Is anyone waiting for Coffee Lake?

[DarkSmith2]

1 minute ago, MageTank said:

What? I didn't quote a post at all when I said they had identical IPC. I think you may have mistaken me for someone else. 

i quoted wrong, sorry

[MageTank]

1 minute ago, JDE said:

base clock

When was the last time base clock really mattered though? For locked chips, they do, but we are talking unlocked chips with platforms that completely ignore them. 

Just now, RadiatingLight said:

Rumored baseclock is 3.7Ghz, no? and boost at 4.3Ghz.

that's lower clocks than the 7700K, so IPC needs to go up for intel to remain the same in single threaded

It's 3.8. I can confirm that much.

 

Also, no, IPC doesn't have to change at all. You people are assuming Intel is locking the CPU's at their base clock when running these tests. They never do. That is how Kaby was technically "faster" than Skylake. Boost table vs boost table, coffeelake has superior clocks all around. 

[DarkSmith2]

Ehrm i dont think there will be an IPC increase to Kabylake either.

[RadiatingLight]

7 minutes ago, MageTank said:

When was the last time base clock really mattered though? For locked chips, they do, but we are talking unlocked chips with platforms that completely ignore them. 

It's 3.8. I can confirm that much.

 

Also, no, IPC doesn't have to change at all. You people are assuming Intel is locking the CPU's at their base clock when running these tests. They never do. That is how Kaby was technically "faster" than Skylake. Boost table vs boost table, coffeelake has superior clocks all around. 

last I checked, 4.5ghz was a bigger number than 4.3ghz

how are coffee lake clocks faster? in what scenario?

[MageTank]

8 minutes ago, DarkSmith2 said:

Ehrm i dont think there will be an IPC increase to Kabylake either.

It's a "process refinement", similar to how Devil's Canyon was relative to Haswell. In fact, Kaby is that exact same thing in regards to Skylake, so Intel has done it twice in a row now. Good news is, Cannonlake will work on the 300 series platform alongside CFL, so we at least have that going for us x.x

2 minutes ago, RadiatingLight said:

last I checked, 4.5ghz was a bigger number than 4.3ghz

how are coffee lake clocks faster? in what scenario?

4.5 is bigger than 4.3, but you are ignoring the 4.7ghz. The 7700k's all-core turbo is 4.4ghz. It's only 4.5ghz if you enable advanced turbo on your motherboard. The 8700k's all core turbo is 4.3ghz, but 4.7ghz if you enable advanced turbo. 4.7 is a bigger number than 4.5. 

[DarkSmith2]

Just now, MageTank said:

It's a "process refinement", similar to how Devil's Canyon was relative to Haswell. In fact, Kaby is that exact same thing in regards to Skylake, so Intel has done it twice in a row now. Good news is, Cannonlake will work on the 300 series platform alongside CFL, so we at least have that going for us x.x

2 minutes ago, RadiatingLight said:

whaaat? wait. Will Cannonlake really work on 300series boards? How can you know?

[MageTank]

Just now, DarkSmith2 said:

whaaat? wait. Will Cannonlake really work on 300series boards? How can you know?

This is the part where I have to do the terrible "anonymous source" shtick. It will work on the 300 series platform, but it will also be releasing with a 370 series platform with features exclusive to CNL. No idea what those features are yet, but that's all I've seen thus far. 

[DarkSmith2]

2 minutes ago, MageTank said:

This is the part where I have to do the terrible "anonymous source" shtick. It will work on the 300 series platform, but it will also be releasing with a 370 series platform with features exclusive to CNL. No idea what those features are yet, but that's all I've seen thus far. 

Well then it might be something similar to Sandy/Ivy Bridge. Sandy Bridge doesnt support PCI-E 3.0 f.e. But Ivy Bridge does and Gen3 Boards where made... Maybe we are closer to PCI-E 4.0 then we think

[MageTank]

Just now, DarkSmith2 said:

Well then it might be something similar to Sandy/Ivy Bridge. Sandy Bridge doesnt support PCI-E 3.0 f.e. But Ivy Bridge does and Gen3 Boards where made... Maybe we are closer to PCI-E 4.0 then we think

Well, I know native USB 3.1 (gen2) is confirmed on the CNL chipset, along with integrated wifi, but aside from that, they've been very quiet on actual chipset features. I did hear AVX3(512) may come to the consumer platform, but that part I highly doubt. They crippled it on X299's lower end CPU's, so I certainly don't expect them to offer it on CNL, but I could be wrong. 

[Nationalist]

21 minutes ago, MageTank said:

It's 3.8. I can confirm that much.

I guess either you are wrong or several news sites are wrong.

[MageTank]

Just now, Nationalist said:

I guess either you are wrong or several news sites are wrong.

What sites are reporting 3.7?

[DarkSmith2]

5 minutes ago, MageTank said:

What sites are reporting 3.7?

yea some sites are reporting 

 

8700 (non k) 3,7 base / 8700k 3,8 base

http://wccftech.com/intel-coffee-lake-8th-gen-core-i7-8700k-performance-specs-leak/

 

and some are reporting

 

8700 (non k) 3,2 base / 8700k 3,7 base

http://www.pcgameshardware.de/CPU-Hardware-154106/News/Line-Up-und-Performance-Vergleich-geleakt-1236145/

[MageTank]

Just now, DarkSmith2 said:

yea some sites are reporting 

 

8700 (non k) 3,7 base / 8700k 3,8 base

 

and some are reporting

 

8700 (non k) 3,2 base / 8700k 3,7 base

That's odd. Both of my people said 3.8, as does Intel's "leaked" slide:

 

Kinda odd where the 3.7 number came from. Wonder if Intel is changing the stepping without telling anyone? It's also possible early leaks used an ES with a lower base, but from everything I heard, the number is 3.8. We will see soon though. 

[DarkSmith2]

2 minutes ago, MageTank said:

Kinda odd where the 3.7 number came from. Wonder if Intel is changing the stepping without telling anyone? It's also possible early leaks used an ES with a lower base, but from everything I heard, the number is 3.8. We will see soon though. 

yea we can wait for the announcement, ill be watching it live

[MyName13]

Everyone who can afford a more powerful CPU than r3 but can't afford r5 1600, i5 8400 will be amazing for 180$, there were some rumors that low end chipset B360 Will come out next year so I guess this means that only unlocked CPUs will come out this year.

[TahoeDust]

If the 8700k's 6 cores clock like a 7700k, has the memory latency of a 7700k, has the IMC performs of my 7820x, and the cost falls inline with previous main stream i7s, it is going to be a nasty little chip for the money.

[RadiatingLight]

1 minute ago, TahoeDust said:

If the 8700k's 6 cores clock like a 7700k, has the memory latency of a 7700k, has the IMC performs of my 7820x, and the cost falls inline with previous main stream i7s, it is going to be a nasty little chip for the money.

It's going to be a crazy good chip. especially if it can OC to higher than the 7700K, which is what @MageTank is claiming.

[MageTank]

Just now, RadiatingLight said:

It's going to be a crazy good chip. especially if it can OC to higher than the 7700K, which is what @MageTank is claiming.

I will say this. 5.2ghz is not out of the question for people with custom loops

[RadiatingLight]

1 minute ago, MageTank said:

I will say this. 5.2ghz is not out of the question for people with custom loops

"Lets disregard IPC and push clocks higher and higher"

- Some CPU company... not too long ago.

[Klovnious]

I have a 4790K and I'm looking forward for CL. But I think I might wait for IceLake (10nm right?) before I build a new system.

[RadiatingLight]

1 minute ago, MageTank said:

I will say this. 5.2ghz is not out of the question for people with custom loops

 

Just now, RadiatingLight said:

"Lets disregard IPC and push clocks higher and higher"

- Some CPU company... not too long ago.

but that is an awesome clockspeed. it'll be a beast CPU.

[MageTank]

Just now, RadiatingLight said:

"Lets disregard IPC and push clocks higher and higher"

- Some CPU company... not too long ago.

I don't think they are disregarding IPC. You have to understand, it's getting harder and harder to pull it out of thin air. They have been on 14nm for quite a while now. Sure, we are getting refined versions of it, but it's still just optimizations, not exactly new die shrinks or entirely new architecture. Cannonlake will bring an IPC boost, that is when you can expect a change from the status quo. How much it will bring, I simply do not know, but it's going to be the new process in their "process, architecture, optimization" cycle. We tend to see IPC boosts on both process and architecture changes. 

 

It should also be noted that these chips might cool easier as well. Having a slightly larger die, you have more surface area of the IHS making contact with the die, resulting in less cooling efficiency being wasted cooling parts of the IHS not making contact with the die. The exact same thing can be said of the 12c+ X299 chips. They have a larger die than the 10c and below SKU's, and actually run cooler as a result.

 

The 8700k will certainly be interesting, I just don't see it as a compelling upgrade for current Skylake/Kabylake i7 users that don't need the additional cores. 

[DarkSmith2]

Well i dont even expect huge gains in IPC from Cannonlake, i expect less powerdraw and if, it will probably come with lower clock speeds in hand. I probably will buy the more mature 14nm and Coffeelake now then Cannonlake in a year. We already know they have trouble producing in 10nm thats why CFL will excist...

[MageTank]

2 minutes ago, DarkSmith2 said:

Well i dont even expect huge gains in IPC from Cannonlake, i expect less powerdraw and if, it will probably come with lower clock speeds in hand. I probably will buy the more mature 14nm and Coffeelake now then Cannonlake in a year. We already know they have trouble producing in 10nm thats why CFL will excist...

Same. I'd say 2-3% at most, similar to Broadwell over Haswell. If we keep getting power efficiency optimizations and better pitch density for improved overclocking, I'll certainly be happy. IPC is not the be all, end all for CPU performance. We barely use half the efficient instruction sets that are given to us. 

[MyName13]

New process doesn't increase IPC, it's the architecture optimisation that changes IPC, new process can increase clock rate (or instructions per second, not per clock).