Coffeelake thread, benchmarks, reviews, input, everything.

[epsilon84]

Pretty amazing if true. 4.7 out the box.

So the next question, what is better? 8/8 or 6/12?

As I posted above, for Cinebench? 8/8 Blender? 6/12 Probably ends up pretty even over a mix of benchmarks, though the 9700K would probably edge it due to higher boost clocks

 

[moonbogg]

I jokingly predicted an 8/8 variant and almost didn't actually believe it could happen to be quite honest. I didn't think even Intel could be that ridiculous. Well...I said it before and I'll say it again; Intel just does Intel. LOL AHAHA! 8/8? Get out of here with that insulting garbage. Should honestly be 4/4, 4/8, 6/12, 8/16 because 4/8 is too close to 6/6 for 6/6 to make sense and 6/12 is too close to 8/8 for 8/8 to make sense. The 4/4 is for those who only need a low power, basic chip. 8/8 is just too funny.

 

[epsilon84]

I have to say one thing though, and it's that if these turbo clocks are true, then Intel is clocking their chips very aggressively - 4.6/4.7 is the point where my 8700K starts needing extra voltage to remain stable.

Either Intel worked a minor miracle and somehow managed to eek out an extra 300 - 400MHz in headroom despite adding two cores, or the more likely scenario where these are clocked very close to their stable ceilings, similar to what AMD did with the 2700X

 

[Zucker2k]

An 8/8 Coffeelake CPU is definitely going to be the new gaming king is all I know.
Edit: It'll have enough core resources to handle all codes out there without the help of 'fake' cores.

 

[epsilon84]

An 8/8 Coffeelake CPU is definitely going to be the new gaming king is all I know.
Edit: It'll have enough core resources to handle all codes out there without the help of 'fake' cores.

TBH there will be basically no difference in gaming performance between 6/12, 8/8 and 8/16. Even 6/6 suffers little to no performance degradation in the majority of current games.

 

[TheGiant]

IMO the 8C/8T CPU is the new i5-2500K from 2011. Will last 5 years without problems...

 

[moonbogg]

The 8/16 chip will be tempting, but I can't see myself going for another 14nm CPU. It won't be long until the new nodes hit.

 

[Zucker2k]

TBH there will be basically no difference in gaming performance between 6/12, 8/8 and 8/16. Even 6/6 suffers little to no performance degradation in the majority of current games.

Except that it'll eliminate any HT penalty in games and two extra cores over the 8600K means it'll be the better processor going forward.

 

[epsilon84]

Except that it'll eliminate any HT penalty in games and two extra cores over the 8600K means it'll be the better processor going forward.

Hmm, HT penalty is still a thing these days? That's news to me. Last I checked a 8700K was either equal to or slightly faster than a 8600K depending on how well threaded the game was

 

[Eug]

In retrospect, I guess 8/8 makes sense from the perspective of binning, now that there is an 8/16 variant.

 

[StinkyPinky]

I have to say one thing though, and it's that if these turbo clocks are true, then Intel is clocking their chips very aggressively - 4.6/4.7 is the point where my 8700K starts needing extra voltage to remain stable.

Either Intel worked a minor miracle and somehow managed to eek out an extra 300 - 400MHz in headroom despite adding two cores, or the more likely scenario where these are clocked very close to their stable ceilings, similar to what AMD did with the 2700X

It will probably still overclock to 5ghz or so, Intel will want to make sure it does so they can make that extra $50 by having that K branding.

 

[Lovec1990]

I wonder what kind of cooling you will need for it to sustain 4.7ghz with all cores.
Otherwise im thrilled we have competition again so AMD and Intel are competing again so better products for Intel and AMD fans

 

[CHADBOGA]

A future Intel chip will have the added benefit of an improved architecture, an improved process and hopefully no slowdowns(however slight) from Meltdown and Spectre.

 

[epsilon84]

I wonder what kind of cooling you will need for it to sustain 4.7ghz with all cores.
Otherwise im thrilled we have competition again so AMD and Intel are competing again so better products for Intel and AMD fans

A decent air cooler will suffice. As long as the stock volts are around the 1.2 - 1.25V mark these won't be power hungry behemoths. It's more when you're overclocking to 5GHz using 1.35V or more, that's when good cooling (high end air or AIO) would be a necessity, especially with 8 cores.

 

[epsilon84]

It will probably still overclock to 5ghz or so, Intel will want to make sure it does so they can make that extra $50 by having that K branding.

I have almost no doubt these will overclock to 5GHz, how far beyond though? Are the 5.2 - 5.3GHz overclocks on the 8086K a sign of things to come, due to an (even further) refined 14nm process, or are those cherry picked 8700Ks and the true ceiling remains at 5.0 - 5.1 like the majority of existing CFL chips?

 

[french toast]

8/8?? LOL that is ridiculous! I guess they will justify it by having higher ACT over a 8700k..plus higher ST clocks.
Also if you believe that leak the only SKU that will have Intel's vaunted hyperthreading will be the top i9 SKU...that's stupid.
Maybe by using 14+++, soldered chips + disabling HT they can get such crazy 4.7 all core turbo's...who knows what such a chip could O/C to? 5.3+?

Probably 8/8 with slightly higher IPC/better latency to men and ring bus plus much higher clocks will best a 2700x across the board for $350.
I don't like it when intel does this kind of thing (artificial segmentation), but the end product might justify the means, plus means coffeelake s won't outshine icelake.

Icelake is probably going to offer 8/16 +15% IPC + fixes+ much better igpu (72 EU?)to $350 market Q3 next year imo.
As clocks will most likely be very similar to coffeelake S the above improvements will be more than enough to encourage 9700k users to upgrade imo.

The 300kg gorilla in the room is of course what AMD can do with zen 2...

 

[The Stilt]

Unless there are some significant process improvements, there is just no way the 9900K can sustain 4.7GHz on all cores at 95 / 119W power limit.
The 8700K can barely sustain 4.3GHz at the same power limit with 25% lower core count.

The claimed specs are impressive and seem to be in the realms what is technically possible to achieve, but not within the claimed power figures.

 

[epsilon84]

8/8?? LOL that is ridiculous! I guess they will justify it by having higher ACT over a 8700k..plus higher ST clocks.
Also if you believe that leak the only SKU that will have Intel's vaunted hyperthreading will be the top i9 SKU...that's stupid.
Maybe by using 14+++, soldered chips + disabling HT they can get such crazy 4.7 all core turbo's...who knows what such a chip could O/C to? 5.3+?

Probably 8/8 with slightly higher IPC/better latency to men and ring bus plus much higher clocks will best a 2700x across the board for $350.
I don't like it when intel does this kind of thing (artificial segmentation), but the end product might justify the means, plus means coffeelake s won't outshine icelake.

Icelake is probably going to offer 8/16 +15% IPC + fixes+ much better igpu (72 EU?)to $350 market Q3 next year imo.
As clocks will most likely be very similar to coffeelake S the above improvements will be more than enough to encourage 9700k users to upgrade imo.

The 300kg gorilla in the room is of course what AMD can do with zen 2...

Yeah I'm not a fan of the artificial segmentation either, especially when it comes to HT. I guess they needed something to differentiate between the 9700K and 9900K, but still, 6/12 to 8/8 feels like a sidegrade to me at best, it's probably why Intel are bumping the clocks up to avoid negative press of a 8700K beating a 9700K...

I have my doubts that a 8/8 9700K can get anywhere close to 2700X levels of MT performance, you only have to look at 8600K and 2600X comparisons to see the gulf in MT performance, and it's not a gulf higher clockspeeds can overcome.

WRT to Icelake, I would be very (happily) surprised if we get such drastic IPC improvements. Then again, Intel has had years now to develop a successor to the Lake series of chips. Let's hope you're right on that front

 

[rainy]

Icelake is probably going to offer +15% IPC

Could you remind me when something like that happened in the last 5-6 years because I don't recall it?

 

[The Stilt]

If we expect that the 9700K would really be able to operate at 4.6GHz on all cores, then its average MT performance would almost exactly match the 2700X.

The SMT yield on Ryzen is < 29% on average and the 2700X typically runs at =< 4.05GHz on all cores.

8 * (4.05 * 1.29) = 41.796

Meanwhile the 9700K would be 8 * (4.6 * 1.14) = 41.952.

1.14 being the typical IPC advantage over Pinnacle Ridge (while expecting that the IPC is unchanged again compared to CFL-S6).

 

[epsilon84]

If we expect that the 9700K would really be able to operate at 4.6GHz on all cores, then its average MT performance would almost exactly match the 2700X.

The SMT yield on Ryzen is < 29% on average and the 2700X typically runs at =< 4.05GHz on all cores.

8 * (4.05 * 1.29) = 41.796

Meanwhile the 9700K would be 8 * (4.6 * 1.14) = 41.952.

1.14 being the typical IPC advantage over Pinnacle Ridge (while expecting that the IPC is unchanged again compared to CFL-S6).

Interesting, though that doesn't explain why even an overclocked 8600K still lags behind a 2600X in most MT workloads?

I mean if your maths is correct then an overclocked 8600K @ 5.0GHz should exceed a 2600X @ 4.1GHz right? But clearly that's not the case, just look at Cinebench scores, for example, or Blender rendering times

 

[The Stilt]

Interesting, though that doesn't explain why even an overclocked 8600K still lags behind a 2600X in most MT workloads?

I mean if your maths is correct then an overclocked 8600K @ 5.0GHz should exceed a 2600X @ 4.1GHz right? But clearly that's not the case, just look at Cinebench scores, for example, or Blender rendering times

In these applications neither the IPC difference or the SMT yield on Ryzen is close to the average.

Blender: ~35% SMT (8.4% IPC), CB15 nT: ~42% SMT (5.6% IPC).

2600X: 6 * (4.1 * 1.35) = 33.21 (Blender), 6 * (4.1 * 1.42) = 34.932
8600K: 6 * (5 * 1.084) = 32.52 (Blender), 6 * (5 * 1.056) = 31.68

 

[french toast]

Could you remind me when something like that happened in the last 5-6 years because I don't recall it?

Valid point ... Call me naive, but I think they have had so long to work on this (4-5 years+ with all the delays) that they could well have gone back to the drawing board and bumped up the uarchitecture, maybe increased the caches or something. Especially once they realised probably by 2016 that 10nm was a shot process and even 10nm+ would not achieve the same clocks as 14nm++ (nevermind 14+++).
I don't know whether the spectre fixes and Ryzen's unexpected competitiveness could have influenced this or whether these events were too late in the game?.
Either way by the time Kabylake was launched as a stopgap they must have had a tough decision to make, that's where 14nm++ & 14nm+++ 6 & 8 core skylake dice came in to hold the fort whilst they Sellotaped 10nm+ together with a rejig of icelake microarchitecture...

Perhaps they had the foresight (alongside some handy information from poached ex AMD employee's) to scrub the IGPU and put on some more cores to the MESH..like 12 for instance.

Just speculation on my part, but I can't fathom how intel is going to compete with zen 2 with an 8/16 icelake, that has only 5-10% IPC and lower clocks than coffeelake can you?

 

[Lovec1990]

Guys how its possible to push 9900k 4,6Ghz all cores with TDP of 95W ?

 

[Shivansps]

Well they could easily add HT to every SKU on the list as well now, considering the 9700 has less cache, HT only on 1 sku makes no sence to me.