Discussion Intel current and future Lakes & Rapids thread

[IntelUser2000]

Cannonlake is not a shrink of Kabylake. This is confirmed because Intel already confirmed that Cannonlake comes with Gen10 graphics. That rules out a shrink of Kabylake.

Technically there's no such thing as a "pure shrink" anymore. It take so much more work. Traditionally though, Cannonlake should see few % improvements in the CPU core. Probably 2%. 😛

Yes, they released at least one picture where 10nm is clearly below 14nm++ in terms of transistor performance, so it's clear that we won't see any desktop SKUs based on the 1st 10nm gen..

Not only that, they said the second iteration of 10nm, the 10nm+ is slightly slower than 14nm++. It'll take 10nm++ to beat it. So overall the 10nm brings even less of a benefit to CPUs than 14nm did. Isn't it great?

Clearly performance is the reason why they don't have it for performance parts. I'd even think based on what they are saying, it'll be the primary driver not having it on high performance parts, not yields.

Yeah the IGP is Gen10. I don't know if there is anything really notable in it however.

Early leaks indicated 40EUs coming to GT2. That'd be nice.

 

[krumme]

Do we have any leaks or rumors for freq or perf?

 

[TheF34RChannel]

Do we have any leaks or rumors for freq or perf?

Which processor family are you asking about?

 

[TheF34RChannel]

Another 6-core Intel Coffee Lake CPU has appeared on the Geekbench database

 

[TeknoBug]

Price of motherboards for socket 2066 is ridiculous.

 

[mikk]

Another 6-core Intel Coffee Lake CPU has appeared on the Geekbench database

Nothing new, already posted some days ago.

 

[formulav8]

Not only that, they said the second iteration of 10nm, the 10nm+ is slightly slower than 14nm++. It'll take 10nm++ to beat it. So overall the 10nm brings even less of a benefit to CPUs than 14nm did. Isn't it great?

Clearly performance is the reason why they don't have it for performance parts. I'd even think based on what they are saying, it'll be the primary driver not having it on high performance parts, not yields.

Didn't know Intel was having such transistor speed issues on 10nm. That's probably the biggest problem with their 10nm process and not yields like you mentioned.

 

[Azuma Hazuki]

How do those Geekbench scores compare to current machines? I'm seeing higher single *and* multithread scores for i7-7700K systems. Though, to be fair, they're 4C/8T at 4.2 Ghz, not 6C/12 at 3.2...

 

[Bouowmx]

Look in aggregated results: https://browser.primatelabs.com/processor-benchmarks

Intel Core i7-7700K scores 5745 single, 18752 multi

2 months left until a supposed August-September launch of Coffee Lake and names are still "Intel 0000". Should there be finalized product names and specifications sooner or later?

 

[TheF34RChannel]

Look in aggregated results: https://browser.primatelabs.com/processor-benchmarks

Intel Core i7-7700K scores 5745 single, 18752 multi

2 months left until a supposed August-September launch of Coffee Lake and names are still "Intel 0000". Should there be finalized product names and specifications sooner or later?

Much later on the time line, these early 'leaks' comprise ES and not necessarily finalized specifications (in terms of (clock*) speeds) nor is the microcode updated with a SKU name/the software might not detect it yet. It's going to be i7-8700K or i7-8760K or something anyway, personally I don't find the name very interesting.

*I would be sorely disappointed if Coffee Lake-S 6C/12T would launch with 3.1/4.2GHz frequencies at 14nm++. Surely the optimized process has got more in it than that even with six cores. Only caring about the boost clock, and leaving manual OC out of the equation, said boost should be nothing short of 4.5GHz, in my opinion.

 

[LTC8K6]

Why is the 7350K 8% faster than the 7700K in single core at the same clock speed? 😀

 

[witeken]

Didn't know Intel was having such transistor speed issues on 10nm. That's probably the biggest problem with their 10nm process and not yields like you mentioned.

Choosing to optimize for power instead of performance is for a large part an intentional decision.

 

[TheF34RChannel]

Why is the 7350K 8% faster than the 7700K in single core at the same clock speed? 😀

Because GeekBench is the most peculiar thing ever and no one seems really sure what to make of any entry? Maybe.

 

[TheF34RChannel]

Choosing to optimize for power instead of performance is for a large part an intentional decision.

Yeah, sadly it's all about that power saving stuff nowadays ugh

 

[mikk]

Why is the 7350K 8% faster than the 7700K in single core at the same clock speed? 😀

Core clock isn't the same.

 

[LTC8K6]

Core clock isn't the same.

Base clock is the same, but we would expect the 7700K to turbo up to 4.5 and beat the 7350K handily.

For some reason CPU World has 4.2Ghz listed for the 7350K and then 4.6Ghz listed as it's "maximum speed".

 

[TheF34RChannel]

Hmm seems off? No idea where they got that boost from, not from the ark:

https://ark.intel.com/products/97527/Intel-Core-i3-7350K-Processor-4M-Cache-4_20-GHz

 

[mikk]

Base clock is the same, but we would expect the 7700K to turbo up to 4.5 and beat the 7350K handily.

For some reason CPU World has 4.2Ghz listed for the 7350K and then 4.6Ghz listed as it's "maximum speed".

Base clock doesn't matter, there are lots of overclocked Geekbench results included in this list.

 

[LTC8K6]

Base clock doesn't matter, there are lots of overclocked Geekbench results included in this list.

It's probably due to overclocking, but I'd expect the 7700K to be overclocked as well?

We can see some massive 7350K single and multi core scores, which are certainly due to overclocks and top speed ram.

https://browser.primatelabs.com/v4/cpu/search?q=7350K

 

[mikk]

It's probably due to overclocking, but I'd expect the 7700K to be overclocked as well?

We can see some massive 7350K single and multi core scores, which are certainly due to overclocks and top speed ram.

https://browser.primatelabs.com/v4/cpu/search?q=7350K

There are more OC 7350K results included relative to 7700k which is why 7350K is above in singlethread there.

 

[Eddward]

Look in aggregated results: https://browser.primatelabs.com/processor-benchmarks

Intel Core i7-7700K scores 5745 single, 18752 multi

2 months left until a supposed August-September launch of Coffee Lake and names are still "Intel 0000". Should there be finalized product names and specifications sooner or later?

Well, the thing is that current rumors are saying about Q4 launch so basically not sooner than October this year... so realistically we are closer to 4 months from launch... so this is maybe one of the reasons why we are still seeing only "Intel 0000"...

 

[TheF34RChannel]

Well, the thing is that current rumors are saying about Q4 launch so basically not sooner than October this year... so realistically we are closer to 4 months from launch... so this is maybe one of the reasons why we are still seeing only "Intel 0000"...

Correct me if I'm wrong, however I think us still seeing Intel 0000 only depends on what individual chip they're using - i.e. there may be chips about that correctly show the name and surely Intel wants to keep a lid on it for now. So in short, it doesn't mean anything perse.

 

[Eddward]

yes, you are right, it's just a speculation, but there is still a possibility that there is some connection... we should see much more in July if the August launch is real

 

[TheF34RChannel]

yes, you are right, it's just a speculation, but there is still a possibility that there is some connection... we should see much more in July if the August launch is real

Absolutely! And from that we basically can guess whether or not it'll be August or October+

 

[Arachnotronic]

Correct me if I'm wrong, however I think us still seeing Intel 0000 only depends on what individual chip they're using - i.e. there may be chips about that correctly show the name and surely Intel wants to keep a lid on it for now. So in short, it doesn't mean anything perse.

Mobile Coffee Lake won't arrive until 1Q 2018, so those are still a ways out.