Intel's Future Chips: News, Rumours & Reviews

[de5_Roy]

Providing a pair of Intel official slides answering your question is not spamming.

sigh. it looks like you do intend to drag it out. directly posting images for absolutely no reason is spamming. simply because it's unwanted and unnecessary. it's not semantics.

I see you didn't pay attention to the slides given to you. If you had done, you would see that Skylake-S includes "enthusiast quad core: 95W" and unlocked processors (overclocking). The slides also show that Skylake-S is coming to the Z170 chipset. I don't know how to say this to you... but your pretension that "Intel isn't launching standard voltage parts" because Skylake S are "low voltage parts with lower TDP" is plain wrong.

this is just semantics sprinked with ad hominem attack.
model numbers mean very little in the context of technology. model numbers are assigned to denote an s.k.u. and may not necessarily point to the hardware characteristics. you are Trolling using a gap in the leaked slides. the gap is that intel didn't specify the tdp and the models associated with model names. haswell, ivy bridge's -S series l.v. processors also fit their own z-series motherboards. i said skylake-s will be l.v. because of intel's past trends of model numbering and the TDP ranges in the leaked slide - which match the tdp ranges accociated with past -s cpus and their specs, there's no "pretense". that is always subject to change, depending on how intel markets them. this is the last time i'll reply as you've clearly resorted to trolling over a trivial matter.

 

[juanrga]

Providing a pair of Intel official slides answering your question is not spamming.

sigh. it looks like you do intend to drag it out. directly posting images for absolutely no reason is spamming. simply because it's unwanted and unnecessary. it's not semantics.

No. The first time I gave only a link to an article. You did make wrong claims about it and then I reproduced a pair of Intel slides. Giving official slides with information relevant to your points is not spamming, please check a dictionary.

I see you didn't pay attention to the slides given to you. If you had done, you would see that Skylake-S includes "enthusiast quad core: 95W" and unlocked processors (overclocking). The slides also show that Skylake-S is coming to the Z170 chipset. I don't know how to say this to you... but your pretension that "Intel isn't launching standard voltage parts" because Skylake S are "low voltage parts with lower TDP" is plain wrong.

this is just semantics sprinked with ad hominem attack.
model numbers mean very little in the context of technology. model numbers are assigned to denote an s.k.u. and may not necessarily point to the hardware characteristics. you are Trolling using a gap in the leaked slides. the gap is that intel didn't specify the tdp and the models associated with model names. haswell, ivy bridge's -S series l.v. processors also fit their own z-series motherboards. i said skylake-s will be l.v. because of intel's past trends of model numbering and the TDP ranges in the leaked slide - which match the tdp ranges accociated with past -s cpus and their specs, there's no "pretense". that is always subject to change, depending on how intel markets them. this is the last time i'll reply as you've clearly resorted to trolling over a trivial matter.

You cannot claim that "intel didn't specify the tdp", because one of the slides I provided mentions the range of TDP. You deleted the slide from your reply, but I will reproduce it once again

See it? Skylake-S includes up to 95W processors. There is no "trolling" neither "ad hominem" nor "spamming" on providing official slides with information that disprove your claims.

 

[Turb0Yoda]

I knew I did SOMETHING wrong

 

[juanrga]

And here we have the first 95W unlocked Skylake processors

 

[cemerian]

the last roar of itanium
http://wccftech.com/intel-itanium-kittson-ia64-processors-32nm/

 

[Alpha3031]

the last roar of itanium
http://wccftech.com/intel-itanium-kittson-ia64-processors-32nm/

Nay, Itanium ends not with a but a whimper.

 

[Cazalan]

At least it still fits the same socket. That's a bonus.

 

[gamerk316]

http://www.techspot.com/news/60419-leaked-specs-upcoming-intel-skylake-cpus-reveal-high.html

6700k @ 4Ghz stock. So it looks like clocks are a tad higher this go around.

 

[-Fran-]

http://www.techspot.com/news/60419-leaked-specs-upcoming-intel-skylake-cpus-reveal-high.html

6700k @ 4Ghz stock. So it looks like clocks are a tad higher this go around.

My take is they have to clock them higher each time to show a performance gain from gen to gen.

Since we're "enthusiasts" stock doesn't matter much, but Max OC and Optimal OC do. And some might say least voltage at stock freq.

Cheers!

 

[juanrga]

http://www.techspot.com/news/60419-leaked-specs-upcoming-intel-skylake-cpus-reveal-high.html

6700k @ 4Ghz stock. So it looks like clocks are a tad higher this go around.

Those specs were given a pair of posts above in table form here

http://www.tomshardware.co.uk/forum/id-1581001/intel-future-chips-news-rumours-reviews/page-15.html#15691805

 

[jdwii]

Will be fun looking at the comparisons between the two. I will want to consider if it will be worth it over my 4790K power consumption is about the same.

 

[cemerian]

http://www.extremetech.com/gaming/203926-analyst-intel-will-adopt-quantum-wells-iii-v-semiconductors-at-10nm-node
hope it's true, would be an very interesting result

 

[juanrga]

More Skylake leaks

About 14% IPC over Haswell

 

[gamerk316]

About 14% IPC over Haswell

yep; comparing the 4790k and 6700k, which have the same clock, number of cores, it's easy to solve IPC, as you simply need to measure the performance difference, which is a hair over 13%.

 

[-Fran-]

About 14% IPC over Haswell

yep; comparing the 4790k and 6700k, which have the same clock, number of cores, it's easy to solve IPC, as you simply need to measure the performance difference, which is a hair over 13%.

** In Cinebench R15.

Cheers!

 

[jdwii]

Guessing on average we will see 10% IPC boost i also wonder how much the GPU improved over the 4600.

Edit
If you really think about it 10-15% is a pretty big boost when most are claiming x86 is pushed to the limits i mean Amd still has a LOT to grow so that is why we expect more from that camp. Intel i really don't expect much for their CPU now a days.

 

[gamerk316]

About 14% IPC over Haswell

yep; comparing the 4790k and 6700k, which have the same clock, number of cores, it's easy to solve IPC, as you simply need to measure the performance difference, which is a hair over 13%.

** In Cinebench R15.

Cheers!

Fair enough. But at least extensions aren't coming into play yet, so, at least in that one benchmark, that 13% is PURELY architectural.

So yeah, just randomly guessing 10-15% IPC gains per gen seems to be doing me quite well with predictions. 😛

 

[juanrga]

This 15% is not per gen. Skylake is ~15% compared to Haswell, not to Broadwell. It is ~7% per gen. Of course the real upgrade will be on AVX-like extensions. Expect 70% IPC gains or so.

 

[-Fran-]

This 15% is not per gen. Skylake is ~15% compared to Haswell, not to Broadwell. It is ~7% per gen. Of course the real upgrade will be on AVX-like extensions. Expect 70% IPC gains or so.

I have a *big* doubt there. When you say "AVX" you mean "from AVX v1 to AVX v2", right?

The context of this goes into current "AVX v1" software. Unless they're using a superset of AVX v2 for v1, there should be no gains in current software, isn't it?

Cheers!

EDIT: Typo.

 

[Cazalan]

According to the rumors Skylake desktop/mobile will not support AVX-512. Only Xeon.

 

[juanrga]

This 15% is not per gen. Skylake is ~15% compared to Haswell, not to Broadwell. It is ~7% per gen. Of course the real upgrade will be on AVX-like extensions. Expect 70% IPC gains or so.

I have a *big* doubt there. When you say "AVX" you mean "from AVX v1 to AVX v2", right?

The context of this goes into current "AVX v1" software. Unless they're using a superset of AVX v2 for v1, there should be no gains in current software, isn't it?

The mentioned huge improvement will be when recompiling AVX2 software to use the new AVX512 and companion extensions. Recompiled AVX software will show bigger gains and legacy binaries will show no gain over Broadwell (except the general architectural ~7%).

 

[-Fran-]

According to the rumors Skylake desktop/mobile will not support AVX-512. Only Xeon.

Given the adoption rate of new tech in the non-enterprise, kinda makes sense.

The mentioned huge improvement will be when recompiling AVX2 software to use the new AVX512 and companion extensions. Recompiled AVX software will show bigger gains and legacy binaries will show no gain over Broadwell (except the general architectural ~7%).

So, if Skylake were to launch today, then we would see software for it around... mid 2016? AVX 1 took a while to be introduced, even though it was given support at the end of 2011 massively (windows, mac and linux IIRC). I remember the first game I saw using AVX was DiRT 3 and that one was released mid 2012. I haven't seen any other piece of software using it explicitly (having a "_avx" executable).

Cheers!

 

[de5_Roy]

Intel to help vendors launch inexpensive tablets, say Taiwan makers
http://www.digitimes.com/news/a20150423PD211.html
The Ridiculously Powerful MicroATX PC
http://www.techspot.com/review/987-ridiculously-powerful-microatx-pc/

 

[juanrga]

Software with specific Haswell extensions was available for testing on reviews of Haswell. I see no reason why AVX512 software don't will be ready for Skylake launch. In fact, I know one dev has been working with AVX512 for a while now.

Of course Windows users will be late to the party... as usual.

 

[-Fran-]

Software with specific Haswell extensions was available for testing on reviews of Haswell. I see no reason why AVX512 software don't will be ready for Skylake launch. In fact, I know one dev has been working with AVX512 for a while now.

Of course Windows users will be late to the party... as usual.

I'll concede this point, since if Cazalan is right and Desktop variants won't get the extension until the E CPUs are released, it won't matter much.

And is there *any* relevant desktop piece of software that uses AVX2?

Cheers!