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.

 

[Cardyak]

Intel: According to my rough stats, Golden Cove has ~7% higher IPC than Zen 3, and I'm banking on only a ~5% IPC increase with Raptor Lake (10% seems a little ambitious to me), and that's only on the big cores. Clock speeds will be higher though, so I reckon a 10% increase in performance overall. Nothing to really get excited about.

AMD: Zen 4 will most likely be a whopper, I'm banking on 25% IPC increase and ~15% clock speed increases, so overall I expect about ~40% performance increase, and that's at the same core counts, hence this applies to ST as well as MT.

Apple: M2/A16 Big core should be a large increase given that Avalanche was focused more on efficiency and clock speed increases over Firestorm. It's been a while now since Apple had a major core revision from a uarch standpoint, I'd hazard a guess at a 15% IPC increase and also a slight clock speed bump. No idea about the little cores.

Winners:

1) Single threaded performance - I'll estimate Zen 4 will edge out Raptor Lake, just. Basically par for par. Apple some way behind due to lower clock speeds.
2) Multithreaded performance - Zen 4. No question.
3) Gaming performance - Never really sure why people treat this like a different workload, every workload is unique and different to others in it's own respects, so not really sure what do about this. Intel might be slight favourites due to monolithic design but really it comes down to how optimised the engine is for multi-threading
4) Performance per watt - Apple. I can 100% guarantee this. No one is within a mile of Apple when it comes to Perf/W, and I don't expect this to change for at least a few years.
5) Overall performance (who wins the majority of applications) - Zen 4 is a solid bet, I just can't quite see Raptor Lake cutting the mustard. Seems like a another "minor" upgrade that Intel puts out during alternating years. (Really reminds me of Tiger Lake/Willow Cove)

 

[Carfax83]

Intel: According to my rough stats, Golden Cove has ~7% higher IPC than Zen 3, and I'm banking on only a ~5% IPC increase with Raptor Lake (10% seems a little ambitious to me), and that's only on the big cores. Clock speeds will be higher though, so I reckon a 10% increase in performance overall. Nothing to really get excited about.

Surely its higher than 7%. As a gamer, IPC to me seems like such a nebulous concept, and really difficult to gauge. In some workloads, Alder Lake straight up destroys Zen 3; for example compiling. I've always heard on tech heavy forums like Anandtech that compiling is a really good gauge of raw CPU performance because it's damn near impossible to spoof.

But then again, maybe some of these large performance advantages for Alder Lake are better explained by its ridiculously good power/frequency scaling.

M2/A16 Big core should be a large increase given that Avalanche was focused more on efficiency and clock speed increases over Firestorm. It's been a while now since Apple had a major core revision from a uarch standpoint, I'd hazard a guess at a 15% IPC increase and also a slight clock speed bump. No idea about the little cores.

Even though I'm an Apple hater, I do hope the M2 is a great revision over the M1 because it will keep both AMD and Intel from becoming complacent. Also, Apple CPUs tend to be very interesting to me from an architectural perspective because they are such a dichotomy to x86-64 CPUs.

 

[JoeRambo]

Raptor Lake: Intel has already built insanely wide and powerful GC core that currently is under performing due to weak memory subsystem, Raptor Lake should fix some of that and should improve IPC somewhat. IPC improvements will range from zero in CB23 to up to 15% or so in apps that are sensitive to memory subsystem.

ZEN4: AMD on the other hand seems to be staying with ZEN3 core just increasing L2 to 1MB, DTLB to make better use of their now massive stacked L3 caches. IPC is gonna be wild guess, AVX512 support is not free and does not help most apps that don't support it. Hard to estimate, cause ZEN1->ZEN2 extracted a lot of IPC with similar changes.
A lot of Zen4 performance hangs on question if they will release stacked L3 or not. Two different beasts with 32MB or 96MB L3.

M2: No idea, probably still very efficient, but with lagging peak perf due to low clocks.

Desktop/laptop CPUs with assumption that Zen4 desktop will release without stacked L3 or said cache will be very exclusive:

1) Single threaded performance - Raptor Lake. I think on Anandtech testing with JEDEC memory ZEN4 will be ahead somewhat, but will scale with memory incredibly.
2) Multithreaded performance - Raptor Lake, unless ZEN4 releases >16C SKU on non HEDT desktop/mobiles. 12900K has no troubles matching 5950x, so things can go only go more wrong for AMD here with 8 more marketing cores and more IPC overall. Not enough AVX512 apps to carry them.
3) Gaming performance - RL or hypothetical Zen4 + stacked L3 SKU
4) Performance per watt - Probably M2 with Zen4 closely behind
5) Overall performance (who wins the majority of applications) - Raptor Lake

 

[moinmoin]

Surely its higher than 7%.

I wouldn't be so sure. Around that amount seems to be Intel's usual gain per gen. ADL just gave an impression of a bigger jump coming from all the previous Skylake derivatives while the new gens stayed mobile only.

ZEN4: AMD on the other hand seems to be staying with ZEN3 core just increasing L2 to 1MB, DTLB to make better use of their now massive stacked L3 caches. IPC is gonna be wild guess, AVX512 support is not free and does not help most apps that don't support it. Hard to estimate, cause ZEN1->ZEN2 extracted a lot of IPC with similar changes.
A lot of Zen4 performance hangs on question if they will release stacked L3 or not. Two different beasts with 32MB or 96MB L3.

I think stacking is barely ready for N5 so I highly doubt consumer chips would come with it by default.

As for IPC increase Zen 1 to 2 is the precedence, and I'm sure the team at AMD is trying to surpass it with Zen 3 to 4. Without X3D.

As for AVX512 support not being free, a lot of IPC gain in Intel chips in the last decade came from AVX improvements whereas Zen was more balanced with integer improvements as well. I'd hope the IPC improvements in all Zen 4, RPL and M2 are balanced.

 

[Carfax83]

Raptor Lake: Intel has already built insanely wide and powerful GC core that currently is under performing due to weak memory subsystem, Raptor Lake should fix some of that and should improve IPC somewhat. IPC improvements will range from zero in CB23 to up to 15% or so in apps that are sensitive to memory subsystem.

Interesting. I'm curious, but what makes you believe that Alder Lake has a weak memory subsystem? Is it the increased cache latency? That's the only thing I recall Anandtech mentioning in their review regarding the cache system, and it was explained as a result of increasing the size.

 

[Carfax83]

I wouldn't be so sure. Around that amount seems to be Intel's usual gain per gen. ADL just gave an impression of a bigger jump coming from all the previous Skylake derivatives while the new gens stayed mobile only.

That 7% figure was meant as a comparison to Zen 3. @Cardyak stated that Golden Cove has a 7% greater IPC than Zen 3.

 

[JoeRambo]

Interesting. I'm curious, but what makes you believe that Alder Lake has a weak memory subsystem? Is it the increased cache latency? That's the only thing I recall Anandtech mentioning in their review regarding the cache system, and it was explained as a result of increasing the size.

Some benchmarks just keep scaling big time with reduced memory latency and increased uncore speeds. So from personal experience, but it is backed by analysis on the web, for example Chips And Cheese found this:




I am fine with L1 region as 48KB of L1 has its own virtues, but currently L2 is questionable, substantially slower when working set is small, and advantages with larger sizes end real fast due to ZEN3's excellent L3. RKL is fixing this by giving 2MB of L2. So AMD will no longer catch up so soon ( who are of course looking into extending that initial lead zone to 1MB ).
Intel's L3 cache is on disaster level, i don't care if it's cause by uncore clocks 3.7Ghz with marketing cores, or ring speed, power, but at 16MB size we are talking about 20 cycles difference. No excuses here, when AMD has more bandwidth and Intel somehow has 50% more latency? RPL is helping in this area even if they don't touch L3 by virtue of hitting L3 way less as private cache is 2MB now and hopefully more L2 cache ways to reduce conflicts even further, the slowness price is already paid in design, server CPUs supposedly already have 2MB of L2)
So when we arrive to memory hit due to missing LLC, Intel is already working with 20 cycle penalty, due to incompetent L3 cache. Except they unlock major retard achievement here too, matching in latency a chip that has IMC on different die, requests crossing multiple clock domains. Hint it's not 20 cycles from slower L3 the problem here, but most likely mobile class memory controller and designed by people with less know-how about memory than people who beat IMC equipped Hammers with MC on NB chip.

 

[Ajay]

Apple will have the fastest processor for Apple products - no one will even come close.
Apple will have the absolute worst processor for x86 products; can't get much lower than zero performance.
Just had to say it...

 

[TheELF]

AMD: Zen 4 will most likely be a whopper, I'm banking on 25% IPC increase and ~15% clock speed increases, so overall I expect about ~40% performance increase, and that's at the same core counts, hence this applies to ST as well as MT.

The only way for AMD to achieve that would be to use 25-40% more transistors per core so that things that can use more instructions per cycle, like cinebench and so on, can use them and be faster, that would seriously cut into the amount of chips they can produce which would mean that their CPUs would have to be seriously more expensive to make up for it.

 

[Rannar]

The only way for AMD to achieve that would be to use 25-40% more transistors per core so that things that can use more instructions per cycle, like cinebench and so on, can use them and be faster, that would seriously cut into the amount of chips they can produce which would mean that their CPUs would have to be seriously more expensive to make up for it.

dont forget 5nm scaling for logic. it's up to 80% so 25-40% more transistors per core would not make chiplet itself bigger then 7nm zen 3 one.

 

[DrMrLordX]

The only way for AMD to achieve that would be to use 25-40% more transistors per core

AMD was able to increase IPC and clocks on N7 moving from Zen2 to Zen3 without significantly increasing CCD size. Not that AMD really has to worry since, as @Rannar indicated, moving to N5 gets them more density. AMD could certainly up transistor count. Given the beefier SIMD units, we should expect such.

 

[TheELF]

dont forget 5nm scaling for logic. it's up to 80% so 25-40% more transistors per core would not make chiplet itself bigger then 7nm zen 3 one.

Yes but that means that they could still make 40% more cores in 5nm if they keep the transistor count the same, it's still 40% more cost or less cpus per wafer.

 

[nicalandia]

My Napkin math numbers say that the 13900K will match or beat the 6950X in Cinebench R23

So the 12900K gets 27,000 points in CBR23 MT we need to add 7% of IPC on top across both P and E cores and 8 more E cores to the mix, I say that the 13900K is getting more than 36,000 points in CBR23 MT.

 

[Henry swagger]

My Napkin math numbers say that the 13900K will match or beat the 6950X in Cinebench R23

So the 12900K gets 27,000 points in CBR23 MT we need to add 7% of IPC on top across both P and E cores and 8 more E cores to the mix, I say that the 13900K is getting more than 36,000 points in CBR23 MT.

Leaks are saying raptor cove is 5.5ghz maybe 5.8ghz.. so i.m predicting 38000 multi.. 2200 single core

 

[DrMrLordX]

So the 12900K gets 27,000 points in CBR23 MT we need to add 7% of IPC on top across both P and E cores

Whatever IPC increase you'll see from the Gracemont cores will be muted by the inevitable reduction in clockspeed that Intel will have to introduce in order to include more of them without the power budget ballooning out of control.

Leaks are saying raptor cove is 5.5ghz maybe 5.8ghz.. so i.m predicting 38000 multi.. 2200 single core

5.8 GHz all-core in CBR23? What kind of power budget are people expecting? It's not like Intel is moving to a new process. There's only so much they can do on 10ESF.

 

[Henry swagger]

N

Whatever IPC increase you'll see from the Gracemont cores will be muted by the inevitable reduction in clockspeed that Intel will have to introduce in order to include more of them without the power budget ballooning out of control.



5.8 GHz all-core in CBR23? What kind of power budget are people expecting? It's not like Intel is moving to a new process. There's only so much they can do on 10ESF.

Nah.. 5.8ghz single core abd 5.1 all core boost.. raptor lake will have new intel technology for voltage it decreases power consumption by 30%

 

[Mopetar]

Whatever IPC increase you'll see from the Gracemont cores will be muted by the inevitable reduction in clockspeed that Intel will have to introduce in order to include more of them without the power budget ballooning out of control.

Having more at lower clock speeds will still be a net gain. The efficiency takes a nose dive with Intel pushing the clocks to absurd levels just to have a slightly bigger bar.

 

[nicalandia]

Whatever IPC increase you'll see from the Gracemont cores will be muted by the inevitable reduction in clockspeed that Intel will have to introduce in order to include more of them without the power budget ballooning out of control.

I really don't see that happening.

 

[defferoo]

Process-wise, M2 should be on TSMC 3nm which should bring power improvements, Zen 4 will also move to TSMC 5nm which will help as well. The outlier is Raptor Lake still on Intel 7.

1) Single threaded performance - Intel, with the architectural improvements in Raptor Lake and their willingness to push high frequencies at high wattage, I think Intel will continue leading here.
2) Multithreaded performance - Hard to see this changing since Intel will just add more cores to Raptor Lake (rumored 8P+16E) and their efficiency cores are pretty performant.
3) Gaming performance - basically see single-threaded performance
4) Performance per watt - likely Apple here, with TSMC 3nm, their lead over Intel increases while AMD moves to 5nm so the advantage is about the same.
5) Overall performance (who wins the majority of applications) - Probably Intel considering their dominance in ST and MT performance.

Let me add one last category, power consumption: Intel wins easily here

 

[Mopetar]

I believe that all evidence we have points to M2 being on the 4nm (refunded 5nm) process and not on 3nm.

 

[nicalandia]

Process-wise, M2 should be on TSMC 3nm which should bring power improvements, Zen 4 will also move to TSMC 5nm which will help as well. The outlier is Raptor Lake still on Intel 7.

1) Single threaded performance - Intel, with the architectural improvements in Raptor Lake and their willingness to push high frequencies at high wattage, I think Intel will continue leading here.
2) Multithreaded performance - Hard to see this changing since Intel will just add more cores to Raptor Lake (rumored 8P+16E) and their efficiency cores are pretty performant.
3) Gaming performance - basically see single-threaded performance
4) Performance per watt - likely Apple here, with TSMC 3nm, their lead over Intel increases while AMD moves to 5nm so the advantage is about the same.
5) Overall performance (who wins the majority of applications) - Probably Intel considering their dominance in ST and MT performance.

Let me add one last category, power consumption: Intel wins easily here

I believe this time Zen4 will match Raptor Lake in frequency , as they were showing a ES/QS running at 5 Ghz all cores while gaming. Alder Lake is within single digit IPC from Zen3. Zen4 will be double digit gains for sure. If I had to bet my house I would say Zen4 would beat Raptor Lake in ST performance quite handily

MT Performance will be close.

Gaming will go to Zen4 as I don't think Alder/Raptor Lake and Zen3 V Cache matching Zen4 in gaming

 

[scineram]

Nah.. 5.8ghz single core abd 5.1 all core boost.. raptor lake will have new intel technology for voltage it decreases power consumption by 30%

Completely delusional.

 

[nicalandia]

Completely delusional.

Alder Lake 12900KS has a 5.5 Ghz ST boost and 5 Ghz all core boost already, how is 5.1 Ghz All cores on 13900K delusional?

Intel Core i9-12900KS Benchmark, Test and specs

Intel Core i9-12900KS benchmark results and review of this cpu with specs including the number of cores, threads, memory bandwidth, pcie lanes and power consumption. Benchmarks in Cinebench R23 and Geekbench 5

www.cpu-monkey.com

 

[defferoo]

I believe that all evidence we have points to M2 being on the 4nm (refunded 5nm) process and not on 3nm.

ah, in that case we might have a much closer race for performance/watt between Apple and AMD