Question Raptor Lake - Official Thread

Page 84 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Hulk

Diamond Member
Oct 9, 1999
4,525
2,519
136
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.
 
  • Like
Reactions: vstar

nicalandia

Diamond Member
Jan 10, 2019
3,331
5,282
136
TLDR:
  • ~12% higher ST / gaming performance
  • ~ 42% higher MT performance at "unlimited mode"
  • Power usage have increased from 236w to 343w to reach performance numbers above

  • P-core IPC increase is basically nothing ~1%
  • E-core IPC increase is ~6%
  • Most of performance increase comes from higher clockspeeds
Seem to be higher memory latency in aida because of the more stops on the ringbus, but the L2 and L3 bandwidth is increased substantially.
Ecore only downlock the ringbus to 4.6ghz instead of 3.6ghz when they are in use

Great TLDR, I appreciate it...!
 

Abwx

Lifer
Apr 2, 2011
11,557
4,349
136
It looks like about a 42% boost when both are in unlimited mode. Though I don't see a mention of how much power the 12900k consumes in that scenario. But out of the box, we'll probably see a smaller performance gap but greater efficiency improvement.

The improvement is mainly brought by the amount of cores, RPL limited to 8 + 8, that is the 13700K, would use about 210W at same MT perf than the 12900K@240W according to Intel s process perf/watt improvement wich is stated at 10-15%.
 
Last edited:

Exist50

Platinum Member
Aug 18, 2016
2,452
3,102
136
The improvement is mainly brought by the amount of cores, RPL limited to 8 + 8, that is the 13700K, would use about 210W at same MT perf than the 12900K@240W according to Intel s process perf/watt improvement wich is stated at 10-15%.
Where's that 10-15% from?
 

Det0x

Golden Member
Sep 11, 2014
1,299
4,234
136
Where's that 10-15% from?
Google translate:

On September 28, 2022, Beijing time, after eleven months, Intel released their new generation of desktop products, Raptor lake-S-Raptor Lake. (CPP) further relaxed from 54 nm to 60 nm to get the ultimate 10nm process, finally offering a Pcore frequency in excess of 0.5 Ghz compared to the Alder lake series. At the same time, Intel doubled the Ecore in i5 i7 i9 products and provided more L2/L3 cache to improve application and game performance, finally taking performance to a whole new level.

...One year after the release of the Alder lake product, Intel obtained a new generation of Raptor lake with higher frequency and better energy efficiency by further relaxing the CPP's 10nm Enhanced Super Fin plus (Intel 7+) process. 作者:ECSM_Official https://www.bilibili.com/read/cv18648273 出处:bilibili
 
  • Like
Reactions: mikk

Exist50

Platinum Member
Aug 18, 2016
2,452
3,102
136
Google translate:

On September 28, 2022, Beijing time, after eleven months, Intel released their new generation of desktop products, Raptor lake-S-Raptor Lake. (CPP) further relaxed from 54 nm to 60 nm to get the ultimate 10nm process, finally offering a Pcore frequency in excess of 0.5 Ghz compared to the Alder lake series. At the same time, Intel doubled the Ecore in i5 i7 i9 products and provided more L2/L3 cache to improve application and game performance, finally taking performance to a whole new level. 作者:ECSM_Official https://www.bilibili.com/read/cv18648273 出处:bilibili
Again, where's the 10-15% efficiency claim coming from?

And as for that bolded quote, iirc Intel's pretty much been exclusively using 60nm CPP since at least Tiger Lake. Don't think there's been any change for Raptor Cove.
 
  • Like
Reactions: Kaluan

Abwx

Lifer
Apr 2, 2011
11,557
4,349
136
Again, where's the 10-15% efficiency claim coming from?

And as for that bolded quote, iirc Intel's pretty much been exclusively using 60nm CPP since at least Tiger Lake. Don't think there's been any change for Raptor Cove.

The slide was undoubtly posted in this very thread, i tried to find it at Computerbase but they dont always publish all slides in their articles even for big events, FI some slides from AMD s presentation were not published, some can be found at AT...
 
  • Like
Reactions: Kaluan

inf64

Diamond Member
Mar 11, 2011
3,865
4,549
136
TLDR:

13900k compared to 12900k:
  • ~12% higher ST / gaming performance
  • ~ 42% higher MT performance at "unlimited mode"
  • Power usage have increased from 236w to 343w to reach performance numbers above

  • P-core IPC increase is basically nothing ~1-3%
  • E-core IPC increase is ~6%
  • Most of performance increase comes from higher clockspeeds
Seem to be higher memory latency in aida because of the more stops on the ringbus, but the L1, L2 and L3 bandwidth is increased substantially.
Ecore only downlock the ringbus to 4.6ghz instead of 3.6ghz when they are in use
Thanks for the summary.

I almost nailed it in my perf. projection here: bit.ly/3PEY8cM
I predicted 15% better ST performance for 13900K vs 12900K and then end result is ~12%. I estimated that MT increase will be 49% and the preview shows 41% which matches intel's info from the last conference they held. Overall, I missed the ST/MT performance by 3/5%.

I'm pretty confident now that ST will be a ~tie between 7950X and 13900K and 7950X should win the MT by 5-7%.
 

inf64

Diamond Member
Mar 11, 2011
3,865
4,549
136
Regarding perf./watt in the "unlimited" mode, in MT workloads it should be 1.41/343W Vs 1/236W. So a tie more or less (~3% difference)
 
  • Like
Reactions: Kaluan

BorisTheBlade82

Senior member
May 1, 2020
680
1,069
136
Regarding perf./watt in the "unlimited" mode, in MT workloads it should be 1.41/343W Vs 1/236W. So a tie more or less (~3% difference)
Doesn't that indicate that they were able to flatten their V/F curve a bit? If so, it should be quite a bit more efficient than ADL at reasonable limits like 251w or even 125w.
 
  • Like
Reactions: Lodix

Abwx

Lifer
Apr 2, 2011
11,557
4,349
136
Both Alder Lake and Raptor Lake ostensibly use the same Intel 7 process. If there're differences, Intel hasn't said.

I m about sure, in CB R23 8 e cores would add about 8800 pts as well as 50W at same process and frequencies, to dial down the whole from 290W to 253W would require 14.6% perf/Watt improvement.

Since the 12900K@240W score 27281 pts this would put RPL at around 36000 pts@253W, that s about what is leaked by some people who have the chip in hands.
 

Hulk

Diamond Member
Oct 9, 1999
4,525
2,519
136
I am dying to see how that compares to a retail 7950x, but in 9 days we should know. That will save arguments.

But I must says, 42% boost for 45% more power and 343 watts total seems insane. How would you cool that ? AIO are only 250 watt I think. Custom WC only ?

It IS insane if you are considering prolonged use at that power level. But if you are only looking at bursty workload that last for a few seconds here and there, think Photoshop, then the power spikes aren't a huge concern. Then again it's easy to cap the power at more sane levels with only a small decrease in performance. But yes I'm certainly not going to compress video for hours on end at 350W!

As you wrote we'll know more in 9 days.
 

Exist50

Platinum Member
Aug 18, 2016
2,452
3,102
136
It IS insane if you are considering prolonged use at that power level. But if you are only looking at bursty workload that last for a few seconds here and there, think Photoshop, then the power spikes aren't a huge concern. Then again it's easy to cap the power at more sane levels with only a small decrease in performance. But yes I'm certainly not going to compress video for hours on end at 350W!

As you wrote we'll know more in 9 days.
I'm not sure why it's even a topic of discussion. The PL2 is ~250W, not 350W. You'd only see numbers that high if you went out of your way to uncap it, in which case you'd presumably know why.
 
  • Like
Reactions: controlflow

Hulk

Diamond Member
Oct 9, 1999
4,525
2,519
136
Very true. I think I responded because sometimes benchmarks are quoted without power levels. Then someone responds, "Yeah but that's at 350W!" And on and on. Trying to get people thinking that there are usage scenarios that might work better for one CPU over another. Zen 4 and Raptor look to be very competitive, it's gonna come down to specific usage and as always price.
 

scannall

Golden Member
Jan 1, 2012
1,960
1,678
136
Very true. I think I responded because sometimes benchmarks are quoted without power levels. Then someone responds, "Yeah but that's at 350W!" And on and on. Trying to get people thinking that there are usage scenarios that might work better for one CPU over another. Zen 4 and Raptor look to be very competitive, it's gonna come down to specific usage and as always price.
The performance metrics that are important to me may not mean anything to someone else. We all have different tastes, quirks, requirements etc. Cool, quiet and my CPU not be the bottleneck. Others may prefer pure throughput, forget the heat. And that's fine, their needs are theirs. As well as the money they spend.

When the reviews come out pay more attention to the things that are important to you, not what is important to the reviewer.
 
  • Like
Reactions: DAPUNISHER

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
26,171
15,317
136
The performance metrics that are important to me may not mean anything to someone else. We all have different tastes, quirks, requirements etc. Cool, quiet and my CPU not be the bottleneck. Others may prefer pure throughput, forget the heat. And that's fine, their needs are theirs. As well as the money they spend.

When the reviews come out pay more attention to the things that are important to you, not what is important to the reviewer.
Yes, and when I comment, many know I am talking about 100% 24/7 MT performance and power usage. Also, for someone to claim (for example) "Raptor lake is king of (insert metric)", but then its at 350 watts, that is just stupid. At todays speeds, aside from what I do, no one needs to be 10% faster for 50% more power usage(example, not necessarily real), thats just wasteful. And yes, heat is also a metric.
 

DrMrLordX

Lifer
Apr 27, 2000
22,065
11,693
136
How would you cool that ? AIO are only 250 watt I think. Custom WC only ?

I think some of the 360mm AiO units can do it, but it might take loud fans. There are some 420mm AiO units. HSF is completely off the table.

It IS insane if you are considering prolonged use at that power level. But if you are only looking at bursty workload that last for a few seconds here and there, think Photoshop, then the power spikes aren't a huge concern.

Honest question, does Photoshop even benefit from e-cores?