Intel deprived us of an exciting and entertaining battle (Ice Lake vs A12X)

[willfr]

The A12X is pretty exciting, but I badly want to know how it stacks up against Ice Lake, and we should be able to do that by now. The Skylake architecture is so old that it's not even fair to compare it to the A12X. Let's hope Intel can ship Ice Lake soon...

 

[happy medium]

Amber lake seems to be a good chip to compare to the Apple A12.

I think there is a one that turbo s to 4.2ghz

 

[Nothingness]

Here is CINT 2006 on a Kaby Lake R i7-8650U gcc 7.3.0 -O3 -march=native

                8650u    A12    A12/8650u
400.perlbench   -        45.38  -
401.bzip2       30.2     28.54  0.95
403.gcc         43       44.56  1.04
429.mcf         33.3     49.92  1.50
445.gobmk       31.5     38.54  1.22
456.hmmer       37.3     43.24  1.16
458.sjeng       32.9     27.97  0.85
462.libquantum  95.5    113.40  1.19
464.h264ref     69.6     66.59  0.96
471.omnetpp     20.5     35.73  1.74
473.astar       24.4     27.25  1.12
483.xalancbmk   47.3     57.03  1.21

400.perlbench failed to compile and I have no time to investigate.

A12 definitely rocks!

 

[name99]

The A12X is pretty exciting, but I badly want to know how it stacks up against Ice Lake, and we should be able to do that by now. The Skylake architecture is so old that it's not even fair to compare it to the A12X. Let's hope Intel can ship Ice Lake soon...

The one performance point we have for IceLake is
https://browser.geekbench.com/v4/cpu/10445533
and this is distincly unimpressive. Ice Lake looks like exactly what you would expect from the name --- a slightly tweaked Skylake targeting a new process, NOT a new micro-architecture.

Now it's possible that this GB4 submission is fake or otherwise very misleading. (God knows there are plenty of such GB4 submissions.) But I think the naming makes it clear that Ice Lake was ALWAYS meant to be nothing more than a slight tweak, no different from Sandy Bridge->Ivy Bridge or Haswell->Broadwell. Pinning your hopes on it as some great leap forward seems likely to end in major disappointment...

 

[mikk]

But I think the naming makes it clear that Ice Lake was ALWAYS meant to be nothing more than a slight tweak, no different from Sandy Bridge->Ivy Bridge or Haswell->Broadwell. Pinning your hopes on it as some great leap forward seems likely to end in major disappointment...

Not sure if the code name is a good indicator for anything, nobody predicted such a huge IPC leap from Goldmont to Goldmont+ based on the naming scheme.

 

[willfr]

The one performance point we have for IceLake is
https://browser.geekbench.com/v4/cpu/10445533
and this is distincly unimpressive. Ice Lake looks like exactly what you would expect from the name --- a slightly tweaked Skylake targeting a new process, NOT a new micro-architecture.

Now it's possible that this GB4 submission is fake or otherwise very misleading. (God knows there are plenty of such GB4 submissions.) But I think the naming makes it clear that Ice Lake was ALWAYS meant to be nothing more than a slight tweak, no different from Sandy Bridge->Ivy Bridge or Haswell->Broadwell. Pinning your hopes on it as some great leap forward seems likely to end in major disappointment...

I was under the impression the Ice Lake is the first new architecture since Skylake. Cannon Lake was supposed to be the 10nm die shrink of Skylake, no?

 

[Arachnotronic]

Here is CINT 2006 on a Kaby Lake R i7-8650U gcc 7.3.0 -O3 -march=native

                8650u    A12    A12/8650u
400.perlbench   -        45.38  -
401.bzip2       30.2     28.54  0.95
403.gcc         43       44.56  1.04
429.mcf         33.3     49.92  1.50
445.gobmk       31.5     38.54  1.22
456.hmmer       37.3     43.24  1.16
458.sjeng       32.9     27.97  0.85
462.libquantum  95.5    113.40  1.19
464.h264ref     69.6     66.59  0.96
471.omnetpp     20.5     35.73  1.74
473.astar       24.4     27.25  1.12
483.xalancbmk   47.3     57.03  1.21

400.perlbench failed to compile and I have no time to investigate.

A12 definitely rocks!

Impressive...most impressive. Can we put to bed the meme that Apple CPUs are "only good in mobile phones"? The A12 CPU cores are excellent!

 

[Nothingness]

Impressive...most impressive. Can we put to bed the meme that Apple CPUs are "only good in mobile phones"? The A12 CPU cores are excellent!

Looking at how people here try to dismiss every benchmark thrown at them, I think that's premature.

 

[Thunder 57]

I was under the impression the Ice Lake is the first new architecture since Skylake. Cannon Lake was supposed to be the 10nm die shrink of Skylake, no?

That was my impression as well. Granted that was back when Intel was still at least somewhat on track. My guess now is that Cannon Lake only appears in a few (at best) low power mobile chips, and Ice Lake is the next true Tick. I think it will be on some kind of revised 10nm. Whether it is 10nm+, or if SemiAccurate would love, more of a "10nm-" who knows. There may be a few percent IPC like there was with Sandy to Ivy or Haswell to Broadwell, but nothing much. I think that whatever comes after that will be the next "Tock" or "Architecture", or whatever Intel calls them these days.

That's all just a wild ass guess though.

 

[rbk123]

Looking at how people here try to dismiss every benchmark thrown at them, I think that's premature.

Only those that don't suit their agenda. The ones that do, they wholeheartedly embrace. And that's true for both/all sides.

 

[Carfax83]

Perhaps Zen 2 can step up?

 

[.vodka]

Impressive...most impressive. Can we put to bed the meme that Apple CPUs are "only good in mobile phones"? The A12 CPU cores are excellent!

Excellent is an understatement. It's ridiculous to have such hardware inside a phone.

Now I wonder how fast the A12/X would run on a proper motherboard with a proper heatsink, past its limitations to run in a phone/tablet... Probably not much more for frequency, but sustained performance at high clocks for what it is could make some heads turn at the results.

Seeing these resuts, I guess it's about time for the x86 -> ARM transition to happen at Apple, if not at the next node available to them. They could pull off PPC -> x86 with a translation layer, so...

 

[Nothingness]

Seeing these resuts, I guess it's about time for the x86 -> ARM transition to happen at Apple, if not at the next node available to them. They could pull off PPC -> x86 with a translation layer, so...

That transition could be done because back then x86 was much faster than PowerPC. Now A12, no matter how good it is, is not fast enough to run a Rosetta like layer without seriously impacting user experience.

FWIW 32-bit x86 translated to ARM code by Microsoft results in about half of speed compared to native. And their translated code looks very good. That gives an idea of how much faster Apple chips would need to be to support translation.

 

[IntelUser2000]

I don't think Intel will be able to make big of an impact as with a turnaround architecture like "Conroe" like they did against last time.

AMD vs Intel seems to be repeating again, but unlike nearly two decades ago, we have Apple. Conroe/Merom would have been far less impressive if Apple executed like today back then. It's like if everyone was competing in the desktop space and Apple had a 25W laptop class chip that performed like top of the line Pentium D and Athlon X2 chips that used 4x the power.

Big PC gaming companies like Blizzard are moving into the Smartphone space. Whether forced or happening naturally, things are really moving towards the mobile space.

 

[DrMrLordX]

Maybe we won't have IceLake vs. A12x. But now we can have Zen2 vs. A12x! Go get your lawnchairs and popcorn kids, this fight could get interesting.

 

[Entropyq3]

That transition could be done because back then x86 was much faster than PowerPC. Now A12, no matter how good it is, is not fast enough to run a Rosetta like layer without seriously impacting user experience.

FWIW 32-bit x86 translated to ARM code by Microsoft results in about half of speed compared to native. And their translated code looks very good. That gives an idea of how much faster Apple chips would need to be to support translation.

The user experience would probably be quite good actually. Apple has laid a good foundation for another architectural transition. They have promoted the use of bit code since early/mid 2015, which means that most apps compiled since then would run natively on the new devices. And of course all work that is achieved by calling system resources would also run natively.
The scenario you envision only applies to those running Windows and Windows applications on their Macs, and only when they do specifically that. I don’t know how many there are who are running absolutely performance critical Windows apps on Macs, but I can’t for the life of me believe there are many (any?) of them.