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.

 

[Henry swagger]

https://twitter.com/x/status/1605324121034063872

25% faster the gaps are getting bigger.. will grow even more with faster ddr5 being cooked ✔️🎮

 

[tamz_msc]

https://twitter.com/x/status/1605324121034063872

25% faster the gaps are getting bigger.. will grow even more with faster ddr5 being cooked ✔🎮

It is well-documented that dual-CCD Zen CPUs crap the bed when it comes to CPU intensive scenes in games with RT enabled. 25% deficit in memory performance doesn't help much either.

 

[Thunder 57]

He doesn't care, he's an Intel shill.

 

[ZGR]

Saw a new ElmorLabs video and they reached 9 GHz on a 13900k!

 

[Henry swagger]

Saw a new ElmorLabs video and they reached 9 GHz on a 13900k!

View attachment 73205

Intel 7nm is a impressive node when it comes to performance ✅😏

 

[lopri]

That is bonkers. Holy moly.

 

[TheELF]

It is well-documented that dual-CCD Zen CPUs crap the bed when it comes to CPU intensive scenes in games with RT enabled. 25% deficit in memory performance doesn't help much either.

Sure, but if an intel hired third party enables game mode to fix this issue everybody gets in an uproar as well....
Well, server CPUs are bad for gaming we know this since ever, and the 7950x is a server CPU downscaled for desktop.

 

[Hitman928]

It is well-documented that dual-CCD Zen CPUs crap the bed when it comes to CPU intensive scenes in games with RT enabled. 25% deficit in memory performance doesn't help much either.

This seems like rather specific testing that I haven't seen data on. Where has this been documented?

 

[Markfw]

Sure, but if an intel hired third party enables game mode to fix this issue everybody gets in an uproar as well....
Well, server CPUs are bad for gaming we know this since ever, and the 7950x is a server CPU downscaled for desktop.

The 7950x is NOT a scaled down server. The chiplets are either manufactured or binned for speed, they only have a 2 channel memory interface and customizable BIOS for speed, power etc. NONE of that matches in any way the server chips, aside from the face that MAYBE the actual layout of the chiplets is the same.

 

[Hitman928]

Sure, but if an intel hired third party enables game mode to fix this issue everybody gets in an uproar as well....

Not to rehash old news, but that is not what happened at all. Principled Tech used game mode on a 2700x (which didn't have multiple CCDs), effectively turning it into a 4 core CPU. Game/creator mode was never intended for Ryzen CPUs and using it on a 2700x had a significant performance penalty in production work and many games at the time. Using game mode on a 2700x was almost never beneficial and only hurt performance. There was no reason to use game mode on the Ryzen CPU.

Edit: This was also not the only issue people had with their testing though it was probably the biggest one. Gamers Nexus made a video on it you can check out for the details.

 

[nicalandia]

The 7950x is NOT a scaled down server. The chiplets are either manufactured or binned for speed

I beg to differ this pint good sir. Every Zen CCD that AMD makes are manufactured the same. There is no special Silicon Wafer for CCD made for EPYC or Ryzen products. The Top 90% CCDs go to make EPYC CPUs. The rest are distributed for client based CPUs.

This has been true since Zen1.

 

[Markfw]

I beg to differ this pint good sir. Every Zen CCD that AMD makes are manufactured the same. There is no special Silicon Wafer for CCD made for EPYC or Ryzen products. The Top 90% CCDs go to make EPYC CPUs. The rest are distributed for client based CPUs.

This has been true since Zen1.

You can't tell me there is NO binning goin on ! And most likely they are the same silicon, but do you KNOW for a fact that there is NOT a production line at TSMC that is a tweaked process for lower power consumption, or speed ?

 

[nicalandia]

You can't tell me there is NO binning goin on !

That's it. Now that they have design a process so modular that they can mix and match the same CCDs into different SKUs. Their Binning process has become very efficient. Yes they do very selective binning.

 

[TheELF]

The 7950x is NOT a scaled down server. The chiplets are either manufactured or binned for speed, they only have a 2 channel memory interface and customizable BIOS for speed, power etc. NONE of that matches in any way the server chips, aside from the face that MAYBE the actual layout of the chiplets is the same.

Multiple CPUs connected with each other in some way, because they only have to run MT and the lag between CPUs doesn't matter with MT, no threads waiting for output of other threads to slow things down.
It's a server CPU.

You can't tell me there is NO binning goin on !

Ohhhhhhhhh is there ever.
You get one good CPU and one not so good one, again because you are only supposed to run MT and you will only be looking at the end result.
(And also they have to save money somewhere)

AMD Ryzen 9 7950X Review - Impressive 16-core Powerhouse

The Ryzen 9 7950X is a monster CPU. When paired with the right workload it will eat even the 12900K for breakfast. As our review shows, the performance uplifts can be massive: +30-50% gen-over-gen is totally possible. What makes things complicated though, is that keeping the beast cool is almost...

www.techpowerup.com

 

[TheELF]

Not to rehash old news, but that is not what happened at all. Principled Tech used game mode on a 2700x (which didn't have multiple CCDs), effectively turning it into a 4 core CPU. Game/creator mode was never intended for Ryzen CPUs and using it on a 2700x had a significant performance penalty in production work and many games at the time. Using game mode on a 2700x was almost never beneficial and only hurt performance. There was no reason to use game mode on the Ryzen CPU.

Edit: This was also not the only issue people had with their testing though it was probably the biggest one. Gamers Nexus made a video on it you can check out for the details.

They did that for consistency which makes sense.
The other option would be to not include the 2700x at all, like when people do benches of overclockable CPUs and leave out all the others.
They didn't bench the 2700x alone with gamemode in a vacuum.

Follow-Up Interview: Principled Technologies on Intel Testing Concerns | GamersNexus

stub After our exhaustive in-person interview with Principled Technologies, published on our YouTube channel, we followed-up via email to clarify some questions that were left unanswered during the initial video. As we noted in the initial interview, we give credit to Principled Technologies for...

www.gamersnexus.net

• Use of "Game Mode" on the AMD Ryzen 7 2700X: Some inquiries we have received concern the use of the Ryzen utility and the number of active cores in the AMD-based systems. Based on AMD's recommendations and our initial testing on the Threadripper processors, we found installing the AMD Ryzen Master utility and enabling Game Mode increased most results. For consistency purposes, we did that for all AMD systems across Threadripper and Ryzen. We are now doing additional testing with the AMD systems in Creator Mode. We will update the report with new results.

 

[dullard]

Would you please move the AMD 7950X discussion to a more appropriate thread? Thanks.

 

[nicalandia]

Would you please move the AMD 7950X discussion to a more appropriate thread? Thanks.

Yeah, nothing like seeing your post about Raptor Lake being buried in post about Zen processor.

 

[Markfw]

Would you please move the AMD 7950X discussion to a more appropriate thread? Thanks.

Talk to theelf who brought it up, as a cut down server cpu.

Edit: and that was in response to tamz_msc's comment about Zen.

 

[Hulk]

I have always wondering at what point in production chips are tested for their v/f curve and how long this test takes?

 

[DrMrLordX]

Intel 7nm is a impressive node when it comes to performance ✅😏

That's a 10nm node, not 7nm. Source: Intel.

 

[Hitman928]

They did that for consistency which makes sense.
The other option would be to not include the 2700x at all, like when people do benches of overclockable CPUs and leave out all the others.
They didn't bench the 2700x alone with gamemode in a vacuum.

Follow-Up Interview: Principled Technologies on Intel Testing Concerns | GamersNexus

stub After our exhaustive in-person interview with Principled Technologies, published on our YouTube channel, we followed-up via email to clarify some questions that were left unanswered during the initial video. As we noted in the initial interview, we give credit to Principled Technologies for...

www.gamersnexus.net

That makes zero sense. They are completely different CPUs and platforms. Turning on game mode on Ryzen isn't consistent, it's idiotic. Also, their consistency argument goes out the window when you look at the coolers they used for each CPU. There's a reason those tests were pretty much universally panned when they came out. Anyway, I'm not going to belabor the point, anyone can watch the GN video and read the follow-up and make up their own minds.

 

[Carfax83]

It is well-documented that dual-CCD Zen CPUs crap the bed when it comes to CPU intensive scenes in games with RT enabled. 25% deficit in memory performance doesn't help much either.

I don't think we can blame it on dual CCDs. Other RT enabled games display similar performance characteristics on Zen 4 regardless of single or dual CCDs. PCGH.de tested Spider-Man Miles Morales a while back and the 13900K had a nearly 50% lead over the 7950x with RT enabled.

The 7950x was also faster than the 7700x in the same test. I think a major component is the memory bandwidth. Building and maintaining BVH structure for RT requires a lot of CPU power and memory bandwidth. Spider-Man Miles Morales also added RT shadows, which increased the load even more which is why the gap is bigger than with Spider-Man Remastered.

 

[Markfw]

I don't think we can blame it on dual CCDs. Other RT enabled games display similar performance characteristics on Zen 4 regardless of single or dual CCDs. PCGH.de tested Spider-Man Miles Morales a while back and the 13900K had a nearly 50% lead over the 7950x with RT enabled.

The 7950x was also faster than the 7700x in the same test. I think a major component is the memory bandwidth. Building and maintaining BVH structure for RT requires a lot of CPU power and memory bandwidth. Spider-Man Miles Morales also added RT shadows, which increased the load even more which is why the gap is bigger than with Spider-Man Remastered.

My last post on this subject, but it IS on-topic, as it compares Raptor lake to Zen 4.

Nobody that I have seen runs their 7950x or any Zen 4 at less than DDR5 6000 EXPO, and they just released (or soon ??) 7200 EXPO memory. I am sure that this whole game situation will change in the near future, due to 3dvcache and memory speeds. AMD is newer to DDR5, and its learning how to adjust.

Bottom line is that I think that Raptor Lake and Zen 4 are competitive in cases where "normal load" is 0- ~12 threads. Once over that the e-cores start to lose to Zen 4 cores.

 

[Carfax83]

My last post on this subject, but it IS on-topic, as it compares Raptor lake to Zen 4.

Nobody that I have seen runs their 7950x or any Zen 4 at less than DDR5 6000 EXPO, and they just released (or soon ??) 7200 EXPO memory. I am sure that this whole game situation will change in the near future, due to 3dvcache and memory speeds. AMD is newer to DDR5, and its learning how to adjust.

Bottom line is that I think that Raptor Lake and Zen 4 are competitive in cases where "normal load" is 0- ~12 threads. Once over that the e-cores start to lose to Zen 4 cores.

Faster DDR5 would definitely increase Zen 4's performance in these types of games, but it would do so for the Raptor Lake CPU as well; as evidenced by the 12900K with DDR5 6400 getting a big boost. I don't think it could make up an almost 50% difference though. Some of that is likely from the cache bandwidth as well (plus maybe the higher OoO window), which Raptor Lake is very strong in as well.

AMD might need the 3D cache versions to catch up in these types of workloads.

 

[tamz_msc]

This seems like rather specific testing that I haven't seen data on. Where has this been documented?

Reviewers usually don't test this, so data is limited, but Computerbase has a comparison, although limited RT titles with a 3090 Ti.

Games with RT:



Games without RT:



Of particular interest is the 7900X vs the 7600X. The effect seems to be more pronounced there.