(Guru3D) kaby Lake i7 is another boring quad

[Silverforce11]

30% user penetration for your latest architecture is a pretty good statistic. Not Apple iOS good, but still pretty good for a hardware manufacturer.

Right, and that's for GPUs which bring huge performance leaps.

What is the % of penetration do you think it's around for CPUs? Half that?

Just on Steam figures, there's still a ton of gamers on C2D.

SO. Zen just needs to be a viable choice when it debuts to whatever Intel sells at the time.

 

[HiroThreading]

Right, and that's for GPUs which bring huge performance leaps.

What is the % of penetration do you think it's around for CPUs? Half that?

Just on Steam figures, there's still a ton of gamers on C2D.

SO. Zen just needs to be a viable choice when it debuts to whatever Intel sells at the time.

Sorry, not quite sure I follow the logic here.

One can infer from this is that even if Zen is amazing it will struggle to sell or achieve any form of significant market penetration because of the stagnant nature of the "gaming" CPU market... Which is exactly what I was saying, and hence why AMD won't waste their time aiming to take back market share amongst the gaming market -- they're going to focus on workstations and servers.

 

[VirtualLarry]

Sorry, not quite sure I follow the logic here.

One can infer for this is that even if Zen is amazing it will struggle to sell or achieve any form of significant market penetration because of the stagnant nature of the "gaming" CPU market... Which is exactly what I was saying, and hence why AMD won't waste their time aiming to take back market share amongst the gaming market -- they're going to focus on workstations and servers.

Not sure I agree. The FX 6/8-"core" line of CPUs, is mostly marketed to gamers these days, not to most "ordinary users".

I expect this line of marketing to continue, with 6/8-core Zen CPUs. (Sure, they'll be marketed at servers too, of course. But the gamer marketing will continue.)

 

[HiroThreading]

Not sure I agree. The FX 6/8-"core" line of CPUs, is mostly marketed to gamers these days, not to most "ordinary users".

Mm... but that's because no one involved with servers and workstations would even think of touching a Bulldozer FX processor. Meanwhile, APUs target non-gamers.

So AMD is left to target gamers similar to how Intel did back in the Pentium 4/D era: "More GHz. More cores. More performance."

I expect this line of marketing to continue, with 6/8-core Zen CPUs. (Sure, they'll be marketed at servers too, of course. But the gamer marketing will continue.)

If Zen delivers on its performance/watt promises (~8 core Haswell or even 8 core Ivy Bridge), then it will priced at a premium and targeted at servers and workstations, eliminating it from the desktop gamer market. AMD has repeatedly stated that they are no longer aiming to be the "cheap option".

 

[Silverforce11]

Sorry, not quite sure I follow the logic here.

One can infer from this is that even if Zen is amazing it will struggle to sell or achieve any form of significant market penetration because of the stagnant nature of the "gaming" CPU market... Which is exactly what I was saying, and hence why AMD won't waste their time aiming to take back market share amongst the gaming market -- they're going to focus on workstations and servers.

The logic is simple. There's a ton of users and gamers on older CPU uarch.

These are the market for Zen. Those already on Haswell to Skylake is NOT the market for Zen, neither are they the market for Kaby Lake.

If AMD does not show up to compete, KL will automatically get all those who are building systems in 2017, as well as all the OEM rigs sold. Thus, Zen just needs to be competitive on performance and price. It doesn't need to have performance that destroy Skylake.

All AMD have to achieve, is Broadwell IPC, in a 8 core SKU and sell it at a similar price to Intel's 4c/8t Skylake/Kaby Lake and AMD will bank it big time, much more than their current situation, for sure.

 

[lopri]

Can someone explain what the difference is between LPDDR3 and DDR3L? (or LPDDR4 and DDR4L) I have thought they are the same things called differently by different folks. Do they have officially (i.e. JEDEC-sanctioned) different qualification parameters?

 

[StrangerGuy]

Can someone explain what the difference is between LPDDR3 and DDR3L? (or LPDDR4 and DDR4L) I have thought they are the same things called differently by different folks. Do they have officially (i.e. JEDEC-sanctioned) different qualification parameters?

LPDDRx are mobile RAM that are directly soldered onto the mobo or on top of the SoC chip as a PoP package like in iPhones.

DDRxL are just regular desktop/laptop DIMMs that offers both normal and low voltage modes.

 

[TeknoBug]

I"m personally waiting for the right time to buy an i7 6700K when the price drops, I anticipate that the Kaby Lake i7 will be priced more than the i7 6700K currently is.

 

[DA CPU WIZARD]

You mean the same AMD that showed this slide to investors/analysts in 2012?

"Focus on power-performance optimized cores"

:sneaky:

 

[Arachnotronic]

:sneaky:

CPU design cycles are 3-5 years, so what you see in 2016/2017 was conceived of in 2012/2013. Nice try, though.

 

[Dresdenboy]

CPU design cycles are 3-5 years, so what you see in 2016/2017 was conceived of in 2012/2013. Nice try, though.

And the Zen project has been started in 2012 or earlier.

 

[lopri]

LPDDRx are mobile RAM that are directly soldered onto the mobo or on top of the SoC chip as a PoP package like in iPhones.

DDRxL are just regular desktop/laptop DIMMs that offers both normal and low voltage modes.

Ty

 

[Azuma Hazuki]

What sort of performance gain from SKL to KBL are we looking at here, compute-wise? The IGP doesn't really interest me, but am I reading those benchmarks right when they seem to say <5% overall?

There doesn't seem to be a point to this generation :/

 

[DrMrLordX]

1-4% if we're lucky, 0% if we're not.

 

[Dresdenboy]

1-4% if we're lucky, 0% if we're not.

But process maturity and perhaps some pwr mgmt improvements will result in somewhat higher base/boost clocks and maybe also longer boost durations.

 

[DrMrLordX]

Maybe. Yields should get better for 14nm in any case.

 

[SAAA]

What sort of performance gain from SKL to KBL are we looking at here, compute-wise? The IGP doesn't really interest me, but am I reading those benchmarks right when they seem to say <5% overall?

There doesn't seem to be a point to this generation :/

1-4% if we're lucky, 0% if we're not.

I used geekbench and made a few comparisons in the Skylake thread, anyway yes closer to 0% rather than 5 in most task... There should be a nice boost in some multimedia and clocking aspects besides little IPC gains anyhow.

A few leaked chips have 100-200MHz boosts at base speeds so expect something like a 4.2+ GHz 7700k with vaguely better overclock, maybe more of the rare 4.9-5 GHz golden chips etc.

 

[Azuma Hazuki]

Pointless. This is going to be an even bigger snooze-fest than Broadwell. I almost wonder if Intel is deliberately throwing Kaby under the bus and focusing on Cannonlake to put the hurt on Zen.

 

[ehume]

I got the impression that KL was not going to be much of a change from Sky Lake, just improved graphics. Seems to be what we're getting, along with improved manufacturing. Like Devil's Canyon, it's more of a hiccup than a new tick or a tock.

 

[IntelUser2000]

I got the impression that KL was not going to be much of a change from Sky Lake, just improved graphics. Seems to be what we're getting, along with improved manufacturing. Like Devil's Canyon, it's more of a hiccup than a new tick or a tock.

"Key Improvements"

Sounds like from leaked benches, the GPU is getting zero and CPU is getting 2-4% boost per clock(maybe higher clocks). I'd assume the improvements are like it was with "Tick" chips in the past. Basically, what we would have expected with Cannonlake, but just with 14nm.

Hopefully this isn't what we are getting.

2 year cadence:
Tick: 5%
Tock: 10%

3 year cadence:
Process: 5%
Architecture: 5%
Optimization: 5%

 

[DrMrLordX]

I'm surprised the GPU is getting no love with Kabylake. Where is Gen10?

 

[frozentundra123456]

I though KL was supposed to be all about igpu improvements. Not so much architecture per se but codecs for 4k playback or whatever.

 

[R0H1T]

I'm surprised the GPU is getting no love with Kabylake. Where is Gen10?

KL was never going to be gen10, it's just a repeat of Devil's Canyon. The more pertinent question is if we'll see a true(r) gen10 with Cannonlake or another repeat of Broadwell i.e. gen8.5 D:

 

[frozentundra123456]

Quit spreading FUD. Skylake is already gen 9.

 

[R0H1T]

Quit spreading FUD. Skylake is already gen 9.

What do you mean by spreading FUD? I know Skylake is gen9 however, just as Broadwell is gen8.5 (with a node shrink) there is a good chance that Cannonlake will be gen 9.5 & not gen10, KL barely qualifies as having any IGP improvement at all.

Next time try reading my posts carefully, without blinkers, otherwise there's no need to respond to them!