Rumour: New Intel chips will have SOLDERED IHS, dropping August 1st

[Nicnac]

Quick Update: I've read through the german article that was quoted by wccftech and they claim up to 5.5ghz for next gen!

Source: PC BUILDERS CLUB

 

 

According to our favourite website wccftech we have some news on the next gen intel chips:

 

Quote

Intel Coffee Lake Refresh “9th Generation” CPUs Rumored To Launch on 1st August, Soldered CPUs Could Be Making Their Way To Consumers PCs

Here's the three high-end models we know of so far:

Quote

• Intel Core i9-9900K (est: 8 core / 16 thread)
• Intel Core i7-9700K (est: 6 core / 12 thread)
• Intel Core i5-9600K (est: 6 core / 6 thread)

 

Surprisingly, current gen Z370 Mobos will support 9th gen cpus:

 

Quote

At the same time, board makers have started adding BIOS support for next-generation CPUs on the current Z370 motherboards so maybe this rumor has some weight to it.

 

 

I think soldered IHS would be great,obviously, and the one thing enthusiasts have always been missing with intel chips.

Other than that, it's gonna be a Coffee Lake refresh, but who knows where the clocks can go with a soldered IHS?

 

Source: WCCFTech

 

 

[Arika]

holy shit, about time

[Humbug]

Wow, Intel.

 

Adding more Cores

Soldering the IHS

 

What next? All unlocked CPUs?

[Amazonsucks]

They had soldered IHS with Sandy Bridge... that wasnt that long ago lol

[AlTech]

3 minutes ago, Humbug said:

What next? All unlocked CPUs?

 

I doubt it.

 

And i3s will likely stay 4C4T which is also another major disappointment,

 

If they use Ring Bus for this refresh then the 8 core i9 will be an interesting SKU to look out for since Ring bus works optimally for up to 6 cores and does worse with higher core count CPUs.

[Majestic]

All this tells me is that the power consumption will be shit.

[iamdarkyoshi]

Its almost like there's been issues with their paste!

 

Gonna miss delidding tbh

[Majestic]

1 minute ago, iamdarkyoshi said:

Its almost like there's been issues with their paste!

 

Gonna miss delidding tbh

 

As someone who's got a delidded chip: I'm not going to miss it.

[Lurick]

1 minute ago, iamdarkyoshi said:

Its almost like there's been issues with their paste!

Nah, that's just rumors by AMD fanboys. The paste was fine

/s

[Trixanity]

8 minutes ago, Amazonsucks said:

They had soldered IHS with Sandy Bridge... that wasnt that long ago lol

7 years is a damn long time in the tech industry.

[TVwazhere]

20 minutes ago, Nicnac said:

Source: WCCFTech

I trust these guys about as far as I can throw them

 

And since WCCFTech is entirely web based and therefore an intangible, digital entity I cannot throw it, which means I cant even physically throw them. 

 

I guess I'll enjoy delidding whatever next Intel CPU comes out, or go to Zen2

[Taf the Ghost]

21 minutes ago, Humbug said:

Wow, Intel.

 

Adding more Cores

Soldering the IHS

 

What next? All unlocked CPUs?

Don't push it, haha. 

 

But the solder was somewhat expected. It was the only way they could keep an 8c Coffee Lake part within TDP of the 300 series platform and keep the clocks the same. I expect it only for the top of this generation, and only for this generation. But it was something logical they could have gone to when they needed the headroom. You can thank AMD for that one.

[Granular]

12 minutes ago, Trixanity said:

7 years is a damn long time in the tech industry.

It would be if x86 chips had become substantially better since then.

[WereCat]

19 minutes ago, iamdarkyoshi said:

Its almost like there's been issues with their paste!

 

Gonna miss delidding tbh

-img-

I will still wait for der8auers deliding results first.

The main issue was always the Z-height (space) between the die and the IHS rather than the paste itself.

But even if this is true then it is definitely a good step forward. They better dont mess up somehow.

[Amazonsucks]

20 minutes ago, Trixanity said:

7 years is a damn long time in the tech industry.

For enterprise chips yes. They get a 10x performance increase every 5 years or so. For desktop x86 CPUs, not so much. The difference between an i7-2600K and 6700K was like 25% in 5 years.

[porina]

34 minutes ago, Nicnac said:

Other than that, it's gonna be a Coffee Lake refresh, but who knows where the clocks can go with a soldered IHS?

For my sample of 8086k, got got to 5.1 on air stock, and 5.2 after delid. Good samples might hit 5.3, maybe water would help, but I don't think soldering by itself is going to radically change this.

27 minutes ago, AluminiumTech said:

If they use Ring Bus for this refresh then the 8 core i9 will be an interesting SKU to look out for since Ring bus works optimally for up to 6 cores and does worse with higher core count CPUs.

Before SKX all HCC CPUs still use ring bus, or worse, nested rings. Was there specific testing to show where the ring runs out of steam? The mesh structure of SKX has its own different problems, in part due to its stock low speed.

21 minutes ago, Majestic said:

As someone who's got a delidded chip: I'm not going to miss it.

+1 to that, I delid because I need to, not because I want to.

9 minutes ago, Taf the Ghost said:

But the solder was somewhat expected. It was the only way they could keep an 8c Coffee Lake part within TDP of the 300 series platform and keep the clocks the same. I expect it only for the top of this generation, and only for this generation.

Shouldn't significantly affect TDP, but it is the temperature that will ultimately limit what performance they get out of it... shall have to watch and see how this plays out. The LCC SKX parts managed with TIM.

[AlTech]

2 minutes ago, porina said:

For my sample of 8086k, got got to 5.1 on air stock, and 5.2 after delid. Good samples might hit 5.3, maybe water would help, but I don't think soldering by itself is going to radically change this.

Before SKX all HCC CPUs still use ring bus, or worse, nested rings. Was there specific testing to show where the ring runs out of steam?

2

Some youtuber (I can't remember which) did testing and found that Ring Bus worked optimally at up to 6 cores and worked less optimally at higher core counts.

 

the Mesh architecture's main issue afaik is latency.

[Trixanity]

14 minutes ago, Granular said:

It would be if x86 chips had become substantially better since then.

 

6 minutes ago, Amazonsucks said:

For enterprise chips yes. They get a 10x performance increase every 5 years or so. For desktop x86 CPUs, not so much. The difference between an i7-2600K and 6700K was like 25% in 5 years.

I think you two need to revise your idea of progress.

Sure, IPC gains are in the 25-35% range over the span of 7 years but plenty of other metrics exist. The benchmarks speak for themselves.

[porina]

3 minutes ago, AluminiumTech said:

Some youtuber (I can't remember which) did testing and found that Ring Bus worked optimally at up to 6 cores and worked less optimally at higher core counts.

Might have to see if I can find it later. Side question would be, how would you even test this? pcper did their own custom software to measure inter-core latency for example, and used it with good effect on Ryzen to examine the impact of CCX, but I'm not sure if they ever used it on Intel.

3 minutes ago, AluminiumTech said:

the Mesh architecture's main issue afaik is latency.

I think the main problem is that software optimised historically for ring doesn't know what to do with it. I like the large inclusive brute force nature of ring, compared to the smaller, slower smarter mesh. Most people in overclocking circles overclock it 50% or so to try and regain some of that deficit.

[TigerHawk]

why does it feel like kaby-lake was the worst time to upgrade my ivy bridge system....just before the generation where they finally moved up from 4 cores.

Sigh.

Well at least my ram will be reusable!

[Herman Mcpootis]

37 minutes ago, iamdarkyoshi said:

Its almost like there's been issues with their paste!

 

Gonna miss delidding tbh

on the bright side, no more voided warranty 

[iamdarkyoshi]

6 minutes ago, Herman Mcpootis said:

on the bright side, no more voided warranty 

Thats the fun part though...

[TVwazhere]

5 minutes ago, Herman Mcpootis said:

on the bright side, no more voided warranty 

Yes but then how will you impress girls who like adventurous and dangerous men? Zooming in on Google maps as an alternative to skydiving only works so may times. (Dont ask me how I know)

[KWelz]

I find it interesting how they are bringing the consumer line and HEDT line close together.  

[AngryBeaver]

This is good news. Obviously for people like me who delid it isn't as important, but it means we will see much more consistent thermals from this line. The TIM process intel was using created too much variance in performance.

 

Now that being said, this is an article for WCCFTECH and they are not known for their accuracy... more for their clickbait. So take this with a grain of salt. Anyways I am game for the 8c/16t for maybe a new build. Hoping to hit 5.4ghz+ on custom water.