Coffeelake thread, benchmarks, reviews, input, everything.

Page 21 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

raghu78

Diamond Member
Aug 23, 2012
4,093
1,475
136
Coffelake puts Intel firmly back in the lead with higher IPC, higher clocks, vastly better ST and gaming perf and very competitive MT perf. The only issue for Intel is going to be to have enough supply for the very high demand they will be facing especially for 8700k. The 8700k is a fantastic chip. Seeing the 8700k beat the 1800x in few MT workloads while absolutely crushing it in ST and gaming perf is a testament to the architecture and process lead which Intel have at the moment. Hopefully AMD's response with Pinnacle Ridge turns out to be very strong.
 
  • Like
Reactions: ozzy702

VirtualLarry

No Lifer
Aug 25, 2001
56,227
9,990
126
So that's Intel's plan... to release as few as possible high-end CFL CPUs into the wild, just to generate advertising buzz and benchmark data, to show that they're faster than AMD's mainstream CPU lineup (for now) (in some areas), while not actually supplying any significant qtys of the CPUs into the market, in order to convince potential customers to "wait for" Intel, rather than choosing Ryzen today?

So, Intel is turning into a "Hype Machine", rather than a real CPU supplier?
 

LTC8K6

Lifer
Mar 10, 2004
28,520
1,575
126
So that's Intel's plan... to release as few as possible high-end CFL CPUs into the wild, just to generate advertising buzz and benchmark data, to show that they're faster than AMD's mainstream CPU lineup (for now) (in some areas), while not actually supplying any significant qtys of the CPUs into the market, in order to convince potential customers to "wait for" Intel, rather than choosing Ryzen today?

So, Intel is turning into a "Hype Machine", rather than a real CPU supplier?
We are not even halfway through day one...

You sure bail out easy...
 

tamz_msc

Diamond Member
Jan 5, 2017
3,710
3,554
136
The effect of Intel not disclosing turbo speeds.
Before returning to the performance averages, we must insist, once again, on the issue of automatic overclocking. As we discussed in the introduction, if we can understand that Intel prefers to hide certain characteristics to avoid comparisons, this has a direct consequence: what is not announced by Intel is de facto vague, and therefore open to interpretation by manufacturers of motherboards. Or more precisely automatic overclocking practiced by some.
As always, we disable the option "Enhanced multi-core" which on an 8700K overclocked all the Turbo ratios at 47. This option was enabled by default by the BIOS and the fact that Intel no longer communicates the Turbo ratios is only Asus confirms this choice. We're not happy about that, but that's not what we wanted to talk about (we'll tell you to be complete that an identical option has appeared in the Gigabyte BIOS, which unfortunately confirms our opinion on the impact of Intel decisions even if, in their case, the option is disabled by default). Because even by disabling the option we noticed two weird things. The first was a rather large overconsumption, linked to a particularly high voltage choice by the motherboard.
Despite the automatic adjustment, we note a voltage of 1.264V under load, or 0.1V more than what was noted on the Gigabyte. The consequence of this high voltage is not neutral, we noted 20 watts of charge difference under Prime95, which increases the consumption above the TDP announced by Intel ... if one has regulated the frequency of its memory . This is one of the subtleties used by Asus to justify its practices, if you set the frequency of your memory modules yourself (in DDR4-2400 for our tests), you activate certain overclocking options automatically, to exceed the TDP. One will smile that if one does not set its memory, the TDP is not exceeded ... but the Turbo coefficients all hearts active under Prime95, instead of being at 43x fall to 41x because of the choice of this voltage high!
The other problem we found is even more embarrassing since it clearly impacts the performance we get, and also plays on the opacity of Intel. We noticed that the frequency of the cache was not regulated in the same way on the two models in our possession. In the case of the Gigabyte motherboard, with a Core i7-8700K, this L3 cache frequency (sometimes referred to as Cache or RingLLC) is set to the base frequency of the processor (3.7 GHz), while at Asus it was set ... at 4.4 GHz. A frequency that does not correspond to anything on a 8700K (the frequency all active hearts is 4.3 GHz for recall). In the BIOS of Asus, this setting of the maximum coefficient is set to "Auto" without the possibility of a default choice (the behavior is observed with or without memory adjustment to be precise).
So we tried to check for ourselves what the expected frequency was for the L3, and to say that once again we lost a lot of time. Special thanks go to the author of the excellent HWiNFO which amply helped us on the issue. The software reports both the maximum frequency and the current frequency (it varies depending on usage, down to 800 MHz at rest for example, in the case of 8700K it varies between 4 GHz and 4.4 GHz in charge). After verification of its side, it turns out that it is not possible to read in a register what would be the "default" frequency. The information read is in registers available read / write and thus modifiable by the BIOS.
We then contacted the three protagonists last weekend, Gigabyte, Asus and Intel last weekend to try to find out more and know how often the L3 is supposed to run on a Core i7-8700K.
  • For its part, Gigabyte sent us back to Intel's ARK page and the BCLK to 3.7 GHz
  • Asus explains that disabling the "Multi-core enhancement" option places the CPU in a default Intel setting (which is not the case as seen above) and did not answer our question
  • Intel replied that "they only communicate the basic frequencies and turbo on a heart, and no longer communicated other frequencies like turbo multi core"
If we dwell on the question, it is that overclocking the cache L3 has an impact in the benchmarks, of course. Unsurprisingly, the applications most sensitive to the memory subsystem benefit most with 6% gain, for an average gain of 3.8% on the index. And in the games necessarily, the impact is even more clear with 5.7% gain on average on the index.
Obviously, our colleagues who have performed the test on an Asus motherboard (or any other model with such overclocking) will find higher deviations from our tests. Can we blame them? Probably not because if we had not had several motherboards in our hands, we probably would not have noticed this abnormal frequency. This is all the more true when we dig deeper into the question, we found that Asus was already playing with this frequency in the last BIOS of the Z170-A that we use for the Skylake / Kaby Lake tests.
Once again we are sending a red card to Asus, but the manufacturer shares it this time with Intel. We are no longer surprised by Asus's choices in terms of "interpretation" of specifications or by not answering our questions on the topic of automatic overclocking. However, by oddly enough, by default , the voltage of 0.1V (for those who would not have read our page overclocking, this voltage allows to hold the 4.7 GHz on all hearts) and overclocking this time very significantly the frequency Ring / LLC (we move from a 100 MHz that we had not noticed to 700!) The manufacturer does more and more, still thinking that a motherboard is selling on performance and that these choices must be applied by default to any the world (and often, especially in the BIOS used by the press when launching a product).

But it is especially the silence of Intel - and its non-response when it is clearly asked the question of this difference - on its specifications that we are frankly embarrassed since we see concrete consequences. Hiding frequencies because we know they would not be supported, what we saw on the big Skylake-X is one thing, but no longer communicate on these frequencies is leaving the door open to builders motherboards to do just about everything and anything. What's more, if we point to Asus today, we know from experience that by letting practice, other manufacturers will most likely end up aligning (the arrival of an option "Enhance multi core" on the Gigabyte BIOS, although disabled by default, is proof of this). And if Intel will not lend itself to bad intentions, see the motherboards manufacturers "innovate" on specifications pushes after all performance up in published tests by creating artificial generational gaps. We were expecting Intel better.
.
No wonder why "IPC" debates on Intel threads attract drones playing the same broken record. A few days back I was flagged for deviating off-topic with my IPC hair-splitting. Now just look at some of these posts bringing AMD back into the same discussion. I hope moderators will take appropriate action.
 

Glo.

Diamond Member
Apr 25, 2015
5,661
4,419
136
OC-3.png

So now its AMD who is the efficient one? o_O

The power consumption after OC goes through the roof! For both companies.
 

CatMerc

Golden Member
Jul 16, 2016
1,114
1,149
136
Coffee Lake launch stock is several orders of magnitude worse than Skylake according to my source who works in retail. They got basically no chips.
Also looking at local retailers in Israel, no one has them listed, while Skylake was listed a couple of days early.

I wouldn't expect being able to buy one at a reasonable price until mid December at the earliest, though 2018 would be a reasonable expectation. I'm hoping I'm wrong because honestly these look enticing. My 6600K looks awful next to it :confused:
 

Glo.

Diamond Member
Apr 25, 2015
5,661
4,419
136
In my country the availability is pretty decent. All of the CPUs released are widely available, apart from 8600K in just one retailer. Already sold out.
 

CatMerc

Golden Member
Jul 16, 2016
1,114
1,149
136
In my country the availability is pretty decent. All of the CPUs released are widely available, apart from 8600K in just one retailer. Already sold out.
Wouldn't be surprised if it highly depends on region. What's your country Glo?
 

Zucker2k

Golden Member
Feb 15, 2006
1,810
1,159
136
Geez, it gets absolutely demolished by even the base 1600 in AES. Why does it perform so poorly?

8700K-8400-33.jpg


4.7Ghz getting beaten badly by 3.2Ghz AMD. I never thought I'd see that. It's like the reverse of bulldozer/sandy bridge
The AMD chips have dedicated AES hardware.
 
  • Like
Reactions: Phynaz

Phynaz

Lifer
Mar 13, 2006
10,140
819
126
A few days back I was flagged for deviating off-topic with my IPC hair-splitting. Now just look at some of these posts bringing AMD back into the same discussion. I hope moderators will take appropriate action.

They will, I reported your Mod callout.
 

tamz_msc

Diamond Member
Jan 5, 2017
3,710
3,554
136
They will, I reported your Mod callout.
"I hope mods take action regarding latest comparisons with AMD, when my doing the same before was off-topic" = mod callout?

WOW.

Biggest tryhard in the world spotted.
 
  • Like
Reactions: Drazick

Bouowmx

Golden Member
Nov 13, 2016
1,138
550
146
Intel Core i7-8700K gets 142% and 128% the score of AMD Ryzen 7 1800X 4.0 GHz in Tom's Hardware AutoCAD Cadalyst 2D and 3D , respectively. Although no test results from Tom's Hardware concerning The Witcher 3 frame rate, other reviews say increased frame rate of Intel Core i7-8700K over AMD Ryzen 7. The extra power is justified.
 
  • Like
Reactions: Zucker2k

coercitiv

Diamond Member
Jan 24, 2014
6,151
11,686
136
Tom's Hardware has a nice thermal and power evaluation for 8700K.

People willing to spend some time tweaking CFL systems should be able to get some impressive results.
 

Glo.

Diamond Member
Apr 25, 2015
5,661
4,419
136
I see that only HardwareCanucks did Overwatch tests:
8700K-8400-55.jpg

Virually no difference between any of those CPUs.

And I will not have High End GPU, but at best GTX 1060 performance class.
 

jpiniero

Lifer
Oct 1, 2010
14,510
5,159
136
The 8700K is a damn fine 'all round' processor. It has great ST, MT and gaming performance. Seems to overclock to 5GHz reliably. I guess the only thing people can complain about now is the power consumption ;) (and lack of widespread availability, I must add!)

It does show how silly those people who are complaining about the Z270 not being compatible with CFL is though. Why on earth would you upgrade from 7700K to this?
 

epsilon84

Golden Member
Aug 29, 2010
1,142
927
136
1280x720 data is nice to have, but ultimately not applicable to many users.

Would you prefer a 4K graph showing all the CPUs being GPU limited instead? 720P testing shows the relative strengths of the CPUs without GPU bottlenecking. This will become more relevant as GPUs get faster, if you run a multi GPU setup, or if you are a single GPU user willing to sacrifice some graphical detail in order to maximise framerates for 144Hz or even 240Hz gaming monitors.
 
Last edited:

Pookums

Member
Mar 6, 2017
32
13
36
Good Performance all things considered, however not without a few bad self-inflicted points:

Why waste $50+ more on the 8700k with base 8700 almost identical in turbo settings, and stock 8700k vs OC 8700k has very minimal performance improvements(only in fully parallel MT workloads). However, there appears to be either bios, thermal, or power throttling issues (or a mix of three) at 4.9+ OCs in those respective MT loads as well(creeps up and down. Avx related??). Perhaps fixable if bios, but even without the occasional odd MT benchmark, the base 8700 being a near identical to stock 8700k, makes it pointless to waste the money. The ST and dual performance is already there for the primary threads of games, and with six cores at those clocks its more than enough to handle additional threads or background processes to prevent dips.

8600k undoubtedly also OCs without any potential issues much better it seems.

I mentioned before that one of the two processors 8600k or 8700k might cannibalize one or the other. IMO its the 8700 non-k combined with 8600k ease of OC should both harm the 8700k sales if the world still makes any sense. Sorry, but the extra money for almost no gain just isn't worth it. Base 8700 should also cannibalize 6 - core skylake-x if anyone was interested in buying that workstation load.

8700 and 8600k and down are all good chips based on the presentation. However, if anyone asks me, I'll tell others to save themselves money and go with one of those two instead over the 8700k. I think this is the most important message. Pure Gamers should get the 8600k. All arounders should get the 8700 non-k. Gamers who want to save some money for performance should get the 8400.
 
  • Like
Reactions: Burpo

Glo.

Diamond Member
Apr 25, 2015
5,661
4,419
136
Looking at power/performance scaling between the CPUs it may actually be possible that Low-power i3 models will "actually" be low-power, but at the same time they might have pretty decent clock speeds.