Intel Skylake / Kaby Lake

Page 535 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

TheGiant

Senior member
Jun 12, 2017
748
353
106
Outside of high FPS competitive gaming, an OC Ryzen 8-core, an OC Sandy Bridge i7 2600K, and an OC i7 7700K all perform within 5% of each other, using setups which reflect the real world, like GTX 1080Ti for 4K, GTX 1080 for 1440p, and GTX 1060/RX 580 for 1080p, in relatively modern AAA games, barring a few exceptions.
this is exactly what is agains the experimental observations if you are sensitive to smoothess and if you actually play the games, there is a big difference in smoothness in the intensive scenes

games are tested in limited scenarios but I've learned an empirical equation- if review shows min fps at 60 the real fps in most sensitive scenes (which at high difficulty matters most) is around 20-40 fps

there is a huge difference between oced 7700k and 4,5GHz 2600K
 

mikk

Diamond Member
May 15, 2012
4,111
2,105
136
Outside of high FPS competitive gaming, an OC Ryzen 8-core, an OC Sandy Bridge i7 2600K, and an OC i7 7700K all perform within 5% of each other, using setups which reflect the real world, like GTX 1080Ti for 4K, GTX 1080 for 1440p, and GTX 1060/RX 580 for 1080p, in relatively modern AAA games, barring a few exceptions.



Nonsense. Not all games are GPU bound. Try playing Astroneer with a large base. Even on a Kabylake 4.4 Ghz it can drop below 60 fps (more cores won't help). Or Planet Coaster, Arm 3. etc.
 
  • Like
Reactions: JoeRambo

Edrick

Golden Member
Feb 18, 2010
1,939
230
106
Ah ok maybe that makes a difference - both are X299 though so shouldn't be too much different. Is there another way to run them other than through the PCH on X299? Regardless I would think on a higher end MOBO, the speeds of modern M2 drives shouldn't be bottlenecked.

Running any drive through the PCH is going to cap your speed to the x4 DMI rate (minus overhead and other things running through it like LAN, USB and SATA). That goes with any board (X299, Z270, Z170, etc.). You will need to buy a PCIe m.2 card and leverage your CPU PCIe lanes in order to get the MAX performance from your m.2 devices. https://www.asus.com/us/Motherboard-Accessory/HYPER_M2_X4_MINI_CARD/

Bottom line is there is nothing wrong with your motherboard or SSD.
 

Eddward

Member
Apr 10, 2012
56
19
81
I expected more cores, slightly higher TDP, lower base clock, and about the same turbo. I think the turbo will be a bit lower than the 7700k turbo, but they might surprise us with the same turbo. Crude math says that a 6 core 4.0 GHz turbo is doable with the numbers that Intel released in the spring, but maybe they are better than that now. Even if not, Intel could set the turbo to 4.5 GHz but the desktop chips probably will down-throttle a lot more. You can't put in 50% more cores in a process that is about 23% more efficient with about the same TDP and expect to be able to stay at full turbo as long.
hmm, so let's say 4.1Ghz plus minus 100Mhz Turbo 2.0 (all cores?) and around 4.4Ghz Turbo 3.0 (1-2 cores) is that possible ? In this case what can be average air cooled OC for all cores ? up to 4.5Ghz, more ?
 

AdamK47

Lifer
Oct 9, 1999
15,168
2,826
126
On X99 the m.2 slots ran off of the CPUs PCI-E 3.0 lanes. For X299 it seems that motherboard makers would rather run it off of the PCH. That's great for me since I only have 16 lanes available from my CPU.
 

Edrick

Golden Member
Feb 18, 2010
1,939
230
106
On X99 the m.2 slots ran off of the CPUs PCI-E 3.0 lanes. For X299 it seems that motherboard makers would rather run it off of the PCH. That's great for me since I only have 16 lanes available from my CPU.

I have noticed that as well. With X99, they did not have a choice as the chipset did not support PCIe 3.0. It suckes that X299 owners have to buy cards to use m.2 at peak performance. But 2.6GB/s is good enough for me for now.
 

ManyThreads

Member
Mar 6, 2017
99
29
51
Running any drive through the PCH is going to cap your speed to the x4 DMI rate (minus overhead and other things running through it like LAN, USB and SATA). That goes with any board (X299, Z270, Z170, etc.). You will need to buy a PCIe m.2 card and leverage your CPU PCIe lanes in order to get the MAX performance from your m.2 devices. https://www.asus.com/us/Motherboard-Accessory/HYPER_M2_X4_MINI_CARD/

Bottom line is there is nothing wrong with your motherboard or SSD.

Interesting. I asked Elmor from Asus and told me he would have to investigate, which suggests it shouldn't be performing as badly as it is. Admittedly I am new to the M2 SSD world though so I'm sure I have a lot to learn as well.

I did find these two reviews - I didn't see anywhere that said their test setup included a separate PCI-E adapter, but maybe I missed something?:

http://www.cdrlabs.com/reviews/samsung-960-pro-512gb-m2-nvme-pcie-solid-state-drive/all-pages.html

https://benchmarkreviews.com/41954/samsung-960-pro-nvme-ssd-review/6/
 

dullard

Elite Member
May 21, 2001
24,998
3,325
126
hmm, so let's say 4.1Ghz plus minus 100Mhz Turbo 2.0 (all cores?) and around 4.4Ghz Turbo 3.0 (1-2 cores) is that possible?
I think those are quite possible. Even slightly better is possible if Intel bins hard or the process is better than what they claimed in the spring.
In this case what can be average air cooled OC for all cores ? up to 4.5Ghz, more ?
I personally think that the Coffee Lake overclocking won't be as good as Kaby Lake unless they solder the lids. And even then, we aren't likely going to hit 5.0 GHz. But, 4.4 GHz to 4.5 GHz seems like a quite reasonable number all cores with good cooling and a good chip.
 

TheF34RChannel

Senior member
May 18, 2017
786
309
136
Coffee Lake-S

Core i7-8700K 6C/12T 3.7 GHz base, 12MB L3, 95W
Core i7-8700 6C/12T 3.2 GHz base, 12MB L3, 65W
Core i5-8600K 6C/6T 3.6 GHz base, 9MB L3, 95W
Core i5-8400 6C/6T 2.8 GHz base, 9MB L3, 65W
Core i3 gets 4C/4T

https://www.cpchardware.com/coffee-lake-approche

Finally someone posted. Should I leak the Turbo clocks? I didn't believe at first. :D:eek:

Make me a happy man and leak those turbos! And tell us what you do for a living...

Canard PC Hardware says Coffee Lake is compatible with 200-series and some 100-series motherboards. Up until late June I heard it would be 300-series only, but great news if true.

https://mobile.twitter.com/CPCHardware/status/886940741599145984

I doubt it in all honesty...

How does better than Core i7-7700K's Turbo sound?

As above, enough teasing :p ha ha

Nice to see the clock speeds confirmed - I'm happy with this and I'll probably go with an i7 8700K for my upcoming build. I'm hoping they won't use crappy TIM but this is Intel...

If they are not using solder on HEDT they're not using it on the mainstream I can assure you. Unfortunately.

I think those are quite possible. Even slightly better is possible if Intel bins hard or the process is better than what they claimed in the spring.

I personally think that the Coffee Lake overclocking won't be as good as Kaby Lake unless they solder the lids. And even then, we aren't likely going to hit 5.0 GHz. But, 4.4 GHz to 4.5 GHz seems like a quite reasonable number all cores with good cooling and a good chip.

It's just me as I don't bench; I don't understand the need for 5GHz personally. It's not going to help my games. For benches I can understand it.
 
  • Like
Reactions: Sweepr

Zucker2k

Golden Member
Feb 15, 2006
1,810
1,159
136
That's a pretty small subset of users though. Not saying it isn't important to the few who are actually comptetive gamers, but really it's a small crowd.
Couldn't be smaller than those who mostly do rendering and encoding?
 

aigomorla

CPU, Cases&Cooling Mod PC Gaming Mod Elite Member
Super Moderator
Sep 28, 2005
20,839
3,174
126
Sigh.... fine make us delid the cpu, im cool with that...
ive always been a "real people pay for toys they break" club... meaning if i break it from over clocking, i pay for it.

However PLEASE and i say PLEASE give us a shim so we can put the naked die over a heat sink and not have to liquid metal / gallium / whatnot and then recycle the IHS.

Is it that hard for other board manufactorers to follow MSI and give us a Die Guard?

If were gonna be forking over 300+ dollars on a board with uber overclocking abilities, is it too much for them to supply us with something like this?
MSI-X99S-Xpower-AC_foto-19.jpg
 
  • Like
Reactions: Zucker2k

DrMrLordX

Lifer
Apr 27, 2000
21,582
10,785
136
I have my NVMe (m.2) issues all sorted. It was a PITA for the first week however.

It was a combination of BIOS fixes and actual testing application updates. Now I am running full speed with 2 Samsung NVMe drives installed.

Hmm, might be worth it to post your findings in the memory/storage forum for x299 owners, or maybe in the motherboard forum for people that own your particular board. If you haven't done that already.
 
  • Like
Reactions: Drazick

Edrick

Golden Member
Feb 18, 2010
1,939
230
106
Hmm, might be worth it to post your findings in the memory/storage forum for x299 owners, or maybe in the motherboard forum for people that own your particular board. If you haven't done that already.

Well, after today's conversation in this thread, I went and ordered a PCIe m.2 card so I can test my 960 Pro using both methods and compare results. I will post once my card arrives (2 days).
 
  • Like
Reactions: Zucker2k

AdamK47

Lifer
Oct 9, 1999
15,168
2,826
126
Sigh.... fine make us delid the cpu, im cool with that...
ive always been a "real people pay for toys they break" club... meaning if i break it from over clocking, i pay for it.

However PLEASE and i say PLEASE give us a shim so we can put the naked die over a heat sink and not have to liquid metal / gallium / whatnot and then recycle the IHS.

Is it that hard for other board manufactorers to follow MSI and give us a Die Guard?

If were gonna be forking over 300+ dollars on a board with uber overclocking abilities, is it too much for them to supply us with something like this?

I remember having a few Coppermine Pentium III socket 370 FC-PGA CPUs. No IHS back then. Just a bare die without any die guard to prevent damage. You had to tighten the cooler just right or else you would end up cracking the corners of the die. The Athlon Thunderbird CPUs had bare dies as well, but they had rubber pads on each corner of the PCB.
 
Last edited:

Nebor

Lifer
Jun 24, 2003
29,582
11
76
These are NOT my results - My results are all around 2/3 of these numbers and in some cases 1/2. The below is what I should be getting but I'm way less on both identical tests. The biggest differences are in Seq Q32T1 and Seq where I am a full 1000MB/s slower.

Samsung-960-PRO-512GB_cdm.jpg


Samsung-960-PRO-Magician-3.png

I'm seeing similar results from my 1tb Samsung 960 Pro. I had better benchmarks on my x99. I don't notice any difference in actual usage, but the only thing on the drive is Windows 10, iTunes, Office, & Steam, with all my games stored on an 8tb HGST drive and the machine only used for gaming. In other words, I don't really utilize the m2 anywhere near it's potential.
 

tamz_msc

Diamond Member
Jan 5, 2017
3,708
3,554
136
Nonsense. Not all games are GPU bound. Try playing Astroneer with a large base. Even on a Kabylake 4.4 Ghz it can drop below 60 fps (more cores won't help). Or Planet Coaster, Arm 3. etc.
A lot of these so-called ST dominant games are more memory latency and/or bandwidth bound than core performance bound. Proof, in case of Arma III:
nv_arma3.png
oc_nv_arma3.png
So I guess people might be looking at it the wrong way if they think that it's the CPU alone that makes the difference in these games. Faster RAM can go a long way in overcoming any deficiencies.

Compare gaming benchmarks across websites - AAA console ports, you know the games most of the review articles test with, are GPU bound in most circumstances.

this is exactly what is agains the experimental observations if you are sensitive to smoothess and if you actually play the games, there is a big difference in smoothness in the intensive scenes

games are tested in limited scenarios but I've learned an empirical equation- if review shows min fps at 60 the real fps in most sensitive scenes (which at high difficulty matters most) is around 20-40 fps

there is a huge difference between oced 7700k and 4,5GHz 2600K
Except in multi-GPU setups, there isn't much of a difference between those two. This generation of console ports are, with certain exceptions, GPU bound in most cases. Going by this article, the best example where the 5GHz 7700K has a big lead over the 4.5GHz 2600K is GTA V maxed out at 1440p with a GTX 1080. Look at what consoles are focusing on after an unprecedented mid-generation refresh - 4K. So of course most of the new AAA titles are going to be made in a way to extract maximum performance out of the GPU. You can't push the CPUs in consoles too hard given how weak they are relative to modern desktop CPUs.

If reviewing gaming performance of CPUs is to be reduced to finding out how well it can cope with bottlenecks, then we should really be testing Blighttown in Dark Souls, because it fits perfectly with your description of a game that can run otherwise at 60fps but plummets to 20-30fps in particular regions. However, I doubt that many people are going to be interested in such a comparison.
 
  • Like
Reactions: Drazick

beginner99

Diamond Member
Jun 2, 2009
5,208
1,580
136
hen use a 1080 TN panel is beyond comprehension. Makes not the least bit of sense.

It does if you don't want to see a blurry mess in nice colors and don't want to spend another $300 to get the same features in a 1440p display and pay another $1000 on GPU to actually use these features at 144 hz.

Can you read the street names here? No? Well I can on my <redacted>TN panel.

Profanity is not allowed here
Markfw
Anandtech Moderator
 
Last edited by a moderator:

wildhorse2k

Member
May 12, 2017
180
83
71
Running any drive through the PCH is going to cap your speed to the x4 DMI rate (minus overhead and other things running through it like LAN, USB and SATA). That goes with any board (X299, Z270, Z170, etc.). You will need to buy a PCIe m.2 card and leverage your CPU PCIe lanes in order to get the MAX performance from your m.2 devices. https://www.asus.com/us/Motherboard-Accessory/HYPER_M2_X4_MINI_CARD/

Bottom line is there is nothing wrong with your motherboard or SSD.

I already posted reviews that tested Z270 and X299 M.2 speed few pages back (separately, but same m.2 type). On Z270 M.2 was faster. If he is getting 2/3 of speed then I doubt its caused by PCH as they use the same.
 
  • Like
Reactions: ManyThreads

Sweepr

Diamond Member
May 12, 2006
5,148
1,142
131
Daily Coffee Lake tease: What if an entry level Core i5 could match the MT performance of the popular Core i7-7700K?
 

coercitiv

Diamond Member
Jan 24, 2014
6,150
11,670
136
Daily Coffee Lake tease: What if an entry level Core i5 could match the MT performance of the popular Core i7-7700K?
i5 7400 has 3.3Ghz MT load turbo, even with same turbo clocks and 50% increased throughput i5 8400 will stay bellow 7700K. (close, but no cigar)

But how would MT turbo stay the same when TDP remains at 65W while core number increases? Do you honestly expect 14nm++ to drop power usage by as much as 35% in the low 3Ghz range?
 

JoeRambo

Golden Member
Jun 13, 2013
1,814
2,105
136
New Daily Coffee Lake tease: What if an entry level Core i5 could match the MT performance of the popular Core i7-7700K?

Please stop torturing fine ppl of the forums :) Leak those turbo clocks already, so we can get a taste of where OCed CL will end up.

I dream about 5Ghz / 4.5AVX clock 6 core beast, hopefully with 4.5Ghz uncore too. Question is, how will it scale with memory speeds? Hexacore could get bottlenecked on dual channel...