Intel Skylake / Kaby Lake

Page 650 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

TahoeDust

Senior member
Nov 29, 2011
557
404
136
That seems a little... egregious? Your post is making me re-think my foray into X299. Do the KBL-X CPUs have similar TDP / power usage issues as Skylake-X?
If power consumption or desire for unnecessarily low temps is a concern for you, then running a high clocked x299 setup is not for you.

That said, 275w sounds a little high...although he could be running an AVX test and his AVX offset may be less than mine.
 

tamz_msc

Diamond Member
Jan 5, 2017
3,726
3,560
136
Exactly. That is why TDP is somewhat meaningless and the minor TDP difference between the 7700k and 8700k is highly misleading. 8700k will use a lot more power. That's why it need a new socket spec and hence a new mobo compared to 7700k. OCed 7700k uses a lot of power. 14nm++ adds some efficiency so the 8700k will likely use another 35% more power (50% more cores minus efficiency gains).

A 7700k can easily use 180w of power when OCed. Add 35% to that and we are at 243W. No wonder it needs a new socket.
I don't think existing VRM designs on Z270 boards have much trouble supplying OC'd 7700Ks with 180-200W, and even if OC'd CFL does consume 250W, I still don't think it'll be a problem.
 

epsilon84

Golden Member
Aug 29, 2010
1,142
927
136
8400 has way too low base clocks at 2.8GHz when the 7500 base clocks are at 3.4GHz, that may be a problem with games though.
I still believe the R5 1600 is better overall but it is also more expensive currently.

At stock clocks the 8400 will be quite a lot faster at ST and games due to a combination of higher IPC and clockspeed. MT performance would probably be slightly behind the R5 1600 due to the lack of HT.

Of course you can overclock an R5 1600 to approx 4GHz to close the ST gap and extend the MT lead, but at a significant cost to power consumption, like any other CPU that is pushing its clock speed ceiling with raised voltages. I wouldn't be surprised to see a 1600 @ 4GHz exceed 150W power draw (depending on voltages). Maybe power efficiency suddenly doesn't matter to AMD fans though ;)
 
  • Like
Reactions: pcp7 and Sweepr

imported_bman

Senior member
Jul 29, 2007
262
54
101
Check the "?" mark on the TDP of any product listed on ark:

Reading the datasheet it says there are PL states, the CPU will only exceed the TDP while in PL2 or above. PL2 is the only state which is sustained for any significant amount of time (up to 100s), PL3 and PL4 disabled by default, PL2 by default is 1.25x the TDP. From watching Prime95 runs on stock configs on Youtube it seems like Intel CPUs have no problem remaining in the all core turbo bin indefinitely, thus they are staying in PL1 (which does not exceed TDP). Perhaps the PL2 state is hit when a single core is loaded. So it seems that TDP is only exceeded intermittently, not in a sustained manner even while turboing.
 

pantsaregood

Senior member
Feb 13, 2011
993
37
91
If power consumption or desire for unnecessarily low temps is a concern for you, then running a high clocked x299 setup is not for you.

That said, 275w sounds a little high...although he could be running an AVX test and his AVX offset may be less than mine.

It's not using AVX. There's some specific combination the Prime95 26.6 blend throws at my CPU that pushes it to higher power draw than even Small FFTs does. Part of the power consumption (though likely insignificant) comes from the memory controller running at DDR4-4000 and the uncore/mesh running at 3.2 GHz.

This particular CPU spirals out of control with power consumption fairly abruptly. 4.6 GHz will run at 1.13V without issue - power draw won't pass 185W. Going from 4.6 GHz to 4.8 GHz pulls an extra 90W.

EDIT: 275W was the absolute peak I saw after hours of Prime95 blend. Average was probably about 215W.
 

coercitiv

Diamond Member
Jan 24, 2014
6,151
11,686
136
That 8700 non-K is impressive, 6 cores with a turbo all the way up to 4.6GHz at only 65W TDP. I hope that Apple offers the non-K version in their 2018 Imacs to avoid the noise issues that the 95W 7700Ks have caused in their 2017 Imacs.
From watching Prime95 runs on stock configs on Youtube it seems like Intel CPUs have no problem remaining in the all core turbo bin indefinitely, thus they are staying in PL1 (which does not exceed TDP). Perhaps the PL2 state is hit when a single core is loaded. So it seems that TDP is only exceeded intermittently, not in a sustained manner even while turboing.
PL2 is never hit with only one core loaded on desktop, and up until now it was not a limit to be reached by most desktop CPUs no matter the load, except with combined CPU + iGPU cases. However, I personally believe this time it will be hit by most 6 core parts, which is not a bad thing considering that was it's purpose to begin with: enable Intel to offer higher turbo clocks as long as power usage is within designed limits.
 
  • Like
Reactions: Phynaz

CHADBOGA

Platinum Member
Mar 31, 2009
2,135
832
136
Introducing 8th gen Intel® Core™ DESKTOP processor Family

Core i7-8700K: $359
Core-i7-8700: $303
Core i5-8600K: $257
Core i5-8400: $182
Core i3-8350K: $169
Core i3-8100: $117

https://newsroom.intel.com/newsroom...es/11/2017/09/8th-gen-intel-core-overview.pdf

Okay I will put my hand up and admit I was wrong about Intel pricing Coffee Lake at effectively Kaby Lake price points, as I thought they would make the i7 8700K be at least $400.

Even though I am not in the market for a new PC/CPU, I still look forward to the benchmarking of these beasts.
 

mikk

Diamond Member
May 15, 2012
4,112
2,108
136
$359 for the 8700K, $257 for the 8600K. Really nice pricing for these chips.


Didn't Anandtech claim 400 USD a week ago? Some people are really owned. It was so dumb, comparing pre-order prices (which are always way higher) to Intels official price list. This is the same with every launch. 4 of 6 SKUs are identical and two are 15 USD and 20 USD higher. That's a fair pricing.

Intel published also some higher resolution pictures of the wafer and die: http://download.intel.com/newsroom/kits/8th-gen/8th-gen-core-series.zip

According to my calculation the die is 154 mm² big. That calculation can't be 100% accurate but it should match roughly the true size of 6C Coffeelake. So it is pretty clear the 149 mm² leak was legit.
 

tamz_msc

Diamond Member
Jan 5, 2017
3,726
3,560
136
Specwpc is one weird collection of software when any of the tested cpus can be best or worst depending on which component was benchmarked.
There's a clear trend, whenever there are render/encoding-type workloads, SKL-X excels as expected. Whenever the workloads depend on inter-core communication and don't scale very well, like convolution, poisson etc. it falls flat on it's face. Comparatively, the 6950X is much more consistent.
 
  • Like
Reactions: Drazick

AtenRa

Lifer
Feb 2, 2009
14,000
3,357
136
At stock clocks the 8400 will be quite a lot faster at ST and games due to a combination of higher IPC and clockspeed. MT performance would probably be slightly behind the R5 1600 due to the lack of HT.

At default the R5 1600 has a base clock of 3.2GHz for 12 Threads and 3.6GHz ST turbo. The Gaming performance difference should be very close (5-10%).

Of course you can overclock an R5 1600 to approx 4GHz to close the ST gap and extend the MT lead, but at a significant cost to power consumption, like any other CPU that is pushing its clock speed ceiling with raised voltages. I wouldn't be surprised to see a 1600 @ 4GHz exceed 150W power draw (depending on voltages). Maybe power efficiency suddenly doesn't matter to AMD fans though ;)

Overclocking the R5 1600 to 4GHz all 12 threads will make it 50%+ faster in MT than the 8400. Yes it will consume more but it will make it a lot faster too.
Also if you really like lower power consumption then you just OC to 3.8GHz and you still have way higher MT performance than 8400 but very close ST performance as well at reasobly power consumption.

8400 will be very popular for OEM SFF desktop PCs with iGPU only, DIY builders and gamers will chose the R5 1600 and 8600K.
 

tamz_msc

Diamond Member
Jan 5, 2017
3,726
3,560
136
At default the R5 1600 has a base clock of 3.2GHz for 12 Threads and 3.6GHz ST turbo. The Gaming performance difference should be very close (5-10%).



Overclocking the R5 1600 to 4GHz all 12 threads will make it 50%+ faster in MT than the 8400. Yes it will consume more but it will make it a lot faster too.
Also if you really like lower power consumption then you just OC to 3.8GHz and you still have way higher MT performance than 8400 but very close ST performance as well at reasobly power consumption.
You're vastly overestimating what SMT can do in applications which aren't canned benchmarks.That's why a 4C/8T 1500X just about matches a Core i5 7500 in general application performance.
 
  • Like
Reactions: Drazick

mikk

Diamond Member
May 15, 2012
4,112
2,108
136
At default the R5 1600 has a base clock of 3.2GHz for 12 Threads and 3.6GHz ST turbo. The Gaming performance difference should be very close (5-10%).


No chance. ST Turbo doesn't matter for games and base clock doesn't matter either for Intel CPUs. i5-8400 supports an 6C Turbo of 3.8 Ghz, so there is a clear clock and IPC advantage which is traditionally way higher in games. The gap must be way higher than 5-10% therefore.
 

Glo.

Diamond Member
Apr 25, 2015
5,661
4,419
136
8400 will be very popular for OEM SFF desktop PCs with iGPU only, DIY builders and gamers will chose the R5 1600 and 8600K.
Underestimation. This is first i5 X400 chip which is worth the money Intel asks for it.
 

coffeeblues

Member
Jun 23, 2017
49
18
36
Okay I will put my hand up and admit I was wrong about Intel pricing Coffee Lake at effectively Kaby Lake price points, as I thought they would make the i7 8700K be at least $400.

The only thing that keeps it under $400 are the leaks from retailer websites and they still have time to set the prices.
Intel released pricing for 1k units, if Kaby 1k-to-retail margins would be applied to Coffee 1k prices the retail would easily reach $400.
 

coercitiv

Diamond Member
Jan 24, 2014
6,151
11,686
136
Overclocking the R5 1600 to 4GHz all 12 threads will make it 50%+ faster in MT than the 8400.
Your scenario will be R5 1600 @ 4Ghz versus i5 8400 @ 4Ghz. Considering people will have to use Z370 boards, it's almost certain they'll be able to set all core turbo to max multiplier, hence effectively overclock locked parts as well

At best R5 1600 will have a 10-15% advantage in multithreaded, throughput oriented loads. In some benchmarks it won't even have that.
 
  • Like
Reactions: Sweepr

eddman

Senior member
Dec 28, 2010
239
87
101
Also gaming performance should be very close between the two (3.8GHz to 4GHz OC for the R5 1600).

If a Kaby Lake i5 7500 is 5% faster in games than a Ryzen 5 1500X with similar clocks and TDP, then the i5 8400 will also be 5-10% faster in games against a similarly clocked 1600/1600X.

Even at 1080, most games are more or less GPU limited. Just compare 720 benchmarks, which are less GPU limited, and you'd see that 7500 is ahead of 1300X and 1500X by quite a margin in a lot of games.

https://www.techpowerup.com/reviews/AMD/Ryzen_3_1300X/11.html

Compare the numbers to higher res results on the following pages.

Yes, almost no one plays at 720, but it's an indication that once the next gen GPUs hit, intel CPUs will probably have a considerable advantage.
 

SpaceBeer

Senior member
Apr 2, 2016
307
100
116
No it won't. Cause when next gen GPUs arrive, games will be more demanding. And GPU will be bottleneck again. I think for vast majority of gamers, GPU is bottleneck. Therefore, investment in GPU is always better solution.

Underestimation. This is first i5 X400 chip which is worth the money Intel asks for it.
True. But I still think B350 + R5 1600X is better than Z370 + i5-8400. So until cheaper MBs become an option, I would rather go for AM4 combo
 

Glo.

Diamond Member
Apr 25, 2015
5,661
4,419
136
True. But I still think B350 + R5 1600X is better than Z370 + i5-8400. So until cheaper MBs become an option, I would rather go for AM4 combo
The difference between B350 itx and Z270 ITX Mobo From ASRock is 9$.

Price difference between 1600X and 8400 will be most likely around 40$(!).

And Z370 MoBo model should cost the same as Z270.
 
Mar 10, 2006
11,715
2,012
126
That 8700 non-K is impressive, 6 cores with a turbo all the way up to 4.6GHz at only 65W TDP. I hope that Apple offers the non-K version in their 2018 Imacs to avoid the noise issues that the 95W 7700Ks have caused in their 2017 Imacs.

"K" CPUs have no business in Apple AIO systems. Not sure what Apple was thinking there.
 
Mar 10, 2006
11,715
2,012
126
8400 has way too low base clocks at 2.8GHz when the 7500 base clocks are at 3.4GHz, that may be a problem with games though.
I still believe the R5 1600 is better overall but it is also more expensive currently.

Base clock is irrelevant. What I want to know is the all-core turbo, but sadly (and inexplicably) Intel is not publishing that info anymore.