Waiting for the 2019 MBP [Merged]

[koyoot]

(Emphasis added.)

Apparently you should take a vocabulary class. "Converted" is significantly different from "offer." Per the Dell website, 6 out of 130 of their 15" laptops use AMD processors, which translates to ~4.6%. That's not "converted," that's a rounding error.

And for server share, AMD just hit 1% (yes, after falling from 24% or so to below 1%), as you can see here.

So your implied suggestion that Apple, or any other company, switch to AMD for CPUs borders on the absurd.

You understand the mechanisms of this industry? There will not be full conversion if you have two equally competing companies. Nobody can allow themselves to offer one company's solution, because they end up relying on that company's technological mercy. If, like Intel's TMG, you f***** up, you can end in killing a company, like SemiAccurate has reported here: https://www.semiaccurate.com/2018/0...down-is-crushing-a-20b-market-cap-tech-giant/ and here https://semiaccurate.com/2018/08/30/update-to-intel-custom-foundry-10nm-customer-meltdown/ or like with all of the delays, and overall stagnation of Apple computers, because Apple is relying on Intel way too much. But that was Apple's decision, and they will have to deal with everything it implies, even if it means that their competition will have for upcoming two years better CPUs, from competing with Intel brand(which is very likely, with process advantage, and possibility that AMD has once again pulled a rabbit from silicon hat).

Server adoption takes years. There was a time, when the sentence "Nobody was fired for choosing IBM" was very accurate, about the attitude in this market. Thankfully, now we have three players in the space. And x86 Server market is undergoing dramatic change.

I just remember so much talk about AMD APU’s from years ago, I just can’t believd they will suddenly take over Intel, they failed so often - not you but a lot of AMD fans used to talk about the Intel killer years ago on overclock.net due to being disgruntled with Intel.

I’d be happy if they did, would increase competition.

Currently AMD's APUs are much better than Intel offerings. They offer very high clock in low power states, and they offer very fast GPU, at the same time, which simplifies the engineering effort if you want to adopt this technology. AMD on the other hand is a company that with Zen 1 cannot give high enough capacity because of relying on GloFo manufacturing capabilities, which are not the greatest in the world, which massively affects the adoption rates of their hardware, but the hardware itself is very good. 7 nm will have much higher capacity. AMD will offer a lot of 7 nm products: CPUs, APUs, GPUs, console SoC's, semi-custom hardware, all of this.

 

[macintoshmac]

If that tweet is accurate, then intel will be using 12nm and calling it 10nm, so I'm not sure they'll be all that forthright in confirming the news.

Wouldn't that be just wrong to claim 10nm when it is actually 12?

 

[gaanee]

Intel would lose anyway if they keep delaying 10 and 7nm process. Apple could choose switching to custom ARM CPU and possibly even Windows OEMs

You really thing Intel's Ego is open enough to do this?

They would lose a LOT if they would use anything else than their own process, for manufacturing of their CPUs. They will manufacture the GPUs on TSMC in 2020, and H310 chipset in 2018/2019.

 

[koyoot]

Intel would lose anyway if they keep delaying 10 and 7nm process. Apple could choose switching to custom ARM CPU and possibly even Windows OEMs

If they will switch to ARM they will straight lose any competitive position against anything apart from Chromebooks.

 

[Naimfan]

You understand the mechanisms of this industry?

From the available evidence, better than you do.

I note your concession of the evidence provided; this alone, even without more, is sufficient to dismiss anything you've said on the subject.

 

[Howard2k]

Wouldn't that be just wrong to claim 10nm when it is actually 12?

The terms are actually kinda loose. They’re strongly indicative of technology but they’re not absolute. Intel’s 10nm is comparable to some other 7nm designs, for example. They were fairly well defined at one point but due to differences in manufacturing processes they’re not all directly comparable these days.

 

[gaanee]

Apple could switch to AMD

If they will switch to ARM they will straight lose any competitive position against anything apart from Chromebooks.

 

[maflynn]

Apple could switch to AMD

They could but will they? Also will AMD handle the volumes and updates in a timely manner? I think the issues of PPC platform are still in the mind of the executives. IBM promised both volume shipments and continual updates, neither of which occurred. AMD is hitting their stride, but they're not has large as intel and could run into issues down the road.

 

[koyoot]

They could but will they? Also will AMD handle the volumes and updates in a timely manner? I think the issues of PPC platform are still in the mind of the executives. IBM promised both volume shipments and continual updates, neither of which occurred. AMD is hitting their stride, but they're not has large as intel and could run into issues down the road.

With Capacity and volume production of 14 nm chips, its Intel who has capacity problems, not AMD .

 

[koyoot]

Next MBP refresh will have at best 8 core Coffee Lake-R.

 

[RumorConsumer]

Next MBP refresh will have at best 8 core Coffee Lake-R.

Guesstimate of when in 2019?

 

[Ma2k5]

They could but will they? Also will AMD handle the volumes and updates in a timely manner? I think the issues of PPC platform are still in the mind of the executives. IBM promised both volume shipments and continual updates, neither of which occurred. AMD is hitting their stride, but they're not has large as intel and could run into issues down the road.

I wouldn't hold my breath for AMD anything.

Just over 2 years ago, there was a whole AMD circle-jerk on this forum https://forums.macrumors.com/threads/amd-zen.1988652/. It is a long thread but in a nutshell, a lot of excitement over nothing - and the same may happen again (before anyone gets their hopes up). The same kind of "evidence" was cited back then too - even talks about how Nvidia should be worried in the mobile GPU market etc. A lot of theories about all sorts really, none of which has surfaced. I've seen AMD crush the hopes and dreams of some of the biggest AMD fanboys you can imagine for so many years that maybe I've just become cynical of them.

I'd be glad to be proven wrong, it would be great for the consumer, but that is what makes me sometimes think it will never happen, markets rarely move to cater to us, always for shareholders...

 

[koyoot]

Actual new Intel roadmap. As you can see, no new products on 10/12 nm till end of 2020.

AMD will have both: architectural, and process advantage over Intel for 18 months!

 

[MBX]

Actual new Intel roadmap. As you can see, no new products on 10/12 nm till end of 2020.

AMD will have both: architectural, and process advantage over Intel for 18 months!

So any chance Apple switches to AMD?

 

[Naimfan]

So any chance Apple switches to AMD?

No.

Particularly if it is looking into switching to ARM.

 

[koyoot]

So any chance Apple switches to AMD?

Only Apple knows what this means for them.

No.

Particularly if it is looking into switching to ARM.

I am having hard time looking at idea of ARM and Professional workspace, and put a straight face on.

Unless Apple wants to make their computers useless toys, for kids - that is not going to happen.

 

[gaanee]

If Intel is struggling catching up to TSMC in the "nm" race then what's more important is they bring LPDDR4 and other improvements to 14nm architecture instead of waiting for 10nm chip in 2019-20. This will also result in better battery life.
With Whiskey Lake they have integrated wifi and I would like to see more such developments.
From consumer point of view these incremental improvements will also add more value than just "nm"

 

[koyoot]

If Intel is struggling catching up to TSMC in the "nm" race then what's more important is they bring LPDDR4 and other improvements to 14nm architecture instead of waiting for 10nm chip in 2019-20. This will also result in better battery life.
With Whiskey Lake they have integrated wifi and I would like to see more such developments.
From consumer point of view these incremental improvements will also add more value than just "nm"

Sure. But imagine that 7 nm chips can clock to 3.6 GHz BASE clock in 35-45W TDP with 8 core chips, with Higher IPC than Intel CPUs, have 3200 MHz IMC, etc. Even Intel 10 nm CPUs were supposed to clock higher, with higher IPC in lower thermal design power, and offer more cores in the same thermal envelope, than 14 nm CPUs, all at the same time. IceLake was supposed to be next generation of Intel CPUs in 100% of this meaning.

Every company on the Leading edge node overpromised, and underdelivered. Intel - 10 nm complete fiasco. GloFo - they stopped development of 7 nm process. TSMC - missed performance and power targets, hence why A12 Bionic on N7 process is only 100 MHz faster than A11 Bionic which was on 10 nm TSMC process. Samsung - their 7 nm process is more like 8 nm process, than anything else.

And those features you talk in the first place? They should've been in Intel CPUs a long time ago. Right now adding those features is just Intel desperately trying to get out this fiasco with a rest of its face.

 

[Ma2k5]

And those features you talk in the first place? They should've been in Intel CPUs a long time ago. Right now adding those features is just Intel desperately trying to get out this fiasco with a rest of its face.

Well...

a) Don't see anyone else offering anything similar now.
b) Those features are actually quite nice. Whether Intel are desparate or not, it is still beneficial to the consumer.

 

[gaanee]

Better be late than never....
With decreasing nm, the industry is slowly hitting the point of diminishing returns - it's more and more expensive using the current materials and processes.
Next phase of innovation will be a switch to a different material or in combination with silicon. They can't simply push the clock speed up with decreasing node size because heat and interference becomes a problem. That's why Intel is now adding more cores and not just advertising base clock speed.

And those features you talk in the first place? They should've been in Intel CPUs a long time ago. Right now adding those features is just Intel desperately trying to get out this fiasco with a rest of its face.

 

[koyoot]

Better be late than never....
With decreasing nm, the industry is slowly hitting the point of diminishing returns - it's more and more expensive using the current materials and processes.
Next phase of innovation will be a switch to a different material or in combination with silicon. They can't simply push the clock speed up with decreasing node size because heat and interference becomes a problem. That's why Intel is now adding more cores and not just advertising base clock speed.

Its funny, but the reason why Intel added more cores in Cannon and IceLake was because 10 nm process could not clock up as high as 14 nm process is clocking, with the IceLake architecture, and they had to from marketing point of view, give some sort of new SKUs for mainstream. ICL was higher IPC, but lower clock speed arch. It would Turbo to at best 4.2 GHz on 10 nm process(slightly higher than Skylake(6th gen). The other side of this is that the process allowed for higher base clocks in lower thermal envelopes. If 35W 6700T has had 2.9 GHz base, 3.7 GHz Boost clock, with 4C/8T, you could've expected that 4C/8T 35W TDP ICL CPU would clock 3.2/4.2 in the same thermal package, while also having higher IPC.

 

[Ploki]

Next MBP refresh will have at best 8 core Coffee Lake-R.

oh dang really? 6years no Core count jump now 2cores extra every year?
getting buyers remorse for 2018 if thats true, could've stuck with the 2012 for a year longer, although its significantly faster than the 2012.

 

[koyoot]

Well, first of all, TSMCs 7N process, on which Apple A12 Bionic is made has 69 mTR/mm2. GloFo 14 nm LP has 25 mTR/mm2, Intel's 14 nm process has 25 mTR/mm2, and TSMC's 7 nm HPC node on which AMD Matisse, Rome and Vega will be made, have 67 mTR/mm2.

I was quite a bit wrong here.

Apple A12 Bionic is 83 mm2. Which equals, with 6.9 Bln transistors to 83 mTR/mm2. Over 3 times higher density we have seen on 14/16 nm processes.

7 nm products will have insane density, and this paves way to not giving up hope on very high clocks on HPC 7 nm TSMC products, because you will have some room to burn those transistors on increasing the core clocks.

 

[gaanee]

For increasing the clock speed, max TDP is also one of the design parameters which couldis an important factor for notebooks.

I was quite a bit wrong here.

Apple A12 Bionic is 83 mm2. Which equals, with 6.9 Bln transistors to 83 mTR/mm2. Over 3 times higher density we have seen on 14/16 nm processes.

7 nm products will have insane density, and this paves way to not giving up hope on very high clocks on HPC 7 nm TSMC products, because you will have some room to burn those transistors on increasing the core clocks.

 

[koyoot]

For increasing the clock speed, max TDP is also one of the design parameters which couldis an important factor for notebooks.

What this means is that if you have 7 nm chips that can clock WITHOUT OC to 5 GHz, within 95W TDP, clocking them at 4 GHz base in 45W will be piece of cake in this scenario, which will be completely different level.

Im not that hopeful for this, however. I will be perfectly content if Zen2 will offer 2-5% higher IPC in games than Skylake/Kaby Lake/Coffee Lake, while be available in 45W TDP with 3.6 GHz base/4.2 GHz boost clock for 4C/8T CPU. That is all I care for desktop CPU.