Discussion Intel current and future Lakes & Rapids thread

[IntelUser2000]

The CPU can contribute in how quickly it can ramp up and down alongside how quickly it can complete the short tasks waking the chip up from sleep, but otherwise agreed.

That concept is known as HUGI, or Hurry Up and Get Idle. A more technical term is Race to Halt.

It's also not as simple as that. For example, they said a low TDP chip like Atom, the fixed system power will dominate and there's less of an incentive to HUGI, because the CPU takes up less, and in that case its better to operate at an optimal frequency. But for high power chips the CPU can finish its task and lower total power use in a greater way.

This level will vary tremendously between loads. And if you are doing something that takes an hour, or playing games, this HUGI concept doesn't matter. In that case the only way to increase battery life is reducing the TDP significantly, which will also have a great impact on performance.

The average power used when they consider the sleep state in between bursty workloads is called, average power.

But people, especially enthusiasts don't like that term. Because the average power varies so much based on workload. TDP is nice, because its a fixed number we can rely on.

 

[DrMrLordX]

That's pretty surprising. I was expecting products shipping soon, not 2 months from now. Especially since it seems like they are blowing most of their 10 nm capacity on this.

I thought products were shipping this month? And that was essentially a two month delay already.

 

[Gideon]

I thought products were shipping this month? And that was essentially a two month delay already.

Yeah, it's dissapointing considering they were hyping Tiger Lake already during CES in january. Then people claimed "it will arrive much earlier than Ice Lake" (August). Hell, in the end of last year somebody even claimed that it will be released Q1 2020.

 

[IntelUser2000]

Yeah, it's dissapointing considering they were hyping Tiger Lake already during CES in january. Then people claimed "it will arrive much earlier than Ice Lake" (August). Hell, in the end of last year somebody even claimed that it will be released Q1 2020.

Yup. I'm pretty sure lockdowns affected the supply chain too. It's not just Tigerlake if you look around.

Still delay's a delay.

 

[coercitiv]

Ian Cutress on the way Intel presented TGL base clocks because TDP isn't relevant anymore.

https://twitter.com/x/status/1301470603132784643

Here's the entire chain of tweets:

So to add to the Intel TDP and frequency confusion. The '12-28W' TGL parts were listed with 28W base clocks. Intel's ARK database confirms this. If anything, we should probably post the 12W numbers to be closer to prev gen and competition. For the '7W-15W' TGL parts, Intel's spec list from the presentation actually gives the '9W' TDP mode nase frequency. Looking on ARK, the numbers for 7W and 15W are different, so we do actually have the 9W base clocks for these CPUs.

Intel said 'TDP isn't useful', which I'm sure isn't a slogan that will be applied to Xeons, where TDP *is* useful. Obfuscating the power curve further just to get a higher number sounds like a 'hope they won't notice' move from comms? Or you know, Hanlon's Razor. We're still figuring out how we should represent this in future, especially as it bifurcates Intel's own specification delivery. I mean, I could just quote turbo and power, call the 1165G7 'a 50W CPU at 4.8 GHz', but that doesn't tell the whole story.

Perhaps we should attach something like
B 1.2 GHz at 12 W
B 3.0 GHz at 28 W
T 4.8 GHz at 50 W

Every time these CPUs are mentioned. If that's the case, 'What's your BBT' should be the new slogan.

Hopefully when we get parts in to test, we'll get the power curves for ourselves, and find out what these CPUs are reporting to the OS. There are easy ways to do this.

And a little cherry on top:

Oh, and just to add more confusion into the mix, Intel didn't have a full list of TGL CPUs at the event. Some of these processors will be offered with and without the IPU, and these are classified as different part numbers.

 

[clemsyn]

I thought products were shipping this month? And that was essentially a two month delay already.

I thought so too. A bit frustrating. I have a feeling it's either related to COVID or yield issues. Any idea when they will get to independent reviewers?

 

[mikk]

Review from JZWSVIC: https://zhuanlan.zhihu.com/p/188224934?utm_source=zhihu&utm_medium=social&utm_oi=754405736406073344


Intel Blueprint Series: 11th Gen Intel Core Processors:

 

[Asterox]

In the past when Intel CPU-s ruled the market, at that time Cinebench/Maxon was on Intel best friends list.Today, hm Cinebench is just a useless and unrealistic test.

 

[mikk]

In the past when Intel CPU-s ruled the market, at that time Cinebench/Maxon was on Intel best friends list.Today, hm Cinebench is just a useless and unrealistic test.

At this time AMD and all fanboys hated Cinebench and now it's their number 1 showcase benchmark everyone should use. But to be honest even in Cinebench Tigerlake is really good, the gap from i7-1165G7 to 4700U is not that big in the first reviews.

 

[lobz]

At this time AMD and all fanboys hated Cinebench and now it's their number 1 showcase benchmark everyone should use. But to be honest even in Cinebench Tigerlake is really good, the gap from i7-1165G7 to 4700U is not that big in the first reviews.

By your very definition, everyone who hates cinebench now, is an Intel fanboy.

 

[eek2121]

By your very definition, everyone who hates cinebench now, is an Intel fanboy.

Does anyone actually hate Cinebench? That would be weird. Also, as Intel chips beat the vast majority of AMD chips in CB ST, that doesn’t make any sense.

 

[lobz]

Does anyone actually hate Cinebench? That would be weird. Also, as Intel chips beat the vast majority of AMD chips in CB ST, that doesn’t make any sense.

Very silly to ask that from me, since I replied to a comment where he used this phrase and I was quoting that for a reason.

 

[IntelUser2000]

Early Q1 for Rocketlake-S and Q3 for Alderlake-S according to Fanlesstech. Though for the latter I suspect it'll be similar to Tigerlake in timeframe.

Process is still a fundamental recipe for success. Maybe its more important than ever. The backport of Sunny/Willow is the problem with Rocketlake. If we had the backport as Coffeelake in 2018 it would have been fine. In 2021 its not.

Also, I have doubts on Alderlake using Foveros. It has a PCH. MCM is ok on desktops. But on laptops it needs more integration, not less. So either Alderlake mobile will use Foveros with PCH as an active interposer and GMCH (graphics/memory controller hub)+ CPU cores on top,* or it'll be monolithic. And of course desktops can do whatever it wants.

*This is the absolute minimum. Because bandwidth between CPU + GMCH is much higher than what's needed using DMI to the PCH. Reality is also more complicated than that. The increased distance seems to translate into something that makes it harder to synchronize power management for the two.

 

[jpiniero]

Also, I have doubts on Alderlake using Foveros. It has a PCH. MCM is ok on desktops. But on laptops it needs more integration, not less. So either Alderlake mobile will use Foveros with PCH as an active interposer and GMCH (graphics/memory controller hub)+ CPU cores on top,* or it'll be monolithic. And of course desktops can do whatever it wants.

I'm imagining it would be like Lakefield, both desktop and mobile. Desktop could also have a secondary discrete PCH, the kind of thing where if it wasn't Intel it would be optional but at the same time popular because OEMs love them some 20 USB ports.

 

[coercitiv]

Early Q1 for Rocketlake-S

Are they really going to let AMD have the holiday season all to themselves? The boards are already out, all they need is a top bin RKL-S launch, on paper if it needs to.

At this rate holiday season will mean AMD CPU + Nvidia GPU for most gamers looking for a meaningful upgrade.

 

[Spartak]

Not sure if it was discussed here already but what's the second die on the Tiger Lake chip? I always assumed it was the graphics but the die shot in the Anandtech article clearly shows the die still has integrated graphics.

 

[Spartak]

Ian Cutress on the way Intel presented TGL base clocks because TDP isn't relevant anymore.

https://twitter.com/x/status/1301470603132784643

And a little cherry on top:

Perhaps we should attach something like
B 1.2 GHz at 12 W
B 3.0 GHz at 28 W
T 4.8 GHz at 50 W

The single core turbo isnt relevant in this "BBT" rating as it doesnt say anything about the non-turbo speeds of the other processors. The T should be the all-core turbo at 4.3 GHz.

I would refrase it as
Base 1.2 GHz at 12 W
Medium 3.0 GHz at 28 W
High 4.3 GHz at 50 W

where every mobile chip could be configured as either B/M or M/H for PL1/PL2. You could throw in even more intermediate steps that are assigned to a letter.

 

[IntelUser2000]

@Spartak The second die is the PCH. If you see a third, then that's eDRAM or HBM.

The last time graphics was on the second die was in 2010 with the first gen Core btw.

 

[DrMrLordX]

Are they really going to let AMD have the holiday season all to themselves? The boards are already out, all they need is a top bin RKL-S launch, on paper if it needs to.

RocketLake-S has been a 2021 CPU for awhile. It was only RKL-U that could have shown up in 2020, and that looks cancelled. Comet was about six months late, so there's no telling how that affects Rocket Lake-S' launch either.

 

[coercitiv]

The last time graphics was on the second die was in 2010 with the first gen Core btw.

Actually...

Spoiler

 

[Spartak]

Well they're right. TDP isn't useful.

Especially now that Intel's base clocks appear to be rated at 28W and 9W (guesstimate) respectively. The one reason why TDP was useful was to give a baseline for base clocks. Now it can't even do that.

You need TDP to match / fit your power supply, it's not just thermals. Especially for laptops / SFF PC's the power supply isnt that buffy. It's really great they allow you to specify PL1/PL2 states, AMD could take a cue form that. But to pretend it's not important for your build is a plain lie.

 

[Spartak]

Everyone who's ego depends on their chosen side winning is willing to come up with extreme use cases to feel like they are in the lead. This is true for both camps. So of course, you do your heavy lifting on ultra-thin laptops! That is, until the other side is in the lead then you would never do heavy lifting on ultra-thin laptops!

It's not about extreme use cases it's about *your* use cases. For me, as a civil3d (autoCAD) user, single threaded performance is important as AutoCAD doesn't scale well over multiple threads. That's like 90% of my heavy lifting usage that's directly impacted by single threaded performance.

 

[eek2121]

You need TDP to match / fit your power supply, it's not just thermals. Especially for laptops / SFF PC's the power supply isnt that buffy. It's really great they allow you to specify PL1/PL2 states, AMD could take a cue form that. But to pretend it's not important for your build is a plain lie.

Incorrect. TDP has nothing to do with power. A CPU can consume 11W of power and still have a TDP of 28W, or it can consume 50W of power and still have a 28W TDP. TDP stands for Thermal Design Power.

 

[lobz]

Early Q1 for Rocketlake-S and Q3 for Alderlake-S according to Fanlesstech. Though for the latter I suspect it'll be similar to Tigerlake in timeframe.

Process is still a fundamental recipe for success. Maybe its more important than ever. The backport of Sunny/Willow is the problem with Rocketlake. If we had the backport as Coffeelake in 2018 it would have been fine. In 2021 its not.

Also, I have doubts on Alderlake using Foveros. It has a PCH. MCM is ok on desktops. But on laptops it needs more integration, not less. So either Alderlake mobile will use Foveros with PCH as an active interposer and GMCH (graphics/memory controller hub)+ CPU cores on top,* or it'll be monolithic. And of course desktops can do whatever it wants.

*This is the absolute minimum. Because bandwidth between CPU + GMCH is much higher than what's needed using DMI to the PCH. Reality is also more complicated than that. The increased distance seems to translate into something that makes it harder to synchronize power management for the two.

A year earlier everyone here wanted me to believe that Rocket lake comes half a year after Comet Lake (Q1 and Q3). I said yeah right. - they could barely launch CML in the middle of Q2, and we all know where RKL stands now.
Now a fanless person wants me to believe again, that intel will launch 2(!) actually new (!!!) uarchs 6 months apart.
My opinion will come as a real shocker: yeah right.
Alder Lake Q1 2022, or MAYBE for Christmas 2021. But I highly doubt that too.

 

[CakeMonster]

Rocket Lake is the new desktop chip, IIRC? This is getting really hard to follow. Is this the one that will add PCIE4 and have 8-16 cores competing with Zen 3 for better gaming performance?