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.

 

[TheELF]

Tell me about: Intel Raptor Lake cpu eco-mode.

How do you go about setting up this mode? Bios? What temperature and wattage specs? % of reduction performance, wattage from normal mode. Thx!

Core i9-13900K, i7-13700K & i5-13600K: Gaming-Könige im Test: Leistungsaufnahme und Effizienz

Intel Raptor Lake im Test: Leistungsaufnahme und Effizienz / Leistungsaufnahme in Anwendungen ab Werk / Der Verbrauch steigt

www.computerbase.de

With a pretty decent amount of productivity apps, if you lock it (the 13900k) to 142W you use less than half the power while loosing like 15% of performance, if you use productivity apps in the first place...
It also shows the power usage in games if the power limit is lifted (unlimited power)

If you are a gamer you can go all the way down to 88W without loosing any noticeable amount of performance, and that's at 720p so at higher resolutions there will be basically no difference at all.

Core i9-13900K, i7-13700K & i5-13600K: Gaming-Könige im Test: Benchmarks in Games

Intel Raptor Lake im Test: Benchmarks in Games / Leistung in Spielen (720p, RTX 3090 Ti) / CPU-Gaming-Leistung im absoluten CPU-Limit

www.computerbase.de

 

[reb0rn]

delete

 

[Asterox]

GN review of the 13600K. Looking very good.

Lol, Gamer Nexus but only four games were tested.

- Steve talks/rants way too much, even the Klingons can see that

A very good example of how it's done, and it tests as many as twenty-five games along the way.

 

[blckgrffn]

So if you set the power limit to 250W, you still get like 98% of the performance do you not? So the question I have to ask is why this is not the default for motherboards? Doesn't seem much point blasting heat throughout your system for very small gains.

You'd think that it would be better for the motherboards too, since there has to be some weaknesses in some of these boards that are going to be a failure point when an unruly CPU gobbles up power and your AIO or tower cooler isn't really cooling the power delivery bits well.

 

[Markfw]

So if you set the power limit to 250W, you still get like 98% of the performance do you not? So the question I have to ask is why this is not the default for motherboards? Doesn't seem much point blasting heat throughout your system for very small gains.

Or lower. I set my 7950x's down to ~142 watts, and they still are WAY faster than my stock 5950x's. If Raptor lake was that good I would consider them.

 

[Markfw]

Lol, Gamer Nexus but only four games were tested.

- Steve talks/rants way too much, even the Klingons can see that

A very good example of how it's done, and it tests as many as twenty-five games along the way.

I LOVE that second video, but would like to see the 13900k and the 7950x compared. In the bottom end one, AMD won gaming by ~6% but Intel won productivity by ~20% due to more cores. (12 threads compared to 20) But on the top end, the thread count is the same, so it would be interesting. But even at the bottom end the Intel took a lot more power.

 

[controlflow]

Core i9-13900K, i7-13700K & i5-13600K: Gaming-Könige im Test: Benchmarks in Games

Intel Raptor Lake im Test: Benchmarks in Games / Leistung in Spielen (720p, RTX 3090 Ti) / CPU-Gaming-Leistung im absoluten CPU-Limit

www.computerbase.de

Computerbase only used a RTX 3090 Ti here but Raptor Lake is dominating pretty well at gaming in their tests.

Generally a 10-20% lead over the 5800X3D and 7950x with bigger leads in RT titles. If you look at 1% lows the differences are even bigger.

When considering this alongside the data from derbauer:

The 13900k is capable of being the top performer in gaming and mixed workloads while matching Zen 4 efficiency or beating it. You can pretty much set the 13900K to 88-90W and keep the top of the chart gaming experience.

Given the cheaper cost of the CPU and the cheaper platform costs, for gaming and mixed workloads it really is making little sense to buy a Zen 4 over RPL. This is especially true for the lower end Zen 4s.

Only the 7950x makes any sense and the only use case for where this applies is for people that run 100% load all the time and especially AVX workloads. Even in this case the 13900k isn't actually too bad but the 7950x is clearly more performant and more efficient here.

Intel Core i9-13900K vs. AMD Ryzen 9 7950X at 125W and 65W | Club386

Prefer to keep temperature and power consumption down to lower levels this winter? Here's what happens when the best CPUs are scaled back.

www.club386.com

There is a lot of hyperbole about the 13th gen being horrifically inefficient, power hungry, hot and burning down your home but I think the objective data shows it has very reasonable efficiency at more sane power limits while keeping almost all of the performance.

 

[aigomorla]

A very good example of how it's done, and it tests as many as twenty-five games along the way.

Why would you need to test so many games unless you have that many variances in graphic engines to test from?
Its to me honestly wasted time.

You would need to select the most commonly used engines and test performance that way, as each game can change by slight texture or optimizations the developers put into it, which would again scale if the same applied on the same engine.

And there are not 25 different engines.

And yes i agree with you, sometime steve just goes on and on and on.... but i trust his test more then anyone's as he has documented all the variances and even the test room is fed constant ambient air temp thoughout the entire test, and not run a bedroom but a real testing designated station.

Also Steve did a very nice comparision on what you'll pay on average for both platforms.
After performance numbers everyone looks at price tag...



You don't see him comparing expensive Z and X boards when it comes to a 13600k.
Well, i don't see how or way a 5800X3D user would not use a X570, but i guess he tried to keep that as cheap as possible.

 

[A///]

Part of that equation is the driver. An AMD card that is around the same performance as a 4090 won't have exactly the same CPU bottlenecks. I'm sure Nvidia will be very focused on lowering driver overhead over the next couple years, although it might be less of an issue for them after Zen4 V-cache is out and reviewers switch over.

@A/// No it happened a few times in the past. Also the driver overhead is relevant at high fps. At low fps it's limited by the GPU and doesn't play a big role.

Resolutions, graphics detail, and refresh rate(to a lesser extent) will continue to increase. Continuing the increasing 3D graphics and CPU requirements.

Right. Guys you're going to have to throw me a bone here with dates or models of cards. My memory isn't exactly what it used to be. I'm aware of NVidia's plans as it was laid out once or twice, but as it stands the launch drivers and the new 522 drivers are causing these issues. I've got no skin in the game because the 4090 is way too expensive for my tastes and I game less than a few hours a month. I'd read the 7950X was causing bottlenecks but I expected the 13900K to deliver. Though my original post was more along the lines of if you could give the 4090 what it wants, how long can it stretch its legs? I don't know much about this particular subject because video cards have always been a unicorn mystery to me as long as I can remember. The whole GPU vs CPU bound resolution and it being separate from individual games is a daze to me.

 

[Markfw]

Core i9-13900K, i7-13700K & i5-13600K: Gaming-Könige im Test: Benchmarks in Games

Intel Raptor Lake im Test: Benchmarks in Games / Leistung in Spielen (720p, RTX 3090 Ti) / CPU-Gaming-Leistung im absoluten CPU-Limit

www.computerbase.de

Computerbase only used a RTX 3090 Ti here but Raptor Lake is dominating pretty well at gaming in their tests.

Generally a 10-20% lead over the 5800X3D and 7950x with bigger leads in RT titles. If you look at 1% lows the differences are even bigger.

When considering this alongside the data from derbauer:

The 13900k is capable of being the top performer in gaming and mixed workloads while matching Zen 4 efficiency or beating it. You can pretty much set the 13900K to 88-90W and keep the top of the chart gaming experience.

Given the cheaper cost of the CPU and the cheaper platform costs, for gaming and mixed workloads it really is making little sense to buy a Zen 4 over RPL. This is especially true for the lower end Zen 4s.

Only the 7950x makes any sense and the only use case for where this applies is for people that run 100% load all the time and especially AVX workloads. Even in this case the 13900k isn't actually too bad but the 7950x is clearly more performant and more efficient here.

Intel Core i9-13900K vs. AMD Ryzen 9 7950X at 125W and 65W | Club386

Prefer to keep temperature and power consumption down to lower levels this winter? Here's what happens when the best CPUs are scaled back.

www.club386.com

View attachment 69594
View attachment 69595


There is a lot of hyperbole about the 13th gen being horrifically inefficient, power hungry, hot and burning down your home but I think the objective data shows it has very reasonable efficiency at more sane power limits while keeping almost all of the performance.

Look at the multicore test. The more power you remove the worse it gets beaten by. And even at say 150 watt, it loses. And thats just cinebench, which it does well on. On some app that uses all the cores and runs for a long time its gets hotter and throttles. I agree that at say 150 watts its not too bad, but its strengths are really gaming, and that with the power set down. Its only has 8 strong cores. And the test I mentioned above with 25 games ? the 7600x wins, by more than chance of error on all 3 resolutions.

Its not bad, but unless you take the power down (which most people either won't or don't know how to do), it is not as good as Zen 4 IMO.

 

[A///]

Well, i don't see how or way a 5800X3D user would not use a X570, but i guess he tried to keep that as cheap as possible.

Some of the B450 boards are phenomenal and don't need to be bumped up to the B550 chipset or X570 for that matter. If going for pure budget gaming, it makes a lot more sense than a big board with a lot of functionality that won't be useful to the end user. To make use of that functionality you need to spend more money.

 

[Hans Gruber]

Intel said they were going to leave AMD in the rearview mirror. Right now it looks like they have a car length lead over AMD. Not enough to put them in the rearview mirror. Intel uses more power, not as efficient as AMD. For whatever reason, AMD motherboards have been jacked up in price over the years. Now you can see why I said AMD needed Zen 4 released at the beginning of 2022.

Whenever AMD releases their 3D v-cache CPU's. They will be the gaming king once again for this current generation of silicon. I do not think it will matter by then because people will be waiting for the next generation hardware or already bought their new hardware. AMD has made several key missteps in their battle against Intel in the past year.

Most people who take an interest in the AMD vs Intel battle are waiting for the Intel 4 (7nm) silicon. I think that is what AMD needs to be worried about most right now. Intel has said they are accelerating their release timeline of new CPU architecture to regain their title as #1.

AMD has not done anything wrong other than letting their product cycle release dates slide by up to a year. AMD is still a winner with power efficiency with the 7950x. It's the lower tier chips that are not quite the value (7600x/7700x they once were. It would have served AMD well to release Zen4 with the 3d-vache included in the CPU's on launch. That way it would make AMD the clear winner in their battle against Intel. There is currently no answer from AMD for the 13600k.

 

[Kocicak]

I got a 13700K, so far it seems it can work with two best cores at 57x and other 6 at 55x.

Here is the comparison with 13900K: draw your own conclusions...

BTW 13700K seems to be a little harder to cool, I suspect that Intel may be using some higher quality TIM in the 13900K. It could be only a sample variance, or my cooler is not optimally mounted now, I do not know...

 

[A///]

Reseat your cooler a few times. I had issues with my 10th gen with the way I was placing the cold plate. Test after each reseat. With me I found I was likely pushing down on one corner earlier than the other three causing bad thermal paste distribution.

 

[Thunder 57]

Intel said they were going to leave AMD in the rearview mirror. Right now it looks like they have a car length lead over AMD. Not enough to put them in the rearview mirror. Intel uses more power, not as efficient as AMD. For whatever reason, AMD motherboards have been jacked up in price over the years. Now you can see why I said AMD needed Zen 4 released at the beginning of 2022.

Whenever AMD releases their 3D v-cache CPU's. They will be the gaming king once again for this current generation of silicon. I do not think it will matter by then because people will be waiting for the next generation hardware or already bought their new hardware. AMD has made several key missteps in their battle against Intel in the past year.

Most people who take an interest in the AMD vs Intel battle are waiting for the Intel 4 (7nm) silicon. I think that is what AMD needs to be worried about most right now. Intel has said they are accelerating their release timeline of new CPU architecture to regain their title as #1.

AMD has not done anything wrong other than letting their product cycle release dates slide by up to a year. AMD is still a winner with power efficiency with the 7950x. It's the lower tier chips that are not quite the value (7600x/7700x they once were. It would have served AMD well to release Zen4 with the 3d-vache included in the CPU's on launch. That way it would make AMD the clear winner in their battle against Intel. There is currently no answer from AMD for the 13600k.

Sigh... You cannot release something that is not ready. Why don't they just go ahead and release Zen 5 at CES?

 

[Markfw]

Intel said they were going to leave AMD in the rearview mirror. Right now it looks like they have a car length lead over AMD. Not enough to put them in the rearview mirror. Intel uses more power, not as efficient as AMD. For whatever reason, AMD motherboards have been jacked up in price over the years. Now you can see why I said AMD needed Zen 4 released at the beginning of 2022.

Whenever AMD releases their 3D v-cache CPU's. They will be the gaming king once again for this current generation of silicon. I do not think it will matter by then because people will be waiting for the next generation hardware or already bought their new hardware. AMD has made several key missteps in their battle against Intel in the past year.

Most people who take an interest in the AMD vs Intel battle are waiting for the Intel 4 (7nm) silicon. I think that is what AMD needs to be worried about most right now. Intel has said they are accelerating their release timeline of new CPU architecture to regain their title as #1.

AMD has not done anything wrong other than letting their product cycle release dates slide by up to a year. AMD is still a winner with power efficiency with the 7950x. It's the lower tier chips that are not quite the value (7600x/7700x they once were. It would have served AMD well to release Zen4 with the 3d-vache included in the CPU's on launch. That way it would make AMD the clear winner in their battle against Intel. There is currently no answer from AMD for the 13600k.

I beg to differ. I have seen at least one video that had 25 games at 3 resolutions (in this thread, not posted by me) that had the 7600x a little over 6% ahead of the 13600k and in MANY of the benchmarks I have seen heavily threaded benchmarks with AMD in the win column. Yes, the Raptor lake is very strong in lightly threaded applications, but overall I say AMD has the advantage AND in efficiency. Even if you manually set the wattage down on both to the same level. And at the lower levels, AMD pulls farther ahead.

 

[reb0rn]

and all other reviews show different picture so you bet that one unknown youtube review is right?
lol so not based

 

[Hulk]

Or lower. I set my 7950x's down to ~142 watts, and they still are WAY faster than my stock 5950x's. If Raptor lake was that good I would consider them.

I know you are open minded regarding what is best/most efficient for all of the DC you do but we both know currently nothing can touch the 7950X for DC in terms of performance and efficiency! I'm kind of an Intel fan but Zen 4 is crazy efficient. As you noted at 142W they are still super fast.

What you need is a 7990X with 64 Zen 4 cores!

 

[Hulk]

Core i9-13900K, i7-13700K & i5-13600K: Gaming-Könige im Test: Benchmarks in Games

Intel Raptor Lake im Test: Benchmarks in Games / Leistung in Spielen (720p, RTX 3090 Ti) / CPU-Gaming-Leistung im absoluten CPU-Limit

www.computerbase.de

Computerbase only used a RTX 3090 Ti here but Raptor Lake is dominating pretty well at gaming in their tests.

Generally a 10-20% lead over the 5800X3D and 7950x with bigger leads in RT titles. If you look at 1% lows the differences are even bigger.

When considering this alongside the data from derbauer:

The 13900k is capable of being the top performer in gaming and mixed workloads while matching Zen 4 efficiency or beating it. You can pretty much set the 13900K to 88-90W and keep the top of the chart gaming experience.

Given the cheaper cost of the CPU and the cheaper platform costs, for gaming and mixed workloads it really is making little sense to buy a Zen 4 over RPL. This is especially true for the lower end Zen 4s.

Only the 7950x makes any sense and the only use case for where this applies is for people that run 100% load all the time and especially AVX workloads. Even in this case the 13900k isn't actually too bad but the 7950x is clearly more performant and more efficient here.

Intel Core i9-13900K vs. AMD Ryzen 9 7950X at 125W and 65W | Club386

Prefer to keep temperature and power consumption down to lower levels this winter? Here's what happens when the best CPUs are scaled back.

www.club386.com

View attachment 69594
View attachment 69595


There is a lot of hyperbole about the 13th gen being horrifically inefficient, power hungry, hot and burning down your home but I think the objective data shows it has very reasonable efficiency at more sane power limits while keeping almost all of the performance.

Granted Raptor holds up pretty well here but it's a little misleading for the "stock" power levels not to be included on the graph. If you don't know better you would assume AMD stock = Intel stock, which I'm pretty sure isn't the case right?

 

[Hitman928]

Granted Raptor holds up pretty well here but it's a little misleading for the "stock" power levels not to be included on the graph. If you don't know better you would assume AMD stock = Intel stock, which I'm pretty sure isn't the case right?

Computerbase sticks to JEDEC standard memory for their tests so RPL has a bit of a memory speed advantage with DDR5-5600 where the Zen 4 CPUs are stuck to DDR5-5200. Speculation, but based upon multiple other review sites, if they retested with both given faster supported memory, the average gaming difference would probably shrink a bit and reflect what other sites are showing.

 

[Markfw]

I know you are open minded regarding what is best/most efficient for all of the DC you do but we both know currently nothing can touch the 7950X for DC in terms of performance and efficiency! I'm kind of an Intel fan but Zen 4 is crazy efficient. As you noted at 142W they are still super fast.

What you need is a 7990X with 64 Zen 4 cores!

I have 5 Rome EPYC 64 core cpus, and 2 Milan EPYC 64 core CPU's, so I am covered there. Thanks for your feedback. I know Zen 4 is faster then Milan, so I had to add some.

Now back to Raptor lake. As for the comments on memory speeds, I have to agree that BOTH should be tested at what is reasonable to each platform. At least 6400 cl34 for RL and 6000 CL30 (per AMD) for Zen 4 . They are not insanely expensive, and I own both Zen 4 and Alderlake, so I know how important memory speed is to benchmarks. Jedec is fine , but not realistic when CPU enthusiast are who ready these benchmarks. EXPO for AMD and XMP for Intel are trusted standards that the companies support.

 

[aigomorla]

What you need is a 7990X with 64 Zen 4 cores!

AMD pricing would put that at over 6000 dollars.... seeing how much a 64core threadripper pro cost.
Can't think of how many people would drop a 6000 dollar cpu.

Even if you manually set the wattage down on both to the same level. And at the lower levels, AMD pulls farther ahead.

Mark i completely agree with you on the concept that the 7950X is a great multithreaded cpu with better efficiency, but gamers do not care about this word called efficiency. To a gamer, overclocking, and pushing that envelope is more important then how efficient the cpu is.
This is why overclockers for the longest time, throw LN2, or try to bring sub ambient into the equation and completely break that 4th wall we call efficient.

You need to think this from a gamers perspective.
Whatever comes first, even tho its strapped to Nitros, and has a MPG of 1, as long as it comes first, is the winner.
This is what intel has done.
They unloaded all the stop gaps, pulled all the restrictions, pulled us back to prescott era, where they gave us something that can do it, as long as we can try to keep it in check.

Completely wrong in my book, but to a gamer, its completely normal.

 

[Hulk]

This is kind of strange. Based on Kocicak's results I calculated that in Cinebench R23 for ST there is no IPC improvement from Golden Cove to Raptor Cove. But for MT Raptor Cove has 4.7% better IPC than Golden Cove.

Gracemont, looking at clusters, is 5.3% better IPC in Raptor as compared to Alder.

I'm thinking that the additional L2 when Raptor is doing HT and for a cluster of Gracemonts is helpful in CB R23 but for ST the additional L2 is not needed?

Based on the 8+8 and 8+16 Raptor scores equations can be simultaneously solved to isolate core throughput/GHz. That is how I acquired the data and here is the data that will calculate Raptor and Alder CB R23 scores based on clocks.

530	Cinebench R23 points per 1GHz for a Raptor Cove core with HT
389	Cinebench R23 points per 1GHz for a Raptor Cove core without HT
255	Cinebench R23 points per 1GHz for a Raptor Lake Gracemont core
506	Cinebench R23 points per 1GHz for a Golden Cove core with HT
389	Cinebench R23 points per 1GHz for a Golden Cove core without HT
242	Cinebench R23 points per 1GHz for a Alder Lake Gracemont core

 

[Markfw]

AMD pricing would put that at over 6000 dollars.... seeing how much a 64core threadripper pro cost.
Can't think of how many people would drop a 6000 dollar cpu.



Mark i completely agree with you on the concept that the 7950X is a great multithreaded cpu with better efficiency, but gamers do not care about this word called efficiency. To a gamer, overclocking, and pushing that envelope is more important then how efficient the cpu is.
This is why overclockers for the longest time, throw LN2, or try to bring sub ambient into the equation and completely break that 4th wall we call efficient.

You need to think this from a gamers perspective.
Whatever comes first, even tho its strapped to Nitros, and has a MPG of 1, as long as it comes first, is the winner.
This is what intel has done.
They unloaded all the stop gaps, pulled all the restrictions, pulled us back to prescott era, where they gave us something that can do it, as long as we can try to keep it in check.

Completely wrong in my book, but to a gamer, its completely normal.

To someone with a $900 a month electric bill power matters a lot. Just giving my feedback, and also in light of Kociaks problems, normal people may have issues if they are not aware of the problems related to high power usage.

 

[Exist50]

This is kind of strange. Based on Kocicak's results I calculated that in Cinebench R23 for ST there is no IPC improvement from Golden Cove to Raptor Cove. But for MT Raptor Cove has 4.7% better IPC than Golden Cove.

Gracemont, looking at clusters, is 5.3% better IPC in Raptor as compared to Alder.

I'm thinking that the additional L2 when Raptor is doing HT and for a cluster of Gracemonts is helpful in CB R23 but for ST the additional L2 is not needed?

Makes sense. Even if a single thread misses the L2, if that's all you're really running, I imagine it has pretty good odds of hitting in the L3. But in multithreaded, you've got a lot more pressure on the cache, and that spills over into pressure on the fabric memory subsystem, which are also improved.