AMD don't really need to be worried. Especially not against Alder Lake and probably not even against Raptor Lake. And that's also the reason why they don't plan to offer more than 16 cores on the mainstream platform. It's just not necessary to compete and keeps production costs as low as possible. But of course they could do at least a 24 core design if needed. It would completely demolish Alder Lake and Raptor Lake.