Intel Skylake / Kaby Lake

[RichUK]

So in summary, Intel > AMD?

 

[MarkPost]

meh all this e-penis brands nonsense is soooo boring

 

[AdamK47]

I tend to consider someones point invalidated when they bring up an argument centering around LN2 clock speeds and benchmark results. People love the ohhh and ahhh factor of big numbers. Numbers far removed from typical everyday usage.

 

[ddogg]

Yeah I agree, LN2 scores are absolutely pointless and so is that 7960x cinebench score. Would like to see real scores when the CPU is released to compare.

 

[Java Sunday]

It's amazing how quickly this thread can derail! Regarding coffee lake, I am wondering what if anything would cause a gaming performance regression from kaby lake? From what has been seen, it seems like the 8700k >=7700k in every aspect. The only things I could imagine would be thermal throttling resulting from having +50% cores, or some fatal design flaw.

I'm very excited for coffee lake as I will be upgrading from an i7 920 (closing in on 10 years of service!)

 

[french toast]

Coffeelake is what kaby should have been.
I can't see it being slower in any scenario.
The gaming CPU to get.

 

[moonbogg]

@TahoeDust
Incase you were curious about the 6800K as well in this test with my 1080Ti.

 

[Charlie22911]

@TahoeDust
Incase you were curious about the 6800K as well in this test with my 1080Ti.

What clocks? GPU, CPU (4.3 I assume?), ring/uncore, ram? That's a nice bump :thumbsup:

 

[PeterScott]

It's amazing how quickly this thread can derail! Regarding coffee lake, I am wondering what if anything would cause a gaming performance regression from kaby lake? From what has been seen, it seems like the 8700k >=7700k in every aspect. The only things I could imagine would be thermal throttling resulting from having +50% cores, or some fatal design flaw.

I wouldn't expect a fatal flaw, but every significant change brings a chance that there will be minor regression in some benchmarks, because of cache is organized slightly different, or just because the previous CPU was so popular it was the target of direct optimization and any small change might no longer be optimal for those optimizations, maybe a bit lower clock speed.

But I am thinking/hoping CL will be great, a 6 core that is provides great IPC and clock speed.

 

[moonbogg]

I wouldn't expect a fatal flaw, but every significant change brings a chance that there will be minor regression in some benchmarks, because of cache is organized slightly different, or just because the previous CPU was so popular it was the target of direct optimization and any small change might no longer be optimal for those optimizations, maybe a bit lower clock speed.

But I am thinking/hoping CL will be great, a 6 core that is provides great IPC and clock speed.

It should be good, but they should have released it sooner when 6 core chips used to be a big deal. Now they are basically entry level chips and nothing to get excited over. Less than a year ago, 6 core was the most horsepower you could get without spending $1000, but now you can get one for $200. Intel held back way too much and kept 6 core mainstream chips out of the sunlight, and now its not going to be a big deal. Its just going to be the fastest gaming chip again (maybe), but still cost $400, lol.
I expect Intel to simply price a lot of people out of Coffee lake. I expect most gamers to simply buy a quad core i3 from now on, and why not? Who the hell needs to blow $400 on a gaming CPU from intel when a 6/12 solution can be had for, yet again, half the price or simply get a quad i3 for $125 or whatever?

 

[Markfw]

@TahoeDust
Incase you were curious about the 6800K as well in this test with my 1080Ti.

Is that benchmark one anybody can run by downloading a file for free ? If so, please link me., maybe I can try it.

 

[PeterScott]

It should be good, but they should have released it sooner when 6 core chips used to be a big deal. Now they are basically entry level chips and nothing to get excited over.

The would have had more cores sooner, if they did't have issues with 10nm not panning out as planned.

 

[moonbogg]

Is that benchmark one anybody can run by downloading a file for free ? If so, please link me., maybe I can try it.

This is a free benchmark as far as I recall. Linky below

https://benchmark.unigine.com/superposition

 

[TahoeDust]

@TahoeDust
Incase you were curious about the 6800K as well in this test with my 1080Ti.

Nice! What were your clocks? CPU & GPU?

 

[LTC8K6]

It should be good, but they should have released it sooner when 6 core chips used to be a big deal. Now they are basically entry level chips and nothing to get excited over. Less than a year ago, 6 core was the most horsepower you could get without spending $1000, but now you can get one for $200. Intel held back way too much and kept 6 core mainstream chips out of the sunlight, and now its not going to be a big deal. Its just going to be the fastest gaming chip again (maybe), but still cost $400, lol.
I expect Intel to simply price a lot of people out of Coffee lake. I expect most gamers to simply buy a quad core i3 from now on, and why not? Who the hell needs to blow $400 on a gaming CPU from intel when a 6/12 solution can be had for, yet again, half the price or simply get a quad i3 for $125 or whatever?

I think the 8700K should be impressive and sell very well.
I think the memory clocks that it will be able to run with will be also be impressive.
The 7740K actually impresses me now that it's out in the real world. Nice high clocks and really good memory numbers.

 

[moonbogg]

The would have had more cores sooner, if they did't have issues with 10nm not panning out as planned.

I doubt they would have. They were absolutely LOVING the quad core milk strategy. And charging $700 for a 6 core chip with extra PCI lanes? Holy Godchild, who wouldn't love that!? No my son, the reason we haven't seen 6/12 on mainstream is because AMD had to do it first and Intel is Intel.

@TahoeDust

CPU was 4.2 or 4.3 (can't remember) and GPU was around 2130. Uncore stock, ram 3200 CAS16.

 

[coffeeblues]

If the 8700K does not best the 7700K in all areas, it will be a disappointment to me.

If I remember correctly we've already had leaks of higher sisoftsandra scores, it's those "other" areas we still don't have information about

It's reasonable to expect that very well-threaded applications will do very well with the extra cores right on release day - rendering, encoding, compiling, simulations, etc.

Now "all areas" is very universal and inclusive, but I think you'll start discovering corner cases once you start running them on 6-core kaby desdendant cpu. If you're willing to consider the best to worst range of scaling multipliers that we could encounter for an application, then there are things to take in to account. The production and release of the chip is already being rushed forward so I don't think everybody will have too much time to fully adopt the cpu in time for release day. Usually you have a higher core ceiling so you can discover how to better utilize the extra available capacity so you can scale better. If you had an application that on 7700k cpu was close to i/o (memory, latency, and what not) limits, then the addition of two cores would expose those limitations and the performance scaling would be less than optimal. An application could exhibit a very conservative 0-10% performance increase even if it seems undeserved given 50% higher core count.

Will 6 core ring bus latency be on par with 4 core ring bus latency? Will Intel clock 6 core ring bus higher to compensate for increased physical distance? Will we have non uniform core to core and to cache latencies in the upcoming 6 core layout whatever that may be?

6 core cpu will provide higher core ceiling but it's application's responsibility to take advantage of it and that's the part that I'm not taking for granted - I'll believe it when I see it and I'm looking forward to people discovering how their software works on new hardware.

Now I don't put much weight in the above considerations, as they have no influence on the future performance results of the chip that will be released, but what I think they indicate is severe lack of information and without it it's just a whole lot of hot air and an emotional hypetrain which amount to just about 0.

 

[Arachnotronic]

I doubt they would have. They were absolutely LOVING the quad core milk strategy. And charging $700 for a 6 core chip with extra PCI lanes? Holy Godchild, who wouldn't love that!? No my son, the reason we haven't seen 6/12 on mainstream is because AMD had to do it first and Intel is Intel.

@TahoeDust

CPU was 4.2 or 4.3 (can't remember) and GPU was around 2130. Uncore stock, ram 3200 CAS16.

It wasn't about "quad core milk strategy" at all.

These chips are designed first and foremost for notebooks, and it wasn't really feasible until 14nm++ to increase the core count while staying within a reasonable power envelope.

 

[TahoeDust]

@TahoeDust

CPU was 4.2 or 4.3 (can't remember) and GPU was around 2130. Uncore stock, ram 3200 CAS16.

Wow!...yeah, that will do it. The other ones in this thread were run @ 2025-2038. I wish my GPU would clock that high. Nice score!

 

[IndyColtsFan]

For me, as impressive as TR is, I think it is down to the 8700K vs 7820K vs 7900K. It has been a couple of months - can any 7900 owners comment on overclocking and the resultant temps and power usage? I've always favored the 7900 but temps scared me off initially.

 

[Zucker2k]

I doubt they would have. They were absolutely LOVING the quad core milk strategy. And charging $700 for a 6 core chip with extra PCI lanes? Holy Godchild, who wouldn't love that!? No my son, the reason we haven't seen 6/12 on mainstream is because AMD had to do it first and Intel is Intel.

@TahoeDust

CPU was 4.2 or 4.3 (can't remember) and GPU was around 2130. Uncore stock, ram 3200 CAS16.

Don't forget the quadcore Skylake/Kabylake + IGP were already at 91watts. 2 additional hyperthreaded cores would've needed about 125watts at same clocks. For a mainstream specced processor, a significant tdp change like that would've been a very unpleasant affair. Plus, we're talking about top-tier mainstream here. An i3 or Pentium is enough for the usage needs of the vast majority of users around the world, imho.

 

[raghu78]

It wasn't about "quad core milk strategy" at all.

These chips are designed first and foremost for notebooks, and it wasn't really feasible until 14nm++ to increase the core count while staying within a reasonable power envelope.

This is you defending Intel. Intel could have launched a Skylake 6C/12T core i7 if they wanted to a couple of years back. Intel's turbo clocking is very sophisticated and they needed to lower base clocks for 6C/12T to fit in the respective power envelopes for notebook and desktop. Since most desktop workloads rarely push all these cores the CPU will be turboing very high at all times. We know very well Intel did not do so because there was no competition from AMD and they were more bothered about maximizing gross margins than providing more cores to the mainstream consumer. AMD are going to bring 4C//8T Raven Ridge APUs with TDP as low as 15w. So why can't Intel provide 6C/12T at 35-45w TDP for notebooks inspite of having a superior process node. Just wait and see how Intel start to ramp core counts up at 10nm. Intel stayed at 4 cores for mainstream from the ancient 45nm to 32nm to 22nm and finally to 14nm node . Lynnfield came with 4 cores for mainstream in late 2009. Thats an eternity in the PC industry. The transition from 4 to 6 cores in mainstream took 8 years. Just watch how quickly Intel move to 8 cores.

 

[coffeeblues]

Any of you Ryzen or Threadripper boys reading want to get in on this?

More benchmarks means more data.

If somebody would volunteer to compile a table of this latest benchmark comparison run, that would be very convenient while trying to follow what's going on with all the scores.

 

[crashtech]

@raghu78 , @Arachnotronic , it's difficult to escape the impression that bringing hexacores to the mainstream is Intel's answer to a renewed competitive atmosphere, and something that was technically feasible before now.

Though, there might be a case made that the previous 1150 and 1151 (as it exists) could not deal with an unlocked hexacore. It will be interesting to see if the updated 1151 socket has formerly reserved pins pressed into service supplying power to the cores.

 

[Arachnotronic]

This is you defending Intel. Intel could have launched a Skylake 6C/12T core i7 if they wanted to a couple of years back. Intel's turbo clocking is very sophisticated and they needed to lower base clocks for 6C/12T to fit in the respective power envelopes for notebook and desktop. Since most desktop workloads rarely push all these cores the CPU will be turboing very high at all times. We know very well Intel did not do so because there was no competition from AMD and they were more bothered about maximizing gross margins than providing more cores to the mainstream consumer. AMD are going to bring 4C//8T Raven Ridge APUs with TDP as low as 15w. So why can't Intel provide 6C/12T at 35-45w TDP for notebooks inspite of having a superior process node. Just wait and see how Intel start to ramp core counts up at 10nm. Intel stayed at 4 cores for mainstream from the ancient 45nm to 32nm to 22nm and finally to 14nm node . Lynnfield came with 4 cores for mainstream in late 2009. Thats an eternity in the PC industry. The transition from 4 to 6 cores in mainstream took 8 years. Just watch how quickly Intel move to 8 cores.

They stayed at 4 cores for mainstream, but they kept throwing in larger iGPUs, integrated additional technologies, improved the media engine significantly, and so on.

These were all features that directly improved the user experience for mainstream PC users far more than throwing in a couple of extra cores would've.

Anyway, I'm glad Intel is going to 6+2 for mainstream, and I hope they will push to even higher core counts and throw in more features too, that's all great stuff for the consumer.

If I can get a 10 core mainstream Ice Lake or Tiger Lake, I'd be really happy