Intel Skylake / Kaby Lake

[eddman]

Why ?? this is a 95W TDP only cpu. Why everyone expecting to have faster than Core i7 7800X (140W TDP) Cinebench MT score ??

... because HEDT and non-HEDT TDPs are not comparable when it comes to comparing performance; the 140W 7800X is less than 10% faster than a 95W 1600X, and that the 112W 7740X is about 3% faster than the 91W 7700K. Also see 7640X vs. 7600K.

 

[AtenRa]

... because HEDT and non-HEDT TDPs are not directly comparable;

I would really like you to provide a technical paper (link) for that because this is the first time i read this.

 

[eddman]

I would really like you to provide a technical paper (link) for that because this is the first time i read this.

I edited my comment to make it clear. I don't need a technical paper when that benchmark slide shows it already.

 

[AtenRa]

I edited my comment to make it clear. I don't need a technical paper when that benchmark slide shows it already.

SKL/KBL and CFL are all the same generation, you can expect that a 112W TDP CPU to be faster than a 91W TDP.
So when we have a 95W TDP very close to a 140W TDP then that is excellent. If the 8700K is at $350 and can OC close to 7700K then it should be just fine.

 

[tamz_msc]

Everyone paying attention to the previous leaks will notice this score is bogus. Earlier leak showed 1.410 pts @ CB R15 MT. Retail systems using fast RAM could do even better.

Why is this bogus? ST score is comparable to your "legit" leak. Could have been any number of background applications affecting MT score. Memory speed doesn't affect a benchmark like CB that much on Intel chips except Skylake-X, so ~1400 cb MT is about what can be expected.

 

[eddman]

SKL/KBL and CFL are all the same generation, you can expect that a 112W TDP CPU to be faster than a 91W TDP.
So when we have a 95W TDP very close to a 140W TDP then that is excellent. If the 8700K is at $350 and can OC close to 7700K then it should be just fine.

7740X's 3% better performance comes from its 100 MHz higher clock, not its TDP. By your logic 7700K with its 19% lower TDP should have no chance of coming close to 7740X's performance, but it does.

The 8700K will perform better than 7800X, with sufficient cooling, for the simple reason that its base and all-core boost clocks are 200 and 300 MHz higher, respectively.

TDP is NOT a performance indicator.

 

[AtenRa]

The 8700K will perform better than 7800X, with sufficient cooling, for the simple reason that its base and all-core boost clocks are 200 and 300 MHz higher respectively.

IF 8700K can keep those higher turbo clocks the same time as 7800X turbo then it will be faster, but at 95W TDP why would it be able to do that ??

 

[PeterScott]

The MT performance seems way off, as if HT is disabled.

Likely an artifact of turbo cores reducing clock speed as all the cores come online.

Once you OC, to get set core speed, you should probably see the typical intel 20- 25% boost for HT on this bench.

Though this will still likely be the best showing for AMD.

AMDs SMT seems to give up to 30-35% boost in Cinebench, where Intel HT does 20-25%.

But SMT/HT are highly variable in the boost they give on different workloads. Cinebench is pretty much a synthetic benchmark from my perpsective. Video rendering is the only heavy multithreaded thing I do, so those are the benchmarks I really want to see.

 

[LTC8K6]

IF 8700K can keep those higher turbo clocks the same time as 7800X turbo then it will be faster, but at 95W TDP why would it be able to do that ??

14nm++ could be why.

 

[coercitiv]

14nm++ could be why.

It's one thing to recognize 14nm++ as more than a marketing gimmick, and another to expect optimization gains to be equal or better than a node jump.

Do you honestly expect CFL cores to use 30% less power than KBL cores at 4.3Ghz?

 

[R0H1T]

It's one thing to recognize 14nm++ as more than a marketing gimmick, and another to expect optimization gains to be equal or better than a node jump.

Do you honestly expect CFL cores to use 30% less power than KBL cores at 4.3Ghz?

Is that a rhetorical question?

 

[coercitiv]

Is that a rhetorical question?

I don't know if it is, people show genuine hope that 8700K will maintain those 6C turbo clocks under heavy loads. I'm merely translating hope into numbers.

 

[R0H1T]

I don't know if it is, people show genuine hope that 8700K will maintain those 6C turbo clocks under heavy loads. I'm merely translating hope into numbers.

I think it's just that, anyone realistically expecting a node (or half node) jump's worth of improvements on the same process is not grounded in reality.

 

[AtenRa]

14nm++ could be why.

7800X is on 14nm+ , the difference alone over 14nm++ is not that much.

 

[LTC8K6]

Well, I have always questioned the reported 8700K clocks.

 

[coercitiv]

I think it's just that, anyone realistically expecting a node (or half node) jump's worth of improvements on the same process is not grounded in reality.

Actually, I think that 14nm to 14nm++ will be comparable in terms of power gains with 22nm to 14nm, at least in the high frequency range.

Well, I have always questioned the reported 8700K clocks.

They may as well be true, just that we may see those clocks in games and other light-medium loads, but I expect them to drop in heavy loads such as CB. And that's mighty fine by me, Intel could have done this with KBL too, free performance for most consumers.

 

[LTC8K6]

What clock speed does the 7700K maintain during Cinebench?

 

[LTC8K6]

If the 7700K maintains 4.4 on all cores during benchmarking, then why is it a stretch to think that the 8700K could maintain 4.2 or 4.3?

Right away, we see the benefits of Intel’s improved 14nm process. Our Core i7-6700K ran at a consistent 4GHz in all our benchmark tests, while our Core i7-7700K preferred a 4.4GHz clock. Intel is pushing a 10% higher clock rate in Prime95, while power consumption for the 7700K is actually down slightly from the 6700K. But — and this is critical — it’s only down slightly.

https://www.extremetech.com/extreme/241950-intels-core-i7-7700k-reviewed-kaby-lake-debuts-desktop

 

[AtenRa]

If the 7700K maintains 4.4 on all cores during benchmarking, then why is it a stretch to think that the 8700K could maintain 4.2 or 4.3?


https://www.extremetech.com/extreme/241950-intels-core-i7-7700k-reviewed-kaby-lake-debuts-desktop

That is fine, both 6700K and 7700K are on the same 91W TDP, 6700K is on 14nm and 7700K is on 14nm+.

Im expecting the same from 14nm++, a 200-300MHz increase of Clocks but at the same TDP, not higher clocks AND lower TDP.

 

[tamz_msc]

If the 7700K maintains 4.4 on all cores during benchmarking, then why is it a stretch to think that the 8700K could maintain 4.2 or 4.3?


https://www.extremetech.com/extreme/241950-intels-core-i7-7700k-reviewed-kaby-lake-debuts-desktop

I think that the turbos are working as intended, the ST turbo on the leaked 196 ST score and 1410 score on MT are both accurate enough for a stock configuration.

 

[LTC8K6]

That is fine, both 6700K and 7700K are on the same 91W TDP, 6700K is on 14nm and 7700K is on 14nm+.

Im expecting the same from 14nm++, a 200-300MHz increase of Clocks but at the same TDP, not higher clocks AND lower TDP.

We don't have a lower TDP, though. We have a 95W TDP for the 8700K along with an improved process and probably a slightly bigger die. We also have a lower all core turbo clock. It all seems to fit in nicely to me.

 

[dullard]

Do you honestly expect CFL cores to use 30% less power than KBL cores at 4.3Ghz?

We only have one set of data from Intel, at 0.7 V and a "high performing device":
https://en.wikichip.org/wiki/File:intel_14nm++.png

At the normalized performance of 1.0 on that graph, 14nm++ uses 27% less power than 14nm+. (you'll have to do the math as the graph numbers compare 14 nm++ to 14 nm).

14nm+ uses 34% less power than 14 nm at that location, so the combined reduction in power is 1-(1-0.27)*(1-0.34) = 52%.

So, according to Intel, we should expect CFL cores to use roughly 27% less power than KBL. Of course, it will vary with voltage, use, optimization, and with device class. But that 27% number isn't too far off of the 30% number being floated around in this thread.

 

[AtenRa]

We don't have a lower TDP, though. We have a 95W TDP for the 8700K along with an improved process and probably a slightly bigger die. We also have a lower all core turbo clock. It all seems to fit in nicely to me.

The lower TDP was vs the 7800X 6C 12T 14nm+ CPU at 140W TDP.

And when comparing the 8700K to 7700K also remember you have anothe 2 Cores or 50% more threads. So do you believe that 14nm++ will be able to have almost the same all core turbo as the 7700K with 50% more cores at a slightly higher TDP of 95W vs 91W ??

 

[AtenRa]

We only have one set of data from Intel, at 0.7 V and a "high performing device":
https://en.wikichip.org/wiki/File:intel_14nm++.png

At the normalized performance of 1.0 on that graph, 14nm++ uses 27% less power than 14nm+. (you'll have to do the math as the graph numbers compare 14 nm++ to 14 nm).

14nm+ uses 34% less power than 14 nm at that location, so the combined reduction in power is 1-(1-0.27)*(1-0.34) = 52%.

So, according to Intel, we should expect CFL cores to use roughly 27% less power than KBL. Of course, it will vary with voltage, use, optimization, and with device class. But that 27% number isn't too far off of the 30% number being floated around in this thread.

In the same graph, the power reduction (at same performance point 1.0 in X axis) from 14nm (Core i7 6700K 4C 8T) to 14nm+ (Core i7 7700K 4C 8T) is higher than 14nm+ (Core i7 7700K 4C 8T) to 14nm++ (Core i7 8700K 6T 12T). So if Core i7 7700K with the same number of cores vs 6700K and same TDP only got 300MHz higher turbo, then its illogical to expect same turbo as 7700K with 50% more Cores/Threads from 14nm++.

 

[LTC8K6]

In the same graph, the power reduction (at same performance point 1.0 in X axis) from 14nm (Core i7 6700K 4C 8T) to 14nm+ (Core i7 7700K 4C 8T) is higher than 14nm+ (Core i7 7700K 4C 8T) to 14nm++ (Core i7 8700K 6T 12T). So if Core i7 7700K with the same number of cores vs 6700K and same TDP only got 300MHz higher turbo, then its illogical to expect same turbo as 7700K with 50% more Cores/Threads from 14nm++.

I can't recall any claim of the same turbo, though.

We have a higher single core turbo for the 8700K and a lower all core turbo for the 8700K.