Coffeelake thread, benchmarks, reviews, input, everything.

[PeterScott]

CFL 8C for next year is not really the question anymore. As for Icelake the earliest realistic launch timeframe is Computex 2019 in June imho. Unless the slides refer to their foundry business which might differ to Intels own 10nm IP products.

It is very questionable, as it seems mostly to be based on wishful thinking. Unless Pinnacle ridge gives a big boost, not much changes from today in the next year.

Even if Pinnacle ridge increased Fmax by 10% with all cores running, it still wouldn't change much from today, and that is about the best case outcome for Pinnacle ridge.

Intel just made it's biggest leap in a decade, they really won't be in a rush to do it again.

 

[ozzy702]

Hello.
Question.....
Will the corsair h75 be enough for an i7 8700k?

Thanks

Absolutely but you'll be limited on how far you can OC. I'd have to think it would be good for 4.6-4.8ghz and reasonable temps, especially with a healthy avx offset.

 

[fastamdman]

Hello.
Question.....
Will the corsair h75 be enough for an i7 8700k?

Thanks

Depends on how high you are trying to over clock it and how good the silicon on the chip. Generally speaking, as long as you aren't going for the highest overclock possible, I don't see it being an issue buddy.

 

[Zucker2k]

It is very questionable, as it seems mostly to be based on wishful thinking. Unless Pinnacle ridge gives a big boost, not much changes from today in the next year.

Even if Pinnacle ridge increased Fmax by 10% with all cores running, it still wouldn't change much from today, and that is about the best case outcome for Pinnacle ridge.

Intel just made it's biggest leap in a decade, they really won't be in a rush to do it again.

I think Intel is not going to sit around and have AMD steal market share from them. It'll be a mistake of gross proportions if they do. They'll love to put AMD back into the pre-Ryzen hole, if you ask me. It's a no brainer, actually.

 

[Steve Van der Velde]

Depends on how high you are trying to over clock it and how good the silicon on the chip. Generally speaking, as long as you aren't going for the highest overclock possible, I don't see it being an issue buddy.

Thanks alot bro. That made my day

 

[PeterScott]

I think Intel is not going to sit around and have AMD steal market share from them. It'll be a mistake of gross proportions if they do. They'll love to put AMD back into the pre-Ryzen hole, if you ask me. It's a no brainer, actually.

If Intel was so worried about have a core count deficit, why isn't 8C Coffee Lake here now?

Also unlike fan wars on forums, companies don't overreact to every change in the market.

Right now there is sort of a tech parity for both companies. People can rationalize choosing either a Ryzen or Coffee Lake CPU, and we are likely in a quiet period for most of the next year where that won't change much.

 

[FIVR]

Who cares about solder vs TIM? 6900K can't clock that high, even though it's soldered. An 8-core CFL would stomp a 6900K into oblivion.

Wrong. Memory bandwidth is extremely important in games and the 6900k has twice the memory channels of coffee lake CPUs. Maybe you are correct in productivity apps, but who overclocks for those anyway?


Also, lots of people care about having processor that isn't dumping heat into the case unnecessarily due to crappy intel TIM . An 8 core coffee lake processor would be a furnace if overclocked and probably wouldn't have much of an advantage over a soldered 6900k with a good cooler and quad channel memory.

 

[deathBOB]

Also, lots of people care about having processor that isn't dumping heat into the case unnecessarily due to crappy intel TIM . An 8 core coffee lake processor would be a furnace if overclocked and probably wouldn't have much of an advantage over a soldered 6900k with a good cooler and quad channel memory.

Yeah that's not how that works. Kind of the opposite actually.

 

[Jan Olšan]

More importantly, i7-8700K didn't really get +2 cores by being much more power efficient. To a large degree it used up the headroom in the K-series Kaby Lake's TDP, and perhaps while simultaneously being aggressively binned (meaning, smaller percentage of chip output will be able to qualify for i7-8700K as opposed to 7700K before).

So eight Coffee cores on 14nm will likely require 125/130W TDP and will fill that number. Or it will require substantially lowered clocks. I still doubt this thing will even come though, there are still no ES sightings, no leaks, no mentions in roadmap. LGA 1151 is not LGA 2066 where they can repurpose bigger server die. They can't add cores on short (= several months) notice, unless they can make an LGA 1151 chip out of a Xeon D die.

 

[Bouowmx]

Once there is sufficient bandwidth, approximately 42.6 GB/s from 2 channels, 2666 MT/s, more bandwidth (but lower latency) does not help gaming performance.

Old test with Intel Core i7-5820K: PCWorld, comparing 2 vs 4 channels, with identical timings
Recent test with Intel Core i7-8700K, NVIDIA GeForce GTX 1080: TechPowerUp

 

[FIVR]

Yeah that's not how that works. Kind of the opposite actually.

The excess heat on the die transfers to the motherboard and is radiated into the case. This is different from how it functions when the die is cooled because it has solder and can easily transfer heat the the cooler or the loop.


But you bring up a good point - Why would anybody want a coffee lake processor that will degrade form being run constantly at high temps when they could get a 6900k that is soldered and will stay cool?

 

[DaveSimmons]

Once there is sufficient bandwidth, approximately 42.6 GB/s from 2 channels, 2666 MT/s, more bandwidth (but lower latency) does not help gaming performance.

Old test with Intel Core i7-5820K: PCWorld, comparing 2 vs 4 channels, with identical timings
Recent test with Intel Core i7-8700K, NVIDIA GeForce GTX 1080: TechPowerUp

Interesting, so you'll get maybe 1 more FPS going from 2666 up to 4000. Faster than 2666 seems like a waste of money for gaming on Coffee Lake CPUs.

 

[mikk]

It is very questionable, as it seems mostly to be based on wishful thinking.

You missed several sources if you call that wishful thinking. The wishful thinking part was Icelake when some OEM talked about 8C for Z390. It wasn't Icelake but Coffeelake 8C.

 

[PeterScott]

You missed several sources if you call that wishful thinking. The wishful thinking part was Icelake when some OEM talked about 8C for Z390. It wasn't Icelake but Coffeelake 8C.

What sources? From what I remember that one offhand forum comment, from support staff at niche laptop company, about Z390 chipset supporting 8 core CPUs, is THE ONLY "reference".

Leaping to the conclusion that we are getting 8C coffee lake, based on that flimsy hearsay is wishful thinking.

 

[LTC8K6]

The excess heat on the die transfers to the motherboard and is radiated into the case. This is different from how it functions when the die is cooled because it has solder and can easily transfer heat the the cooler or the loop.


But you bring up a good point - Why would anybody want a coffee lake processor that will degrade form being run constantly at high temps when they could get a 6900k that is soldered and will stay cool?

GPUs must degrade pretty fast if running at high temps causes chip degradation.

 

[FIVR]

GPUs must degrade pretty fast if running at high temps causes chip degradation.

Are you trying to suggest that high temperatures do not cause degradation of semiconductors?


Odd that they even bother cooling them then, right? I mean why bother?

 

[CHADBOGA]

Wrong. Memory bandwidth is extremely important in games and the 6900k has twice the memory channels of coffee lake CPUs. Maybe you are correct in productivity apps, but who overclocks for those anyway?


Also, lots of people care about having processor that isn't dumping heat into the case unnecessarily due to crappy intel TIM . An 8 core coffee lake processor would be a furnace if overclocked and probably wouldn't have much of an advantage over a soldered 6900k with a good cooler and quad channel memory.

How much does a 6900k cost?

 

[LTC8K6]

Are you trying to suggest that high temperatures do not cause degradation of semiconductors?


Odd that they even bother cooling them then, right? I mean why bother?

Yes, because running at 95C 24/7 is exactly the same as running at 80C 24/7...

But wait, who runs their CPU wide open 24/7 anyway?

What was I saying?

Oh yeah...

Yeah, why? Why bother to hook all those radiators and coolant lines and even refrigerators up to the chips? What's the point, really?

I'm suggesting that the manufacturers of said semiconductors know what they are doing when they set the TDP ratings and the throttle thresholds.

 

[Jan Olšan]

GPUs must degrade pretty fast if running at high temps causes chip degradation.

GPUs (graphics cards) actually fail a lot, much much more than processors. So those 80-97C temperatures that people suddenly started to consider normal thanks to TIM probably do have some effects in fact. Maybe some of those failures are blown VRMs but I'd say lots of it is going to be failed bumps/soldering and other thermal wear.

And that's with lifetimes that are much shorter than for most CPUs, because GPUs fall out of grace much faster as games' requirement climb up.

Actually - aren't high-power GPUs like the most likely to die part in a todays' PC? Maybe some OCers blow up their motherboards, but dead GPUs are the thing I hear about most often. The fact that mining causes more failures, manufacturers don't want to cover it with warranty and second-hand buyers consider it a hazard also tells you something. And that's with those cards being taxed continuously, so it is not the thermal cycling, but purely the temperature and heat itself.

Few years back, we thought that ~55C temperatures were good, and going higher was bad. people settling for 90C today...

 

[SangamonTaylor]

How much does a 6900k cost?

2 to 2.5 coffee lakes.

 

[LTC8K6]

GPUs (graphics cards) actually fail a lot, much much more than processors. So those 80-97C temperatures that people suddenly started to consider normal thanks to TIM probably do have some effects in fact. Maybe some of those failures are blown VRMs but I'd say lots of it is going to be failed bumps/soldering and other thermal wear.

And that's with lifetimes that are much shorter than for most CPUs, because GPUs fall out of grace much faster as games' requirement climb up.

Actually - aren't high-power GPUs like the most likely to die part in a todays' PC? Maybe some OCers blow up their motherboards, but dead GPUs are the thing I hear about most often. The fact that mining causes more failures, manufacturers don't want to cover it with warranty and second-hand buyers consider it a hazard also tells you something. And that's with those cards being taxed continuously, so it is not the thermal cycling, but purely the temperature and heat itself.

Few years back, we thought that ~55C temperatures were good, and going higher was bad. people settling for 90C today...

Well, no one said that temperature doesn't degrade semiconductors.
I don't think people run their CPUs at the same temps and times they let their GPUs get to.
People generally don't let their CPU run very hot, so it's never a problem.
If the CPU does get very hot, it's briefly during a benchmark run or such.
GPUs tend to be pushed hard for longer periods at higher temps.

 

[TahoeDust]

Also, lots of people care about having processor that isn't dumping heat into the case unnecessarily due to crappy intel TIM .

How does that work? Wouldn't the same amount of heat go into the case either way? The use of solder just makes it easier to get the heat away from the CPU. It takes roughly the same amount of energy solder or not. If anything solder should actually make getting the heat from the CPU to the case easier. Even if it is transferred to the motherboard, energy is energy.

 

[jpiniero]

What sources?
Leaping to the conclusion that we are getting 8C coffee lake, based on that flimsy hearsay is wishful thinking.

Don't remember, but specifically mentioned 8+2 CFL. It does makes sense; it's extremely unlikely that Intel would be able to deliver a 10 nm die that big in 2018, and Intel's going to want to have something new to sell this time next year.

 

[PeterScott]

Don't remember, but specifically mentioned 8+2 CFL. It does makes sense; it's extremely unlikely that Intel would be able to deliver a 10 nm die that big in 2018, and Intel's going to want to have something new to sell this time next year.

I have been following this saga all along. You probably don't remember any specifics, because there were no other sources. People just took that one bit of forum hearsay, and speculated wildly, mixing in their desires...

 

[ozzy702]

Wrong. Memory bandwidth is extremely important in games and the 6900k has twice the memory channels of coffee lake CPUs. Maybe you are correct in productivity apps, but who overclocks for those anyway?


Also, lots of people care about having processor that isn't dumping heat into the case unnecessarily due to crappy intel TIM . An 8 core coffee lake processor would be a furnace if overclocked and probably wouldn't have much of an advantage over a soldered 6900k with a good cooler and quad channel memory.

None of that makes any sense. If quad channel memory mattered in gaming then the 6900k would be besting the 8700k but it doesn't come close in anything but possibly Ashes. 8700k has significantly lower latency than 6900k, higher IPC and faster clock speeds, the delta would only widen with two additional cores.

Power usage is power usage. A CPU using 200w with TIM will "dump heat into the case" at roughly the same rate as a 200w CPU with solder. Yes, if you're using a large water cooler it may be able to dump slightly more heat out of the case with a soldered CPU but the difference is negligible. The fact of the matter is that at max OC the 8900k would use less power than the 6900k (~215W @ 4.3ghz) and be significantly faster in every single way.