Intel Coffee Lake Leaked News

[Bleedingyamato]

1 hour ago, Suika said:

I wouldn't say I was expecting an IPC boost from Coffee Lake, rather the existing increase to IPC that was seen moving to Broadwell and then to Skylake. That combined with higher frequency chips would tempt me to upgrade, so if what you're saying is true, I may jump to Z370 then.

Wouldn't there be an IPC increase for Coffee lake?

 

It's 2 generations past Skylake and after last year's no improvement from Kaby along with the threat of Ryzen Intel would be insane to release new CPUs with no IPC improvement wouldn't they?

 

29 minutes ago, Princess Cadence said:

@TheRandomness
Coffee Lake is what Kaby Lake should've been, I am ignoring it even if motherboard compatibility is a thing, I'll only replace my i7 7700 for a 10nm processor, or who knows if performance bump is still meaningless and Intel keeps only caring for power efficient on the mainstream segment I'll wait until 7nm altogether lol.

Do you think we'll actually see an actual increase in performance?    

 

Intel is having trouble perfecting smaller nm CPUs I think?    So to get to even 10nm might take a while right?   Even longer for 7nm?

[Jito463]

16 minutes ago, Bleedingyamato said:

Wouldn't there be an IPC increase for Coffee lake?

 

It's 2 generations past Skylake and after last year's no improvement from Kaby along with the threat of Ryzen Intel would be insane to release new CPUs with no IPC improvement wouldn't they?

If they're far enough along in development, they probably can't make significant changes to the design.  They had to have started development on these long before Ryzen actually came out (they would have known about it from AMDs roadmap, but not known the performance level).

[Clanscorpia]

I feel like if Intel really did push forward as everyone here wants them to have done AMD would have never caught up. Just think about that a little bit. I think in the end, Intel being complacent actually helped the market because it allowed AMD to catch up

[Ravager911]

I will only buy this, if Intel names it Beerlake.

[Clanscorpia]

3 minutes ago, Ravager911 said:

I will only buy this, if Intel names it Beerlake.

Figured lol

[Jito463]

12 minutes ago, Clanscorpia said:

I feel like if Intel really did push forward as everyone here wants them to have done AMD would have never caught up. Just think about that a little bit. I think in the end, Intel being complacent actually helped the market because it allowed AMD to catch up

I understand what you're saying - and I'm glad that AMD did catch up - but it still doesn't excuse Intel stagnating technology, and charging inflated prices for minor improvements.

[Bleedingyamato]

24 minutes ago, Jito463 said:

If they're far enough along in development, they probably can't make significant changes to the design.  They had to have started development on these long before Ryzen actually came out (they would have known about it from AMDs roadmap, but not known the performance level).

Even without major design changes there should be a performance increase anyway.  Can you imagine the uproar if Intel released a second new CPU generation with no improvement?

 

Correct me if I'm wrong but there was no improvement with Kaby over Skylake was there?

[Drak3]

4 minutes ago, Bleedingyamato said:

Correct me if I'm wrong but there was no improvement with Kaby over Skylake was there?

Minor overclocking improvements, better speedshift, and a hardware decoder capable of 4k

[Clanscorpia]

13 minutes ago, Jito463 said:

I understand what you're saying - and I'm glad that AMD did catch up - but it still doesn't excuse Intel stagnating technology, and charging inflated prices for minor improvements.

The inflated prices were the worst outcome of it, but as other people said, improving significantly was a bad business decision.

[Bleedingyamato]

56 minutes ago, Drak3 said:

Minor overclocking improvements, better speedshift, and a hardware decoder capable of 4k

Ok.  

 

I'm curious if coffee lake will be any better.  

[Jito463]

4 minutes ago, Bleedingyamato said:

I'm curious if coffee lake will be any better.  

I'm of the mind they may actually need to slightly drop clock speeds, to handle increased heat from the 2 additional cores.  At best, it'll be the same clock speeds as KL.  I don't see a clock speed increase happening, and given there was really no IPC gain to speak of from SL to KL, I don't foresee that happening from KL to CL.

[That_PC_Kid]

2 hours ago, Jito463 said:

Having only recently upgraded from the FX-8370 to Ryzen, the FX served my needs quite well, considering.  I was able to game on it without any issues.  It was performing pretty much between the i3 and i5 (based on 4th gen), and that was without doing any overclocking.  It may not have been a great architecture, but it wasn't horrible, either.

It was a horrible architecture. How do you manage to put 8 cores on a CPU and have the performance of a low-end quad core and a dual core?

[Jito463]

Just now, That_PC_Kid said:

How do you manage to put 8 cores on a CPU and have the performance of a low-end quad core and a dual core?

Because - while it has 8 physical cores - each pair of cores shared an FPU (along with some other features shared).  Also, many games were optimized for Intel's way of doing things, and there wasn't enough market share for AMD to justify most developers going in and optimizing for the Bulldozer configuration.  The extra cores actually did help when multi-threading, but it was behind considerably in single-threaded applications (as most games are, and is only recently beginning to change).  When all cores were being utilized, then BD was actually pretty decent.

[That_PC_Kid]

8 minutes ago, Bleedingyamato said:

I'm curious if coffee lake will be any better.  

If it does end up having 6C/6T on the i5 and 6C/12T on the i7 they will be way better. The multi-threaded performance will be way better than Ryzen's. At that point Ryzen will just become the budget CPU to get.

[Bleedingyamato]

17 minutes ago, Jito463 said:

I'm of the mind they may actually need to slightly drop clock speeds, to handle increased heat from the 2 additional cores.  At best, it'll be the same clock speeds as KL.  I don't see a clock speed increase happening, and given there was really no IPC gain to speak of from SL to KL, I don't foresee that happening from KL to CL.

You don't think in 2 years they'd increase IPC at all?    That's a long time to stand still.  

 

I was was hoping to upgrade to 8th gen and stay with that for at least 4 years unless I'd need to upgrade sooner to play newer games.

 

But if there really would be no performance increase I might have to wait.

 

Which would likely be fine since I'm using a 6700K which should be plenty good for years to come.  

 

12 minutes ago, That_PC_Kid said:

If it does end up having 6C/6T on the i5 and 6C/12T on the i7 they will be way better. The multi-threaded performance will be way better than Ryzen's. At that point Ryzen will just become the budget CPU to get.

Not everything can take full advantage of multiple threads though right?

 

I just mean that single threaded performance and general IPC gains are still important I assume?

[Jito463]

8 minutes ago, Bleedingyamato said:

You don't think in 2 years they'd increase IPC at all

I don't think it would be very significant (just a few percentage points), considering everything from SB until now has had very minimal improvements in IPC.  Most of the gains have come from clock speed increases.

[Drak3]

19 minutes ago, Bleedingyamato said:

You don't think in 2 years they'd increase IPC at all?    That's a long time to stand still.

Outside of an instruction set or two, there isn't much ground for improvement.

[That_PC_Kid]

25 minutes ago, Bleedingyamato said:

Not everything can take full advantage of multiple threads though right?

 

I just mean that single threaded performance and general IPC gains are still important I assume?

Single threaded performance is still very important. Intel CPUs have single threaded performance due to the great IPC. That great IPC mixed in with a higher core count would destroy AMD. The extra cores will be good for YouTubers, anyone who does editing, encryption, decryption, streaming, multi-tasking, encoding, programming, and many more. 

[MageTank]

4 hours ago, Suika said:

I wouldn't say I was expecting an IPC boost from Coffee Lake, rather the existing increase to IPC that was seen moving to Broadwell and then to Skylake. That combined with higher frequency chips would tempt me to upgrade, so if what you're saying is true, I may jump to Z370 then.

Most of the IPC gain we saw from Broadwell to Skylake, came from the memory subsystem swapping to DDR4. Most of the tests that showed a tangible gain in IPC, were tests that coincidentally scaled with memory bandwidth.

 

Now, if CoffeeLake brings Skylake-X's mesh cache structure, then I can totally see a difference in IPC coming, but the fact that the L3 cache is still that large, tells me it's likely not going to include that cache restructure.

 

I am personally expecting zero IPC gains, and zero clock speed gains (from overclocking, stock turbo may be higher). We are getting more cores after all, so to expect higher overclocking on top of having more cores, is a bit too optimistic. I could be wrong though, given how massive the 8700k's stock turbo boost is. I suppose we will find out soon enough. 

[Bleedingyamato]

17 minutes ago, Jito463 said:

I don't think it would be very significant (just a few percentage points), considering everything from SB until now has had very minimal improvements in IPC.  Most of the gains have come from clock speed increases.

I thought prior to Skylake there'd been better IPC increases generation to generation?

 

10 minutes ago, Drak3 said:

Outside of an instruction set or two, there isn't much ground for improvement.

So Intel has no potential for IPC increase without some sort of large scale change to their current CPUs or to create a new design that would allow for better year to year performance increases?

 

2 minutes ago, That_PC_Kid said:

Single threaded performance is still very important. Intel CPUs have single threaded performance due to the great IPC. That great IPC mixed in with a higher core count would destroy AMD. The extra cores will be good for YouTubers, anyone who does editing, encryption, decryption, streaming, multi-tasking, encoding, programming, and many more. 

I pretty much just use my computer for iTunes and games.  ?  

[Drak3]

1 minute ago, Bleedingyamato said:

I thought prior to Skylake there'd been better IPC increases generation to generation?

From Sandy to Kaby, around 15% total improvement I believe.

 

2 minutes ago, Bleedingyamato said:

So Intel has no potential for IPC increase without some sort of large scale change to their current CPUs or to create a new design that would allow for better year to year performance increases?

IPC doesn't have much more potential, unless newer instruction sets aren't quite up to snuff. What Intel has been focusing on is efficiency and clock speed, mostly on mobile.

[MageTank]

1 minute ago, Bleedingyamato said:

I thought prior to Skylake there'd been better IPC increases generation to generation?

 

So Intel has no potential for IPC increase without some sort of large scale change to their current CPUs or to create a new design that would allow for better year to year performance increases?

 

I pretty much just use my computer for iTunes and games.  ?  

Most of Intel's major IPC yields, came from the die shrinks associated with those architectures at the time. We have not had a die shrink since going from Haswell to Broadwell (22nm down to 14nm). Even with that shrink, we only saw a 3-5% difference in IPC (depending on instruction set used). Broadwell to Skylake only yielded another 2-3% (again, depending on instruction set) being 14nm-14nm, but the memory subsystem changed from DDR3 to DDR4. Instruction sets that benefit heavily from memory (take AVX for example) were bound to show an improvement up until the point in which the cores themselves are the bottleneck.

 

Now, Skylake to Kaby yielded zero IPC gains. Fundamentally the same architectures, with some slight changes as @Drak3 pointed out previously. Coffeelake is still considered an "Optimization" in Intel's "Process, Architecture, Optimization" pattern. Kaby was also considered an "optimization", while Skylake was considered an "architecture". Broadwell was considered a "process", if that helps any.

 

We've hit diminishing returns when it comes to gaining IPC. Part of this is the fact that die shrinks are getting harder and harder to achieve (and silicon is not going to be applicable for 7nm and below), while the rest of it is due to the shift in computing in general. We are starting to favor more cores now, rather than having fewer, faster cores. The consumer has evolved to demand even more multi-tasking, and so the product must evolve to deliver that need. As you throw in more cores, you have to compromise on something. In this case, that something is clock speeds. 

[Bleedingyamato]

8 minutes ago, Drak3 said:

From Sandy to Kaby, around 15% total improvement I believe.

 

IPC doesn't have much more potential, unless newer instruction sets aren't quite up to snuff. What Intel has been focusing on is efficiency and clock speed, mostly on mobile.

Was Sandy 2nd gen or 3rd generation?

 

That's not much improvement over that many generations.  

 

1 minute ago, MageTank said:

Most of Intel's major IPC yields, came from the die shrinks associated with those architectures at the time. We have not had a die shrink since going from Haswell to Broadwell (22nm down to 14nm). Even with that shrink, we only saw a 3-5% difference in IPC (depending on instruction set used). Broadwell to Skylake only yielded another 2-3% (again, depending on instruction set) being 14nm-14nm, but the memory subsystem changed from DDR3 to DDR4. Instruction sets that benefit heavily from memory (take AVX for example) were bound to show an improvement up until the point in which the cores themselves are the bottleneck.

 

Now, Skylake to Kaby yielded zero IPC gains. Fundamentally the same architectures, with some slight changes as @Drak3 pointed out previously. Coffeelake is still considered an "Optimization" in Intel's "Process, Architecture, Optimization" pattern. Kaby was also considered an "optimization", while Skylake was considered an "architecture". Broadwell was considered a "process", if that helps any.

 

We've hit diminishing returns when it comes to gaining IPC. Part of this is the fact that die shrinks are getting harder and harder to achieve (and silicon is not going to be applicable for 7nm and below), while the rest of it is due to the shift in computing in general. We are starting to favor more cores now, rather than having fewer, faster cores. The consumer has evolved to demand even more multi-tasking, and so the product must evolve to deliver that need. As you throw in more cores, you have to compromise on something. In this case, that something is clock speeds. 

Intel is already slipping from their changed improvement pattern?    Two optimization generations is not good.  

 

Basibally they hit a brick wall for improvement and haven't figured out how to get around it yet?

[That_PC_Kid]

6 minutes ago, MageTank said:

We are starting to favor more cores now, rather than having fewer, faster cores. The consumer has evolved to demand even more multi-tasking, and so the product must evolve to deliver that need. As you throw in more cores, you have to compromise on something. In this case, that something is clock speeds. 

Yeah, this is very true. Especially the compromise of having more cores and less base clock frequency. The only reason higher core count CPUs have low clock frequency is because of power efficiency.

[MageTank]

1 minute ago, Bleedingyamato said:

Was Sandy 2nd gen or 3rd generation?

 

That's not much improvement over that many generations.  

 

Intel is already slipping from their changed improvement pattern?    Two optimization generations is not good.  

 

Basibally they hit a brick wall for improvement and haven't figured out how to get around it yet?

The issue is, 10nm was not ready yet for Cannonlake. In fact, it's been delayed repeatedly. Rather than sit and deliver nothing to the consumers, they decided to deliver a higher core count on the consumer platform. Now, you can blame Ryzen for this if you'd like, or you can see it as the market trend in general changing to kick off the "core race 2.0". 

 

If you would like more reading material on Intel's release patterns, Wikipedia has a pretty accurate page on the subject: https://en.wikipedia.org/wiki/Tick-tock_model

2 minutes ago, That_PC_Kid said:

Yeah, this is very true. Especially the compromise of having more cores and less base clock frequency. The only reason higher core count CPUs have low clock frequency is because of power efficiency.

The biggest problem with the consumer, is that they don't understand compromise. They want it all. More cores, while also being faster than fewer cores. Since we cannot defy the laws of physics that easily, it's unlikely they will get both. I am expecting 5ghz on these 6 core SKU's (after delidding). That's about 100-200mhz lower than the 4c/8t designs, and is a pretty decent compromise assuming any efficiency yields allowed for lower volts to handle the surplus in thermals. 

 

I could be wrong, and 5ghz might be out of the question, but I certainly don't expect 5.2ghz while also having 50% more cores.