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.

 

[inf64]

PCGH now uses only officially supported memory speeds which hurts Zen architectures significantly more than RPL compared to what pretty much any gamer will run. It also explains why the 5800x3d does better against Zen 4 in their results compared to most other reviewers.

Exactly this. Good thing is that Vcache models don't really need DDR5-6000, so it will not matter much in their testing.

 

[Carfax83]

PCGH now uses only officially supported memory speeds which hurts Zen architectures significantly more than RPL compared to what pretty much any gamer will run. It also explains why the 5800x3d does better against Zen 4 in their results compared to most other reviewers.

You don't strike me as a spring chicken young type of dude, so you should know that testing with official standardized memory frequencies was the status quo for years for the vast majority of review sites, including Anandtech.

The shift towards using overclocked memory as a review standard is quite recent and was done specifically to put Zen 4 in the best possible light. In fact, several review sites tested Zen 4 with DDR5 6000 at launch and ADL with DDR5 4800 due to pressure from AMD.

Pcgh.de isn't the only one to still adhere to that review standard either. Computerbase.de, the other big German review site also does that, and so does Tomshardware and many others. Anandtech also went back to using official memory standards for their Zen 4 non X review.

It's amusing to see what was once (and still is to a degree) a widespread review practice now become stigmatized because AMD pressured reviewers into using overclocked settings in their reviews.

We've had this discussion before. I personally have no issues with reviewers using OC settings, but they need to be differentiated from stock settings. Also, many reviewers are pairing both Zen 4 and Raptor Lake with DDR5 6000, knowing full well that Raptor Lake can run DDR5 at much higher speeds and benefit from it.

If I were conducting a CPU review, I would run both RPL and Zen 4 at stock memory speeds and list OC optimized settings as well, DDR5 6000 for Zen 4 with 1:1 ratio and DDR5 7600 for RPL with 1:2 ratio.

 

[Carfax83]

Normally PCGH is a pretty good review site but their results don't seem to jive with most other reviews.

This isn't true. Computerbase.de also has similar results as PCGH.de. I think the reason has less to do with the memory frequencies used, and more to do with the settings and locations within the games.

A lot of reviewers stick to using canned benchmarks or non CPU limited areas. PCGH.de and Computerbase.de don't do that, they run custom benchmarks where they find the most CPU demanding areas and then do their test runs.

If you're willing to look for example, you can find areas in Cyberpunk 2077 which totally decimate Zen 3 and Zen 4 by combining crowd density with RT effects. I posted a YouTube video in a separate thread where you can see a 5800X3D getting totally demolished in this one area and drops below 60 FPS regardless of settings and is CPU bottlenecked, while a Raptor Lake rig is also CPU bottlenecked in that same area but manages to sustain triple digit frames per second:

 

[Markfw]

If I were conducting a CPU review, I would run both RPL and Zen 4 at stock memory speeds and list OC optimized settings as well, DDR5 6000 for Zen 4 with 1:1 ratio and DDR5 7600 for RPL with 1:2 ratio.

Would you also bench at the OC'ed memory settings you just said ? If so, I think that would be fair. We could also see how much benefit each platform can benefit from higher memory speed.

 

[Carfax83]

Would you also bench at the OC'ed memory settings you just said ? If so, I think that would be fair. We could also see how much benefit each platform can benefit from higher memory speed.

Yes, there's nothing wrong with providing OC and optimized memory results as the vast majority of enthusiasts and gamers will be using those settings and hardware. But, they should also differentiate those results from stock settings.

Tomshardware actually does this in their reviews now I've noticed. They used DDR5 6800 for the OC Intel results, and DDR5 6000 for the PBO Zen 4 results and then used the stock memory frequencies for both as well in the same graph.

 

[Timorous]

City skyline is very single threaded not that intensive.. hub showed it

Hub tested FPS which is not the important metric for Skylines or Stellaris or CK3 etc. Tic rate is the important metric.

 

[Thunder 57]

Yes, there's nothing wrong with providing OC and optimized memory results as the vast majority of enthusiasts and gamers will be using those settings and hardware. But, they should also differentiate those results from stock settings.

Tomshardware actually does this in their reviews now I've noticed. They used DDR5 6800 for the OC Intel results, and DDR5 6000 for the PBO Zen 4 results and then used the stock memory frequencies for both as well in the same graph.

Pretty sad that Tom's has surpassed Anandtech in probably every way except the forums. Hell, they still review GPU's! Ryan's forest fire excuse fort of lost credibility years ago.

 

[Markfw]

Yes, there's nothing wrong with providing OC and optimized memory results as the vast majority of enthusiasts and gamers will be using those settings and hardware. But, they should also differentiate those results from stock settings.

Tomshardware actually does this in their reviews now I've noticed. They used DDR5 6800 for the OC Intel results, and DDR5 6000 for the PBO Zen 4 results and then used the stock memory frequencies for both as well in the same graph.

Only problem with that is mixing CPU OC with ram OC. For example. I have no problem using EXPO memory, which is supported by BIOS, 6000 CL30 in my case to be exact, but no way would I OC the CPU, as the wattage would skyrocket, I actually use PCO CO = -25 and max temp = 85. I think a lot of people here have done the same.

I want to see ram at EXPO or XMP and CPU stock. I think that the most common use case.

 

[Hitman928]

You don't strike me as a spring chicken young type of dude, so you should know that testing with official standardized memory frequencies was the status quo for years for the vast majority of review sites, including Anandtech.

The shift towards using overclocked memory as a review standard is quite recent and was done specifically to put Zen 4 in the best possible light. In fact, several review sites tested Zen 4 with DDR5 6000 at launch and ADL with DDR5 4800 due to pressure from AMD.

Pcgh.de isn't the only one to still adhere to that review standard either. Computerbase.de, the other big German review site also does that, and so does Tomshardware and many others. Anandtech also went back to using official memory standards for their Zen 4 non X review.

It's amusing to see what was once (and still is to a degree) a widespread review practice now become stigmatized because AMD pressured reviewers into using overclocked settings in their reviews.

We've had this discussion before. I personally have no issues with reviewers using OC settings, but they need to be differentiated from stock settings. Also, many reviewers are pairing both Zen 4 and Raptor Lake with DDR5 6000, knowing full well that Raptor Lake can run DDR5 at much higher speeds and benefit from it.

If I were conducting a CPU review, I would run both RPL and Zen 4 at stock memory speeds and list OC optimized settings as well, DDR5 6000 for Zen 4 with 1:1 ratio and DDR5 7600 for RPL with 1:2 ratio.

Way back when, sure, only officially supported memory speeds were used, but the platforms were very different back then. After the advent of XMP, it became fairly common to use above official spec'd memory because XMP was a one click change in the bios and was reliably stable. I have said it before multiple times, but I am perfectly fine with reviewers testing memory at officially supported specs, however, pretty much everyone who is actually a PC gamer shouldn't rely on those results because they won't be running such slow memory and it can make a big difference in the results.

I don't remember seeing any reviewers only using OC'd memory on the AMD side but leaving the Intel side at stock but I also don't search all of the test sites for every release. Do you have a link to such a review?

 

[TheELF]

Here you go, 175w sustained = 40.7k score
View attachment 74026

https://imgur.com/a/tCBVe9B

Where are your screenshot btw ?
*forum insanely slow*

It's not your fault, you can only show what the CPU tells you...

The CPU is drawing much more than what PPT tells you.
Just saying...

Starting with the peak power figures, it's worth noting that AMD's figures can be wide off the mark even when restricting the Package Power Tracking (PPT) in the firmware. For example, restricting the socket and 7950X to 125 W yielded a measured power consumption that was still a whopping 33% higher.

A Lighter Touch: Exploring CPU Power Scaling On Core i9-13900K and Ryzen 9 7950X

www.anandtech.com

 

[Hitman928]

It's not your fault, you can only show what the CPU tells you...

The CPU is drawing much more than what PPT tells you.
Just saying...

A Lighter Touch: Exploring CPU Power Scaling On Core i9-13900K and Ryzen 9 7950X

www.anandtech.com

I think whoever did this Anandtech article is just confused or the setting they are changing isn't doing exactly what they think it is. Their power limits to peak power deltas match consistently with them having set the TDP target for the CPU where it is allowed to exceed that power by a certain amount at every step (e.g. setting a 65W TDP will allow the CPU to use up to 88 W and may peak just a hair above that).

Either way, @Det0x is clearly showing actual power readings rather than relying on a set power limit to determine power usage, so I don't really see what the point you are trying to make is.

 

[Carfax83]

Way back when, sure, only officially supported memory speeds were used, but the platforms were very different back then. After the advent of XMP, it became fairly common to use above official spec'd memory because XMP was a one click change in the bios and was reliably stable. I have said it before multiple times, but I am perfectly fine with reviewers testing memory at officially supported specs, however, pretty much everyone who is actually a PC gamer shouldn't rely on those results because they won't be running such slow memory and it can make a big difference in the results.

I agree with this. The best approach is to have both stock memory frequencies and overclocked memory frequencies.

I don't remember seeing any reviewers only using OC'd memory on the AMD side but leaving the Intel side at stock but I also don't search all of the test sites for every release. Do you have a link to such a review?

I remember there were a slight few that I came across that were like that. It's been a while though so I can't specifically remember which ones. Doing a quick Google search lead me to Ars Technica's review where they used DDR5 6000 for Zen 4, and DDR5 5800 for ADL i9 and DDR5 4800 for ADL i5:

Ryzen 7600X and 7950X review: Zen 4 starts off expensive but impressive | Ars Technica

There were a few others though, but I just can't recall as there were a ton of reviews.

 

[Carfax83]

Pretty sad that Tom's has surpassed Anandtech in probably every way except the forums. Hell, they still review GPU's! Ryan's forest fire excuse fort of lost credibility years ago.

Yep, I have to agree. Tomshardware and Anandtech were the biggest review sites back in the old days, but while Tomshardware has maintained their stature, Anandtech has shrunk into near irrelevance......especially after Dr. Ian Cuttress and the other guy (forgot his name) left. Anandtech was the best for doing deep technical dives into CPUs, but now I guess ChipsandCheese has taken over that niche.

 

[Carfax83]

Only problem with that is mixing CPU OC with ram OC. For example. I have no problem using EXPO memory, which is supported by BIOS, 6000 CL30 in my case to be exact, but no way would I OC the CPU, as the wattage would skyrocket, I actually use PCO CO = -25 and max temp = 85. I think a lot of people here have done the same.

I want to see ram at EXPO or XMP and CPU stock. I think that the most common use case.

So AMD PBO overclocks the CPU? I didn't know that. I thought it was just putting the memory controller and memory frequency in a 1:1 ratio and setting the IF to the optimized frequency.

 

[Hitman928]

I agree with this. The best approach is to have both stock memory frequencies and overclocked memory frequencies.



I remember there were a slight few that I came across that were like that. It's been a while though so I can't specifically remember which ones. Doing a quick Google search lead me to Ars Technica's review where they used DDR5 6000 for Zen 4, and DDR5 5800 for ADL i9 and DDR5 4800 for ADL i5:

Ryzen 7600X and 7950X review: Zen 4 starts off expensive but impressive | Ars Technica

There were a few others though, but I just can't recall as there were a ton of reviews.

They didn't use stock memory speeds here though. They used the same memory as on the AMD board but it was EXPO memory which wasn't stable on the Intel platform so they had to lower the speed to achieve stability. This is just the reviewer being cheap/clueless and far from AMD pressuring a reviewer to bias the review as you claimed.

Other specs are outlined in the table below; slower RAM speeds on the Intel board are because the AMD EXPO RAM profile can't currently load on the board we used without becoming unstable.

 

[Hitman928]

Yep, I have to agree. Tomshardware and Anandtech were the biggest review sites back in the old days, but while Tomshardware has maintained their stature, Anandtech has shrunk into near irrelevance......especially after Dr. Ian Cuttress and the other guy (forgot his name) left. Anandtech was the best for doing deep technical dives into CPUs, but now I guess ChipsandCheese has taken over that niche.

Anandtech and Toms are both owned by the same company now which I don't really understand, but it seems like they are being given budgets for different things with Toms getting the money for PC review type articles.

 

[Thunder 57]

Yep, I have to agree. Tomshardware and Anandtech were the biggest review sites back in the old days, but while Tomshardware has maintained their stature, Anandtech has shrunk into near irrelevance......especially after Dr. Ian Cuttress and the other guy (forgot his name) left. Anandtech was the best for doing deep technical dives into CPUs, but now I guess ChipsandCheese has taken over that niche.

I always thought Anandtech was superior. It was dealt a blow when Anand left. Since Ian left it has only gotten worse. Pretty sure the other one you are thinking of was Andrei. They used to go very in depth with CPU's, not anymore. And GPU's, forget it. I guess I can say it because the forum is its own entity but Ryan has ruined AT. Chips & Cheese is pretty good from what I've seen so far.

So AMD PBO overclocks the CPU? I didn't know that. I thought it was just putting the memory controller and memory frequency in a 1:1 ratio and setting the IF to the optimized frequency.

I guess it shows you haven't used Zen . PBO can do a few things. It is commonly used to allow higher power limits but can also be used to allow a higher clock speed or a lower voltage curve. It doesn't have to OC the CPU but it can.

 

[Markfw]

I always thought Anandtech was superior. It was dealt a blow when Anand left. Since Ian left it has only gotten worse. Pretty sure the other one you are thinking of was Andrei. They used to go very in depth with CPU's, not anymore. And GPU's, forget it. I guess I can say it because the forum is its own entity but Ryan has ruined AT. Chips & Cheese is pretty good from what I've seen so far.



I guess it shows you haven't used Zen . PBO can do a few things. It is commonly used to allow higher power limits but can also be used to allow a higher clock speed or a lower voltage curve. It doesn't have to OC the CPU but it can.

Edit @Carfax83
Yes, PBO can lower the curve optimizer or raise it, and can change all sorts of other settings. It OC to the moon or make it a 35 watt laptop type efficiency king. I choose to make it more efficient, as it still retains 95% of the true power, but HALF the wattage, and lower temps.

But I also use high end air and AIO's to allow it to go higher than using a cheap cooler at the same settings.

 

[itsmydamnation]

It's not your fault, you can only show what the CPU tells you...

The CPU is drawing much more than what PPT tells you.
Just saying...

A Lighter Touch: Exploring CPU Power Scaling On Core i9-13900K and Ryzen 9 7950X

www.anandtech.com

i down voted you for being both patronising and wrong

 

[DrMrLordX]

The CPU is drawing much more than what PPT tells you.

That's TDP, not PPT. Hurf durf.

PPT is what the CPU is actually pulling. Bring up HWiNFO64 (or similar) on a modern AMD system and you will see. PPT maxes out at 1.35 x TDP. The math here is not complicated.

 

[Timur Born]

A hint for those doing per core OC on Intel CPUs with enabled E cores and any power plan/mode other than "Best Performance": You need to check your P core stability in combination with fully loaded E cores, because something like this can happen in practice:

 

[jpiniero]

Per AT, the 13900KS is now available for a very low price of $699

 

[Hitman928]

First 13900KS review. It was 2% faster in their gaming suite test compared to the 13900K.

 

[igor_kavinski]

First 13900KS review. It was 2% faster in their gaming suite test compared to the 13900K.

That's lame performance for an extra $100 or so. Maybe that guy has no idea how to test the chip properly

 

[igor_kavinski]

TEC Makes a User-friendly Comeback as the Cooler Master ML360 Sub-Zero Evo AIO CLC

Cooler Master is working with Intel to bring TEC cooling back to the PC mainstream under an Intel-designed standard. Back in the 2010s, the professional overclocking scene was pervaded with various sub-zero cooling methods including liquid-nitrogen or dry-ice evaporators, refrigerated cooling...

www.techpowerup.com

This might be what the KS needs.