14th Gen "Raptor Lake Refresh"

[humbug]

So thats why im asking you to run your 5800x at 45w,and ill run my 12900 with ecores OFF at 75w. As the article suggests, your 5800 will outperform my 12900. So lets test it. I bet a paycheck, it will be a massacre...

Prove your claim that golden cove cores are less efficient

Go on, run it then, 20 minute R23 loop, recorded.

 

[Bencher]

Go on, run it then, 20 minute R23 loop, recorded.

Why? You want me to go through trouble for no reason, since it wont change yours or any others mind, so I have absolutely nothing to gain. Even when the 12900 at 75w smashes the 5800x at 45w to smitheerens, youll still keep talking about how inefficient it is, so its absolutely pointless.

And of course, you are now dodging cause you know the 8 zen 3 cores will get smacked.

 

[humbug]

Why? You want me to go through trouble for no reason, since it wont change yours or any others mind, so I have absolutely nothing to gain. Even when the 12900 at 75w smashes the 5800x at 45w to smitheerens, youll still keep talking about how inefficient it is, so its absolutely pointless.

And of course, you are now dodging cause you know the 8 zen 3 cores will get smacked.

Talk is cheap.

 

[Bencher]

Talk is cheap.

Exactly, i already uploaded a 15k+ score at 34w and you still claim its inefficient. Facts and proof wont change your mind, so there is no point for me to even try. I mean just now, you agreed with an article saying 8 zen 3 cores at half the wattage outperform 8 alder lake cores. Lol, do you honestly believe that?

 

[humbug]

You're going round in circles again, i'm done with this particular argument.

Yes i'm going to keep on because too many tech journalists are saying the same thing., you can tune your CPU to get a screenshot but that takes 5 seconds, i can run mine 24/7 with 45 watts cores power and score over 14K just by limiting the package power to 70 watts and do nothing else to it.

PS: that is less than 6 watts per core.

 

[Bencher]

You're going round in circles again, i'm done with this particular argument.

Yes i'm going to keep on because too many tech journalists are saying the same thing., you can tune your CPU to get a screenshot but that takes 5 seconds, i can run mine 24/7 with 45 watts cores power and score over 14K just by limiting the package power to 70 watts and do nothing else to it.

PS: that is less than 6 watts per core.

I can run mine 26/7.
Tldr, the article is a hot potato and 8 gc cores at 75w scorch the heck out of 8 zen3 cores at 45w. Its not even close

 

[Bencher]

Just tested, 12900k with E cores OFF scores over 17k at 75 watts. It outperforms the 5800x at stock 105w TDP, but SOMEHOW, the article suggests that if you drop it down to 45w, it will somehow score better. OKAY bro

 

[Armageus]

Just tested, 12900k with E cores OFF scores over 17k at 75 watts. It outperforms the 5800x at stock 105w TDP, but SOMEHOW, the article suggests that if you drop it down to 45w, it will somehow score better. OKAY bro

We'll await the proof, but until then probably better to try and get back on topic?

 

[Bencher]

We'll await the proof, but until then probably better to try and get back on topic?

Sure, here you go, high and low priority @ 77watts. The article was talking about 76.7, but I don't think it will make that big a difference

Spoiler: 12900 ecores off

 

[Dave2150]

Some info on Meteor Lake here though not much that's new. I did find this interesting though:

Data gathered by Mercury Research(opens in new tab) indicated that Intel chips accounted for 74.4% of all traditional CPU sales at the end of 2021.

Source: https://www.creativebloq.com/news/intel-meteor-lake

Considering Alder Lake only launched in November, it would have had limited impact on these figures. I'd guess Intel are closer to 80% of CPU sales at this point!

 

[Wolverine2349]

E cores doesn't work as they should, mostly they supplement p cores in MT workloads like cinebench, but they fail to impress even when doing background processes, ryzen 6000 beats them in efficienty and battery. Not to mention gaming where it is better to turn them off and only use P cores. It would be better for them to focus just on P cores, and because each P core is much stronger you don't need as much as small ones, and also software optimization would be much easier, and games can utilize all of them without degredation. We can conclude that Big/Little is fail in x86 world.

I could not agree more. Intel if they want to get back in the game and as big as they are and resources they have, they need to focus on P cores only at least in desktop segment. The e cores stink except for some uplifting to assist the super strong p cores in MT workloads like Cinebench when in reality they are crap for much else including background tasks as software has no clue how to use them properly as they are so behind, Plus you need so much more of them to even compete with less P cores. Its a shame really Intel is going this direction.

 

[Wolverine2349]

Games that can utilize more than 8 cores will have regression because as soon as it needs to access from big to small cores there will be latency. It would be better to just create all big cores so less work for developers, no regression, incompability etc, and each big core is much more powerful than bunch of those e cores.

Exactly well said. The whole hybrid arch was a massive radical change and not right for X86. The X86 ecosystem of software is so massive going back 2 decades and beyond. Yes changes in PCIE gens, chipsets and CPU archs happen all the time, but it is not near as radical of a change as trying to stick 2 different CPU types on same arch since multi processing systems have always used same types of CPUs going back to dual Pentium Pro in the 90s.. But the one radical constant has been low level Windows scheduler regarding multi- threading as such.

The hybrid arch is a nightmare for software developers and ecosystem as it deviates form normal multi processing that has been the base for development for over 20 years.

And I normally liked Intel better than AMD except during the Athlon 64 X2 era prior to Conroe. But since Conroe, I was Intel all the way even through Zen 2. But with Intel going hybrid arch route and no more than 8 p cores, I am now now going AMD. Well maybe maybe Intel HEDT is an option assuming it beats or equals Zen 4??

 

[Dave2150]

Intel's P cores (as seen on Alder Lake since 12900k release date in November 2021) are currently the most powerful cores available on X86. This lead will be extended with Raptor Lake, thanks to many improvements in the architecture.

Meteor Lake will improve this further, exciting time ahead

Hybrid designs are here to stay, E cores clearly useful and have allowed Intel to return to leaderships status across the majority of workloads, whether you accept/like this or not.

 

[arknor]

I'd rather have more P cores and less E cores but I guess TDP or whatever doesn't permit it.
in reality you only really need like 2-4 ecores max to be doing background windows stuff.
I Imagine it would be the kinda thing consoles would benefit from most

 

[=XDC=FluphyBunny]

Ecores are there to incr

Exactly well said. The whole hybrid arch was a massive radical change and not right for X86. The X86 ecosystem of software is so massive going back 2 decades and beyond. Yes changes in PCIE gens, chipsets and CPU archs happen all the time, but it is not near as radical of a change as trying to stick 2 different CPU types on same arch since multi processing systems have always used same types of CPUs going back to dual Pentium Pro in the 90s.. But the one radical constant has been low level Windows scheduler regarding multi- threading as such.

The hybrid arch is a nightmare for software developers and ecosystem as it deviates form normal multi processing that has been the base for development for over 20 years.

And I normally liked Intel better than AMD except during the Athlon 64 X2 era prior to Conroe. But since Conroe, I was Intel all the way even through Zen 2. But with Intel going hybrid arch route and no more than 8 p cores, I am now now going AMD. Well maybe maybe Intel HEDT is an option assuming it beats or equals Zen 4??

You need to read up on ecores/pcores and troll again.

 

[Wolverine2349]

I'd rather have more P cores and less E cores but I guess TDP or whatever doesn't permit it.
in reality you only really need like 2-4 ecores max to be doing background windows stuff.
I Imagine it would be the kinda thing consoles would benefit from most

I do not have a problem with some amount of hybrid arch especially in the laptop space where e cores belong. But choosing to make it so you cannot get more than 8 P cores on the enthusiast desktop is not a good move. That's why I am going all AMD. Intel can win me back by just releasing a desktop CPU with 10-12 Golden Cove/Raptor Cove or whatever they are called in future P cores that equal or beat AMD and you got a buyer in me.

And it would not of made sense to stay AMD back before the Ryzen era cause even though AMD offered high core counts, those Bulldozer and Pilediver cores were so beyond pathetic it was not even funny. If AMD was still like that today, I would go Intel still despite the 8 P core cap. But Zen 3 cores while a little behind Golden Cove cores are still very very good cores. In fact the gap of Golden Cove over Zen 3 at same clock speed is not even as great as gap between K8 and Conroe across the board.

 

[Bencher]

I do not have a problem with some amount of hybrid arch especially in the laptop space where e cores belong. But choosing to make it so you cannot get more than 8 P cores on the enthusiast desktop is not a good move. That's why I am going all AMD. Intel can win me back by just releasing a desktop CPU with 10-12 Golden Cove/Raptor Cove or whatever they are called in future P cores that equal or beat AMD and you got a buyer in me.

And it would not of made sense to stay AMD back before the Ryzen era cause even though AMD offered high core counts, those Bulldozer and Pilediver cores were so beyond pathetic it was not even funny. If AMD was still like that today, I would go Intel still despite the 8 P core cap. But Zen 3 cores while a little behind Golden Cove cores are still very very good cores. In fact the gap of Golden Cove over Zen 3 at same clock speed is not even as great as gap between K8 and Conroe across the board.

You are missing the point. Without the E cores, there won't be any P cores either. The reason the P cores are so much faster and more efficient than the zen 3 cores is because they are huge in actual die space. So in order to have competing multithreaded performance they had put some efficient cores in their design as well. The hybrid design is the future, there is no reason not to do it.

 

[Wolverine2349]

You are missing the point. Without the E cores, there won't be any P cores either. The reason the P cores are so much faster and more efficient than the zen 3 cores is because they are huge in actual die space. So in order to have competing multithreaded performance they had put some efficient cores in their design as well. The hybrid design is the future, there is no reason not to do it.

Yes in a way that is true. Problem is why is there a maxing out of 8 P cores. Ok for Alder Lake I get it. Alder Lake was first Big.Little arch on X86. But future gens, it is like Big. More little then Big. Post Alder Lake and each subsequent gen it appears per Intel roadmap, Big.More and more and more little, and no more Big.

Like with Alder Lake, there is 8+8 for highest SKU, 8+4, 6+4, and so forth and it made sense as a starting point.

With future gens, it does not feel like Big. Little evolving. Instead it is Big the same count with only maybe IPC improvements and More and more little. Like why not with Raptor Lake then Meteor Lake and future gens, a 10+6 config, 12+8 config, 10+10 config or 12+12 config and such. Instead it is 8+ more and more little. That seems very odd and concerning. With Alder Lake there was more Big or an equal amount of little at most which actually made more sense for a balanced Big.Little approach starting out. Future gens it appears all about increasing e cores without any increase in P cores.

Hopefully Intel once they get to new process node changes focus and keeps Big.Little, but makes it more balanced and actually has 10+6 and 12+6 and 10+10 configs and such for example. Everyone will be a happy camper in that case cause even those who do not like the e cores can disable them and still get more than 8 P cores in future gens and the areas where Big.Little shines, it will shine even more with a more balanced P and e core count rather than just 8 P cores plus more and more e cores.

 

[Robert896r1]

We have no solid data for MTL skus yet.

RPL is just an optimization/refresh of ADL.

If you think of it as ADL+ that’s probably fair.

 

[Journey]

If you think of it as ADL+ that’s probably fair.

Probably the best, most realistic, assessment I have seen suggested yet around here so far.

Meteor Lake looks promising though, from what little info we have so far and I really like the tile design. I do wonder how what the equivalent of IF will be like for it, and how well optimised it will be at the first iteration.