Coffee Lake or Ice Lake?

[jimbothejester]

Is there any reason I shouldn’t go with an i7 8700k right now and wait for Ice Lake? Just curious. I’m trying not to fall in the hole of waiting for the next new tech to release in order to justify my purchase.

 

[InvalidError]

Surely an 8 core chip would not make sense until at least 2019 or later?

AMD's 1700/1800 launched in 2017 and the 2700/2800 refresh which can be expected to be up to 17% faster on heavily threaded workloads based on leaks are coming in about two months. Intel may really need an octo-core mainstream CPU to remain competitive.

 

[Gon Freecss]

Surely an 8 core chip would not make sense until at least 2019 or later?

AMD's 1700/1800 launched in 2017 and the 2700/2800 refresh which can be expected to be up to 17% faster on heavily threaded workloads based on leaks are coming in about two months. Intel may really need an octo-core mainstream CPU to remain competitive.

Yeah, you got a source on that? I'm gonna guess no.

AMD will likely have less than 10% performance increase overall with Pinnacle Ridge (clocks and other improvements). I'd argue you'd see a wall >4.4GHz with it, while it takes up to >5.4GHz to see a wall with Coffee Lake. Also, Coffee Lake has a ~20% IPC advantage over AMD overall. They won't catch up to Intel at all. That's why in a lot of computing tasks, the 8700K rivals the 1800X, and even beats it, despite being a 6C vs an 8C.

Icelake is pushing for an end of year release, and Q1 2019 at the latest. If OP could wait that long, then by no means. If not, the 8700K is a monster.

 

[razamatraz]

An 8 core mainstream chip within such a short space of Coffee Lake would seem a little ridiculous.. I'm just about to move on an 8700K myself. Surely an 8 core chip would not make sense until at least 2019 or later?

It's possible that Intel may choose to leave that in its HEDT space, but given that you can get an R7 1700 for $230 that easily clocks to 4.0 and there are Zen+ chips due soon Intel may not have a choice. The move to 10nm also makes sense as this is the first real process node change since Broadwell and in theory that gives them room on the same size package to place more more cores.

Usually when moving to a new process node Intel has had trouble maintaining the clock speeds so compensating with more cores also seems reasonable. I think it would be too risky for Intel not to release an 8 core if they are able to make one in the 1151v2 socket. The 8700K is a great CPU but forcing you to buy a much more expensive platform to get into 8 core seems risky when you can do it on a $90 MB with AMD, ad althoguh I don;t think the 1800X replacement will get up to the IPC and sheer max clocks of an 8700K it's multicore performance will be far better (an 1800X is already better). If they can do it economically I'd say 90% they do do it. You never know with Intel though; it's notorious for not giving anyone what they want.

 

[razamatraz]

Surely an 8 core chip would not make sense until at least 2019 or later?

AMD's 1700/1800 launched in 2017 and the 2700/2800 refresh which can be expected to be up to 17% faster on heavily threaded workloads based on leaks are coming in about two months. Intel may really need an octo-core mainstream CPU to remain competitive.

The 2700/2800 refresh is not Zen2 it is Zen+ and at absolute best will have 5% IPC improvements (although it really is a refresh with better RAM support so probably 0% RAM speed held equal.) and maybe 5% Clock speed improvements. Zen2 is due later and 17% overall is possible but certainly not guaranteed.

 

[InvalidError]

Yeah, you got a source on that? I'm gonna guess no.

The 2700/2800 refresh is not Zen2 it is Zen+ and at absolute best will have 5% IPC improvements (although it really is a refresh with better RAM support so probably 0% RAM speed held equal.) and maybe 5% Clock speed improvements.

If you spent all of 5 seconds doing a google search for "pinnacle ridge performance leak" (that's Ryzen+, not Zen 2), you'll find plenty of coverage about the Ryzen 1600 vs 2600 leak.
http://lmgtfy.com/?q=pinnacle+ridge+performance+leak

 

[razamatraz]

Yeah, you got a source on that? I'm gonna guess no.

The 2700/2800 refresh is not Zen2 it is Zen+ and at absolute best will have 5% IPC improvements (although it really is a refresh with better RAM support so probably 0% RAM speed held equal.) and maybe 5% Clock speed improvements.

If you spent all of 5 seconds doing a google search for "pinnacle ridge performance leak" (that's Ryzen+, not Zen 2), you'll find plenty of coverage about the Ryzen 1600 vs 2600 leak.
http://lmgtfy.com/?q=pinnacle+ridge+performance+leak

And there's your problem. You believe the absolute best case shown by the leaks without bothering to question why. if you bothered to read any of those leaks you would see that the entire speed change is likely coming from increased clock speeds, not IPC. No one is surprised that the stock base and boost are higher, but who cares, the old Zen chips already overclocked far above those boost clock speeds. Unless the new ones overclock higher than the old ones overclocked there will be no speed change.

If you believe that a stock and installed to standard 2600 is going to be 31% faster than a stock and installed to standard 1600 based on one geekbench run you are delusional. One geekbench score showing that sure...because some guy undervolted and downclocked his 1600 and ran it with 2133 RAM in a case with no airflow using a turnip for a cooler and the tester with the 2600 OC'ed it to 4.3 and Ran 3400MHz RAM. A 7% difference between an OC on the 2600 and an OC on the 1600...sure, and I hope so or better...otherwise what's the point? It's not like there's a VEGA GPU on the 2600.

Come on, the 2600 will be X% faster than the 1600 where X is its clockspeed divided by the 1600's clockspeed minus 1 and then multipied by 100 and then fudged for RAM speed. It's the same chip on 12nm, this is just like the move from Sandy to Ivy or Haswell to Broadwell (Although cache changes actually helped braodwell IPC there the clock speeds went down from Devil's Canyon)

Geekbench scores are collected from randoms....the range they fall into even for the same chip is huge.

 

[InvalidError]

And there's your problem. You believe the absolute best case shown by the leaks without bothering to question why. if you bothered to read any of those leaks you would see that the entire speed change is likely coming from increased clock speeds, not IPC.

Who cares how the performance gains are achieved? kol12 asked why Intel should bother releasing 8-core mainstream CPUs in 2018 and AMD bumping its CPUs' stock performance by 15-17% would definitely put the heat on Intel regardless of the combination of clock frequency bumps, memory controller and fabric latency improvements (the main reason why Ryzen is so sensitive to memory clock frequency), minor IPC gains from core/cache tweaks, better boost management, etc. that AMD may have used to get there if the leaks turn out to be accurate.

 

[razamatraz]

AMD bumping its CPUs' stock performance by 15-17% would definitely put the heat on Intel regardless of the combination of clock frequency bumps, memory controller and fabric latency improvements (the main reason why Ryzen is so sensitive to memory clock frequency), minor IPC gains from core/cache tweaks, better boost management, etc. that AMD may have used to get there if the leaks turn out to be accurate.

The leak probably is accurate, but the leak is that in one test of a 2600 one one benchmark suite, with unknown clock speed there was a 7% performance gain in single thread over the high end of results from the 1600 and 31% over the low end multithreaded results from a 1600. That's reasonable and expected. It doesn't generalize to mean that the 2600 is 31% faster than 1600. If AMD is leaking 2600 numbers you know they are running it as fast as they reasonably can while the low 1600 numbers are from a poorly setup 1600. Clock for clock absolute best is 5% gains. Clock speed absolute best is 5% gains. Absolute best Ryzen 2 does is 10% better than Ryzen (Ryzen 2 is not Zen 2) when you put the same effort into making them perform. I suspect the 7% number is closer.

Zen 2 might actually get 20% over Zen but this is not that.

Don't get me wrong here; I would buy AMD over Intel for 90% of build scenarios right now....they have built some awesome gear with Zen....expecting a process shrink to boost the same design by the highest gap from a leak though is wishful thinking. Regardless, we'll find out in a few weeks or less. I suspect reviewers have them in hand by now; maybe where the leak came from. Then the numbers will tell us who was right....feel free to throw it in my face if a 2600 is really 31% faster than a similar setup 1600 across multi threaded applications...if it is I need a new computer, although it'll be a 2700/2800.

 

[razamatraz]

The one thing that could throw a wrench in the whole theory is if they've somehow managed to separate the infinity fabric clock from the RAM clock...imagine if 2133 DDR4 could suddenly run Zen withing 5% of 3200....that'd be a big deal.

 

[Gon Freecss]

Yeah, you got a source on that? I'm gonna guess no.

The 2700/2800 refresh is not Zen2 it is Zen+ and at absolute best will have 5% IPC improvements (although it really is a refresh with better RAM support so probably 0% RAM speed held equal.) and maybe 5% Clock speed improvements.

If you spent all of 5 seconds doing a google search for "pinnacle ridge performance leak" (that's Ryzen+, not Zen 2), you'll find plenty of coverage about the Ryzen 1600 vs 2600 leak.
>leak]http://lmgtfy.com/?q=pinnacle+ridge+performance+leak

>leak

Your argument is invalid. Also, those comparisons aren't valid cause they use different versions of the application.

 

[kol12]

Surely an 8 core chip would not make sense until at least 2019 or later?

AMD's 1700/1800 launched in 2017 and the 2700/2800 refresh which can be expected to be up to 17% faster on heavily threaded workloads based on leaks are coming in about two months. Intel may really need an octo-core mainstream CPU to remain competitive.

An 8 core mainstream chip within such a short space of Coffee Lake would seem a little ridiculous.. I'm just about to move on an 8700K myself. Surely an 8 core chip would not make sense until at least 2019 or later?

It's possible that Intel may choose to leave that in its HEDT space, but given that you can get an R7 1700 for $230 that easily clocks to 4.0 and there are Zen+ chips due soon Intel may not have a choice. The move to 10nm also makes sense as this is the first real process node change since Broadwell and in theory that gives them room on the same size package to place more more cores.

Usually when moving to a new process node Intel has had trouble maintaining the clock speeds so compensating with more cores also seems reasonable. I think it would be too risky for Intel not to release an 8 core if they are able to make one in the 1151v2 socket. The 8700K is a great CPU but forcing you to buy a much more expensive platform to get into 8 core seems risky when you can do it on a $90 MB with AMD, ad althoguh I don;t think the 1800X replacement will get up to the IPC and sheer max clocks of an 8700K it's multicore performance will be far better (an 1800X is already better). If they can do it economically I'd say 90% they do do it. You never know with Intel though; it's notorious for not giving anyone what they want.

My question is, from a gaming perspective is an 8 core Intel CPU going to matter? This is more of a case of Intel remaining competitive with Ryzen rather than Intel releasing the next 8 core "gaming" chip right... It's quite possible the 8700K would still be the better gaming chip...

If Intel can release an 8 core option without another silly new chipset I think a lot of people would be very happy with that.. I am guessing though that an 8 core chip may be aimed more as a production chip rather than gaming?

 

[InvalidError]

If Intel can release an 8 core option without another silly new chipset I think a lot of people would be very happy with that.. I am guessing though that an 8 core chip may be aimed more as a production chip rather than gaming?

A very small number of games can make use of 6C12T right now and that number is likely to grow slowly but steadily in the future. Throw in any amount of multi-tasking while gaming and the usefulness of extra cores today becomes far more evident, the most common use being live-streaming your own gaming where you need the CPU to do audio/video processing and transcoding for the best streaming video quality. At the moment, most competitive gamers who stream their games rely on a separate system for video processing to prevent streaming from bogging down their gaming machine.

 

[kol12]

If Intel can release an 8 core option without another silly new chipset I think a lot of people would be very happy with that.. I am guessing though that an 8 core chip may be aimed more as a production chip rather than gaming?

A very small number of games can make use of 6C12T right now and that number is likely to grow slowly but steadily in the future. Throw in any amount of multi-tasking while gaming and the usefulness of extra cores today becomes far more evident, the most common use being live-streaming your own gaming where you need the CPU to do audio/video processing and transcoding for the best streaming video quality. At the moment, most competitive gamers who stream their games rely on a separate system for video processing to prevent streaming from bogging down their gaming machine.

Is it possible that the 8700K might still be a better gaming chip than the next Intel 8 core chip? Most benches show the 8700K having better single and multithreaded performance for games.

 

[InvalidError]

Is it possible that the 8700K might still be a better gaming chip than the next Intel 8 core chip?

Does the 3770k have better performance than the 2600k? Does the 4770/4790k have better performance than the 3770k? Does the 6700k have better performance than the 4790k? Etc.? Yes. So, barring the odd regression here and there, the 9700k or whatever it ends up being called will likely be incrementally better than the 8700k at a most things.

We'll see when it launches, if desktop Cannonlake doesn't end up delayed to the point of getting skipped much like Broadwell was. (Which would have the knock-on effect of turning the 370-series chipsets into single-generation boards just like LGA1156.)

 

[kol12]

It makes sense for only an increment of performance in this space of time. I am under the impression that the 8700K performs better than most 8 core counterparts in gaming, especially Ryzen. My question is can Intel's 8 core actually be as good a gaming chip as the 8700K? More cores likely result in lower boost clocks and overclocking potential.

The 8700K as far as I can tell performs very similar to the 8 core i7 7820X with the 8700K gaining a few extra fps in certain titles. I guess the real advantage from an Intel 8 core mainstream will be great multitask as well as gaming performance. I just wonder if it might fall short of the 8700K gaming performance due to some of the 8 core limitations. Never know what Intel have up their sleeves though!

I have been contemplating an 8700K upgrade for the last 2-3 months, I now have the funds and literally have the parts in my cart ready to go! With all of the 390Z, Ice lake stuff still being very hazy but possibly just on the horizon it's a little hard to know what to do. Intel's current barrage of CPU releases sure is unusual, a chipset (370Z) that literally last's a 4-5 months? I really hope these theoretical 8 core chips are compatible with 370Z as well as 390Z!!

 

[InvalidError]

It makes sense for only an increment of performance in this space of time. I am under the impression that the 8700K performs better than most 8 core counterparts in gaming, especially Ryzen.

First-gen Ryzen has a significant clock frequency and somewhat of an IPC handicap in many workloads, can't really compare it 1:1 to Intel and call it anywhere near representative of what 8-core 10nm Cannonlake might be. Intel's mainstream 8-core CPU (if it actually happens) will be native 8-core design instead of cut-down 12-core design like the i9-78xx/79xx, which means that much less less latency between cores and resources. Throw in three generations worth of architectural refinements since Skylake (Kaby-Coffee-Cannon) and the 10nm shrink, the differences are likely to be significant, so you can't really use Skylake-X as a reference point either.

There is also the Spectre/Meltdown patches to consider. Updated numbers for the updated patches haven't come out yet and I'd wait a while longer to see whether any more issues come up before considering those patches and performance numbers final.

 

[kol12]

@InvalidError

Some people are claiming that Meltdown will be completely fixed at the OS level and that's it Spectre that will need a hardware fix. A finalised hardware fix for Spectre could be quite a few CPU's away... I have briefly read that the current Meltdown patch performance hit on more recent gen CPU's is 1%??

RE: 8 core comparison, that's why I chose the 7820X as a theoretical 8 core comparison, but I agree the architectural refinements since Skylake could make for a significant performing 8 core Icelake chip or whatever it will be.

Has there been any hint that these new CPU's will be compatible with the Z370 chipset or is it looking like a new chipset within the space of 4-5 months?

 

[InvalidError]

Has there been any hint that these new CPU's will be compatible with the Z370 chipset or is it looking like a new chipset within the space of 4-5 months?

I suspect that depends entirely on whether Cannonlake will be a hit or a miss / Broadwell-Take-II.

Intel has never used the same unchanged mainstream socket for more than two consecutive generations. The 370-series chipsets and motherboards were for Coffee Lake and Cannonlake. Also, given the number of lawsuits Intel is facing due to launching new products after learning about Spectre and Meltdown, I would not be surprised if Intel "pulled another Broadwell" by skipping Cannonlake for mainstream. If Intel ends up effectively scrapping Cannonlake by making it nearly unobtainable, then the 370-series will likely die with Coffee Lake.

 

[kol12]

They're talking about a new line of Coffee lake "S" processors coinciding with Z390. What the heck could be difference between these "S" processors and the current range of Coffee lake processors?

https://www.tweaktown.com/news/59851/intels-next-gen-z390-ready-coffee-lake-1h-2018/index.html

Most of the rumors seem confident that the release of Z390 is for the purpose of an 8 core chip. To date no 8 core chip can clock anywhere near as high as an 8700K. Should be a solid chip nonetheless.

 

[InvalidError]

They're talking about a new line of Coffee lake "S" processors coinciding with Z390. What the heck could be difference between these "S" processors and the current range of Coffee lake processors?

I'd say packing two more cores in the mainstream die than first-gen coffee lake would qualify as a substantial difference. Seems pretty weird to bump core count in a refresh generation though.

 

[kol12]

They're talking about a new line of Coffee lake "S" processors coinciding with Z390. What the heck could be difference between these "S" processors and the current range of Coffee lake processors?

I'd say packing two more cores in the mainstream die than first-gen coffee lake would qualify as a substantial difference. Seems pretty weird to bump core count in a refresh generation though.

It does seem weird. Maybe an 8 core was supposed to be part of the original Coffee lake line up but they've had some delay and released the Z370 in the interim to get those chips out. A lot of unknown. The question will be whether the 8 core is exclusive to the Z390. Not really sure what else to think at this stage.

 

[mdd1963]

If Intel can release an 8 core option without another silly new chipset I think a lot of people would be very happy with that.. I am guessing though that an 8 core chip may be aimed more as a production chip rather than gaming?

A very small number of games can make use of 6C12T right now and that number is likely to grow slowly but steadily in the future. Throw in any amount of multi-tasking while gaming and the usefulness of extra cores today becomes far more evident, the most common use being live-streaming your own gaming where you need the CPU to do audio/video processing and transcoding for the best streaming video quality. At the moment, most competitive gamers who stream their games rely on a separate system for video processing to prevent streaming from bogging down their gaming machine.

Is it possible that the 8700K might still be a better gaming chip than the next Intel 8 core chip? Most benches show the 8700K having better single and multithreaded performance for games.

Absolutely...

If the all core/MCE clocks on the mythical 9700K are running less than the 8700K (routine all 6 cores at 4.6 GHz with good cooling) is capable of sustaining, I'd fully expect the 8700K to remain king for a few years....short of some revolutionary new game that favors 8 cores/16threads at lesser speeds over 6 faster cores.

 

[kol12]

If Intel can release an 8 core option without another silly new chipset I think a lot of people would be very happy with that.. I am guessing though that an 8 core chip may be aimed more as a production chip rather than gaming?

A very small number of games can make use of 6C12T right now and that number is likely to grow slowly but steadily in the future. Throw in any amount of multi-tasking while gaming and the usefulness of extra cores today becomes far more evident, the most common use being live-streaming your own gaming where you need the CPU to do audio/video processing and transcoding for the best streaming video quality. At the moment, most competitive gamers who stream their games rely on a separate system for video processing to prevent streaming from bogging down their gaming machine.

Is it possible that the 8700K might still be a better gaming chip than the next Intel 8 core chip? Most benches show the 8700K having better single and multithreaded performance for games.

Absolutely...

If the all core/MCE clocks on the mythical 9700K are running less than the 8700K (routine all 6 cores at 4.6 GHz with good cooling) is capable of sustaining, I'd fully expect the 8700K to remain king for a few years....short of some revolutionary new game that favors 8 cores/16threads at lesser speeds over 6 faster cores.

I agree. To date no 8 core has been able to clock anywhere near as high as a 4c/8t CPU. It is a first for a 6 core with the 8700k but I am not sure an 8 core would be at that capability yet. I am guessing an 8 core Coffeelake will be a great productive chip to compete with Ryzen but that the 8700K will still be king in gaming. I wan't sure if or how many games could even take advantage of 8 cores but apparently we are only at the beginning of that with 6 cores...

 

[InvalidError]

It does seem weird. Maybe an 8 core was supposed to be part of the original Coffee lake line up but they've had some delay and released the Z370 in the interim to get those chips out.

I'd wager that mainstream Coffee Lake was never intended to be more than hex-core. Botched availability of CPUs and chipset through most of its 2017 market life shows that it clearly wasn't ready for full-scale launch back then - lack of B300 and H300 series chipsets to this date indicate that it still isn't quite ripe for full-scale even today. To me, that clearly looks like Coffee Lake got launched several months premature entirely due to Ryzen turning out to be a credible threat to its quad-core CPUs.

The octo-core consumer Coffee Lake refresh was likely never meant to exist but on-going delays and complications on 10nm have forced it to introduce another set of in-between CPUs to prevent AMD from closing the gap any further with Ryzen+.

 

[kol12]

It does seem weird. Maybe an 8 core was supposed to be part of the original Coffee lake line up but they've had some delay and released the Z370 in the interim to get those chips out.

I'd wager that mainstream Coffee Lake was never intended to be more than hex-core. Botched availability of CPUs and chipset through most of its 2017 market life shows that it clearly wasn't ready for full-scale launch back then - lack of B300 and H300 series chipsets to this date indicate that it still isn't quite ripe for full-scale even today. To me, that clearly looks like Coffee Lake got launched several months premature entirely due to Ryzen turning out to be a credible threat to its quad-core CPUs.

The octo-core consumer Coffee Lake refresh was likely never meant to exist but on-going delays and complications on 10nm have forced it to introduce another set of in-between CPUs to prevent AMD from closing the gap any further with Ryzen+.

Well said. I can't imagine they are making a Z390 chipset for just one 8 core processor though, there must be more?

Some people are saying that Cannonlake will be mobile only while others seem to think there will be a desktop line in the 300 series chipset. I found this image which could indicate some truth in that:

https://i.imgur.com/wbbW1Bl.jpg