Question Raptor Lake - Official Thread

[Hulk]

Since we already have the first Raptor Lake leak I'm thinking it should have it's own thread.
What do we know so far?
From Anandtech's Intel Process Roadmap articles from July:

Built on Intel 7 with upgraded FinFET
10-15% PPW (performance-per-watt)
Last non-tiled consumer CPU as Meteor Lake will be tiled

I'm guessing this will be a minor update to ADL with just a few microarchitecture changes to the cores. The larger change will be the new process refinement allowing 8+16 at the top of the stack.

Will it work with current z690 motherboards? If yes then that could be a major selling point for people to move to ADL rather than wait.

 

[Markfw]

I am not a fan of the Big.Little arch at all and do not like the e-cores at all and I will always shut them off and use them as 8 core powerhouse chips as I am a huge fan of the p cores.

I wonder how much of an IPC increase the P cores will have in Raptor Lake compared to Alder Lake. A lot of consensus said very little and it was mostly higher clock speeds.

Though the below article says 12% IPC improvement over Alder Lake P cores at same clock speed.

https://www.tomshardware.com/news/core-i9-13900k-early-review-shows-big-gains-over-core-i9-12900k

If that is really true, that would scream awesome if it can really clock much higher all cores.

Basically I am hoping for a 13900K, shut off e-cores and clock all P cores to 6GHz all core on a Noctua NH-D15 cooler. That would be abig boost without IPC gain and oh my with 1`2% IPC uplift would be huge.

Seems maybe 6GHz all core could be doable on NH-D15 with e-cores off given the samples out there that overclock to 8.2GHz on LN2. And the current 12900K hits like 7.4GHz on LN2 and cooling it on NH-D15 at 5/5.1GHz is easy form my experience. Even 12700K hits 5GHz easily to cool on air at 1.275VCORE with e-cores off.

Maybe a lot more clock head room with Raptor Lake P cores??

I though wonder how a Raptor Lake 8 P core part e cores off with ring clock high and P cores 6GHz will do in gaming compared to upcoming 7700X3D?? Given the heat and thermal issues regular Zen 4 is facing already hard to imagine AMD will not have to downclock it a bit so hard to say.

So I see you joined today. Welcome. To fill you in, I own a 12F00F with the 4 E-cores turned off, and right now I am participating in a contest for primegrid where avx-512 is used. My 12700F has not had a bios update to shut off avx-512, and its using it. I also have 2 7950x chip that also use avx-512. This project is configured (for me) to use 8 threads per task and a maximum of 2 tasks per cpu. Sooo. The 7950x task and the 12700F task are evenly matched in horsepower. Sort of. The 7950x is still beating it by about 18%. This means that Raptor lake, being 13% (I think) faster than alder lake would still get beat by the 7950x by 5% for the 8 P-cores. if you turn off the E--cores (16 of them) then the 7950x will beat your chip by 210%. This is with everything at stock. Also, this is the best I can tell by all pre-release benchmarks, I could be wrong.

I am just saying keep an open mind on your buying choice.

 

[inf64]

There is almost no IPC uplift with Raptor Cove's P cores, at least not in the leaks we had thus far. It's just clock and process improvements, plus faster E cores.

 

[Wolverine2349]

So I see you joined today. Welcome. To fill you in, I own a 12F00F with the 4 E-cores turned off, and right now I am participating in a contest for primegrid where avx-512 is used. My 12700F has not had a bios update to shut off avx-512, and its using it. I also have 2 7950x chip that also use avx-512. This project is configured (for me) to use 8 threads per task and a maximum of 2 tasks per cpu. Sooo. The 7950x task and the 12700F task are evenly matched in horsepower. Sort of. The 7950x is still beating it by about 18%. This means that Raptor lake, being 13% (I think) faster than alder lake would still get beat by the 7950x by 5% for the 8 P-cores. if you turn off the E--cores (16 of them) then the 7950x will beat your chip by 210%. This is with everything at stock. Also, this is the best I can tell by all pre-release benchmarks, I could be wrong.

I am just saying keep an open mind on your buying choice.

Well 7950X and sometimes even the 7900X/5950X/5900X is going to beat the 12700K and even 12900K as it has 12-16 good cores where as the Intel parts stop at 8 strong cores. Yes Intel 8 strong cores IPC is better. But then there is a drop off to the e-cores which are so much weaker than the P cores and a lot weaker than even AMD's original Zen cores when you factor in how bad latency is for them.

I wish Intel would make more than 8 good cores on a ring bus like they did with Comet Lake and Broadwell-E and Haswell-E.

The thing about AMD CPUs with more than 8 cores with Zen 3 and Zen 4 and, for gaming, they are not all on same CCD/ring and there is a big latency when switching to another CCD and gaming threads are so dependent on fast communication with one another for best performance. For many productivity things and such, it does not matter and always more cores whether on same CCD or even separate sockets all else equal more cores better for programs that can scale to infinite threads.

Zen 5 I think will be the next chip that will have more than 8 strong cores on same CCD/ring.

Though for now, fortunately very few games effectively take advantage of more than 6/12 CPU let alone an 8/16 CPU. Though that may start to change, but I have read that programming a game for more and more threads is almost impossible and that faster and faster 6/12 and 8/16 CPUs will be much more useful for gaming down the road than slightly slower 10+ core CPUs.

 

[Wolverine2349]

There is almost no IPC uplift with Raptor Cove's P cores, at least not in the leaks we had thus far. It's just clock and process improvements, plus faster E cores.

Yeah based on benchmarks that seems true though they were engineering samples. Even with the launch date

Though they are increasing L2 cache on the P cores from 1MB to 2MB. Will that help IPC at all or was it for something else??

Could Intel be sandbagging as Toms Hardware article said IPC increases were 12% on P cores.

 

[Exist50]

Will that help IPC at all or was it for something else??

Intel had this diagram from their launch event.

Suffice it to say, looks like the extra cache provides only marginal benefits to single thread, at least in SPECint. There'll probably be some use cases that benefit significantly more than others (e.g. gaming might), but they'll be the exception.

 

[zir_blazer]

Intel had this diagram from their launch event.

Suffice it to say, looks like the extra cache provides only marginal benefits to single thread, at least in SPECint. There'll probably be some use cases that benefit significantly more than others (e.g. gaming might), but they'll be the exception.

Is bars are accurate, P-Core IPC improvements are about 5% or so.

As things are, Raptor Lake being possible to use with DDR4 and cheaper Motherboards would probably make it a more cost effective purchase even if Zen 4 on AM5 tends to be faster. Power efficiency is likely not that bad either if tweaked, both Intel and AMD seems to be having fun with aggressive turbos that aren't THAT necessary.

 

[Wolverine2349]

Is bars are accurate, P-Core IPC improvements are about 5% or so.

As things are, Raptor Lake being possible to use with DDR4 and cheaper Motherboards would probably make it a more cost effective purchase even if Zen 4 on AM5 tends to be faster. Power efficiency is likely not that bad either if tweaked, both Intel and AMD seems to be having fun with aggressive turbos that aren't THAT necessary.

What is actually faster at same core count and same clock speed. I know its hard to do that comparison because the core types are different when going above 8 cores for each as AMD has more Zen 4 cores where as Intel has the e-cores.

Though at same clock speed and same P core count like what would be faster. A Ryzen 7700X all 8 cores at 5GHz or a Core i7 13700K all P cores clocked at 5GHz with e-cores disabled so they are effectively both 8 core parts.

I have read different sources that say IPC is higher in Zen 4 and some say higher in Raptor Cove and even Golden Cove still ahead.

Zen 4 also had an underwhelming IPC uplift but not 5% underwhelming though does it for both all depend on the software being used??

 

[deasd]

I have read different sources that say IPC is higher in Zen 4 and some say higher in Raptor Cove and even Golden Cove still ahead.

Zen 4 also had an underwhelming IPC uplift but not 5% underwhelming though does it for both all depend on the software being used??

Some leakers already tested RaptorCove SPEC IPC and has minimal changes to ADL Cove. Zen4 SPEC IPC is coincidently identical to Coves. In realworld it all depends on what type of workload.

 

[inf64]

Some leakers already tested RaptorCove SPEC IPC and has minimal changes to ADL Cove. Zen4 SPEC IPC is coincidently identical to Coves. In realworld it all depends on what type of workload.

Yeah, pretty much the same average IPC in SPECT 1T for all 3 cores. Minor difference in SpecFP and it will depend on type of workload and whether it supports AVX512 or not.

 

[poke01]

Some leakers already tested RaptorCove SPEC IPC and has minimal changes to ADL Cove. Zen4 SPEC IPC is coincidently identical to Coves. In realworld it all depends on what type of workload.

Next major IPC jump for Intel will be Arrow Lake. Lion Cove will indeed be the king of the jungle even more likely is if Intel 3 is potent.
We don't know the details but it's coming in 2024. I know MLID said Allow Lake won't be enough to compete with Zen 5 but it will.

Intel's Raptor Cove(Intel 7) is competing with Zen 4 which is on bleeding edge node 5nm family from TSMC. So I have no doubt that Arrow Lake will be better than Zen 5(yes, I said it).

As Exist50 said Arrow and Lunar are not "Royal Core" so the fact remains whatever comes after Lion Cove is unknown.

 

[Wolverine2349]

Yeah, pretty much the same average IPC in SPECT 1T for all 3 cores. Minor difference in SpecFP and it will depend on type of workload and whether it supports AVX512 or not.

So are you saying Zen 4, Golden Cove and Raptor Cove all have really close IPC??

And how accurate of a test is CPU-Z single thread score for measuring IPC at same clock speed.

I tried it myself at fixed 4.7GHZ locked frequency on a Ryzen 5900X and Core i7 12700K fixed at 4.7GHz a few months back

CPU single thread score

Core i7 12700K 4.7GHz: 767

Ryzen 9 5900X CCD1 at 4.7GHz: 642

642/767 is 0.837

So per CPU-Z single threaded score of Ryzen 5900X at 4.7GHz is between 83 and 84 percent as fast as Core i7 12700K

So therefore Golden Cove P cores of Alder Lake are 16-17% better IPC than Zen 3??

I do not have a Zen 4 CPU to test but have read and IPC uplift says overall 13% though some say only 8-10 over Zen 3%.

SO Golden Cove IPC still a little ahead of even Zen 4 by like 3-8%??

Or is CPU-Z benchmark not a good barometer.

I also got a very similar result when I tried with Cinebench R23 single core test calculating Golden Cove IPC 16-17% above Zen 3.

12700K 4.7GHz Cinebench R23 single core score: 1855

Ryzen 5900X 4.7GHz CInebench R23 single core score: 1548

How accurate are both for IPC test.

 

[igor_kavinski]

How accurate are both for IPC test.

If it's possible to restrict both Zen 3 and Alder Lake to a single core at iso-speed, run PCMARK10 on both PCs and share the screenshots/links. Thanks.

Another benchmark with varied tests is PassMark Performance Test but I would prefer PCMARK10 (free version is fine).

 

[Wolverine2349]

If it's possible to restrict both Zen 3 and Alder Lake to a single core at iso-speed, run PCMARK10 on both PCs and share the screenshots/links. Thanks.

Another benchmark with varied tests is PassMark Performance Test but I would prefer PCMARK10 (free version is fine).

What is ISO Speed. And I no longer have the Zen 3 chip as I sold it a few months ago.

 

[igor_kavinski]

What is ISO Speed.

Same speed. 4.7 GHz. It's a pity that you don't have the Zen 3 anymore. Would you please explain how you ran Zen 3 at 4.7 GHz so maybe someone here can volunteer to test PCMARK10 with a single core enabled?

 

[Wolverine2349]

Same speed. 4.7 GHz. It's a pity that you don't have the Zen 3 anymore. Would you please explain how you ran Zen 3 at 4.7 GHz so maybe someone here can volunteer to test PCMARK10 with a single core enabled?

I have done a lot of part swapping and decided I would rather have the better IPC than more cores as gaming barely touches 6 cores let alone 8. Though I overthought things a lot and thought well maybe more strong cores is more future proof. Then I realized the reality is more than 8 cores on AMD CPUs are not on same CCD which means a latency penalty. Not an issue for productivity apps at all, but for game threads not good.

SO anyways to answer your question, I am a manual static old school overclocked and like static frequency and that is just how I am. So I statically set CCD1 to 4.7GHz and CCD2 to 4.6GHz at 1.275VCORE Then I did stability testing wit OCCT, Prime95 and such and they passed. Had PBO and all boost functions disabled and used Windows 10 Ultimate Power Plan.

I had ran test prior to changing my mind and swapping rigs to get score.

Then when I swapped to Alder Lake 12700K and settled on 5GHz stable overclock, I decided I am curious about IPC and I knew my 5900X CCD1 was 4.7GHz, so I set temporarily 12700K to 4.7GHz to see how IPC was and my above results are what I found. 16-17% IPC better than Zen 3 at least per CPU-Z and CInebench R23.

 

[Hitman928]

I have done a lot of part swapping and decided I would rather have the better IPC than more cores as gaming barely touches 6 cores let alone 8. Though I overthought things a lot and thought well maybe more strong cores is more future proof. Then I realized the reality is more than 8 cores on AMD CPUs are not on same CCD which means a latency penalty. Not an issue for productivity apps at all, but for game threads not good.

SO anyways to answer your question, I am a manual static old school overclocked and like static frequency and that is just how I am. So I statically set CCD1 to 4.7GHz and CCD2 to 4.6GHz at 1.275VCORE Then I did stability testing wit OCCT, Prime95 and such and they passed. Had PBO and all boost functions disabled and used Windows 10 Ultimate Power Plan.

I had ran test prior to changing my mind and swapping rigs to get score.

Then when I swapped to Alder Lake 12700K and settled on 5GHz stable overclock, I decided I am curious about IPC and I knew my 5900X CCD1 was 4.7GHz, so I set temporarily 12700K to 4.7GHz to see how IPC was and my above results are what I found. 16-17% IPC better than Zen 3 at least per CPU-Z and CInebench R23.

If all you care about is gaming, why not wait for Zen4 with 3d cache? It should be the clear gaming king and have a large lead in gaming IPC and performance per watt. The cheaper am5 boards should be out then too.

 

[deasd]

Raptorlake in blender database

https://twitter.com/x/status/1579484009671036928

Looks like it's not only the E-core spam works well (at least in rendering) against mid range competitor like 7700X, but the most problematic thing is it also standing well against its team mate like 12700k, which is 8P+4E....... looks like E-core spam is also hurting themselves....

 

[Exist50]

Raptorlake in blender database

https://twitter.com/x/status/1579484009671036928

Looks like it's not only the E-core spam works well (at least in rendering) against mid range competitor like 7700X, but the most problematic thing is it also standing well against its team mate like 12700k, which is 8P+4E....... looks like E-core spam is also hurting themselves....
View attachment 68893

I don't think Intel's terribly upset about their 13th gen upstaging 12th gen. If anything, it's great we're getting so much progress from a refresh.

 

[Wolverine2349]

If all you care about is gaming, why not wait for Zen4 with 3d cache? It should be the clear gaming king and have a large lead in gaming IPC and performance per watt. The cheaper am5 boards should be out then too.

Is that really going to work well though given that regular Zen 4 runs at 95C especially to give it high enough clocks to handily or better beat 6GHz Raptor Cove??? Just imagine adding cache on top of it and how are they going to keep thermals ok at good clocks??

Do you think Zen 4 with 3D cache will handily beat a manually well tuned Raptor Lake 13900K for gaming or trade blows or be very close??

 

[igor_kavinski]

View attachment 68893

Good news for anyone who was planning to get a 13600K anyway.

 

[Markfw]

Is that really going to work well though given that regular Zen 4 runs at 95C especially to give it high enough clocks to handily or better beat 6GHz Raptor Cove??? Just imagine adding cache on top of it and how are they going to keep thermals ok at good clocks??

Do you think Zen 4 with 3D cache will handily beat a manually well tuned Raptor Lake 13900K for gaming or trade blows or be very close??

You can set Zen4 to run at whatever temp you want in bios. As far as how high the Zen 4 with cache runs, why not post in the Zen 4 thread and talk about it ? And wait until it comes out before saying how bad its going to be ?

 

[igor_kavinski]

Do you think Zen 4 with 3D cache will handily beat a manually well tuned Raptor Lake 13900K for gaming or trade blows or be very close??

It will depend on the game. Some games love frequency+bandwidth. Those will absolutely fly with a 6 GHz Raptor Lake. But plenty others love huge cache. It will sure be an interesting battle to see unless AMD totally decimates Intel with a 7800X3D at 5.5 GHz boost clocks.

 

[Hitman928]

Is that really going to work well though given that regular Zen 4 runs at 95C especially to give it high enough clocks to handily or better beat 6GHz Raptor Cove??? Just imagine adding cache on top of it and how are they going to keep thermals ok at good clocks??

Do you think Zen 4 with 3D cache will handily beat a manually well tuned Raptor Lake 13900K for gaming or trade blows or be very close??

The way AMD stacks the dies, it does cache over cache only, so the hotspots in the actual computational part of the CPU aren't covered by active devices. Instead, they put only silicon over those parts that has as good or better thermal conductivity than the silicon in the non-3d versions. With that said, there are always inefficiencies when you have a materials interface introduced so thermals will probably be a little worse on the 3d version, but only a little. This is exactly what we saw with the 5800x vs. the 5800x3d. Then despite the slightly worse thermals and clock speed, the x3d version was able to greatly outperform the non-3d version in games (~25% faster). The 7700x already matches if not slightly outperforms a 12900K in gaming, so yes, I think a Zen4-3d CPU is the best bet to be the gaming king upon release.

 

[Wolverine2349]

You can set Zen4 to run at whatever temp you want in bios. As far as how high the Zen 4 with cache runs, why not post in the Zen 4 thread and talk about it ? And wait until it comes out before saying how bad its going to be ?

I have no comment on it being bad. Just concerned seeing the 95C on the current Zen 4 parts that just came out even with good cooling when they run at stock speeds. I hope they can make it great and can clock it high and it topples everything.

 

[Hulk]

So are you saying Zen 4, Golden Cove and Raptor Cove all have really close IPC??

And how accurate of a test is CPU-Z single thread score for measuring IPC at same clock speed.

I tried it myself at fixed 4.7GHZ locked frequency on a Ryzen 5900X and Core i7 12700K fixed at 4.7GHz a few months back

CPU single thread score

Core i7 12700K 4.7GHz: 767

Ryzen 9 5900X CCD1 at 4.7GHz: 642

642/767 is 0.837

So per CPU-Z single threaded score of Ryzen 5900X at 4.7GHz is between 83 and 84 percent as fast as Core i7 12700K

So therefore Golden Cove P cores of Alder Lake are 16-17% better IPC than Zen 3??

I do not have a Zen 4 CPU to test but have read and IPC uplift says overall 13% though some say only 8-10 over Zen 3%.

SO Golden Cove IPC still a little ahead of even Zen 4 by like 3-8%??

Or is CPU-Z benchmark not a good barometer.

I also got a very similar result when I tried with Cinebench R23 single core test calculating Golden Cove IPC 16-17% above Zen 3.

12700K 4.7GHz Cinebench R23 single core score: 1855

Ryzen 5900X 4.7GHz CInebench R23 single core score: 1548

How accurate are both for IPC test.

Benchmarks are a curious thing. We tend to live and die by them here on this forum.
I tend to use benchmarks as follows.
First, CPUz, Cinebench, etc.. I find useful to get a general impression of where a CPU stands next to the competition. If the results are close, say within 10 or 15% I don't really draw any conclusions about which chip is better.
Next, the validated benches coming from applications that I actually use on a day-to-day basis. When CPU's are close in benchmarks I need to know how they actually perform on computational tasks that are important to me.

BTW, thanks for posting the results of your ISO tests. Very informative.