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.

 

[nicalandia]

OK, what is going on? How come Puget Systems had 13900KS in 2nd place in 10 minute test but Techspot show it in first place?

Also I would like to point out that they used a contact frame that improved the all core boost when compared to the same loop without it.

 

[Hitman928]

On a different topic...
Is there one programmed VID for a CPU or is there a VID for each core? I thought there was one voltage plane for Alder/Raptor but in the BIOS you can apparently set voltage offsets by core?

I don't have ADL or RPL but Zen CPUs have been able to due per core voltage offsets since at least Zen 3. I wouldn't be surprised if Intel CPUs can do the same thing.

 

[Rigg]

What in particular? It's been recited ad nauseam here. Everything from individual games performing drastically worse on Intel (and only Intel) vs other publications, to purposely choosing test conditions to favor AMD (no W11, no RT, etc), to dishonest price comparisons. It's obvious to anyone who's been following them that they warp their testing and even reporting to fit a preconceived outcome in line with their brand preferences. Why do you think that they, of all outlets, come up so often on these threads?

Provide specific examples. @Hitman928's observations mirror my own.

I've seen a few posters repeat that they are biased/bad ad nauseam but haven't seen anyone provide any actual proof or concrete examples. Do you have solid examples to point to that show their bad testing methodology? They have used W11 since Alderlake dropped specifically for Intel and very few reviewers turn on RT during CPU reviews because it introduces such a large GPU burden (and I have yet to see any real evidence that it is wise to turn it on for CPU reviews). The only one I see pushing for RT in CPU tests is @Carfax83 because he thinks (again, without real evidence) that Intel CPUs are better at it than AMD CPUs.

 

[Rigg]

And the circle jerk continues........

The results are bunched up together because there is very little if any CPU isolation in those tests. Had they turned on RT for those games that supported it, the results likely would have been different.

We've covered this.

And RT isn't an "edge case scenario." Like it or not, RT is here to stay and the amount of games that support it will only continue to grow.

I didn't say that RT is an edge case scenario. Playing video games on a 4090 with 'officially supported' mem speed/JEDEC timings at 720p is an edge case irrespective of whether or not RT is enabled. 1080p gaming on a 4090 is arguably an edge case.

Their YouTube video actually showed snippets of the areas they are testing. Out of the all the games they tested, I only have Plague Tale Requiem and I can confirm that the early town area isn't very CPU intensive at all. Now the later town areas are a bit more CPU intensive (or when there are large amounts of rats on screen), but generally speaking, Plague Tale Requiem is much more GPU limited.

For HZD he's using the canned benchmark and for CBP 2077, it looks like crowd density is turned down.

So in other words you are speculating that they aren't testing in adequately CPU demanding areas of these games. Why is this criticism only leveled at HUB?

That's because the RTX 3090 Ti was probably tapped out. Spider-Man Miles Morales increased the lead Raptor Lake had over Zen 4 substantially despite the engine being the exact same as the original Spider-Man Remastered, because it added RT shadows in addition to the RT reflections that were in the original game. According to the PCGH.de results, Zen 4's performance diminished in Spider Man MM despite using a faster RTX 4090. Raptor Lake took a hit as well, but not nearly as much.

Spider-Man: Miles Morales - CPU Benchmarks (pcgameshardware.de)

None of the Zen 4 CPUs can break the 60 FPS barrier in the Witcher 3 Next gen, which uses RT global illumination, RT reflections, RT shadows, RT ambient occlusion and increased draw distances.

Ryzen 7000 Non-X: Benchmarks, new basics (pcgameshardware.de)



It varies across reviewers, but most of the more established reviewers tend to use stock memory settings.

Since they don't provide data for the same games with and without RT we can't isolate RT as a variable. The pcgameshardware.de testing is a much better example of how testing at officially supported memory speeds biases the data against Zen 4. Lets compare Plague Tale Requiem data between HUB and pcgameshardware.de.

The data for the raptors and x3D is nearly a mirror image. Zen 4 not so much.

 

[Exist50]

Provide specific examples. @Hitman928's observations mirror my own.

Which in particular do you claim hasn't already been shown in this thread? But let's start with something very basic. You don't find something odd about this discrepancy?







I just chose a selection of the biggest publications, but no matter how you attempt to spin it, HWUB is always a huge outlier in AMD's favor. But all this is obvious if you've been following reviews from multiple sources.

 

[Rigg]

Which in particular do you claim hasn't already been shown in this thread? But let's start with something very basic. You don't find something odd about this discrepancy?

View attachment 74610

View attachment 74611

View attachment 74612

I just chose a selection of the biggest publications, but no matter how you attempt to spin it, HWUB is always a huge outlier in AMD's favor. But all this is obvious if you've been following reviews from multiple sources.

No I don't find it odd at all. It's obvious that if you dig into the test system configurations used by these 3 examples that any discrepancies in these results are 100% related to different memory configurations. Multi game averages are also going to vary because the data isn't directly comparable.

 

[nicalandia]

No I don't find it odd at all.

Come February the amount of Copium on this thread will be off the charts...!

 

[Exist50]

No I don't find it odd at all. It's obvious that if you dig into the test system configurations used by these 3 examples that any discrepancies in these results are 100% related to different memory configurations. Multi game averages are also going to vary because the data isn't directly comparable.

So it's always moving the goalposts. Weren't you just disputing my claim about their results being irregular? Now you're coming up with excuses for why that might be, and on it goes.

Come February the amount of Copium on this thread will be off the charts...!

This is terribly ironic given how badly some predictions aged for Zen4X3D. How many people insisted it would be ~30% better? For some people, it seems insufficient to be happy that their favorite brand makes a good product. They have to insist it's absurdly, unrealistically better than anything else anywhere, rather than settling for more moderate praise.

 

[Hitman928]

Which in particular do you claim hasn't already been shown in this thread? But let's start with something very basic. You don't find something odd about this discrepancy?

View attachment 74610

View attachment 74611

View attachment 74612

I just chose a selection of the biggest publications, but no matter how you attempt to spin it, HWUB is always a huge outlier in AMD's favor. But all this is obvious if you've been following reviews from multiple sources.

HWUB matches well to other outlets that don't limit testing to JEDEC speed memory and test a decent amount of games:

Techpowerup




Techdeals


Jarrod'sTech



You can also look at Eurogamer which doesn't give a multi-game average, but the individual games show Zen4 CPUs with much more favorable results than computerbase or pcgh which both only use JEDEC memory speeds.





If you look at the last one especially, Eurogamer shows that when both are using JEDEC memory speeds, the 13900k is 11.3% faster than the 7950x. However, when both are using faster memory, the gap shrinks such that the 13900k is only 3.5% faster. Again, Zen4 typically benefits much more from faster memory than RPL which is one big reason why computerbase and pcgh results show Zen4 trailing RPL by so much compared to HWUB and other outlets. HWUB results are not an outlier or unusual when taking into account memory speeds.

Edit: originally had the wrong screenshot for COD.

 

[Hitman928]

I just calculated it out and Eurogamer (which I guess is run by digital foundry?) actually has the 7950x as 2.4% faster on average across their game tests than the 13900k when both are using DDR6000 memory. So there's that. Keep in mind that it's only 8 games though.

Page 4 | Intel Core i9 13900K and Core i5 13600K review: an effective redoubt against AMD's Ryzen 7000 advances

Digital Foundry tests the Intel Core i9 13900K and Core i5 13600K, two Raptor Lake processors that offer more cores, hi…

www.eurogamer.net

 

[Exist50]

HWUB matches well to other outlets that don't limit testing to JEDEC speed memory and test a decent amount of games:

Techpowerup

View attachment 74619


Techdeals
View attachment 74620

Jarrod'sTech

View attachment 74621

You can also look at Eurogamer which doesn't give a multi-game average, but the individual games show Zen4 CPUs with much more favorable results than computerbase or pcgh which both only use JEDEC memory speeds.

View attachment 74624

View attachment 74623

If you look at the last one especially, Eurogamer shows that when both are using JEDEC memory speeds, the 13900k is 11.3% faster than the 7950x. However, when both are using faster memory, the gap shrinks such that the 13900k is only 3.5% faster. Again, Zen4 typically benefits much more from faster memory than RPL which is one big reason why computerbase and pcgh results show Zen4 trailing RPL by so much compared to HWUB and other outlets. HWUB results are not an outlier or unusual when taking into account memory speeds.

Edit: originally had the wrong screenshot for COD.

Huh? Let's take TechPowerUp, the first one you quote.



Radically different results than HWUB. Look at the 7600X vs 13600K, for example. TPU shows the 13600K winning by 8-9%. HWUB, meanwhile, shows the 7600X as ~3% better. Again, that is from your own choice of source. So what else do you think can explain that?

And then when half of the rest are random tech tubers no one's heard of before they posted benchmarks some people found convenient...

You can also look at Eurogamer which doesn't give a multi-game average, but the individual games show Zen4 CPUs with much more favorable results than computerbase or pcgh which both only use JEDEC memory speeds.

DRAM certainly makes a difference. No one's arguing that. But the results HWUB show are anomalous even accounting for that. Eurogamer is great, and there's obviously some value in testing the same memory config, but that config is a greater overclock on Raphael than Raptor Lake. What would the results be like with, say, DDR5-6400?

And as I said originally, it's not just that their results are odd. These "discrepancies" show up in plenty of other places, like that price comparison I linked.

 

[Exist50]

Techdeals

Oh, and I just gave this one a second look. They literally use worse memory on the Intel system than the AMD system. If you're going to defend using >stock configs, which I'm fine with, then how on earth would you defend that choice? Are you going to tell me you think Raphael's memory support is better than Raptor Lake's?

 

[Hitman928]

Huh? Let's take TechPowerUp, the first one you quote.

View attachment 74625

Radically different results than HWUB. Look at the 7600X vs 13600K, for example. TPU shows the 13600K winning by 8-9%. HWUB, meanwhile, shows the 7600X as ~3% better. Again, that is from your own choice of source. So what else do you think can explain that?

And then when half of the rest are random tech tubers no one's heard of before they posted benchmarks some people found convenient...


DRAM certainly makes a difference. No one's arguing that. But the results HWUB show are anomalous even accounting for that. Eurogamer is great, and there's obviously some value in testing the same memory config, but that config is a greater overclock on Raphael than Raptor Lake. What would the results be like with, say, DDR5-6400?

Yes, in TPU's original review with more limited games, the Zen architectures were a little further behind, but when they did their 53 game suite, the 5800x3d went from 13% behind to only 6.2% behind. This is actually a better result for the 5800x3d than HWUB showed which is why I included it. TPU's standard game suite obviously includes games that are a little more favorable towards Intel (at least compared to the 5800x3d) but when a very large amount of games are tested, the overall average becomes tighter. Game choices obviously play a significant role which is why I typically don't rely on a single source/website for a conclusion. Two different outlets can get two different results and both still be correct when you don't have infinite time to test every game and every scenario in each game.

Also, Jarrod's Tech and Tech Deals are not random youtubers. They both have ~half a million subscribers and have been making videos for years. You have yet to present any reason or evidence why the results from these other outlets can't be trusted or are invalid. It seems no matter what evidence is presented that Zen4 is pretty close to RPL in gaming, you and carfax will just dismiss it with no real reasoning other than it doesn't look right to you because the 13900k isn't winning by enough. The funny thing is that no one is even arguing that the 13900k isn't the fastest gaming CPU currently. It's just not 100% faster under heavy CPU usage as carfax has claimed and the results when both are using overclocked memory are a lot closer than some apparently want to admit when they are only willing to point to sites using JEDEC memory.

For the 13900k scaling with faster memory, just look at the graph you posted. It shows less than 1% gain for the 13900k going from DDR5-6000 memory to DDR5-7400 memory. Eurogamer's own results basically show the same on average. Faster memory does not help RPL, at least not in general.

 

[Hitman928]

Oh, and I just gave this one a second look. They literally use worse memory on the Intel system than the AMD system. If you're going to defend using >stock configs, which I'm fine with, then how on earth would you defend that choice? Are you going to tell me you think Raphael's memory support is better than Raptor Lake's?

They didn't use worse memory, they used the same memory for both systems. The AMD system was just stable at slightly tighter timings. The difference is not that significant and certainly does not invalidate their results. I would also argue that it's more relevant and useful than results where only JEDEC standard memory was used.

Edit: maybe you're thinking the Intel system was using the DDR4 memory, but it wasn't, it was using the same DDR5 memory as the Zen4 system. The DDR4 memory shown was for the Zen 3 system.

 

[Exist50]

Yes, in TPU's original review with more limited games, the Zen architectures were a little further behind, but when they did their 53 game suite, the 5800x3d went from 13% behind to only 6.2% behind.

So then where are you seeing the same results for Zen 4? Or are you just assuming that benefit is inherent to AMD, instead of Zen3 X3D?

You have yet to present any reason or evidence why the results from these other outlets can't be trusted or are invalid.

I consider demonstrating a consistent bias in coverage and results radically different from everyone else to be reasons enough by themselves. I've noticed no one's even tried to defend anything other than their benchmark results.

It seems no matter what evidence is presented that Zen4 is pretty close to RPL in gaming, you and carfax will just dismiss it with no real reasoning other than it doesn't look right to you because the 13900k isn't winning by enough.

I've given reasoning in the past few comments alone. But you know what, let's say all of those results are perfectly valid. Why are you only using those sources, or even particular game tests, rather than the one's I've presented? And I supposed to believe it's just a coincidence that they all happen to be outliers in AMD's favor?

 

[Exist50]

They didn't use worse memory, they used the same memory for both systems. The AMD system was just stable at slightly tighter timings.

A distinction without a difference. Again, are you going to seriously claim to me that you think it's a fair comparison to run the AMD system with strictly better memory than the Intel one, despite Intel generally supporting better settings? Is that representative of what real users will face? I don't think so.

I have no issue with comparing beyond JEDEC timings, including AMD's recommended DDR5-6000 for Raphael. But you can't push one system to the limit and then barely bump the other beyond JEDEC. Get a half-decent set of Hynix M-die, and see what each can do.

 

[Hitman928]

So then where are you seeing the same results for Zen 4? Or are you just assuming that benefit is inherent to AMD, instead of Zen3 X3D?

Again, I included that result to show that HWUB isn't just some pro AMD outlier when TPU's testing shows better relative performance for the 5800x3d when running their full test suite. Maybe Zen4 wouldn't improve relative to RPL in their full suite like Zen3d does, but that wasn't my point.

I consider demonstrating a consistent bias in coverage and results radically different from everyone else to be reasons enough by themselves. I've noticed no one's even tried to defend anything other than their benchmark results.

What consistent bias in coverage and results has been shown? The only thing I've seen pointed to is that you and carfax think their Zen4 results look better than they should.

I've given reasoning in the past few comments alone. But you know what, let's say all of those results are perfectly valid. Why are you only using those sources, or even particular game tests, rather than the one's I've presented?

What reasons have you given other than they are different than others you think look correct for some unknown reason? Seriously, list them out because I must have missed them.

And I supposed to believe it's just a coincidence that they all happen to be outliers in AMD's favor?

You do realize that this question is basically self-defeating, right?

 

[lopri]

Sticking to JEDEC standard might disadvantage Raptors, not Raphaels. Raphaels are bottlenecked by Infinity Fabric.

 

[Exist50]

Again, I included that result to show that HWUB isn't just some pro AMD outlier when TPU's testing shows better relative performance for the 5800x3d when running their full test suite.

You claimed it was merely memory settings that explained the gap, yet the TPU data I linked, from a source you explicitly cited as having acceptable methodology, shows an incredible ~10% difference vs HWUB, similar to many other reviewers. That makes them an outlier by definition.

What consistent bias in coverage and results has been shown?

I linked a very recent example of a price comparison above, for instance. Or stuff like their stance on ray tracing, DDR5, etc. Seriously, are you new to them?

What reasons have you given other than they are different than others you think look correct for some unknown reason?

So you're going to not only insist that all of these results are perfectly representative, but that they're also better than other outlets? Why else would you be quoting them almost exclusively vs more established outlets?

 

[Hitman928]

A distinction without a difference. Again, are you going to seriously claim to me that you think it's a fair comparison to run the AMD system with strictly better memory than the Intel one, despite Intel generally supporting better settings? Is that representative of what real users will face? I don't think so.

I have no issue with comparing beyond JEDEC timings, including AMD's recommended DDR5-6000 for Raphael. But you can't push one system to the limit and then barely bump the other beyond JEDEC. Get a half-decent set of Hynix M-die, and see what each can do.

Again, it's not strictly better memory, it is the exact same memory. The AMD system is more stable at tighter timings with this memory. There's nothing unusual or invalidating about this. You do realize, any time XMP memory was used on both AMD and Intel systems, which has been done for years, the same thing was happening in reverse, right? Would it be better to use specific memory for Intel and AMD systems? Yes. But that still doesn't make their comparison invalid at all and the results will be much more representative of an actual end user experience for both system than JEDEC standard memory. If we're only go by either JEDEC standard memory or both have to be pushed to the edge, then no one who has gone past JEDEC standard memory has done it right, according to you as AMD can go above DDR5-6000 memory with positive results but no one ever seems to test this way.

If you want to feel better and give the Intel results a 1% increase in FPS , go right ahead.

 

[Hitman928]

You claimed it was merely memory settings that explained the gap, yet the TPU data I linked, from a source you explicitly cited as having acceptable methodology, shows an incredible ~10% difference vs HWUB, similar to many other reviewers. That makes them an outlier by definition.

That's only for 1 particular SKU. TPU actually has the 13900k as 8.6% faster than a 7950x at 1080p. HWUB has the 13900k as 6.4% faster at 1080p. Pretty close results I'd say. Why does the 13600k/7600x comparison have a bigger difference? Not sure, you'd have to look specifically at the games involved. Yet again, HWUB is not the only one showing the 7600x as leading the 13600k though, multiple other outlets show the same thing.

I linked a very recent example of a price comparison above, for instance. Or stuff like their stance on ray tracing, DDR5, etc. Seriously, are you new to them?

What is their stance on ray tracing, DDR5, etc. that you think is biased? Seriously, specifics are needed if you are going to make accusations. I must have missed the price comparison you didn't like and a quick search for it didn't yield any recent results. What was this price comparison that showed their bias?

So you're going to not only insist that all of these results are perfectly representative, but that they're also better than other outlets? Why else would you be quoting them almost exclusively vs more established outlets?

Where did I say they are better? I specifically chose these outlets because they don't just stick to JEDEC speed memory and include a decent to large amount of games in their tests. I never once tried to invalidate or argue against computerbase's or pcgh's results, all I said was that they aren't representative of what the vast, if not basically all, PC gamers will get because they only use JEDEC speed memory. You and carfax were the ones who are saying HWUB's results couldn't be trusted. Then I showed results from several outlets who don't stick to JEDEC memory and tested a decent amount of games that overall agree pretty well with HWUB's results but for some reason, none of them count because reasons.

Which, "more established outlets," would you like me to quote that haven't been quoted already?

 

[Hitman928]

I just checked Gamer's Nexus (one of the biggest tech youtubers out there in case there was a feeling they weren't well enough established), and they have the 13900k leading the 7950x by 6% on average at 1080p, even less of a gap than HWUB. Granted, they only tested 7 games which is why I didn't include them originally, but it's another result that seems to show HWUB's results aren't some crazy biased outlier as has been suggested.

 

[Exist50]

Again, it's not strictly better memory, it is the exact same memory.

Tighter timings = better settings/memory. I shouldn't have to point out something so obvious. If you think 6000 CL32 is where Raptor Lake maxes out, you frankly shouldn't be commenting on testing methodology.

It's incredible that you continue to defend such blatantly flawed setups, and it demonstrates my point perfectly. These gymnastics remind me of the excuses people used to use for userbenchmarks.

If we're only go by either JEDEC standard memory or both have to be pushed to the edge, then no one who has gone past JEDEC standard memory has done it right, according to you as AMD can go above DDR5-6000 memory with positive results but no one ever seems to test this way.

For the majority of Raphael chips, you can't go past 6000 in gear 2 (i.e. with a benefit), but Raptor Lake can do that easily. Even HWUB doesn't limit their system that low.

 

[Exist50]

That's only for 1 particular SKU

It's not the only matchup with such gaps.

What is their stance on ray tracing, DDR5, etc. that you think is biased?

They basically have said ray tracing doesn't matter to justify excluding it from GPU coverage, claimed DDR5 didn't matter up until Raphael required it (and yet left DDR4 out of the price comparison I linked), etc.

If you're going to defend them, at least be familiar with their behavior.

Where did I say they are better? I specifically chose these outlets because they don't just stick to JEDEC speed memory and include a decent to large amount of games in their tests.

And yet you ignored TPU's Zen 4 results entirely, and quote that hilarious TechGage graph with crippled RPL. What is that if not cherry picking?

 

[Hitman928]

Tighter timings = better settings/memory. I shouldn't have to point out something so obvious.

Using slower memory and using the tightest timings each CPU can get with the same memory isn't the same thing. I shouldn't have to point out something so obvious.

If you think 6000 CL32 is where Raptor Lake maxes out, you frankly shouldn't be commenting on testing methodology.

Never claimed any such thing. You tend to try to put words in my mouth to discredit my arguments. Please stop.

It's incredible that you continue to defend such blatantly flawed setups, and it demonstrates my point perfectly. These gymnastics remind me of the excuses people used to use for userbenchmarks.

Claiming the same thing over and over without evidence doesn't make something true. I've asked for evidence and concrete examples multiple times and all you do is repeat the original claims.

For the majority of Raphael chips, you can't go past 6000 in gear 2 (i.e. with a benefit), but Raptor Lake can do that easily. Even HWUB doesn't limit their system that low.

What data do you have on Zen4 chips to make this statement?