Intel "Coffee Lake" (2017/2018, 14 nm)

I am an end user. I feel that Intel does care about me. These are not the days of the Pentium4->Core transition, when there were real gains to be had from upgrading the CPU two generations.
I lose the ability to upgrade, which is basically useless to me. I gain a more modern platform (no need for backwards compatibility) and cheaper (no costs from backwards compatibility) platform.
Just like I think Intel was correct to not change the socket notches. Yes, there may be one person in a million that chooses to ignore instructions and fries the CPU/board. OTOH, my new motherboard is now cheaper, because Intel does not have to qualify a new notch location.

I will admit that I am disappointed by the release timing of the cheap(er) chipsets.

 

Actually, this time Intel had good reasons to change the socket, the 6 core parts needed more power, maybe the high end chipsets could accommodate such requirements, but the low end B150 and B250? Those were likely never a good fit for the 8700K and the 8600K, possibly the med range chispets as well (H170, H270). Intel can't risk ignorant users frying their CPUs on low durability motherboards.

 

Edit.

 

That's a fair point, regarding the lower end chipsets. Though it could have been solved by not updating the BIOSes of the problematic chipsests with support of the newer CPUs. Which has a drawback , avereage Joe may buy the lower end chipsets expecting them to work with a 6 core CPU.

But maybe we're (or just me) looking too much into this. At least in the hardware.fr (donno why I'm visiting that site after Damien's gone.. Ah, I love french and don't like AT that much anymore) review, they point out how this is a rushed launch. They claim that Intel didn't plan for Coffee to happen this year (z390 chipset being late is a good argument for that). It's better that we got Cofee now with this socket incompatibility rather than later, I think

 

Better that people buy same chipset twice now, and then the real new chipset comes out in what, eight months or something and obsoletes their new purchase? Covfefe lake is nothing short of a desperate move to piss on AMD's parade. They release with seemingly basically no availability and will stay that way for several months according to rumors, and with a regurgitated chipset with a new label sprayed-on.

But at least the reviews are out there, where it shows their chip stomping on ryzen, holding people back from buying AMD even though they can't get the intel chip. It's a shitty move, but what can you do. Such is the way of capitalism.

 

The power is not a good reason. every single motherboard need to be able to deliver the max TDP that the socket can handle and this CPUs has the same TDP than previous gem so they consume equal amount of power.

 

And since both Z390 and mainstream boards are arriving Q1 2018 (right before Pinnacle Ridge launches) the only thing Intel will shit is their own pants.
It's Coppermine all over again.

 

You're thinking of the bugged (probably overstrained) 1133MHz chip meant to one-up the then-fastest K6?

 

No, the initial shortages of Coppermine P3's.

 

Okay, well you gotta remember moving from aluminium to copper wiring was a difficult step for the whole semiconductor industry at that time IIRC, AMD included. It's not an entirely analogous situation, covfefe lake is on essentially same process as kaby.

 

s

Intel does care about end users. It's just that end users mostly care for price/performance, not some technical details about sockets and form factors - these only matter for OEMs and enthusiasts.


The only vote that counts is with your money - you need to actually buy products to affect product decisions. No forum posts, blog entries, or online product surveys would count - only the hard earned cash does. This is just how it works, and that's why I put such an emphasis on sales figures.

As soon as AMD had CPUs offering significantly better price/performance , they have almost immediately eaten back a substantial part of the market share, even surpassing Intel in some territories.
http://digiworthy.com/2017/09/02/amd-surpass-intel-cpu-sales/

This is in spite of AMD not offering any backward compatibility for their new AM4 and TR4 platforms, so these processors necessarily include a new motherboard purchase (though these come with a vastly improved chipset, bootable M.2 PCIe x4 slots, USB 3.1 Gen2 Type-C ports, and high-durability polymer condensers).


And this is the sole reason why Intel announced of 4-core Core i3 and 6-core (and soon 8-core) Core i5/i7 processors for the mass market, in an apparently rushed decision.

OTOH until now, existing users of Core i5/i7 and AM3/AM3+ processor didn't have any good reason to spend $200-300 on mid-level CPU 'upgrades' with basically no changes to the architecture, because they offered had no significant performance gains - so it didn't even matter if they had the same socket or a new socket which required a new motherboard.

Oh, I apologize, your suggestion that Intel should have introduced a new mechanically incompatible socket does not really appear in that bingo table...

 

But Coppermine P3 used aluminium.
It was simply a rushed kneejerk reaction to OG Athlon (K7).

 

Sure, we'll have to see how availability progresses. You may be spot on. However, it's not unconcieveable that there will be a time window where Coffee will be available at launch prices and in some quantities, and the new chipset is not launch. If that time window exists, Intel's efforts today are not entirely without merit

 

It is not possible to damage the CPU. The revised pins were previously either reserved (not connected) or connected to ground - see my post above.

Also the new socket moves PROCC_DETECT#/SKTOCC# pin to a new location (AC38 instead of AB35), so Z370 would not detect an earlier processor and won't even start supplying VCC to the socket.

I wonder if 'real' 300-series Cannonlake chipsets (that is B360, Q310/Q350, H310/H370, and Z390) would be modified to detect and start Skylake/Kaby Lake processors...

Exactly. The enthusiast-level LGA2066 platform was designed to scale from 4 to 18 cores, 2 to 4 memory channels, and 16 to 44 PCIe lanes in multiple configurations, all within a single chipset, and that would require a lot of feature detection and configuration pins. LGA1151 doesn't seem to have anything like that, even with the high end chipsets.

Then why would Intel add more power/ground pins and update the electric specification?

TDP (heat dissipation) alone does not guarantee that actual processors would work reliably with existing power and ground rails...

 

It's not impossible that the electrical specification changes were really important and needed, but it's not likely either. If they can put Kaby - X and Skylake X on the same socket, two chips with identical architecture and same tdp won't pose much of a challenge

We wouldn't know why they performed the exact pin changes until we ask them. But it's engineering, small little incremeatal improvements are common place. Let's asume that's the case, until we get more data.

 

To justify breaking backwards compatability, of course.

 

7.2.1 Processor Power Rails DC Specifications

Table 7-2. Processor IA Core (Vcc) Active and Idle Mode DC Voltage and Current Specifications

Operating Voltage:
Min 0.55 V
Max 1.52 V

Idle Voltage:
Min 0 V
Max 0.55 V

Icc MAX :
<6th gen (Skylake)>
S(35W)-dual Core GT2/1 - Max 40 A
S(51W)-dual Core GT2/1 - Max 45 A
S(54W)-dual Core GT1 - Max 58 A
S(25W)-quad Core GT2/0 - Max 55 A
S(35W)-quad Core GT2 - Max 66 A
S(45W)-quad Core GT0 - Max 70 A
S(65W)-quad Core GT2 - Max 79 A
S(80W)-quad Core GT2/0 - Max 82 A
S(95W)-quad Core GT2 K-SKU - Max 100 A

<7th gen Kaby Lake)>
S-Processor Line
S(35W) Dual Core GT2/GT1 - Max 40 A
S(51W) Dual Core GT2/GT1 - Max 45 A
S(54W) Dual Core GT1 - Max 58 A
S(35W) Dual Core GT2 - Max 66 A
S(65W) Dual Core GT2 - Max 79 A
S(80W) Dual Core GT2/0 - Max 82 A
S(95W) Dual Core GT2 K-SKU - Max 100 A

<8th gen (Coffee Lake)>
S-Processor Line
(65W) Quad Core GT2 - Max 79 A
(95W) Quad Core GT2 - Max 100 A
(65W) Hexa Core GT2 - Max 133 A
(95W) Hexa Core GT2 - Max 138 A

Notes:
4. Processor IA core VR to be designed to electrically support this current.

Table 7-3. Processor Graphics (Vcc GT) Supply DC Voltage and Current Specifications

Operating Voltage:
Min 0.55 V
Max 1.52 V

Idle Voltage:
Min 0 V
Max 0.55 V

Icc MAX_GT :
<6th gen (Skylake)>
S(35W)-dual Core GT2/1 - Max 48 A
S(51W)-dual Core GT2/1 - Max 48 A
S(54W)-dual Core GT1 - Max 48 A
S(25W)-quad Core GT2 - Max 35 A
S(35W)-quad Core GT2 - Max 35 A
S(65W)-quad Core GT2/1 - Max 45 A
S(80W)-quad Core GT2 - Max 45 A
S(95W)-quad Core GT2 K-SKU - Max 45 A

<7th gen (Kaby Lake)>
S-Processor Line
(35W) Dual Core GT2/GT1 - Max 48 A
(51W) Dual Core GT2/GT1 - Max 48 A
(54W) Quad Core GT2 - Max 45 A
(35W) Quad Core GT2 - Max 35 A
(65W) Quad Core GT2 - Max 45 A
(95W) Quad Core GT2 K-SKU - Max 45 A

<8th gen (Coffee Lake)>
S-Processor Line
(65W) Quad Core GT2 - Max 45 A
(95W) Quad Core GT2 - Max 45 A
(65W) Hexa Core GT2 - Max 45 A
(95W) Hexa Core GT2 - Max 45 A

 

Another year, another round of CPUs is released, and 2600k lives to die another day.

i5 8400 and B360 will be a good mid-range combo it seems.

 

Well in the reviews the 8700k consume around 86W which is on pair with the 7700k isn't it?

 

No. I hope these Intel non-hyperthreaded CPUs would die already.