Because performance per core has gone nowhere so 6 of any cores will be pretty close in performance to any other.
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
Intel to launch 6 core Coffee Lake-S CPUs & Z370 chipset 5 October 2017
- Thread starter AthlonXP1800
- Start date
Because performance per core has gone nowhere so 6 of any cores will be pretty close in performance to any other.
I have an Intel cpu too, so I think personally all those that have an amd cpu should leave this thread and allow us Intel cpu people to discuss Intel cpu business
Especially those who have recently brought amd and have come to regret it.
What if you have both? Do you get stuck in some kind of infinite loop?
Cannon Lake is mobile only. Z390 is roadmapped for mid-2018 and supports 8-core. Given core-count pressure from AMD, current rumours are that Ice Lake will land late 2018; about a year after Coffee Lake.
That's their code name for 10nm shrink stage, no? They moved from Tick-Tock to Process-Architecture-Optimization(s), so for 10nm they have Cannonlake (Process, shrunk Coffee Lake), Icelake (Architecture) and Tigerlake (Optimization).
It usually takes them 1 year from the shrink to introduce the 'new' architecture (Ivy to 1st Haswell, Broadwell to Skylake), so I don't think Icelake will come in 2018.
It usually takes them 1 year from the shrink to introduce the 'new' architecture (Ivy to 1st Haswell, Broadwell to Skylake), so I don't think Icelake will come in 2018.
They probably won't since they've already started switching their bigger dies to polymer TIM.
For the overwhelming majority of their customers they only need the polymer TIM to ensure proper thermal conductivity at stock speeds, enthusiasts who will push the chips to the point where the polymer TIM becomes an issue are just an insignificant % of the market for them.
Intel's materials science engineers have a few papers like this one published which point out the issues with using indium-solder as TIM, namely reliability after intensive thermal cycling, the indium-solder will develop various defects that can even damage the die. Polymer on the other hand has much better long term reliability but at the cost of significantly worse thermal conductivity, indium solder has a thermal conductivity of around 80 W/(m-K) while the polymer TIM they use is at 6~8 W/(m-K).
Not to mention the process of actually soldering a silicon die to a nickel plated copper heat spreader isn't easy or gentle on die. The whole process would need temperatures up to 170 C and this is how many layers you'd need:
As their process gets denser (they market their 10nm as reaching 100 million transistors per sq. mm), it's quite possible soldering the die becomes unfeasible. More sensitive dies would just get destroyed in packaging, so from their point of view they have a cheaper alternative to soldering the chips that won't have a negative impact in packaging but with the cost of much lower thermal conductivity. As long as the chips run to spec, it doesn't really matter to them much and they can probably save a bit of money.
Additionally a thin enough polymer TIM layer can perform similarly to a thicker indium-solder sheet and it's quite possible that at one point they'll get there with their packaging. der8auer has a video where he delids and removes the indium solder sheet from a 6950X and you can see he gets a pretty hefty temperature decrease after he replaces the thick solder sheet with liquid metal.
So we probably won't ever see solder anymore for any of Intel's consumer dies, best get used to delidding.
I'm not sure about AMD, but they'll probably get there at one point once they start using denser processes (GloFo/Samsung's 14nm is closer to Intel's 22nm density wise), or they won't and will keep soldered dies as an advantage.
For the overwhelming majority of their customers they only need the polymer TIM to ensure proper thermal conductivity at stock speeds, enthusiasts who will push the chips to the point where the polymer TIM becomes an issue are just an insignificant % of the market for them.
Intel's materials science engineers have a few papers like this one published which point out the issues with using indium-solder as TIM, namely reliability after intensive thermal cycling, the indium-solder will develop various defects that can even damage the die. Polymer on the other hand has much better long term reliability but at the cost of significantly worse thermal conductivity, indium solder has a thermal conductivity of around 80 W/(m-K) while the polymer TIM they use is at 6~8 W/(m-K).
Not to mention the process of actually soldering a silicon die to a nickel plated copper heat spreader isn't easy or gentle on die. The whole process would need temperatures up to 170 C and this is how many layers you'd need:
As their process gets denser (they market their 10nm as reaching 100 million transistors per sq. mm), it's quite possible soldering the die becomes unfeasible. More sensitive dies would just get destroyed in packaging, so from their point of view they have a cheaper alternative to soldering the chips that won't have a negative impact in packaging but with the cost of much lower thermal conductivity. As long as the chips run to spec, it doesn't really matter to them much and they can probably save a bit of money.
Additionally a thin enough polymer TIM layer can perform similarly to a thicker indium-solder sheet and it's quite possible that at one point they'll get there with their packaging. der8auer has a video where he delids and removes the indium solder sheet from a 6950X and you can see he gets a pretty hefty temperature decrease after he replaces the thick solder sheet with liquid metal.
So we probably won't ever see solder anymore for any of Intel's consumer dies, best get used to delidding.
I'm not sure about AMD, but they'll probably get there at one point once they start using denser processes (GloFo/Samsung's 14nm is closer to Intel's 22nm density wise), or they won't and will keep soldered dies as an advantage.
Last edited:
Done the same thing, My 2500k was giving me stutter-jitters in BF1 put in a 3770k and all was smooth once again.
Define "much". 5%? 10% at max clocks?
non delided, custom loop with 240 rad, gigabyte board, ambient 29-32
Last edited:
Soldato
- Joined
- 19 Feb 2011
- Posts
- 5,849
Very nice, but what kinda temps will that jump to when its gaming on something like BF1 64 player? Id be happy for 5GHZ on all cores and 60c or so when gaming tbh
Soldato
- Joined
- 16 May 2007
- Posts
- 3,221
Very nice, but what kinda temps will that jump to when its gaming on something like BF1 64 player? Id be happy for 5GHZ on all cores and 60c or so when gaming tbh
That will be the interesting part, the 7700k runs hot already, with two extra cores it would needs serious cooling most likely.
Source? Becaus it might be a 7800X not 8700K
Soldato
- Joined
- 10 Nov 2006
- Posts
- 8,565
- Location
- Lincolnshire
Is 6 cores enough these days for gaming?, I mean - 6 cores are main stream now being supported by beta games like PUBG and lots of other AAA titles, I'm itching to upgrade from my [email protected] but want it to last me another 3/4 years like this chip has. 6 cores to me feels so 2013, I only game but I'm feeling like it should be 10+ for top end gaming in 2017? When I got my 5930k big titles just started supporting 4 cores and there wasn't many.
You're fine with that 5930K for the foreseeable future, games spanning across more than 4-6 threads doesn't mean they'll actually use those threads fully.
You'd probably notice more of an upgrade if you waited for Zen 2 or Intel's 10nm 8 cores.
You'd probably notice more of an upgrade if you waited for Zen 2 or Intel's 10nm 8 cores.