Intel Skylake / Kaby Lake

[Enigmoid]

http://www.legitreviews.com/wp-content/uploads/2015/05/6700k_skylake_geekbench.png

Who's right?

The improvement over broadwell is pretty sad if that is the case. Like 1-3%.

Edit: especially when you consider how much skylake's score is boosted by memory.

 

[Arachnotronic]

The improvement over broadwell is pretty sad if that is the case. Like 1-3%.

Edit: especially when you consider how much skylake's score is boosted by memory.

It's Geekbench. Let's wait to see final silicon in more "real world" applications. That said, feeling pretty darn good about my 4790K @ 4.8GHz right now

 

[SAAA]

If these are any indication of real world performance Skylake is shaping up to be pretty sad. Then again geekbench is a terrible measurement.

The improvement over broadwell is pretty sad if that is the case. Like 1-3%.

Edit: especially when you consider how much skylake's score is boosted by memory.

I still find really weird that other leaks (sysoft) give upward of 20-30% better integer scores while here it doesn't improve that much save for some damn crypto score.
And yes the memory subscore is granting most of the increase, few tests show a 10% difference or so... maybe it's time for a Geekbench4?

Heck Broadwell should be 5%: this can't be less so 10% (total) over Haswell is mathematically granted for this tock, anything else is just welcome.

 

[LTC8K6]

The improvement over broadwell is pretty sad if that is the case. Like 1-3%.

Edit: especially when you consider how much skylake's score is boosted by memory.

It makes me think something is not right with the reported benchmarks, and that Skylake will be better than it seems right now.

 

[Sonikku]

The improvement over broadwell is pretty sad if that is the case. Like 1-3%.

Edit: especially when you consider how much skylake's score is boosted by memory.

Intel sure has been phoning it in over the years.

 

[myocardia]

Intel sure has been phoning it in over the years.

I know, right? Maybe one of these years, they'll finally catch up with AMD, and maybe, if they're lucky, they'll even be able to catch up with Via...

 

[-batab-]

BTW, is it me or the numbers here are a bit different than the ones previously posted? ...usually a bit higher

http://www.technikaffe.de/cpu-intel_core_i7_6700k-518
http://www.technikaffe.de/cpu_vergleich-intel_core_i7_4790k-411-vs-intel_core_i7_6700k-518

Also, don't think it was already posted (and if yes i'm sorry), but it seems they are still using grease
http://www.coolaler.com/showthread.php/327877 <-- got this link from here: http://en.chhimg.com/article/35140

 

[MrTeal]

BTW, is it me or the numbers here are a bit different than the ones previously posted? ...usually a bit higher

http://www.technikaffe.de/cpu-intel_core_i7_6700k-518
http://www.technikaffe.de/cpu_vergleich-intel_core_i7_4790k-411-vs-intel_core_i7_6700k-518

Also, don't think it was already posted (and if yes i'm sorry), but it seems they are still using grease
http://www.coolaler.com/showthread.php/327877 <-- got this link from here: http://en.chhimg.com/article/35140

It's too bad they didn't clean it (or learn to use their camera), but that die looks tiny. Maybe it's just a visual artifact of the rectangle in the center being so much shinier.

 

[Sonikku]

I know, right? Maybe one of these years, they'll finally catch up with AMD, and maybe, if they're lucky, they'll even be able to catch up with Via...

That's kind of my point. Ever since they pulled ahead of AMD they just don't give a fudge any more. They don't care to push the envelope. They just... I don't know. Put everything on cruise control. I guess that's how you run a good business? Then again, that's what everyone is doing now.

Never mind me. I'm just kind of bummed that the new and exciting times of the 90's have long since given way to the old and tired era of stagnation of today. :'(

 

[Arachnotronic]

It's too bad they didn't clean it (or learn to use their camera), but that die looks tiny. Maybe it's just a visual artifact of the rectangle in the center being so much shinier.

Haswell 4+2 is 177mm^2. I'd expect Skylake 4+2 to be ~100mm^2.

 

[frozentundra123456]

That's kind of my point. Ever since they pulled ahead of AMD they just don't give a fudge any more. They don't care to push the envelope. They just... I don't know. Put everything on cruise control. I guess that's how you run a good business? Then again, that's what everyone is doing now.

Never mind me. I'm just kind of bummed that the new and exciting times of the 90's have long since given way to the old and tired era of stagnation of today. :'(

I dont think it is from lack of trying. It is just that all the easy gains have been made already and the nodes are becoming so small that each shrink becomes more and more difficult and expensive and yields less gains. Also the focus has shifted from absolute performance to performance per watt. For instance, top end performance from ivy bridge to today has increased maybe 20 to 30 percent, but my wife's 35 watt ivy bridge laptop would probably be beaten in most cpu benches by 15 watt broadwell U with better igpu performance thrown in as well. The only thing I think Intel could do to get an easy performance increase would be to increase core counts and hyperthreading availability across the core line-up. Something like Celeron-dual core no HT, Pentium-dual core + HT, i3 quad core, no HT, i5 quad with HT, and i7-hex core, with and without HT. I really wish they would do this instead of continually adding bigger and bigger igps, but I dont see it happening.

 

[Ajay]

<snip> Also the focus has shifted from absolute performance to performance per watt. For instance, top end performance from ivy bridge to today has increased maybe 20 to 30 percent, <snip>

^^This!

Edit: This is key to Intel's success in the data center, which is their most profitable sector (IIRC).

 

[ShintaiDK]

^^This!

Edit: This is key to Intel's success in the data center, which is their most profitable sector (IIRC).

And mobile.

People need to remember: Server. Mobile, Atoms, Desktops. Thats the priority. And the first 3 wants performance/watt without question.

 

[kimmel]

That's kind of my point. Ever since they pulled ahead of AMD they just don't give a fudge any more. They don't care to push the envelope.

18 Core Haswell @ 661 mm^2 isn't pushing the envelope? Oh you mean they should pander to the people who don't care to buy desktops? There is plenty of envelope pushing. Just for people who are willing to purchase chips. Before you say, it's because of the stagnation that people aren't buying chips, it's the exact opposite. Money stops flowing, priorities then shift. Not the other way around.

 

[Arachnotronic]

18 Core Haswell @ 661 mm^2 isn't pushing the envelope? Oh you mean they should pander to the people who don't care to buy desktops? There is plenty of envelope pushing. Just for people who are willing to purchase chips. Before you say, it's because of the stagnation that people aren't buying chips, it's the exact opposite. Money stops flowing, priorities then shift. Not the other way around.

QFT!

 

[tential]

Lol at people saying Intel is on cruise control because there hasn't been massive IPC improvements. Like Intel didn't improve every single other aspect of their processor from igpu to power consumption to God knows what with mobile. Intel has done a ton of innovation. It wasn't directly for you as a gamer. Get over it.

 

[ShintaiDK]

IPC improvement have pretty much been stable increase since Conroe.

And with Skylake 65W is the norm.

 

[inf64]

I'm thinking about that i5 6600 (non K). For a 65W chip it should be a great performer and very power efficient. Hopefully they are available in August and not for a premium price.

 

[Walter E Kurtz]

Yes it's obvious priorities have shifted from increasing raw performance, to an increase of performance at reduced power draw. Which is a great achievement in itself. no doubt, but in the end no great leaps forward were made. I mean I haven't even upgraded my 4.1Ghz W3570 on socket 1366 from 2010. Single thread performance is still good enough to not waste money on newer chips with IGP and power saving characteristics I don't need.

Coming from the old overclocking days of A64 and Conroe where great raw performance AND huge overclocks came hand in hand - it's just not fun anymore. My old home Xtremsystems is dead. Because there are hardly any extreme overclocks worth doing anymore.

People being happy about Intel making nice laptop chips. It's sad really. Intels huge technical achievements aside, it's just sad.

 

[frozentundra123456]

Yes, but stock clocks now are close to what the overclocks were back in the "good old days", or even faster. How many core 2 chips could get to 4ghz, which is the stock clock for 4790k? Even for SB, which is probably the golden age of cpus, you could no longer take a cheap hundred dollar chip and overclock the hell out of it.

I too was hoping for more improvement from Skylake that it looks like we are going to get, but we really need to quit expecting improvements in performance to keep increasing at the same rate as they were before the product became mature. It is pretty much the same fallacy that we see in sports, or the stock market, or a host of other areas: the idea that a trend should somehow continue forever.

And yes, Shintai, I know have seen your ipc graph, but back then we were getting increases in both ipc and clockspeed, so the actual single thread improvement was much more significant from generation to generation.

 

[Walter E Kurtz]

How many core 2 chips could get to 4ghz, which is the stock clock for 4790k?

I owned an E8600 that did 4.4Ghz on air. And a Q6600 that did close to 4Ghz on water.

I too was hoping for more improvement from Skylake that it looks like we are going to get, but we really need to quit expecting improvements in performance to keep increasing at the same rate as they were before the product became mature. It is pretty much the same fallacy that we see in sports, or the stock market, or a host of other areas: the idea that a trend should somehow continue forever.

Choices are not the same as actual limitations. I bet that if Intel wanted to, they could cram the space now used for IGP with another 8 cores just to name something. Or improve heat transfer to the IHS for better overclocking etc. etc. When there's a will, there's a way. And I personally just don't like laptop oriented design choices.

 

[frozentundra123456]

Exactly. Right now you can get a G3258 that most likely will do that 4.4 ghz on air, with better ipc, and at far less cost than the E8600. And a 4790K does 4.2ghz all core turbo stock, without water cooling.

Yes, I would like a mainstream hex core, and TBH, I think Intel is making a mistake in not bringing one out, especially if Zen comes out on time and lives up to the hype (neither of which is a sure bet by a long shot). But other than that, I just dont think Intel is holding back or not devoting enough effort to improvements. Now maybe they are not doing a good job, or their priorities are different than yours, or they have simply hit the wall where further improvements are extremely difficult, or a combination of all of these, but I think the insinuation that intel is just coasting along does a dis-service to all the hard working engineers and technicians in the company who are undoubtedly trying their best to compete.

 

[witeken]

Didn't bother to create a new thread, but here's an achievement by Imec. It seems they're far behind Intel.

http://www2.imec.be/be_en/press/imec-news/vlsi2015-germanium-multigate-gaanw-finfet.html

 

[mikk]

http://myeventagenda.com/sessions/0B9F4191-1C29-408A-8B61-65D7520025A8/7/5


IDF August will have several Skylake related sessions as expected.

 

[Ajay]

http://myeventagenda.com/sessions/0B9F4191-1C29-408A-8B61-65D7520025A8/7/5


IDF August will have several Skylake related sessions as expected.

There seems to be less in the ARC topics than I would have expected for a TOCK.