Monday, March 22nd 2021

Intel 12th Generation Alder Lake Platform Reportedly Brings 20% Single-Threaded Performance Uplift

Intel only just announced their 11th generation Rocket Lake-S desktop processors last week but we are already receiving information about the next generation Alder Lake-S platform which will finally make the jump to 10 nm. Intel slides for the upcoming family of processors have been leaked and they reveal some interesting information including a claimed 20% single-threaded performance increases from the new Golden Cove core design and 10 nm SuperFin node. The processors will feature Intel Hybrid Technology with a mix of small low-performance cores and large high-performance cores with a maximum of eight each for sixteen total cores. The processors will also include the latest connectivity with both PCIe 4.0 and PCIe 5.0 support along with DDR4 and DDR5 4800 MHz compatibility.

Intel will also be launching a new socket type called LGA1700 with a new package size which will render existing cooling solutions for LGA115X and LGA1200 sockets incompatible. The processors will also come with the launch of a new 600 Series chipset with PCIe 3.0 and PCIe 4.0 support along with the usual complement of USB, SATA, and networking. The entry-level 600-series motherboards will only support DDR4 memory at up to 3200 MHz while high-end Z690 motherboards will include DDR5 support. Intel has confirmed that they intend to launch Alder Lake later this year but it is yet to be known if they are referring to the desktop or mobile series.
Source: VideoCardz
Add your own comment

108 Comments on Intel 12th Generation Alder Lake Platform Reportedly Brings 20% Single-Threaded Performance Uplift

#51
ZoneDymo
Vayra86
Yeah, no... the marketing was disingenuous and the architecture was a massive design failure. The performance just wasn't there so AMD effectively oversold it with misleading marketing. And they got burned for it, rightly so. Just stop playing the poor puppy AMD card, because really, its so sad and we know better.
The lawsuit wasnt really about the architecture being a "massive design failure" or the performance "not being there".
It was purely about the concept of a core in a cpu, what a core is, what can be called a core.

And tbh, Im not sure that was rightly settled/decided, regardless if it was AMD or any other company.
Posted on Reply
#52
londiste
TheinsanegamerN
Yeah, just like Rocket Lake was supposed to feature "double diget increases" in IPC, yet as anandtech showed in many consumer applications its barely any faster, and sometimes slower, then comet lake. Far cry from the jumps from zen 2 to zen 3 provided.

This will likely have "up to 20% in AVX 512, then 3% everywhere else. And it will have to contend with zen 4, which should feature significant improvements of its own.
IPC improvements are usually defined by manufacturer from specific benchmarks or applications. In Intel's case that has been SPEC.
For single core performance Anandtech tests show 11700K score improvement over 10700K +15.5% in SPECint2017 and +19.6% in SPECfp2017. This is with a frequency deficit, Anandtech calculates IPC improvement as +18.5% and +22% respectively.
Multicore results suffers badly, +5.8% and +16.2%. They did not specify the frequencies there but it is probably safe to say Rocket Lake runs at lower frequency.

For comparison - with Zen3 over Zen2, Anandtech found +19% in the same SPEC2017 for single core and +8-9% for multicore.
Posted on Reply
#53
Makaveli
Caring1
Basically yes, all the consumer sheep have been brain washed by years of marketing that Intel inside is good, they don't look for specs.
The majority of the market is computer illiterate. They just buy based on recommendations from friends or what sales people tell them. Most don't know anything about the processors, memory, and all the technical details. People like us that read these forums while very vocal and a super tiny part of the market and the non technical group is much larger. This group will also believe marketing without questioning anything.
Posted on Reply
#54
R0H1T
londiste
IPC improvements are usually defined by manufacturer from specific benchmarks or applications. In Intel's case that has been SPEC.
For single core performance Anandtech tests show 11700K score improvement over 10700K +15.5% in SPECint2017 and +19.6% in SPECfp2017. This is with a frequency deficit, Anandtech calculates IPC improvement as +18.5% and +22% respectively.
Multicore results suffers badly, +5.8% and +16.2%. They did not specify the frequencies there but it is probably safe to say Rocket Lake runs at lower frequency.

For comparison - with Zen3 over Zen2, Anandtech found +19% in the same SPEC2017 for single core and +8-9% for multicore.
This was from the original review ~

In our testing, we saw the following:
  • Single thread floating point: +19.0%
  • Multi-thread floating point: +19.5%
Sounds great, right?
  • Single thread integer: +13.0%
  • Multi-thread integer: +7.3%
Just a reminder those are estimates as SPEC scores have to be submitted for validation(?) & Andrei hasn't done that AFAIK.
Posted on Reply
#55
PapaTaipei
This looks totally trash we don't need small cores for desktop high performance gaming PC. Intel once again slowing down the whole industry.
Posted on Reply
#56
defaultluser
R0H1T
This was from the original review ~

In our testing, we saw the following:
  • Single thread floating point: +19.0%
  • Multi-thread floating point: +19.5%
Sounds great, right?
  • Single thread integer: +13.0%
  • Multi-thread integer: +7.3%
Just a reminder those are estimates as SPEC scores have to be submitted for validation(?) & Andrei hasn't done that AFAIK.
Yup, the target IPC increases haven't been a hit with Ice Lake., or in it's bastard child for Rocket Lake

Tiger Lake is finally clocked high enough for it to impress, but 8 months in and there is still n sign of Tiger Lake 8-core. They promised Q1 of this year, but obviously missed it.

www.anandtech.com/show/16384/intels-8-core-mobile-tiger-lake-h-at-45-w-to-ship-in-q1

The hopeful idiots will continue to believe Intel BS marketing machine, that somehow Intel will magically launch 4x as many cores as they are current;y shipping in consumer 10nm parts, combined with 2 completely new architectures! Meanwhile here in reality, you won't even be able to buy Rocket Lake in retail until sometime in May!

You'll be lucky to see Alder Lake shipping before May of next year.
Posted on Reply
#57
Why_Me
defaultluser
www.anandtech.com/show/16384/intels-8-core-mobile-tiger-lake-h-at-45-w-to-ship-in-q1
The hopeful idiots will continue to believe Intel BS marketing machine, that somehow Intel will magically launch 4x as many cores as they are current;y shipping in consumer 10nm parts, combined with 2 completely new architectures! Meanwhile here in reality, you won't even be able to buy Rocket Lake in retail until sometime in May!

You'll be lucky to see Alder Lake shipping before May of next year.
I'm going to go out on a limb here and assume you live South Africa or Australia because those cpu's are selling in the US, UK, Canada and Germany.

www.newegg.com/p/pl?N=100007671%20601361546%20601361550
Posted on Reply
#58
efikkan
If these chipset specs are accurate, then buying Alder Lake for "future proofing" for future SSDs is futile. Those that feel this is relevant should look at Sapphire Rapids instead then.
Dredi
20% more ST compared to what? Tigerlake or skylake? If that’s on top of tigerlake (which was around 19% faster than skylake), then these could finally be pretty fast.

Funnily in the slide it is written ”up to 20% performance”, which indicates a 80% performance regression. ;)
No you got it all wrong.
Up to 20% ST performance means exactly what it says, that's the best case scenario, but says nothing about the average or the minimum improvements.

There is a huge difference in up to 20% better performance and 20% better average performance. Take for instance Rocket Lake's 19% IPC gain; IPC is average instructions per clock, so this doesn't mean that every workload will get a 19% gain, the worst get ~0% performance gains, the best ~40% performance gains. So Rocket Lake has 19% better IPC but up to 40% better performance per clock. So if this claim of "up to 20% better ST performance" for Alder Lake is accurate, it probably means the average is much lower, and would make it a fairly small generational upgrade.

But keep in mind these claims are probably bogus. No one, even at Intel, does know the clock speeds until the final stepping has gone through testing. So at this point Intel only have performance targets, not precise performance figures for specific models.
TheinsanegamerN
This will likely have "up to 20% in AVX 512, then 3% everywhere else. And it will have to contend with zen 4, which should feature significant improvements of its own.
20% faster AVX-512 is unlikely. While it probably will be a few percent faster due to less throttling on 10nm, a larger improvement like an extra FMA set would yield much higher maximum performance than that, and in such cases they would surely boast about "up to 50% higher performance", not 20%.
Posted on Reply
#59
Operandi
Vayra86
Yeah, no... the marketing was disingenuous and the architecture was a massive design failure. The performance just wasn't there so AMD effectively oversold it with misleading marketing. And they got burned for it, rightly so. Just stop playing the poor puppy AMD card, because really, its so sad and we know better.
The marketing was bad but it was not disingenuous. Bulldozer was weird design but was 8 cores; 2 integer units with 1 floating point unit with shared front end. There is no reason why 2 execution units can't share a front end and technically you don't even need to have a floating point unit so no matter how you look at it its 8 cores.

It was an odd design but it was supposed to crush Intel i7s SMT integer performance and concede floating point but ended up barely being able to compete in integer. It was a big expensive core and AMD was swinging for the fences and struck out; they even had 16 core Opterons. You don't have to feel sad for them it was bad design, but Intel has much longer track record of bullshit marketing not to mention their market manipulation.
Posted on Reply
#60
DAWMan
Such bold claims only shows their lack of confidence in Rocket Lake sales.
New engineers and CEO won’t fix 2019/20 and 21 embarrassments.
2022 is Intel’s year to win/lose.
All 14nm and 10nm designs will lead to a surplus and more market loss to AMD.
Posted on Reply
#63
Why_Me
DAWMan
Im hoping so.
Hate seeing the King stumble over and over.
Intel is winning the gaming build market so there's always that ... other than there's no gpu's to be had. =/
Posted on Reply
#64
watzupken
This is supposedly an "up to" 20% performance uplift. How they derived at it, is questionable at this point. I suspect part of it may be due to running benchmarks/ applications that may benefit from the faster DDR5 memory that comes along with Alder Lake. And looking at the way they gimped the existing i7 and below Rocket Lake processors, I feel you can only expect the top i9 to run up to 20% faster. In other words, it will not make a difference to 98% of the people buying a computer with an Alder Lake processor.
Posted on Reply
#65
BouziGolouM
R0H1T
The only reason why PCIe 5.0 & especially now v6 exists is because of one-upmanship by Intel. That's the (only) major point they can tout against the Zen juggernaut & frankly even enterprises won't care!
I beg to differ... Been like 3 years im screaming that intel will release PCI-e 5.0 in 2021 (it was in one of their older roadmap)... Btw, PCI-e 5.0 was already being tested before we even saw pci-e 4.0... Intel planned to skip it... The only reason we saw PCI-e 4.0 was BECAUSE AMD wanted to tout intel...
Posted on Reply
#66
Mussels
Moderprator


intel definitely need a performance uplift
Posted on Reply
#67
Dredi
efikkan
There is a huge difference in up to 20% better performance and 20% better average performance.
Nowhere on the slide does it say that it is 20% better, average or up to. It only states that it is up to 20% performance, which means at least 80% performance regression.

But yeah, assuming that someone can’t write english and it’s a typo of sorts, you are correct in that up to 20% improvement means jack and shit without specifics.
Posted on Reply
#68
R0H1T
BouziGolouM
Btw, PCI-e 5.0 was already being tested before we even saw pci-e 4.0
Being tested doesn't mean it was ready, I'd give you two thumbs up (:peace:) to show me how many working PCIe 5.0 based hardware we have today?

As opposed to PCIe 4.0 gear back in 2019 or even 2018 end.
BouziGolouM
The only reason we saw PCI-e 4.0
Can't have PCIe 5.0 without 4.0 ever being released, it was delayed a lot for a number of reasons & that's why the relative overlap with v5 though v6 is totally useless as of now!
Posted on Reply
#69
Makaveli
R0H1T
Being tested doesn't mean it was ready, I'd give you two thumbs up :)peace:) to show me how many working PCIe 5.0 based hardware we have today?

As opposed to PCIe 4.0 gear back in 2019 or even 2018 end.

Can't have PCIe 5.0 without 4.0 ever being released, it was delayed a lot for a number of reasons & that's why the relative overlap with v5 though v6 is totally useless as of now!
PCIe 5.0 will be much bigger in the server space than consumer also.

And i'm not sure you will even see PCIe 5.0 devices this year, so totally agreed with you.
Posted on Reply
#70
efikkan
Dredi
Nowhere on the slide does it say that it is 20% better, average or up to. It only states that it is up to 20% performance, which means at least 80% performance regression.

But yeah, assuming that someone can’t write english and it’s a typo of sorts, you are correct in that up to 20% improvement means jack and shit without specifics.
Good point. All the more reasons why this is fake. :)
Posted on Reply
#71
Mussels
Moderprator
Intel didnt release new PCI-E generations because the CPU designs were delayed, remember its part of the new chip design.

They likely wanted to release a new socket and PCI-E gen every ~2 years to justify a new platform and then got stuck with skylake++++ and waited and waited
Posted on Reply
#72
Minus Infinity
Dredi
20% more ST compared to what? Tigerlake or skylake? If that’s on top of tigerlake (which was around 19% faster than skylake), then these could finally be pretty fast.

Funnily in the slide it is written ”up to 20% performance”, which indicates a 80% performance regression. ;)
According to Morre's Law is Dead, it's 20% single core IPC uplift over skylake and up to 100% multithreaded uplift. Apparently the Gracemont cores are pretty strong too, I think he said they are about 70% the performance of skylake cores.
Posted on Reply
#73
Nephilim666
Minus Infinity
According to Morre's Law is Dead, it's 20% single core IPC uplift over skylake and up to 100% multithreaded uplift. Apparently the Gracemont cores are pretty strong too, I think he said they are about 70% the performance of skylake cores.
No offense to anyone who follows that Youtube channel but the guy just talks out of his ass.
Posted on Reply
#74
Dredi
Minus Infinity
According to Morre's Law is Dead, it's 20% single core IPC uplift over skylake and up to 100% multithreaded uplift. Apparently the Gracemont cores are pretty strong too, I think he said they are about 70% the performance of skylake cores.
Only ”up to” 100% in multithreaded applications? Even rocket lake manages a lot more than that in the ”up to” scenario. Around 530% uplift to be specific. Seems more like performance regression to me.
Posted on Reply
#75
Mussels
Moderprator
So intels focused on AVX performance this time around?
Posted on Reply
Add your own comment