Intel's response to RyZen.

[dahorns]

Intel's 14nm+ actually gave no discernible ipc improvements, mobile or desktop.
The sole reason the mobile chips do better on benches is their ability to stay within their TDP envelope while boosting to higher clocks for longer periods.

Did someone assert differently?

 

[LTC8K6]

Did someone assert differently?

Post 224.

 

[LTC8K6]

PC Games Hardware was able to find out from its own sources that the subject matter, that the Core i7-7740K and i5-7640K actually should appear, but not as a timely Ryzen counter-attack and also not as Kaby Lake-S. Rather, it is the 112-Watt chips with code names Kaby Lake-X, which are to appear to Gamescom for the new high-end platform with base 2066. These are the same S-Dies, in which only the GPUs are deactivated and the TDPs are raised. The heat spreader is not soldering to Intel, but is based on its "thermal interface material". However, the Intel i7-7740K does not want to spend more than 4.5 GHz on its current state - the i7-7740K has a 100 MHz increased basistakt compared to the i7-7700K, but needs the same 4.5 GHz turbo .

100Mhz base clock increase only.

Google translate from this:

http://www.pcgameshardware.de/Kaby-...ntels-Antwort-Core-i7-7740K-i5-7640K-1220177/

 

[jpiniero]

It doesn't look like there is much more clock headroom for Intel desktop chips with the 14nm process.
KL-X is not expected to have much of a clock boost, and it's TDP at 112W would seem to be a further indication that it's at the limit.

Yeah, voltage/clock once you get beyond 4.2 starts to go haywire. It may be a fundamental issue with Core though and not specific to 14 nm. But at mobile they've been able to make gains; the 15W 7600U can full core turbo to 3.9 although probably not long. If Intel could somehow fix the issue you could see clocks increase quite a bit.

 

[dahorns]

Post 224.

Here is what he said.

Intel's best HEDT offerings are based on Broadwell arch and first gen 14nm process. Raven Ridge will have to compete with Kaby Lake Refresh-U/Coffee Lake built on a third generation 14nm process. 14nm+ already gave huge improvements in notebooks, and I think 14nm++ will continue that.

I really think you've confused the context of the post. He was responding to someone asserting that Raven Ridge will provide meaningful competition in the mobile/laptop market and using the fact that Zen apparently matches or beats the efficiency of Broadwell HEDT to indicate performance/watt leadership in the mobile/laptop segment. Arachnotronic's point is that the comparisons to Broadwell HEDT is misleading, as it is an older design on an less refined 14mm node. On the other hand, we've already seen large gains in low tdp performance improvement thanks to 14nm+. He asserts that we can expect similar gains with 14nm++ in the lower TDP space. I think that is a reasonable assertion.

 

[LTC8K6]

Here is what he said.



I really think you've confused the context of the post. He was responding to someone asserting that Raven Ridge will provide meaningful competition in the mobile/laptop market and using the fact that Zen apparently matches or beats the efficiency of Broadwell HEDT to indicate performance/watt leadership in the mobile/laptop segment. Arachnotronic's point is that the comparisons to Broadwell HEDT is misleading, as it is an older design on an less refined 14mm node. On the other hand, we've already seen large gains in low tdp performance improvement thanks to 14nm+. He asserts that we can expect similar gains with 14nm++ in the lower TDP space. I think that is a reasonable assertion.

I don't see much more in the way of clocks, though.
And he can defend himself.

 

[coercitiv]

And he can defend himself.

Should he be required to defend against something he did not say?

 

[LTC8K6]

Should he be required to defend against something he did not say?

It's hard to discuss something 3rd party...

We got about 200mhz with the KL mobile chips. They don't maintain the high clocks above that for very long.
I just can't see much more than that in any more refinements of 14nm. We've already had the big refinement, imo.

Notebook check:

As always, the maximum clocks are just one aspect – the processor can only maintain the specified maximum values if it does not surpass the TDP. Looking at the Core i7-7500U, this is only the case for single-threading applications, where we can see the maximum 3.5 GHz all the time. The clocks will drop from initial 3.5 GHz to 3.2-3.3 GHz after 30 seconds in multi-core scenarios. All in all, we get a respectable 15 percent advantage over a Core i7-6500U in both scenarios.

 

[dahorns]

I don't see much more in the way of clocks, though.
And he can defend himself.

Defend? Were you accusing him of something? This is a discussion regarding CPUs. You said something inaccurate, and I wanted to clarify it.

 

[SAAA]

It's hard to discuss something 3rd party...

We got about 200mhz with the KL mobile chips. They don't maintain the high clocks above that for very long.
I just can't see much more than that in any more refinements of 14nm. We've already had the big refinement, imo.
Notebook check:

As always, the maximum clocks are just one aspect – the processor can only maintain the specified maximum values if it does not surpass the TDP. Looking at the Core i7-7500U, this is only the case for single-threading applications, where we can see the maximum 3.5 GHz all the time. The clocks will drop from initial 3.5 GHz to 3.2-3.3 GHz after 30 seconds in multi-core scenarios. All in all, we get a respectable 15 percent advantage over a Core i7-6500U in both scenarios.

Just coming here to clarify... what was Intel statement about 14nm+? 12% better process, then the slide stating 15% better performance with 6→7 generation. It's mobile yes, but that looks pretty much spot on with the review results for me.

If Coffelake does bring 15% too either 14nm++ makes mobile chip reach 4-4.4GHz single core on most parts (7600U does already 3.9, 7y75 can peak at 3.6!) or they reach the same with IPC gains and a little better clocks.
I don't see mobile parts with the same turbo as desktop i7 possible so I'm keening on IPC gains mostly.
Either way 10% up minimum... we already know 14nm+ is good at up to 4.8-5.2GHz for overclocks, another bump would make 5GHz the norm on 90% of chips, 5.5 not a pipe dream and any extra IPC would be just the icing on the cake. There's the matter of core count increase, that should eat into stock parts so I'd expect 8700K or whatever it will be called to be 4.0GHz base at most, with possibly high turbo similar to 7700K (single core isn't TDP limited).

 

[LTC8K6]

Defend? Were you accusing him of something? This is a discussion regarding CPUs. You said something inaccurate, and I wanted to clarify it.

Accuse? What are you on about?

People defend and support their statements on discussion boards. Or not.

Accusations are rarely tolerated, as they would be personal.

You have asked me to defend myself, but I don't think you have accused me of anything except being confused.

You mean you think I said something inaccurate...that's your opinion.

Let's stay on topic and refrain from discussing our opinions as if they were facts.

I don't think we will get much more clock out of Intel's 14nm, mobile or desktop.
I believe we have already seen the big refinement.

 

[dahorns]

You mean you think I said something inaccurate...that's your opinion.

Let's stay on topic and refrain from discussing our opinions as if they were facts.

I don't think we will get much more clock out of Intel's 14nm, mobile or desktop.
I believe we have already seen the big refinement.

Fine, I just didn't appreciation the suggestion (good-natured as it may have been) that I shouldn't be allowed to clarify the conversation. I wouldn't be surprised if you're right regarding clocks on high-end desktop, but I expect there is still quite a bit of room in the 5-35 watt range.

Accuse? What are you on about?

People defend and support their statements on discussion boards. Or not.

Accusations are rarely tolerated, as they would be personal.

You have asked me to defend myself, but I don't think you have accused me of anything except being confused.

This is also a good example of confusion in terminology. As a lawyer I see a pretty big distinction between defending oneself and defending an opinion. My remark was meant mostly in jest (as was yours, I know).

 

[scannall]

I'm not sure they even have to respond. ARM is costing them more than AMD will anytime soon. Even if Ryzen is a home run, and it may well be. I'll be buying one to retire an OC'd 5660, because I need cores. But I have always been unwilling to pay the Intel tax for them. It's not a sale Intel would have made anyway, so they didn't lose any money there. AMD gets the sale, because they will be selling what I want, at a price I am willing to pay.

 

[A///]

Well, let's not forget Naples. That's going to hit Intel below the beltline, too.

 

[.vodka]

Well, let's not forget Naples. That's going to hit Intel below the beltline, too.

Yeah, that's where it'll hurt the most. Hit right in their cashcow. Hopefully they're successful here, going from virtually 0% share to even 1-5% is a monumental achievement.

Desktop is peanuts compared to the server market, but it holds the mindshare that sways future customers, so it's also important to be in.

 

[Phynaz]

I still don't get why anyone would be surprised by this. I've said again and again that I buy more AMD based equipment than I buy Intel based equipment.

 

[Ajay]

I wonder is SKL-X will be priced more competitively when it comes out. Could be a win-win for everyone (of course, Intel HEDT motherboards also usually start out on the expensive side). Still, hard to see Intel coming down to AMD pricing levels for 6 or 8 core CPUs.

 

[jpiniero]

I wonder is SKL-X will be priced more competitively when it comes out. Could be a win-win for everyone (of course, Intel HEDT motherboards also usually start out on the expensive side). Still, hard to see Intel coming down to AMD pricing levels for 6 or 8 core CPUs.

I think they are going to have to do something now. Maybe go 8/8/10/12.

Still pretty stunned there hasn't been any recent Skylake-X "leaks".

 

[jihe]

Intel doesn't cut prices on existing products, not even when they're being discontinued.

They cut Q6600 price by half when AMD was still competitive.

 

[Ajay]

I think they are going to have to do something now. Maybe go 8/8/10/12.

Still pretty stunned there hasn't been any recent Skylake-X "leaks".

I'm guessing Intel is waiting till Ryzen is officially selling before they start leaking to create some FUD. That, or they really want to get yields up so they can goose the clock speeds as a counter to AMD (so FUD would be premature).

 

[IntelUser2000]

It's hard to discuss something 3rd party...

We got about 200mhz with the KL mobile chips. They don't maintain the high clocks above that for very long.
I just can't see much more than that in any more refinements of 14nm. We've already had the big refinement, imo.

Notebook check:

We got better than 200MHz to get 10-15% gain in real world. KBL mobile reviews still indicate that it can't always maintain all core Turbo. If they increase all core and max Turbo by another 200MHz while increasing real world consistency, well get another 10-15%.

The question is: What happens to mobile chips after it reaches thermal-related clock speed limits like desktop chips?

Still pretty stunned there hasn't been any recent Skylake-X "leaks".

Companies are extremely reluctant to release specs nowadays due to competitive pressure. Releasing info about SKL-X means much info about their server line being released. I guess they want to keep that a secret.

 

[Arachnotronic]

We got better than 200MHz to get 10-15% gain in real world. KBL mobile reviews still indicate that it can't always maintain all core Turbo. If they increase all core and max Turbo by another 200MHz while increasing real world consistency, well get another 10-15%.

The question is: What happens to mobile chips after it reaches thermal-related clock speed limits like desktop chips?

Next step is more cores, then you use process improvements to make sure multi-thread performance goes up.

At the same time, keep trying to push single core IPC.

 

[A///]

GB 3 on engineering samples: http://www.cpu-monkey.com/en/cpu-amd_ryzen_7_1800x-705

They cut Q6600 price by half when AMD was still competitive.

Actually, they cut it down to mid $500s and then half again to around $275, IIRC.

Releasing info about SKL-X means much info about their server line being released. I guess they want to keep that a secret.

I'm not all too familiar with this. Could you explain it in simpler terms?

 

[IntelUser2000]

I'm not all too familiar with this. Could you explain it in simpler terms?

If you are not up to date with Intel platforms, the upcoming server chip based on Skylake is supposed to be the biggest change in what they call since - Nehalem. Nehalem on serverside brought huge performance, feature, and scability changes. Up until then Intel server chips were really PC chips with extra sauce on them. The much vaunted Conroe core didn't do it so much for server chips.

Skylake-X is a enthusiast platform chip that's a derivative of server Skylake. Do I think Intel views worth it to sacrifice server for niche market in a niche market? That is, for extremely expensive HEDT chips? No.

I am pretty sure they are well aware that the competitive landscape is more varied than ever. Also if you were keeping with tech news companies are just more secretive. The advancements are slowing down, so I figure they view every bit of gain as precious.

 

[A///]

If you are not up to date with Intel platforms, the upcoming server chip based on Skylake is supposed to be the biggest change in what they call since - Nehalem. Nehalem on serverside brought huge performance, feature, and scability changes.

I am pretty sure they are well aware that the competitive landscape is more varied than ever. Also if you were keeping with tech news companies are just more secretive. The advancements are slowing down, so I figure they view every bit of gain as precious.

I remember the days of Nehalem. Amazing stuff.

Yeah, I drifted away from tech for a few years. I've been catching up the past month. That makes a lot of sense. I can't wait to see what server side will be like once Naples goes live.