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 .
Sancus said:
The biggest thing helping Ryzen competitiveness is honestly the continuing supply issues with Coffee Lake. For lower end builds that would be getting an i5-8400 it's hard to recommend to people that they fight over newegg and amazon in-stock notifications instead of just ordering Ryzen.
Click to expand...

I would have to disagree. What is helping Ryzen is that it is a really good processor, and good value for what you get.

Once AMD can get those clocks higher lets say 400-500mhz, then Intel might be in a little bit of trouble.

Some people actually do need more cores over higher IPC.
 
Sancus said:
The biggest thing helping Ryzen competitiveness is honestly the continuing supply issues with Coffee Lake. For lower end builds that would be getting an i5-8400 it's hard to recommend to people that they fight over newegg and amazon in-stock notifications instead of just ordering Ryzen.
Click to expand...
That and the higher motherboard cost. No need for a z370 for a locked chip. If they would have just allowed them on b250 boards ryzen might be harder to justify. You can get a decent 6c/12t chip with motherboard for $250 (or less) by going ryzen.
 
How are the benefits of lapping these days?

D5wjSVb.png
 
Just build my 8700k rig. Wow what a difference in battlefield 1 multiplayer. Coming from i5 6600k to i7 8700k is noticable.

Quick question. When i was playing battlefield 1, i notice the cpu is clocked at 3.8 ghz only. Why is that? I have not touch any settings yet in bios or overclocking yet.
 
EvoFreak said:
Just build my 8700k rig. Wow what a difference in battlefield 1 multiplayer. Coming from i5 6600k to i7 8700k is noticable.

Quick question. When i was playing battlefield 1, i notice the cpu is clocked at 3.8 ghz only. Why is that? I have not touch any settings yet in bios or overclocking yet.
Click to expand...

3.8ghz is weird. Checked the temps? Most likely just some weird default mobo setting.
 
So how are your OC's holding?
Mine is Prime stable only all the way up at 1.27V @ 4..8 GHz, uncore 46, AVX -2.
It boots and runs mostly fine even at 1.24V, passes Cinebench and such, but a non-AVX Prime95 Blender run makes it BSOD after about 30 minutes. Temps are ~75C without AVX, ~82C with it on hottest core (small FFT's).
 
kalston said:
3.8ghz is weird. Checked the temps? Most likely just some weird default mobo setting.
Click to expand...

Yeah i only had few hours to test the new rig. Temps are good so far. 50c at load. But thats only with 3.8ghz. Still have to play with the settings for sure when i get the free time.
 
stashix said:
So how are your OC's holding?
Mine is Prime stable only all the way up at 1.27V @ 4..8 GHz, uncore 46, AVX -2.
It boots and runs mostly fine even at 1.24V, passes Cinebench and such, but a non-AVX Prime95 Blender run makes it BSOD after about 30 minutes. Temps are ~75C without AVX, ~82C with it on hottest core (small FFT's).
Click to expand...
Will still have to play with the settings before starting to oc. Hopefully i have a nice overclocking chip. Its been a while since i overclock a cpu. I read somewhere 5.0 ghz is possible on this chip? What cooling do you have?
 
Shintai said:
There is a reason why the Ryzen line had a huge price cut up to 36%. And its called CFL. And its not over yet, another round comes in Q1 ;)
Click to expand...

doubt it'll release it in Q1 tbh it'll just kill off CFL. waiting for 8 cores.. 1 more year.. sigh

also LOL at i9 BGA. i want desktop cpu in laptop only.
 
Last edited:
Looks like the Rockit88 site is planning a cyber monday sale, I think I'll order my kit then.
Unless there are other sites selling it? It's more expensive on Amazon.
 
EvoFreak said:
Will still have to play with the settings before starting to oc. Hopefully i have a nice overclocking chip. Its been a while since i overclock a cpu. I read somewhere 5.0 ghz is possible on this chip? What cooling do you have?
Click to expand...
Noctua NH-D15S plus an extra fan. 5 GHz should be no problem with a good AIO.
 
Most chips will do 5 GHz around 1.3-1.35V. If you get a good one it's easy w/o a delid. Average or Poor chip and you will need a delid.
Also depends on if you're using air or water on your CPU/GPU, and your case airflow, in my experience so far.

Anything over 1.3V is a struggle for air cooling (D15).

There's a bunch of results in this thread:
 
Dayman said:
There is wild speculation that the 9000 series parts in this is a CFL Refresh, well according to Fanless Tech it is not(these guys usually know about procs before their released or leak some stuff):


Click to expand...

thats excellent man, that means 9800k is 8 cores 10nm+ end of 2018. no.. not talking about 9600k..

http://www.guru3d.com/index.php?ct=articles&action=file&id=34771 Shintai juanrga
pls explain the ipc how is it here they tested showing ryzen so much above haswell sounds like bs to me. it should be around the same.
 
Communism said:
AMD has always preferred to invest first in advertising tricks and last in actual engineering and development.

AMD sold tons of FX-8150s as well to people dumb enough to believe their lies, the same cycle is replaying today.

Even the dumb-asses that buy the advertising tricks at first realize they've been duped eventually, as we saw with Phenom, and then again with Bulldozer, and now with Ryzen.
Click to expand...

Why the name communism?
 
TaintedSquirrel said:
Most chips will do 5 GHz around 1.3-1.35V. If you get a good one it's easy w/o a delid. Average or Poor chip and you will need a delid.
Also depends on if you're using air or water on your CPU/GPU, and your case airflow, in my experience so far.

Anything over 1.3V is a struggle for air cooling (D15).

There's a bunch of results in this thread:
Click to expand...

Meh these all need delids unless you don't OC of course.
 
Dayman said:
There is wild speculation that the 9000 series parts in this is a CFL Refresh, well according to Fanless Tech it is not(these guys usually know about procs before their released or leak some stuff):


Click to expand...

Icelake comes in that exact time frame. So not sure why even people thought CFL to begin with ;)
 
oleNBR said:
thats excellent man, that means 9800k is 8 cores 10nm+ end of 2018. no.. not talking about 9600k..

http://www.guru3d.com/index.php?ct=articles&action=file&id=34771 Shintai juanrga
pls explain the ipc how is it here they tested showing ryzen so much above haswell sounds like bs to me. it should be around the same.
Click to expand...

Cinebench is single cycle SSE loops at this point, not much to improve. Its like running 10-15 year old code and wonder why there isn't any real benefit. If you ran the latest Cinema engine in the commercial products, you see Haswell etc far ahead of Ryzen.

So anything else is related to cache, memory bandwidth etc. And you start to compare something like 1600Mhz vs 2667Mhz memory and different cache sizes and uncore clocks. Same reason you see lower Ryzen SKUs perform less, despite same IPC. And why you see different results over the same Skylake cores 6700K vs 7740K etc.

Here Ryzen is at SB performance again. And you wouldn´t say a 1300X performs like a 4670K, despite having much faster memory. As in 67% more bandwidth.
upload_2017-11-25_10-45-30.png
 
Last edited:
Shintai said:
Cinebench is single cycle SSE loops at this point, not much to improve. Its like running 10-15 year old code and wonder why there isn't any real benefit. If you ran the latest Cinema engine in the commercial products, you see Haswell etc far ahead of Ryzen.

So anything else is related to cache, memory bandwidth etc. And you start to compare something like 1600Mhz vs 2667Mhz memory and different cache sizes and uncore clocks. Same reason you see lower Ryzen SKUs perform less, despite same IPC. And why you see different results over the same Skylake cores 6700K vs 7740K etc.

Here Ryzen is at SB performance again. And you wouldn´t say a 1300X performs like a 4670K, despite having much faster memory. As in 67% more bandwidth.
View attachment 44276
Click to expand...

hold on, thats GPU involved though i mean it says fps at the bottom. also i think 95% of software out there are still legacy software which is why windows is most popular, which would mean CB single cycle SSE should be more accurate representation of IPC for majority of usage out there. guru3d's review is new, but they didnt include the memory speed of each bench system even if they are all clocked at 3.5ghz.
 
oleNBR said:
hold on, thats GPU involved though i mean it says fps at the bottom. also i think 95% of software out there are still legacy software which is why windows is most popular, which would mean CB single cycle SSE should be more accurate representation of IPC for majority of usage out there. guru3d's review is new, but they didnt include the memory speed of each bench system even if they are all clocked at 3.5ghz.
Click to expand...

Cinebench is not using a lot of instructions and isn't a representation of anything much outside how fast you can run certain older SSE loops. Even for a 2013 release it was ancient and not even a representation of their commercial products that got AVX support. So its not an indication in any way just like other similar limited benchmarks isn't either. Also you have different uncore and memory speeds to mix in. It wasn´t some equal compare. Try mix Sandy Bridge and Haswell with 2667Mhz instead of 1333 and 1600Mhz. Just as Skylake for example gets benefit all the way past 4000Mhz memory for the same reason. Uncore clocks is the same problem and one of the reason SKL-X loses to SKL in applications that likes that, despite the core being superior in all aspects.

Legacy software is a lot less popular than you think. Just look at your own PC. How much software do you run that's 10 years old or even 5 years old. And since we talk Ryzen and SKL/KBL/CFL you can in reality start by running Windows 10.
 
Last edited:
I just learned how to OC my 8700K and so far Im at 5ghz with 1.260v. Im still using 2400 mhz ram that i was using from my previous skylake build. Im thinking of upgrading the ram for faster ones. Will it benefit me upgrading the ram?
 
EvoFreak said:
I just learned how to OC my 8700K and so far Im at 5ghz with 1.260v. Im still using 2400 mhz ram that i was using from my previous skylake build. Im thinking of upgrading the ram for faster ones. Will it benefit me upgrading the ram?
Click to expand...

Quite a bit.
 
Is Realtemp still relevant to monitor temps? Or do you guys recommend something else?

What about Stress Testing programs to test my cpu OC stability?
 
oleNBR said:
thats excellent man, that means 9800k is 8 cores 10nm+ end of 2018. no.. not talking about 9600k..

http://www.guru3d.com/index.php?ct=articles&action=file&id=34771 Shintai juanrga
pls explain the ipc how is it here they tested showing ryzen so much above haswell sounds like bs to me. it should be around the same.
Click to expand...

Compared to rest of reviews, one can see that Guru3d obtains higher scores for RyZen chips and/or lower scores for Intel chips, which explains how they reduce the IPC gap between both.

For instance, guru3d obtains 146cb for 1800X in the link you give, but PcPer and Ars get 139 and 140 respectively; guru3d obtains 146 for Broadwell-E, but PcPer and Ars get 151 and 153 respectively. So guru3d obtains about 4% higher score for AMD and another 4% lower score for Intel, which is equivalent to reducing the IPC gap by about 8%.

It is not random variation in benchmarking, Guru3d sistematically obtains lower gaps between AMD and Intel.

How guru3d gets those scores is the million dollar question. In some cases they are testing overclocked RyZen chips but labeling them as stock in graphs, in other cases they are testing engineering samples of Intel chips, but labeling them as stock in the graphs. I recall that for the i9-7900X review guru3d used an engineering sample and beta BIOS. One would expect them to add an "ES" label to the graphs or something, but they label the chip as if was a retail chip. It is only when on pay attention to CPU-Z screens that one see they are using a engineering sample.

Also guru3d only uses CineBench to measure IPC, whereas other reviews use a combination of different workloads to obtain the IPC. Cinebench is a favorable case for RyZen. On CineBench, AMD is only 10% behind Intel clock-for-clock. On other workloads AMD is 25% behind.

clock-audacity.png



Overall CoffeeLake IPC is a good 15--20% ahead of RyZen. That is how 6C CoffeeLake is able to beat 8C RyZen in some multithread workloads.
 
Last edited:
TaintedSquirrel said:
...Also because Intel's chips are clocked a lot higher. I thought the IPC was closer to 10%.
Click to expand...

well according to Shintai and juanrga they are saying IPC is greater when using software that is more advanced to take advantage of more advanced architecture and cpu extension, such as skylake-x's cache rework but unfortunately with the lowly clocked uncore ruined it. however imho what they didnt explain or perhaps they disagrees to, is that ryzen actually has better and more advanced architecture in certain areas like caching and it's hyperthreading more superior than intels, while being crippled by infinity fabric. in a way it is true because intel is still on core build since forever.

evidence show ryzen lower score in CB15 than intel but when multi threaded it can do real well which gives two thing. ryzen's multi threaded workload takes full advantage of well optimized (SSE but not new extension) software, or the software isnt shuffling workload between the ccx for the fabric penalty to show. otherwise juanrga example using another software to show IPC should be greater would be that specific software is more optimized for intel than AMD, that or its not as optimized in multi threaded workload than CB.
 
oleNBR said:
well according to Shintai and juanrga they are saying IPC is greater when using software that is more advanced to take advantage of more advanced architecture and cpu extension, such as skylake-x's cache rework but unfortunately with the lowly clocked uncore ruined it. however imho what they didnt explain or perhaps they disagrees to, is that ryzen actually has better and more advanced architecture in certain areas like caching and it's hyperthreading more superior than intels, while being crippled by infinity fabric. in a way it is true because intel is still on core build since forever.

evidence show ryzen lower score in CB15 than intel but when multi threaded it can do real well which gives two thing. ryzen's multi threaded workload takes full advantage of well optimized (SSE but not new extension) software, or the software isnt shuffling workload between the ccx for the fabric penalty to show. otherwise juanrga example using another software to show IPC should be greater would be that specific software is more optimized for intel than AMD, that or its not as optimized in multi threaded workload than CB.
Click to expand...

Ryzen does not have superior SMT. It is balanced differently, and is overcoming a resource usage deficit in the single thread case which Intel does not have to the same degree.

CB is just a really simple program, from an instruction use standpoint. It does not really resemble typical program flow very well.
 
PhaseNoise said:
Ryzen does not have superior SMT. It is balanced differently, and is overcoming a resource usage deficit in the single thread case which Intel does not have to the same degree.

CB is just a really simple program, from an instruction use standpoint. It does not really resemble typical program flow very well.
Click to expand...

i honestly donno the details and how the software tells cpu use its resources. i can only make guesses that infinity fabric isnt an issue inside of CB and possibly blender test only. either that or those software, even though as simple as they may be, are well optimized and would show ryzen's strength, but as we all know 95% of the software aren't even remotely optimized hence reflect so many poor test on ryzen. another pov would be, the workload inside ryzen during CB/blender isnt shuffled between the threads, no shuffle = no data going between CCX then no penalty, which i donno if thats possible given how L3 is shared among all cores.
 
TaintedSquirrel said:
...Also because Intel's chips are clocked a lot higher. I thought the IPC was closer to 10%.
Click to expand...

The IPC difference is more than 33%. We can sit and find cherry picked benchmarks with low IPC loads or bottlenecks elsewhere. CFL nuked the price of Ryzen for the same reason because you can´t excuse it anymore. We can also run AVX2 and AVX512 loads and conclude there is a 180% or 250% IPC difference as well, but its quite useless as a real metric.
 
Last edited:
oleNBR said:
i honestly donno the details and how the software tells cpu use its resources. i can only make guesses that infinity fabric isnt an issue inside of CB and possibly blender test only. either that or those software, even though as simple as they may be, are well optimized and would show ryzen's strength, but as we all know 95% of the software aren't even remotely optimized hence reflect so many poor test on ryzen. another pov would be, the workload inside ryzen during CB/blender isnt shuffled between the threads, no shuffle = no data going between CCX then no penalty, which i donno if thats possible given how L3 is shared among all cores.
Click to expand...

Infinity fabric is a massive bottleneck. But if you chose to use applications that scales extremely well and doesn't have any main threads. Then you dont see it as an issue (CB uses tile based render). That's why companies with weak interconnect likes to demo these. There is a reason why you can run faster by disabling a cluster or increase the infinity fabric. Just look at EPYC, there are more latency on a single EPYC chip than a 4 node Xeon. And one of the reasons why EPYC have sold 100-200 chips and Xeons have sold 1M+

There is a reason why Ryzen already got huge price cuts and more to come. And then we dont have to talk about the RMA hell and bugs that plaque Ryzen with machine halts and data corruption because someone skipped close to a year of QA.
 
Last edited:
Shintai said:
The IPC difference is more than 33%. We can sit and find cherry picked benchmarks with low IPC loads or bottlenecks elsewhere. CFL nuked the price of Ryzen for the same reason because you can´t excuse it anymore.
Click to expand...

Shintai said:
There is a reason why Ryzen already got huge price cuts and more to come. And then we dont have to talk about the RMA hell and bugs that plaque Ryzen with machine halts and data corruption because someone skipped close to a year of QA.
Click to expand...

This may of been true if CFL was always available to everyone at MSRP. But.... it wasn't

The R7 competes performance wise with the upper end i5s and lower end i7s. 1700 and 1700x prices have dropped to below those.
Both the R5 and the i5 8400 are competitive. Only a small price drop for the R5, which was already a great deal
The R3 destroys the i3 and still had a price drop.

Why? Because AMD has an opportunity to pick up market share. And, perhaps they are less dickish.

I realize things are different in Europe. Here, CFL availability has been scarce. When they are available, prices are inflated. That is a fact.


Oh and please stop making reliability claims based on a forums you have read. All of the Ryzen CPUs have received high satisfaction ratings on all venues. Growing pains may have come with 3rd party motherboards or Ram, but it is often the case with a new architecture. Pretty sure x299 had plenty of issues as well and less of those have been sold than the mainstream chips.
 
Nightfire said:
This may of been true if CFL was always available to everyone at MSRP. But.... it wasn't

The R7 competes performance wise with the upper end i5s and lower end i7s. 1700 and 1700x prices have dropped to below those.
Both the R5 and the i5 8400 are competitive. Only a small price drop for the R5, which was already a great deal
The R3 destroys the i3 and still had a price drop.

Why? Because AMD has an opportunity to pick up market share. And, perhaps they are less dickish.

I realize things are different in Europe. Here, CFL availability has been scarce. When they are available, prices are inflated. That is a fact.


Oh and please stop making reliability claims based on a forums you have read. All of the Ryzen CPUs have received high satisfaction ratings on all venues. Growing pains may have come with 3rd party motherboards or Ram, but it is often the case with a new architecture. Pretty sure x299 had plenty of issues as well and less of those have been sold than the mainstream chips.
Click to expand...

Riiiight, because AMD is a charity company that now for their blue eyes sake made 30-35% price cuts with more to come after "milking the FX upgraders". The 1800X is nowhere near a 8700K or even a 8700 non K. Unless you wish to apply the FX style "benchmarking".

AMD already missed the opportunity to gain market share. Whats next, the APU as excuse? or will it be 12LP after that with Zen+? Or maybe Zen2 in 2019? There is a reason why they got a lot lower revenue in Q4 unlike the other 2 companies. Despite just launching the fabled Zen APU and the forever wait for EPYC takes off.

And no MCE/data corruption issues, you should tell AMD so they dont have to RMA and send out hand picked and verified chips back.

This is just one of the issues:
https://community.amd.com/thread/215773?start=0&tstart=0

Not to mention the current EPYC line is pretty much cancelled due to bugs. That's what happens when you skip something like a year of QA in desperation.

R5 isn't competitive with 8400, not even close ;)
 
Last edited:
Surprise, surprise. You make the 1800x argument. A CPU for those that want to use an a320 board or simply do not like to o/c. If you haven't been paying attention lately, the 1700x/1700 are about $100 cheaper than the i7 competitors... with an affordable mb. Those of course match or beat the 8700k in cinebench, blender, crona and others. ;);)
But they suck because bad 1080p gaming.

The R5 easily beats the 8400 in several multi thread benchmarks and overclock capable with a cheaper motherboard (no b360, what a joke) ;););):p:p:p
But they suck because bad 1080p gaming.

Your measure of market share: steam users and hwbot submissions. amiright?
 
You must log in or register to reply here.
Back
Top