Intel Raptor Lake Refresh reportedly launching in Q3/Q4 2023 with enough performance to "tread water against Ryzen 7000"

[Redaktion]

Intel could launch refreshed Raptor Lake CPUs later this year if the latest report from leaker Moore's Law Is Dead is accurate. The leaker alleges that RPL-R could be competitive against the Ryzen 7000 processor but this reprieve might be short-lived as Zen 5 CPUs are possibly launching before desktop Intel Arrow Lake (ARL-S).

https://www.notebookcheck.net/Intel-Raptor-Lake-Refresh-reportedly-launching-in-Q3-Q4-2023-with-enough-performance-to-tread-water-against-Ryzen-7000.702146.0.html

[Anonymousgg]

How much is 7950X3D/simulated 7800X3D ahead of the 13900KS in reviews? 5% on average at most? DLVR and some other tweaks might be enough to make up that ground and get a technical tie.

Everything after it (Meteor/Arrow/etc.) is more important, but it is what it is.

[Truther]

Soo.... To only tread the water.... But not swim....

[h4]

MLID also reports that RPL-R may feature "notably better MT performance" if Intel can take advantage of better binning, DLVR, etc. Digital Linear Voltage Regulator (DLVR) is an Intel technology that reportedly has the potential to reduce power consumption by up to 25%.

If you read patent for DLVR, it is not difficult to understand that it is intended to improve the battery life of laptops slightly by making power delivery somewhat more efficient for moderately fluctuating loads, and is not effective at sustained maximum loads such as while benchmarking.

It is easy to calculate from the general power-performance curve that technology that can reduce power consumption by 20% will only raise maximum MT performance by about 10%. MLID may have a poor understanding of engineering, I think every time his story comes around. I guess he is only a gossip-monger.

[Anonymousgg]

If you read patent for DLVR, it is not difficult to understand that it is intended to improve the battery life of laptops slightly by making power delivery somewhat more efficient for moderately fluctuating loads, and is not effective at sustained maximum loads such as while benchmarking.

It is easy to calculate from the general power-performance curve that technology that can reduce power consumption by 20% will only raise maximum MT performance by about 10%. MLID may have a poor understanding of engineering, I think every time his story comes around. I guess he is only a gossip-monger.

That's notable for a REFRESH that could only be expected to raise performance by a couple of % normally. Not to mention the reduced power consumption in lighter loads like gaming. You are a nitpicker.

[h4]

That's notable for a REFRESH that could only be expected to raise performance by a couple of % normally. Not to mention the reduced power consumption in lighter loads like gaming. You are a nitpicker.

MLID says that DLVR can "notably increase MT performance", even though it is a pure power-cutter and cannot be a booster (see image above). RPL uses enhanced Intel7 and has improved MT performance by about 10% at the same power/#cores compared to ADL, so it is possible if further improvements are made to the Intel7 nodes as well, but DLVR is of a completely different nature. He is confusing it.

The MLID information is often questionable. For example, in the video discussed in Noteboockcheck's 11/15/2022 article, he states that "Redwood Cove is targeting a 15-25% [IPC] increase over Raptor Cove."

However, Semianalysis' Intel VisiON coverage (May 2022) already pointed out that Redwood cove is not much different from the Golden Cove. Photos of the exhibited die suggested that transistors (excluding cache area) will only increase by about 10%. This is similar to the change from Zen3 to Zen4, and it would be a more likely guess that IPC will only improve up to 10%, as OneRaichu (PhD of chemical engineering) says.

 

[Mr Majestyk]

That's notable for a REFRESH that could only be expected to raise performance by a couple of % normally. Not to mention the reduced power consumption in lighter loads like gaming. You are a nitpicker.

MLID says that DLVR can "notably increase MT performance", even though it is a pure power-cutter and cannot be a booster (see image above). RPL uses enhanced Intel7 and has improved MT performance by about 10% at the same power/#cores compared to ADL, so it is possible if further improvements are made to the Intel7 nodes as well, but DLVR is of a completely different nature. He is confusing it.

The MLID information is often questionable. For example, in the video discussed in Noteboockcheck's 11/15/2022 article, he states that "Redwood Cove is targeting a 15-25% [IPC] increase over Raptor Cove."

However, Semianalysis' Intel VisiON coverage (May 2022) already pointed out that Redwood cove is not much different from the Golden Cove. Photos of the exhibited die suggested that transistors (excluding cache area) will only increase by about 10%. This is similar to the change from Zen3 to Zen4, and it would be a more likely guess that IPC will only improve up to 10%, as OneRaichu (PhD of chemical engineering) says.

 

You are assuming Golden cove is same architecture as Raptor cove, and basically a die shrink. It's a major architectural change, not just a Raptor cove+.

[h4]

You are assuming Golden cove is same architecture as Raptor cove, and basically a die shrink. It's a major architectural change, not just a Raptor cove+.

The difference between Redwood cove and Golden cove is roughly estimated to be about the difference between Sunny Cove and Willow Cove (average IPC improvement up tp 10%).


Fawad, you write in today's post:

ARL-S is indeed H2 2024 and RPL-R is scheduled for Q3/Q4 2023, Intel may only be competitive with AMD for a few months before the launch of Zen 5. As a refresher, the AMD Zen 5 is rumored to enjoy up to a 30% IPC uplift in addition to other architectural improvements.

Also, you wrote in your 11/15/2022 article citing MLID:

Compared to the 14th gen Redwood Cove, the Lion Cove (Note: Arrow Lake) performance core will allegedly bring an enormous 22-34% IPC uplift.

The two articles together indicate that Arrow lake and Zen5 are supposed to have similar IPC improvements, but today's article makes me think you don't remember what you wrote in your previous articles.

Incidentally, Zen5 is aiming for IPC improvement, as publicly explained at AMD's investor meeting, and as for Arrow lake, if it is manufactured at 20A, it is likely that there will be significant design changes, including IPC improvement, since backside power delivery will change the design rules significantly.