Intel 8700k 11% faster than 7700k

[done12many2]

16 minutes ago, Morgan MLGman said:

I don't think Intel had HCC parts planned at all. AMD pushed them to create them, also part of the reason why we got Skylake-X parts up to 10C earlier and 12C+ later.

If Zen turned out to be a complete flop, I'm almost 100% certain that 7900X would be the top-end chip (the 12C part at best) that would retail for 1500$...

12 minutes ago, done12many2 said:

 

So you think that Intel just created the 12c through 18c parts that fast, right?  Trust me, they were already in the works.  They were most definitely being developed for a later release, but to say that AMD forced them to "create them" is really reaching.  Even Intel can't work miracles.   

 

There's always a second half to Intel Extreme platforms.  x99 was Haswell-E first half and Broadwell-E second half.  

9 minutes ago, Morgan MLGman said:

Intel can't work miracles? So explain to me the reason behind releasing a 4th gen 5960X for 1100$ with 8 cores, going 6950X next gen with 10 cores for over 1700$ and suddenly next gen after that has a 999$ 10 core and an 18 core for nearly the same price of a 10-core from previous gen? This just seems to sketchy to be a coincidence, no matter how you look at it.

 

 

 

How in any way does that have anything to do with what we were just talking about?

 

Man every time you and I talk, you switch the topic mid game.  It's hard to focus on what you were originally claiming.  

 

So to get back on topic, you said that AMD forced Intel to "create" Skylake-X 12c-18c parts.  My response was, trust me, even Intel can't just create new chips that fast.  They were obviously brought over from the server line and weren't magically created.  

 

I agreed that they would not have released the 12c-18c chips if TR wouldn't have happened, but that's not to say that they weren't already planned.  The obviously weren't just "created".  Take note of my 1st half and 2nd half mention.  

 

 

5 minutes ago, The Benjamins said:

Well they made them for Xeons and then responded by rebranding them as x299 chips. This is the first time they used the MCC dies on a extreme edition CPU.

 

This ^^

 

Far cry from developing / creating a new chip.

[The Benjamins]

Just now, spat55 said:

You don't think that IPC will improve by 10% and clocks speeds improve by 10% too with stronger IMC?

Zen+ is a 14nm process improvement. Zen2 is when they go to 7nm and may have some die changes.

[NumLock21]

26 minutes ago, done12many2 said:

 

It's a trend on LTT, hype even the possibility from one company and cast doubt all over the other.  We all do it to some degree so I understand.  

Then those who takes, something like this way too seriously, needs to, as Linus sometimes say, chillax! 

 

 

When to be serious?

Driving

Crossing the street, stop looking down at your freakin phone!

Buying your first house

[ravenshrike]

10 minutes ago, done12many2 said:

So you think that Intel just created the 12c through 18c parts that fast, right?  Trust me, they were already in the works.  They were most definitely being developed for a later release, but to say that AMD forced them to "create them" is really reaching.  Even Intel can't work miracles.

The 12 core probably was slated for a later release at some point. 14, 16, and 18 almost certainly weren't and had to be designed on new silicon.

 

http://www.anandtech.com/show/11464/intel-announces-skylakex-bringing-18core-hcc-silicon-to-consumers-for-1999/2

 

[Morgan MLGman]

1 minute ago, done12many2 said:

How in any way does that have anything to do with what we were just talking about?

 

Man every time you and I talk, you switch the topic mid game.  It's hard to focus on what you were originally claiming.  

 

So to get back on topic, you said that AMD forced Intel to "create" Skylake-X 12c-18c parts.  My response was, trust me, even Intel can't just create new chips that fast.  They were obviously brought over from the server line and weren't magically created.  

 

I agreed that they would not have released the 12c-18c chips if TR wouldn't have happened, but that's not to say that they weren't already planned.  The obviously weren't just "created".

I didn't mean they created them from nothing, they obviously re-used and fine tuned server parts as they already had 22C chips in their previous Broadwell-E lineup...

I meant that they would never appear for consumers in 2017/2018 if not for Zen/TR and this itself is the biggest proof of what Intel has been doing the past few years...

 

I understand competition, market etc etc, but competiton is not healthy if it's at the expense of your own customers.

[done12many2]

5 minutes ago, ravenshrike said:

The 12 core probably was slated for a later release at some point. 14, 16, and 18 almost certainly weren't and had to be designed on new silicon.

 

http://www.anandtech.com/show/11464/intel-announces-skylakex-bringing-18core-hcc-silicon-to-consumers-for-1999/2

 

Read the source you provided man.  They did not create or develop new silicon for the 14c, 16c, and 18c parts.  That would literally take years to do.

 

Here's a quote in the opening of your source.

Quote

Intel is bringing the high core count silicon from the enterprise side down to consumers.

 

 

 

3 minutes ago, Morgan MLGman said:

I didn't mean they created them from nothing, they obviously re-used and fine tuned server parts as they already had 22C chips in their previous Broadwell-E lineup...

I meant that they would never appear for consumers in 2017/2018 if not for Zen/TR and this itself is the biggest proof of what Intel has been doing the past few years...

 

I understand competition, market etc etc, but competiton is not healthy if it's at the expense of your own customers.

 

Agreed.  That's the only point I was ever making.  Don't know where the rest of the stuff came from.  

[DXMember]

4 hours ago, Dylanc1500 said:

IBM has a POWER8 CPU that would clock at 5ghz. That neither here nor there though lol.

@porina AMD already sold FX9590 which was first commercial consumer CPU with 5GHz boost clock

[Taf the Ghost]

26 minutes ago, ravenshrike said:

The 12 core probably was slated for a later release at some point. 14, 16, and 18 almost certainly weren't and had to be designed on new silicon.

 

http://www.anandtech.com/show/11464/intel-announces-skylakex-bringing-18core-hcc-silicon-to-consumers-for-1999/2

 

The 12c through 18c parts are all from the same base die, which has 18c. These are actually binned Xeons, but are, hyper technically, from Intel's Semi-Custom line. Which is what all of the Skylake-X parts are. (At least, if you sort through the spec sheets, that's the logical conclusion.)

[DXMember]

8 minutes ago, Taf the Ghost said:

The 12c through 18c parts are all from the same base die, which has 18c. These are actually binned Xeons, but are, hyper technically, from Intel's Semi-Custom line. Which is what all of the Skylake-X parts are. (At least, if you sort through the spec sheets, that's the logical conclusion.)

I've read that 18 core chip is a cut down 22c die

[Taf the Ghost]

1 minute ago, DXMember said:

I've read that 18 core chip is a cut down 22c die

I hadn't seen any notes the 18c was a cut down from the XCC base. Though it wouldn't be the most shocking.

[ravenshrike]

7 minutes ago, Taf the Ghost said:

The 12c through 18c parts are all from the same base die, which has 18c. These are actually binned Xeons, but are, hyper technically, from Intel's Semi-Custom line. Which is what all of the Skylake-X parts are. (At least, if you sort through the spec sheets, that's the logical conclusion.)

According to the article the 12 cores are on the same die as the 6-10. The 14, 16, and 18 are going to be on a separate HCC die.

[DXMember]

4 minutes ago, Taf the Ghost said:

I hadn't seen any notes the 18c was a cut down from the XCC base. Though it wouldn't be the most shocking.

hmm... this what I google imaged...

 

and this is also I google imaged...

one suggests 18 core design, but displays what looks like 20 core/cache blocks

the other displays 24 core design

[Taf the Ghost]

@DXMember

 

Yeah, for the Server Products (SP) stack, there's 3 base designs. Intel ships something like 50 SKUs from them. Plus you can get "semi-custom" ones with certain things turned on at lower core counts. Stuff like that. Unless you're someone like Google, then you can pay to have super specialized versions.

[Guest]

5 hours ago, Dan Castellaneta said:

I'm trying to figure out the clock speeds of the chips in the slide. I'm wondering if it can be chalked up to that or improvements in Coffee Lake.

probably already answered, but it looks like the single core turbo varies by 100 MHz in favor of the 8700k.

 

so less than 11% increase, but still a relatively standard bump that we have been seeing from intel.

[Taf the Ghost]

27 minutes ago, Alexokan said:

probably already answered, but it looks like the single core turbo varies by 100 MHz in favor of the 8700k.

 

so less than 11% increase, but still a relatively standard bump that we have been seeing from intel.

We don't have any confirmation yet, but I'm almost certain the DDR4 baseline is going from 2400 to 2666. That'll get you a single-core bump in certain contexts.

[Biggerboot]

If the 8700k retails at the standard rate for mainstream i7's this will be a win.  I wasn't even expecting an IPC bump but if they can do both I'm sold.

[Coaxialgamer]

Nothing has been "confirmed". The article itself states that this is a leaked alide. 

No details have been given on core performance either, it just says +11% vs 7th gen. Intel did the same sh*t with kaby lake, and if i had to guess it's because of improved on-board graphics or the like, with no ipc bump. 

Especially given that coffeelake can't be anything more than refreshed kabylake, given the time constraints imposed on the architecture. 

[Coaxialgamer]

3 hours ago, done12many2 said:

 

So you think that Intel just created the 12c through 18c parts that fast, right?  Trust me, they were already in the works.  They were most definitely being developed for a later release, but to say that AMD forced them to "create them" is really reaching.  Even Intel can't work miracles.   

 

There's always a second half to Intel Extreme platforms.  x99 was Haswell-E first half and Broadwell-E second half.  

They didn't create anything. Intel already had those chips prepped for their server platforms, and just rebranded them as HEDT. 

[done12many2]

Just now, Coaxialgamer said:

They didn't create anything. Intel already had those chips prepped for their server platforms, and just rebranded them as HEDT. 

 

I was well aware of that, but thanks for letting me know in case.  

[AnonymousGuy]

4 hours ago, Taf the Ghost said:

@DXMember

 

Yeah, for the Server Products (SP) stack, there's 3 base designs. Intel ships something like 50 SKUs from them. Plus you can get "semi-custom" ones with certain things turned on at lower core counts. Stuff like that. Unless you're someone like Google, then you can pay to have super specialized versions.

Fun fact, eBay for whatever reason were interested in a 300W TDP part years ago.  It was funny to scroll through the SKU list and then see that number jump out.  No idea if they ever put it in production, but they were looking at it.  Pretty much every major player requests semi-custom chip settings.  Dell, HP,ebay, facebook, google, amazon, etc etc.  Boy do I want to be on those meetings when they're like "yeah we'll give you $200M in business if you do us a solid and makea 300W part."

[BingoFishy]

12 hours ago, porina said:

That would blow a big hole in my 5 GHz theory... but then would imply there is some other IPC gain.

The Intel CPUs are already really hot... unless they suddenly got a lot more efficient idk how you could get it up to 5GHz. IPC boost seems the most likely.

[Drak3]

@done12many2 How much you want to bet that Intel has been trying to push more cores on the enthusiast platform for years, but they couldn't meet goals/requirements of their own platform with potential higher core counts than we got?

[Belgarathian]

I'm calling it - 11% single-thread performance improvement by boosting up to 2C only. 

[Taf the Ghost]

5 hours ago, AnonymousGuy said:

Fun fact, eBay for whatever reason were interested in a 300W TDP part years ago.  It was funny to scroll through the SKU list and then see that number jump out.  No idea if they ever put it in production, but they were looking at it.  Pretty much every major player requests semi-custom chip settings.  Dell, HP,ebay, facebook, google, amazon, etc etc.  Boy do I want to be on those meetings when they're like "yeah we'll give you $200M in business if you do us a solid and makea 300W part."

If it got added to the SKU list, then it must have been a really big purchase. My assumption would be that Ebay needed a specific amount of computing ability in a 2U chasis, and that was the only way to get it in that generation. They'd bite the 300W TDP that generation and then replace with more efficient parts in the next cycle. 

 

As for the other parts of the server market, it's actually quite spread out. Google runs its servers from Mountain View, amazingly enough. Their backbone system pretty much just accepts more computing Nodes when each is added. Google doesn't care too much about specifics, they just need Volume of servers. But they do pay to get the first runs. That's why they got their first Skylake-SP servers in December 2016. Rest of the world? October 2017.

[LAwLz]

I wonder what benchmark those numbers are based on.

It's hard to see the frequency numbers, but an 11% single core performance increase and a 51% multicore performance increase indicates to me like the benchmark doesn't take advantage of all cores. I mean, if we assume the benchmark scaled linearly then there would be a 50% performance increase just from the extra cores alone. If you add the 11% per core performance (is it per core or per clock?) then you should be well above 60%.

 

I am so happy that Intel is finally pushing out some real improvements between generations again.