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.

 

[R81Z3N1]

I also have an Alder Lake CPU 12600kf, I know this a a raptor lake thread but was commenting on Win 11 Scheduler. This is a fresh install of Win 11 pro, and I can say that the E-cores create micro stutter in some games, if you have back ground threads running.

In my research about the advantages of E-cores I really did not take into account how I actually use my computer. My computer use is up 99% of the time I crunch with Boinc when not gaming. I leave it on when gaming if possible, but I ran into problems with micro stutter. Depending on game I have to suspend Boinc just to avoid it.

I might game for 4-5 hours, so it feels bad that to enjoy my game I have to suspend an application that is supposed to work regardless of the work load because of E-cores.

I also ran into the problem that these E-core's are not optimal for virtualization. sad thing about it is if you search you really don't hear much about benchmarks of any 12th, or 13th gen intel performance guides for virtualization.

I am old school, and might try vmware, which also lack documentation on performance of E-cores as it relates VM's. What I used to do, is run vmware, use that as my desktop and enjoy, and then pop into windows for gaming, and other tasks.

 

[coercitiv]

This is a fresh install of Win 11 pro, and I can say that the E-cores create micro stutter in some games, if you have back ground threads running.

In my research about the advantages of E-cores I really did not take into account how I actually use my computer. My computer use is up 99% of the time I crunch with Boinc when not gaming. I leave it on when gaming if possible, but I ran into problems with micro stutter. Depending on game I have to suspend Boinc just to avoid it.

Before buying the new hybrid CPU, did you leave Boinc running while gaming and never observed micro stutter? It's hard to pinpoint this on the scheduler when a compute intensive workload can sabotage gaming performance by simply trashing the L3 or using too much RAM relative to the needs of the game.

For better or worse, the Win 11 Scheduler is optimized for this scenario, in the sense that it keeps P cores free and available exclusively to the foreground application. Boinc should be pushed to the E cores and isolated there. You can test scheduler influence by disabling the E cores and checking performance again in the same scenario.

 

[R81Z3N1]

I am using Linux for the most part, and system does have 32GB of memory, for the most part games just work and only experience problems in one or two games.

At first I thought it was GPU related but did not notice much change when using lower settings.

For the most part I would use Steam on Linux, but I have a few games that use battle.net, and Ubisoft version they say they can ban you due to Linux install. Now with steam deck that might not be the case but, I already have two many computers as it is so just using one dedicated to windows is not really a big deal. I just know not to depend on it as much.

 

[CropDuster]

correct but windows 11 performance seems to be half baked a year and a half later. windows 12 should be the golden ticket.

And the checks in the mail !

Yes, win 11 was supposed to work, but so far no. And win 12 ? who knows. I will stick with tried and true (Zen 4) for now. It is the fastest in everything.

I've had no issues with it on my 12700k *shrug* I mainly play DCS in VR these days with discord, live web browsers, etc open in the background. Latest DCS beta can scale to 32 threads

 

[amenx]

Re e-cores, so far not noticed any issues on 13600k in both win 10 and 11 (dual boot). Found a TPU test of 53 games with e-cores disabled and enabled in Win 10. Apologies if posted as only recently joined this thread and its a bit big to search through.

RTX 4090 & 53 Games: Core i9-13900K E-Cores Enabled vs Disabled Review

In this week's TPU50 Megabench we're testing whether you can unlock additional gaming performance by disabling the E-Cores on a Raptor Lake Core i9-13900K processor. For our benchmarks we used the mighty GeForce RTX 4090, with 53 games at three resolutions.

www.techpowerup.com

 

[Hulk]

I find Windows 11 to generally get the thread assignments correct for most applications but every now and then it will go to an HT logical core or an E core while a physical P still seems available. I write still because often times the threads are moving from core-to-core so quickly it's hard to identify what is going on exactly with thread allocation.

 

[A///]

Actually @Hulk that reminds me of the e core behavior I wrote off months ago is coming back to be a nice blessing. e core engagement on window minimisal is wicked good on windows. IDK about Linux. It makes more sense to me now than it did in the past. I couldn't fall asleep last night and spent most of the night in bed thinking about 12th gen and 13th gen and just how intel came up with the concept. It makes sense. Minimize to encode large videos and play a game at the same time without dealing with any issues, just like you would on a *950x processor. The drawback being that Intel's power usage is higher than AMd's for now.

 

[igor_kavinski]

I couldn't fall asleep last night and spent most of the night in bed thinking about 12th gen and 13th gen and just how intel came up with the concept.

You need better fodder for your thoughts

How Intel came up with the concept? Thank AMD. A world without Zen 3 16C/32T would not have pushed Intel to go this route. Mobile maybe but not on desktop. Why bother putting so much effort into making heterogeneous cores work well together when it's far easier to just have homogeneous cores? Intel would happily be selling us max 10 P-core only CPUs if it weren't for AMD. They could have done the heterogeneous core thingy with just P-cores too. Have one set of P-cores optimized for hitting high frequencies while the other set is the mobile version optimized for power efficiency. That way, they could have both AVX-512 and HT enabled.

 

[A///]

You need better fodder for your thoughts

How Intel came up with the concept? Thank AMD. A world without Zen 3 16C/32T would not have pushed Intel to go this route. Mobile maybe but not on desktop. Why bother putting so much effort into making heterogeneous cores work well together when it's far easier to just have homogeneous cores? Intel would happily be selling us max 10 P-core only CPUs if it weren't for AMD. They could have done the heterogeneous core thingy with just P-cores too. Have one set of P-cores optimized for hitting high frequencies while the other set is the mobile version optimized for power efficiency. That way, they could have both AVX-512 and HT enabled.

Intel doesn't need to thank AMD. Heterogeneous cores weren't a new concept in 2021 and haven't been for many years. Most of us old fogies will remember reading in magazines back in the era before the internet and computers being remotely commonplace and playboys being available up for snatch and runs from news stands agents that some company, can't remember what but it rhymed with cyst had developed a heterogeneous design. It meant nothing to me then and I'd only just gotten into computers around that time. I briefly posted a while back how my interest in the damn things began thanks to papa. Now AMD pushed Intel into increasing core counts when it wasn't viable to them. Intel was happy as a clam bake at a Connecticut sea market to not increase core counts because they at the time were struggling with 10nm, furthermore intel possibly had been exploring mass production of chiplet but knew the risks. the lack of a good interconnect meant they may be heavily set back. AMD offering high core counts for cheap for 1st gen put a fire under intel's tushtush. Going on AMD knew if they kept core counts high and increasing each generation it would force intel's hand to do something because amd wasn't stupid, they knew the ever increasing core counts would be a financial folley for intel on either 14nm or 10 nm due to defect rates. Even if 14nm was pristine it wasn't easy for intel to produce such a big die. Remember something like the 9900k or 10900k used to be high end hedt or low end xeons in pure core count. You could implement everything else on the chip but the core and associated cache was the big blunder that could mess them up ifinancially. Intel likely began alderlake research around 2015 when AMD announced Ryzen at hot chips much to the chagrin and eyes rolls of industry titans because hey amd had been sucking its dang thumb for years now back then. You p-core processor example is just a regular 10 core intel processor. No one really knows when or how Intel is going to do something and it's speculation even on my part, but I don't believe AMD were the reason for their shift. Intel knew the writing on the wall and I suspect their consideration began even further before due to apple wanting something faster and better. Apple didn't develop their m1 chipes in the general 4-5 year timeline for a new processor design. They'd been at it for a good decade. Intel's next gen approach with chiplets may address that little ht and avx512 issue. At that point I'd say AMD is in trouble unless they figure out a way to include more than 16 cores on their consumer processors, have avx512 as well as SMT On all core types and at a low power use compared to Intel.

 

[Hulk]

Without AMD Intel would still be selling us 8 core Cypress/Sunny Cove parts built on a 10nm process from 2 years ago as top-of-the line for $999.

 

[A///]

Without AMD Intel would still be selling us 8 core Cypress/Sunny Cove parts built on a 10nm process from 2 years ago as top-of-the line for $999.

I think you mean backported designs on 14nm+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++.

Intel would have been very casual about getting 10nm to work with no competition. Even Apple aren't real competition. You'll have your mainstay companies getting on the new M processors but high end professional software companies are taking their sweet time getting their software to run on the M processors. Autodesk too over 2 years to get some of their most used software to run native on the M processors, with many other highly anticipated software waiting in limbo to when they'll be available.

 

[Timur Born]

A clarification: Anandtech review is really done with unlimited power enabled for 13900K, not at stock as they stated. A 13900K @ stock (253W) scores ~39000. The ~40500 score is with unlimited power enabled, ergo, its overclocked.

This may be mainboard/BIOS/sample dependent. Usually a 13900K only scores around 36k+ at 253W and needs at least 280-290W for 40k+ scores. My sample can be undervolted to run CB23 stable down to 240W and below, but in order to keep other crazy load like Folding@home stable - even with a 253W power limit - my sample needs to run CB23 at 253W+ (depending on temp).

You also have to make sure that your DC LL fits your LLC setting, else the Package Power reading is wrong (and thus the limit based on said reading).

 

[A///]

This may be mainboard/BIOS/sample dependent. Usually a 13900K only scores around 36k+ at 253W and needs at least 280-290W for 40k+ scores. My sample can be undervolted to run CB23 stable down to 240W and below, but in order to keep other crazy load like Folding@home stable - even with a 253W power limit - my sample needs to run CB23 at 253W+ (depending on temp).

You also have to make sure that your DC LL fits your LLC setting, else the Package Power reading is wrong (and thus the limit based on said reading).

Same issue with amd boards and processors. In the past, long before some members here were even born board variation didn't behave like that but how far you could oc a particular chip. Binning was a lot worse then ofc. It's no surprise today you might save money on a cheaper well accessorized board but get smacked upside the head with a trowel when yoou realize your affordable board isn't hitting the scores others are. it's the tech version of being a eunuch.

 

[Alexium]

Same issue with amd boards and processors. In the past, long before some members here were even born board variation didn't behave like that but how far you could oc a particular chip. Binning was a lot worse then ofc. It's no surprise today you might save money on a cheaper well accessorized board but get smacked upside the head with a trowel when yoou realize your affordable board isn't hitting the scores others are. it's the tech version of being a eunuch.

Why wouldn't it be hitting the scores, short of VRM overheating?

 

[A///]

Why wouldn't it be hitting the scores, short of VRM overheating?

artificial board limits. Intel or AMD and their partners would love for you to buy a b650 and higher end not the a620 or whatever Intel is selling as a budget board now. it isn't a new concept and it goes back a long time. price differences used to be a lot less long ago.

The entire pc market has always been a racket to convince you to spend more and more for the next higher product than the one you were looking at and keep going if you can justify the price in your head.

 

[Henry swagger]

https://twitter.com/x/status/1643172802273374214

Demanding games show the real ipc of architectures 😯💻

 

[Hulk]

https://twitter.com/x/status/1643172802273374214

Demanding games show the real ipc of architectures 😯💻

Unless the frequency of each CPU during the benchmark is known it is impossible to make IPC comparisons.

Also, even if this information is known this would only detail the "real" IPC, or more correctly "throughput" of each CPU in this specific game.

 

[DAPUNISHER]

Unless the frequency of each CPU during the benchmark is known it is impossible to make IPC comparisons.

Also, even if this information is known this would only detail the "real" IPC, or more correctly "throughput" of each CPU in this specific game.

The patches are coming so fast, it's hard to know how things will shake out. They just released a huge 14GB patch yesterday https://feedback.naughtydog.com/hc/...08-The-Last-of-Us-Part-I-v1-0-2-0-Patch-Notes

It's supposed to help take some of the load off of the CPU for texture streaming. Consequently, I wouldn't put to much stock in benchmarks at the moment. Like Cyberpunk'd, this game needs more polish before it can give us a good look at CPU performance. It loves threads though, so Raptor should certainly be doing well regardless.

 

[Henry swagger]

Unless the frequency of each CPU during the benchmark is known it is impossible to make IPC comparisons.

Also, even if this information is known this would only detail the "real" IPC, or more correctly "throughput" of each CPU in this specific game.

That's why i said demanding games 😏

 

[igor_kavinski]

That's why i said demanding games 😏

Sure by demanding games you don't mean games optimized for specific architectures?

 

[IEC]

That's why i said demanding games 😏

Is Microsoft Flight Simulator not a demanding game?

 

[Henry swagger]

Sure by demanding games you don't mean games optimized for specific architectures?

What are you implying ? 😃

 

[Henry swagger]

Is Microsoft Flight Simulator not a demanding game?

It is

 

[Hitman928]

It is

So a 7950x3d has way higher real IPC than a 13900k?

The 7950X3D’s performance in Microsoft Flight Simulator 2021 is almost unbelievable — the X3D chip is 53% faster than its vanilla counterpart, the 7950X, and 43% faster than the 13900K at stock settings.

AMD Ryzen 9 7950X3D Review: AMD Retakes Gaming Crown with 3D V-Cache

AMD takes a bite out of Raptor Lake.

www.tomshardware.com

 

[Mopetar]

Guys, knock it off. The kind of contortions he's going to need for the mental gymnastics to square that circle will probably cause a brain aneurysm.