Intel Core i9-9900K and i7-9700K to feature soldered IHS

[HEXUS]

Move would appease enthusiast demand, and deliver the best thermal performance.

Read more.

[watercooled]

If true it's probably out of necessity - 8 Skylake cores at close to 5GHz means a lot of heat to shift and would probably throttle quite significantly. It would be quite funny to see if the 8 core can get away with a basic air cooler when the 6 core struggles regardless of cooler used!

[GinoLatino]

it's probably out of necessity

Agreed

[outwar6010]

Isn't the bigger story no more hyper threading on i7s?

[watercooled]

Assuming it's true, it's just a way of differentiating between the different tiers.

[CAT-THE-FIFTH]

TBF,it would not surprise me if the 8C/8T CPU runs games better than a 6C/12T CPU,since games tend to scale better with cores than virtual threads.

Edit!!

It also means people who were defending Intel not using TIM for XYZ reasons,ie,its for the environment,it will damage the chip,etc can see the real reason - cost.

[Corky34]

If Intel's primary reason for changing from a solder to a grease base TIM was cost then now would be a bad time to suddenly change tack, if anything they'd be trying to keep production costs low so they could maintain higher margins while cutting prices.

[watercooled]

They may not have a choice now. An 'i9' with lower clocks than previous CPUs probably wouldn't be too attractive, if they're up against a thermal wall as I suspect. What they have over Ryzen is their perceived superiority in some games - lowering clocks closes that gap again, and the extra cores probably wouldn't help. Their ability to keep high clock speeds is critical at the moment, given they're still on 14nm Skylake with no indication of that changing any time soon.

[CAT-THE-FIFTH]

If Intel's primary reason for changing from a solder to a grease base TIM was cost then now would be a bad time to suddenly change tack, if anything they'd be trying to keep production costs low so they could maintain higher margins while cutting prices.

There were tons of reasons people on forums were giving for Intel not to use solder - everything from it damages the CPUs made on their FINFET process,to Intel protecting the environment,to solder having a shorter lifespan,etc. It was always about cost - the solders are expensive. Intel was just maximising margins since they knew all the "enthusiasts" would delid them and everyone else would have to pay extra in cooling.

Ryzen caught Intel unaware,so they actually thought they could stick with their crappy TIM based solution,which is horrendous.

The only way Intel can try and outdo AMD now is via sheer clockspeed - remind you of a certain CPU released just before the Athlon 64??!!

Now 10NM seems to be still in a crap state,and Intel is now on its 4th iteration of Skylake and still on 14nm,they have to use solder now,since AMD is on track to release its 7NM CPUs next year. They can't get away with cost cutting now,since they have to use solder otherwise the CPU will probably cook itself. 8 cores on 14NM at 5GHZ.

How amazing Intel has gone from 4C/8T with TIM KL last year to 8C/16T with solder within 18 months,and then attempted to use HEDT CPUs with X299 and felt flat on their face.

The 8C CFL is basically Intel's CFL Emergency Edition CPUs - it proves that there was no real reason NOT to use solder.

Its not only GF 7NM AMD is using,but TSMC 7NM for next year's Epyc:

https://www.anandtech.com/show/13122...fabbed-by-tsmc



Sampling this year with introduction next year. If 7NM Epyc is close to sampling that means 7NM consumer Ryzen chips are also close to sampling too since AMD uses the same basic core design.

So 14NM CFL against 7NM Ryzen 3 next year.

Even if AMD can't hit 5GHZ on Ryzen 3,they will probably increase core counts by at least 50% or 100%,improve IPC,cache latency and power consumption,whilst using a denser set of process nodes.

Next year is going to be VERY interesting indeed!

They may not have a choice now. An 'i9' with lower clocks than previous CPUs probably wouldn't be too attractive, if they're up against a thermal wall as I suspect. What they have over Ryzen is their perceived superiority in some games - lowering clocks closes that gap again, and the extra cores probably wouldn't help. Their ability to keep high clock speeds is critical at the moment, given they're still on 14nm Skylake with no indication of that changing any time soon.

Also apparently the reason the mesh interconnect was introduced was apparently since the current ring bus has issues past 8 to 10 cores. All the noise is indicating that 10NM cannot clock higher than 14NM in its current state,so unless the 10NM successor to CFL has a huge IPC increase,I suspect single threaded performance will stagnate,and they might be on the edge of how many cores they can add unless they move to the mesh interconnect.

However,too many games seem to not be as well optimised for the CPUs using the mesh interconnect(which to a degree also affects Ryzen).



Remember the Core i7 7800X is using SKL cores,has more memory bandwidth than the Ryzen 5 2600 and higher clockspeeds.

[watercooled]

In theory CNL was to be a 'tick' of SKL, so nothing major, if anything, in terms of IPC increases. I'm not sure where the internal roadmap stands now but on early rumours of 10nm troubles I was expecting something more like Broadwell where the first iteration on the node didn't make it to desktop where high clocks are desired, but was still useful for mobile where its power savings desirable.

WRT ring bus - I think even AMD have basically said that; the reason for building Ryzen the way they did is due to scaling problems past a few cores.

[CAT-THE-FIFTH]

In theory CNL was to be a 'tick' of SKL, so nothing major, if anything, in terms of IPC increases. I'm not sure where the internal roadmap stands now but on early rumours of 10nm troubles I was expecting something more like Broadwell where the first iteration on the node didn't make it to desktop where high clocks are desired, but was still useful for mobile where its power savings desirable.

WRT ring bus - I think even AMD have basically said that; the reason for building Ryzen the way they did is due to scaling problems past a few cores.

Providing GF/TSMC 7NM process nodes don't hit issues during production,it will be really interesting if AMD ends up with 7NM CPUs against Intel 14NM ones!

AMD has only leapfrogged Intel once when it came to new nodes IIRC.

[Corky34]

They may not have a choice now. An 'i9' with lower clocks than previous CPUs probably wouldn't be too attractive, if they're up against a thermal wall as I suspect. What they have over Ryzen is their perceived superiority in some games - lowering clocks closes that gap again, and the extra cores probably wouldn't help. Their ability to keep high clock speeds is critical at the moment, given they're still on 14nm Skylake with no indication of that changing any time soon.

You're always up against a thermal wall it just depends on where you want that wall to be, if they do use solder it will certainly give them more headroom.

There were tons of reasons people on forums were giving for Intel not to use solder - everything from it damages the CPUs made on their FINFET process,to Intel protecting the environment,to solder having a shorter lifespan,etc. It was always about cost - the solders are expensive. Intel was just maximising margins since they knew all the "enthusiasts" would delid them and everyone else would have to pay extra in cooling.

And those reasons are all more sound than it was because of cost, but don't take my word for it, there's plenty of evidence that it was for reasons other than cost but i suspect you already know that as you, like myself, have had this discussion with many people, on many forums over the years.

Ryzen caught Intel unaware,so they actually thought they could stick with their crappy TIM based solution,which is horrendous.

The only way Intel can try and outdo AMD now is via sheer clockspeed - remind you of a certain CPU released just before the Athlon 64??!!

Now 10NM seems to be still in a crap state,and Intel is now on its 4th iteration of Skylake and still on 14nm,they have to use solder now,since AMD is on track to release its 7NM CPUs next year. They can't get away with cost cutting now,since they have to use solder otherwise the CPU will probably cook itself. 8 cores on 14NM at 5GHZ.

How amazing Intel has gone from 4C/8T with TIM KL last year to 8C/16T with solder within 18 months,and then attempted to use HEDT CPUs with X299 and felt flat on their face.

The 8C CFL is basically Intel's CFL Emergency Edition CPUs - it proves that there was no real reason NOT to use solder.

Its not only GF 7NM AMD is using,but TSMC 7NM for next year's Epyc:

https://www.anandtech.com/show/13122...fabbed-by-tsmc



Sampling this year with introduction next year. If 7NM Epyc is close to sampling that means 7NM consumer Ryzen chips are also close to sampling too since AMD uses the same basic core design.

So 14NM CFL against 7NM Ryzen 3 next year.

Even if AMD can't hit 5GHZ on Ryzen 3,they will probably increase core counts by at least 50% or 100%,improve IPC,cache latency and power consumption,whilst using a denser set of process nodes.

Next year is going to be VERY interesting indeed!

Sorry but i only caught the start and end of that part as i struggle to find the relevance in a great big images that I've seen a hundred times and claims that something is crappy and horrendous, it's to much subjectivity for me to handle and a little fanboy'ish.

Personally every year is a very interesting year for me when it comes to tech, it's why it interests me so much as unlike a lot of fields its not stagnated to the same degree.

[CAT-THE-FIFTH]

You're always up against a thermal wall it just depends on where you want that wall to be, if they do use solder it will certainly give them more headroom.



And those reasons are all more sound than it was because of cost, but don't take my word for it, there's plenty of evidence that it was for reasons other than cost but i suspect you already know that as you, like myself, have had this discussion with many people, on many forums over the years.

Yet,AMD has had no issues as have other companies and now to save face you are randomly quoting an enthusiast overclocking site. Funny despite all those rather flimsy excuses,Intel now seems to find using solder perfectly fine and it wouldn't surprise me one bit other SKUs use it. BTW,all the people on other forums who tried quoting that site before,are now all suddenly quiet elsewhere. Wonder why!?

Funny how decent solder costs more than TIM though!

Its like when people believed Apple when they said people were holding the iPhone wrong!

Sorry but i only caught the start and end of that part as i struggle to find the relevance in a great big images that I've seen a hundred times and claims that something is crappy and horrendous, it's to much subjectivity for me to handle and a little fanboy'ish.

Personally every year is a very interesting year for me when it comes to tech, it's why it interests me so much as unlike a lot of fields its not stagnated to the same degree.

Sorry,I only caught the start and end of what you said as i struggle to find the relevance in hurling person insults at people since you want to attack them and call them fanbois - the biggest fanbois on forums attack people calling others fanbois instead of discussing what is said. Yay,we can play fanboi wars?? Rather silly,right?

The fact you have to resort to calling people childish names,says a lot about how you feel about your own stance. You do it all the time.

The funny thing is you didn't even read the link and instantly reacted,when the link was posted TODAY which makes you look an utter tool TBH.First time its posted on Hexus. Yes,a 100 times. INDEED!



The problem with people like you is that you end up using tired internet cliches to bury stuff and not discuss anything.

I have criticised Ryzen countless times due to its issues in games like Fallout 4,or its issues at launch,especially with the motherboards,and if you haven't seen that is just internet selective memory on your part.

So I have had people call me an AMD fanboi and an Intel fanboi at different times on different forums,which makes me laugh.

So I will link you to some other news:

https://www.extremetech.com/computin...-14nm-products

When asked about the future of Intel’s 14nm (we’re up to 14nm+++ at this point, if Intel continues to use that metric), Murthy notes:

[W]e found tremendous intra-node capability within our 14-nanometer process. In fact from the very first generation of our 14-nanometer to the latest generation of 14-nanometer product, we’ve been able to deliver over 70% performance improvement as a result of those intra-node modifications and desirable changes. And that’s quite frankly Harlan has given us the ability to make sure that we get 10-nanometer yields right before we go into mainstream production. And so, therefore we’re comfortable with the 14-nanometer roadmap that will give us leadership products in the next 12 to 18 months, as we seek to optimize the cost structure and yields of our 10-nanometer portfolio.

So in May 2018 they said they will be using 14NM for the next 12 to 18 months on "leadership" products. So that means between May and end of 2019 where they will still be using 14NM.

But isn't 8C 14NM CFL being released next month or the month after,ie,August or September. So unless 8C CFL is going to last only six months,9 to 12 months at least seems a good guess at least until it is replaced.

AMD said they would be keeping to 12 month cadences,so let me see,that would mean March to April 2019 for 7NM products,dual sourced at TSMC and GF.

They said they are sampling in 2H 2018,so give a few months for production and see where that lands!

But,but,but FANBOIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII!

[Corky34]

Yet,AMD has had no issues as have other companies and now to save face you are randomly quoting an enthusiast overclocking site. Funny despite all those rather flimsy excuses,Intel now seems to find using solder perfectly fine and it wouldn't surprise me one bit other SKUs use it. BTW,all the people on other forums who tried quoting that site before,are now all suddenly quiet elsewhere. Wonder why!?

Funny how decent solder costs more than TIM though!

Its like when people believed Apple when they said people were holding the iPhone wrong!

And yet AMD use silicon TIM on their low end smaller chips, if as you say Intel did it to penny pinch how much extra margin do you think using silicon based TIM added to their 60 odd percent margin?

Also how is posting a link to an article by what most people consider a highly experienced overclocker a random quote, how about you say why you believe der8auer is wrong instead of trying to just discounting what he says as 'random'.

And of course solder costs more than a silicon based TIM, a full water cooling loop cost more than air and LN2 cost more than water, your point is? Just because something cost more or less than something else doesn't mean cost is the only factor in using one thing over another.

Sorry,I only caught the start and end of what you said as i struggle to find the relevance in hurling person insults at people since you want to attack them and call them fanbois - the biggest fanbois on forums attack people calling others fanbois instead of discussing what is said. Yay,we can play fanboi wars?? Rather silly,right?

The fact you have to resort to calling people childish names,says a lot about how you feel about your own stance. You do it all the time.

The funny thing is you didn't even read the link and instantly reacted,when the link was posted TODAY which makes you look an utter tool TBH.First time its posted on Hexus. Yes,a 100 times. INDEED!

Sorry but what insult have i hurled? You're the one throwing insults around by calling things crappy and horrendous, and now resorting to posting silly videos that have no relevance whatsoever.

And yes i read the link so please stop throwing accusations around like confetti, so far the only person throwing around insults has been you IMO.

The problem with people like you is that you end up using tired internet cliches to bury stuff and not discuss anything.

I have criticised Ryzen countless times due to its issues in games like Fallout 4,or its issues at launch,especially with the motherboards.

So I have had people call me an AMD fanboi and an Intel fanboi at different times on different forums,which makes me laugh.

So I will link you to some other news:

https://www.extremetech.com/computin...-14nm-products




So in May 2018 they said they will be using 14NM for the next 12 to 18 months on "leadership" products. So that means between May and end of 2019 where they will still be using 14NM.

But isn't 8C 14NM CFL being released next month or the month after,ie,August or September. So unless 8C CFL is going to last only six months,9 to 12 months at least seems a good guess at least until it is replaced.

AMD said they would be keeping to 12 month cadences,so let me see,that would mean March to April 2019 for 7NM products,dual sourced at TSMC and GF.

They said they are sampling in 2H 2018,so give a few months for production and see where that lands!

But,but,but FANBOIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII!

The problem with people like me? What exactly is my 'problem' then?

And what does how long Intel is going to be using a particular node size got do do with silicon or solder TIM, and you say I'm the one deflecting?

[[GSV]Trig]

Isn't the bigger story no more hyper threading on i7s?

This, I wonder how they'll pitch the i3/5/7/9, are we going back to the days of giving CPU's names that dont really make sense..

[watercooled]

You're always up against a thermal wall it just depends on where you want that wall to be, if they do use solder it will certainly give them more headroom.

Not really, you'll run into power, stability, binning or perf/watt issues before you hit the point where temps stop you clocking any higher. It's only fairly recently that we've seen stock desktop CPUs throttling back due to hitting temperature limits with an OEM cooler. CPUs with a decent thermal path like Piledriver, Nehalem, etc could dissipate many hundreds of watts without hitting any real thermal walls. My 7700 routinely hits 80C+ at stock with a 212 cooler. That's *really* close to the thermal limit, and it's not even the highest clocked bin.

Given the lack of headroom with KBL and more so with CFL, and with even higher clocks expected on the next gen, constantly bumping off the throttle point with even a partial load is a real possibility.

are we going back to the days of giving CPU's names that dont really make sense..

Did we ever leave those days?