Intel's 8th Generation Core Family - Coffee Lake (LGA 1151, 6C/12T)

Where do you expect Core i7-8700K's Turbo to land?

  • 3.8/3.9 GHz

    Votes: 0 0.0%
  • 4.0/4.1 GHz

    Votes: 3 23.1%
  • 4.2/4.3 GHz

    Votes: 6 46.2%
  • 4.4/4.5 GHz

    Votes: 3 23.1%
  • 4.6/4.7 GHz

    Votes: 1 7.7%

  • Total voters
    13
  • Poll closed .
Some very impressive productivity/synthetics results from i7-8700K. Looks like matching/beating R7 1800X in many apps at stock will be no issue.

10017885_5_thumb.jpg


10017885_4_thumb.jpg


10017885_9_thumb.jpg


10017885_11_thumb.jpg


10017885_7_thumb.jpg


10017885_14_thumb.jpg

Tomb Raider seems to like more cores...

10017885_14_thumb.jpg
 
Last edited:
If fifty to a hundred bucks in higher RAM costs is going to stop your entire Coffee Lake build, best that you don't look at GPU pricing...
Nice try but the 32 gigs of fast ( for ddr3 ) ram that I bought nearly two years ago was only a hundred thirty bucks and it's several hundred dollars more for 32 gigs of fast ddr4. And I already have a 1080 ti so obviously I'm not concerned about current GPU prices.
 
Last edited:
I was getting excited to do an 8700k build until I looked at the DDR4 prices...
Same. Price goes up too quickly when looking at faster sticks.
Answers the question about the 7800X. The 8700K kills it in gaming while they're about equal in productivity and benchmarks. No surprise there.
Nice try but the 32 gigs of fast ( for ddr3 ) ram that I bought nearly two years ago was only a hundred thirty bucks and it's several hundred dollars more for 32 gigs of fast ddr4. And I already have a 1080 ti so obviously I'm not concerned about current GPU prices.
It's not so bad if you stay within spec. The 32GB in my sig was $230 four years ago, while the "same" set from Corsair in DDR4-2666 flavor today is $260. But based on earlier tests on memory bandwidth and speed with Skylake I'd want to go with at least DDR4-3200, and now we're talking $360.
Sweet pricing. Im def picking up an 8700k.

When do these actually go on sale? I am going to pay microcenter a visit.

http://wccftech.com/intel-8th-gen-coffee-lake-desktop-processors-official/
Current information says retail availability is next Thursday.
 
Last edited:
There is something I don't understand completely:

What is up with 8600k's TDP? It clocks lower than a 8600(non-K), has no HT yet it has 50% more TDP?
 
There is something I don't understand completely:

What is up with 8600k's TDP? It clocks lower than a 8600(non-K), has no HT yet it has 50% more TDP?

1. Its unlocked so they raise the TDP

2. Wheres the 8600?
Intel-Coffee-Lake-8th-Gen-Desktop-Processors_13.png

7IawQlp.png



I see no 8600
 
There is something I don't understand completely:

What is up with 8600k's TDP? It clocks lower than a 8600(non-K), has no HT yet it has 50% more TDP?

This is always the case, 7600K is 91W TDP. But AFAIK, TDPs on unlocked CPUs on overclocking motherboards are just a target and not particularly meaningful. My guess is that Intel sets the TDP for unlocked chips significantly higher so that OEMs include sufficient cooling, also note that the base clocks on the unlocked CPUs in this case are quite a bit higher than the ones on the locked ones.

If a fully loaded i7-8700 at stock clocks needs 65W of cooling, it makes sense that Intel bumps the TDP up by 50% for the chips that are already gonna be running all cores15% faster (and probably use 20-25% more power as a result, leading to higher TDP), cause if people turn on multi-core enhancement(which locks all cores to the highest turbo speed) and consider OCing further, it's not too hard to see how the required cooling will be MUCH higher than the locked CPU's 65W.
 
Sorry fellows! I meant the 8600k against the 8700(non-key).

As per Shintai's comment... ok, if its simply an Intel definition I get it.
 
advertised TDPs are just that anyways, they don't mean anything other than being a guideline for what coolers to use at a minimum on a given CPU. K SKU cpu's have always had elevated TDP ratings due to being unlocked to provide overclocking headroom.
 
Sorry fellows! I meant the 8600k against the 8700(non-key).

As per Shintai's comment... ok, if its simply an Intel definition I get it.

Official TDPs come in thermal slots. Instead designing a cooler for each CPU in existence, coolers are designed for some few slots: 140W, 95W, 65W, 35W,... and then CPUs are rated to the closer slot by rounding up.

The 8600K would have higher TPD than the 8700, because has higher base clock. In a first computation I get about 82W, but then it is labeled as 95W because this is the close thermal slot available.
 
Same. Price goes up too quickly when looking at faster sticks.

Answers the question about the 7800X. The 8700K kills it in gaming while they're about equal in productivity and benchmarks. No surprise there.

It's not so bad if you stay within spec. The 32GB in my sig was $230 four years ago, while the "same" set from Corsair in DDR4-2666 flavor today is $260. But based on earlier tests on memory bandwidth and speed with Skylake I'd want to go with at least DDR4-3200, and now we're talking $360.

Current information says retail availability is next Thursday.
Can still buy 16gb(2x 8gb) of ddr4 3000mhz for $130 so 2 sets( sticks) and a little OC is $260

But 16gb stick are expensive
 
Can still buy 16gb(2x 8gb) of ddr4 3000mhz for $130 so 2 sets( sticks) and a little OC is $260

But 16gb stick are expensive
That's not even remotely close to the speeds that I'm looking at for my Ram. And that's the whole point that I was making because the 2133 Ram that I have was pretty much near the very top ddr3 made which is like getting 4133 Ram now which is beyond insanely expensive. 3600 will be the absolute minimum that I will personally go with.
 
Why?

Can you show (for yourself!) that the extra speed will do anything for your workloads?
I'm not going to get into a debate here but there's plenty of proof and benchmarks that show faster Ram most certainly helps out. I went through the same nonsense with people back when I got 2133 DDR3 RAM and luckily sites like techspot and digitalfoundry actually started testing this and completely validated what I said back then and that holds true now for faster ram. Of course back then though the fastest Ram was hardly any more expensive than the slowest Ram unlike now.
 
That's not even remotely close to the speeds that I'm looking at for my Ram. And that's the whole point that I was making because the 2133 Ram that I have was pretty much near the very top ddr3 made which is like getting 4133 Ram now which is beyond insanely expensive. 3600 will be the absolute minimum that I will personally go with.
If you think you need that then just get quad channel.

It all depends on cas x mhZ

My ddr3 works well at 1866mhz and cas 9 and 1t. Even though it will run much faster if I go to cas 11 but shows the same in benches
 
Last edited:
Some very impressive productivity/synthetics results from i7-8700K. Looks like matching/beating R7 1800X in many apps at stock will be no issue.

10017885_5_thumb.jpg


10017885_4_thumb.jpg


10017885_9_thumb.jpg


10017885_11_thumb.jpg


10017885_7_thumb.jpg


10017885_14_thumb.jpg

Tomb Raider seems to like more cores...

10017885_14_thumb.jpg

good to see some leaks but as expected, no ipc increase over 7700k. CB15 ST show similar results within error margin. all we can get out of this is more cache, more cores at same price. and hope that 14nm++ will allow average higher overclock.
 
good to see some leaks but as expected, no ipc increase over 7700k. CB15 ST show similar results within error margin. all we can get out of this is more cache, more cores at same price. and hope that 14nm++ will allow average higher overclock.
Intel is bragging about improved OC in their slides (power delivery). Hopefully this is a Devils Canyon situation.
 
Re: price
The i7-8700K costs $359, up from the $305 launch price of the i7-7700K, while the i5-8600K costs $257, up from the $217 of the i5-7600K. Bear in mind that these are that these per chip in a 1000 unit order prices—expect retail to be higher still.
If that's accurate, we're looking at a $400 chip.
 
good to see some leaks but as expected, no ipc increase over 7700k. CB15 ST show similar results within error margin. all we can get out of this is more cache, more cores at same price. and hope that 14nm++ will allow average higher overclock.

not sure why you'd expect to see IPC increases its still skylake architecture 14+ brought a pretty good clock increase over 14 so 14++ along with the added 2 cores is the performance improvement. 4.7 ghz single core turbo is the fastest stock clock intel has ever put on a cpu and reports so far are these 6 core cpus are hitting 4.8+ ghz on all cores which is pretty impressive. after this is icelake next year (10nm+ arch jump) then 2019 is slated to be tiger lake (10nm+ or better) and is a semi-tock so it will be the next cpu with potential IPC gains from intel.
 
Whatever it is, it's only using 6 cores.

I mean it's still a video game and so it needs high single thread performance for its main thread regardless of DX12 draw call magic or game dev work done to make it take advantage of multiple cores. But yeah all benchmarks I've seen indicate that even DX12 doesn't make use of more than 6 cores. DX11 doesn't even get that far, of course.
 
2. Wheres the 8600?
I see no 8600

Maybe they'll release it with Z390 and it will only be compatible with Z390 and not backward compatible with Z370. Sounds like something Intel would do at this point.
 
I'm not going to get into a debate here but there's plenty of proof and benchmarks that show faster Ram most certainly helps out.

Faster than stock, sure, I'm using DDR4-3000 myself. But how much faster? What's it going to buy you that another $100 into GPU (etc.) won't give?
 
Intel is bragging about improved OC in their slides (power delivery). Hopefully this is a Devils Canyon situation.

yea if 7700k average was around say, 4.8 to at 1.25v, then 8700k with 14nm++ hopefully be around 4.9 or even 5ghz at 1.25v with 4 good cores and 2 cores disabled. with 2 more cores, it goes by the worst core but if they are all similar in quality we should expect to see 4.9 or 5ghz all 6 cores around there is what im hoping.


not sure why you'd expect to see IPC increases its still skylake architecture 14+ brought a pretty good clock increase over 14 so 14++ along with the added 2 cores is the performance improvement. 4.7 ghz single core turbo is the fastest stock clock intel has ever put on a cpu and reports so far are these 6 core cpus are hitting 4.8+ ghz on all cores which is pretty impressive. after this is icelake next year (10nm+ arch jump) then 2019 is slated to be tiger lake (10nm+ or better) and is a semi-tock so it will be the next cpu with potential IPC gains from intel.

i was expecting it because of rumours, we can always dream and be hopeful even though our logical side knew it wont be any different.
 
Can somebody explain how PCIe lanes are counted? The CPU itself has 16, but there are more lanes on the chipset? How does this work?
 
Can somebody explain how PCIe lanes are counted? The CPU itself has 16, but there are more lanes on the chipset? How does this work?

1151 cpus have 20 CPU lanes but 4 of those are dedicated to the DMI 3.0 Link to the PCH (chipset) those 4 lanes get multiplexed by the chipset to create the additional 24 lanes you see with Z270 and now Z370.

for consumer use nobody ever hits full bandwidth saturation in this setup still, it basically takes crazy NVME raid setups to do it (see HEDT/Server) one could argue quad gpu needs it but quad gpu is dead besides on older cards, even SLI 1080 Ti has no issues with running perfectly fine at 8x/8x direct to cpu thanks to the SLI bridge helping out with cross communication, if and when dx12 mGPU ever is really a thing it could however become a problem but the fact that tech is extremely reliant on game developers to implement and optimize and I honestly don't see it ever taking off but only time will tell.
 
Thank you for the explanation. Which components get the direct to CPU link and which ones are on the multiplexed lanes? Are most to the 16 non-MUX lanes for CPU module communication?
 
Thank you for the explanation. Which components get the direct to CPU link and which ones are on the multiplexed lanes? Are most to the 16 non-MUX lanes for CPU module communication?

typically on most motherboard designs the primary x16 slot gets the 16 CPU lanes, everything else is off the 4 lanes to the PCH. alternatively on most boards if there is two GPU's in the correct slots the lanes get split 8x to primary gpu 8x to secondary, rest stays the same.

  • 985 MB/s (×1)
  • 15.75 GB/s (×16)
that is the bandwidth of PCI-E 3.0 so on your typical platform you have 15.75 GB/s available for GPU(s) and 3,940 MB/s split up for storage/network/audio/whatever expansion card you might have in the system (these days usually none)
 
  • Like
Reactions: FPGA
like this
Back
Top