The Intel Atom Thread

[cbn]

That's the issue they have. They want to keep those margins, and before the mobile explosion, the pricing was justified(at least somewhat) because they had that performance difference. Bottom end Intel chips were probably 10x faster than the fastest ARM based ones.

Now the justification for their pricing is because there's no alternative in the Windows space and the only higher performing chips are pricey Core ones. People didn't have that thinking when Medfield was out. They were expecting that they'd be #1 in mobile space or at least #2 and performance improvement of ARM competition would level out short time after. Then Apple came with A7, forcing everyone to admit that Intel would have zero chance going into Apple devices. They'd reason "oh they'll be the next Qualcomm". And they'd say it doesn't matter Apple chips are better since "its not direct competition".

Now, non-Apple top end ARM chip performs almost as well as the Core M chip(if not there already depending on how do you view Intel C compilers), with cost, battery life, and form factor of an Atom chip/device

I can't help but admit that Intel is in a serious quandry here. They want to satisfy shareholders and keep those massive revenues/margins, they want to be dominant in mobile, yet they can't do much about it. Ideally Atom would perform as well as Apple A9 chips(or current Core M), and regular Core would be still 2x above that. But you run into "real world" issues like you are pointing out how unrealistic it is expecting such big cores to happen. They need a "big core" chip not just because they want to improve PC sales, but to keep the pricing/margins and allow lower cores like Atom to grow so they can be competitive in mobile(I know its not just CPU, but the entire SoC is behind).

Whether they aren't capable of doing it due to valid reasons like not being able to make cores that perform another level above Core, or they don't want to do it due to financial reasons, they are in big trouble. Technical troubles will eventually manifest themselves financially. Perhaps their time is past though, just like disk/cassette rental places.

One advantage I see x86 having over ARM (at this time) is standardization.

http://forums.anandtech.com/showpost.php?p=38144160&postcount=21

Nice thing about x86 is that (unlike ARM) it is a standardized platform.

This means, unlike ARM, we should be able to update our OSes like we do our PCs. This should make for a much more secure system for those keeping their phones long term.

Its too bad most consumers (especially non enthusiast ones) wouldn't even realize this when factoring in their buying decision....and yet it is so important.

This, in contrast, to the typical Android phone that doesn't get updates.

P.S. Beyond Android, I actually would like to see Ubuntu used on a low end x86 phone. Its a great versatile OS that we can update ourselves. It can also dock and be used as a desktop.

 

[coercitiv]

One advantage I see x86 having over ARM (at this time) is standardization.

This means, unlike ARM, we should be able to update our OSes like we do our PCs. This should make for a much more secure system for those keeping their phones long term.

Its too bad most consumers (especially non enthusiast ones) wouldn't even realize this when factoring in their buying decision....and yet it is so important.

So, are they updated? Are Clover Trail based products still receiving OS / security updates?

 

[cbn]

So, are they updated? Are Clover Trail based products still receiving OS / security updates?

Clover Trail is for Windows so yes that does get updated, maybe you mean Clovertrail+ for Android?

P.S. With Ubuntu (for x86) we should just be able to do the update ourselves....and even use something like this for docking:

http://www.tomshardware.com/news/nexdock-dumb-laptop,31613.html

https://www.youtube.com/watch?time_continue=21&v=ZVIhtbfLfpk

 

[coercitiv]

Not sure, but with Ubuntu we should just be able to do the update ourselves...

What does this have to do with the average customer? Are OEMs selling x86 Ubuntu phones nowadays?

 

[cbn]

What does this have to do with the average customer?

Security, for one thing.

Are OEMs selling x86 Ubuntu phones nowadays?

No, but maybe Intel will have the right SoC to make this happen soon.

 

[Shivansps]

Clover Trail is for Windows so yes that does get updated, maybe you mean Clovertrail+ for Android?

P.S. With Ubuntu (for x86) we should just be able to do the update ourselves....and even use something like this for docking:

http://www.tomshardware.com/news/nexdock-dumb-laptop,31613.html

https://www.youtube.com/watch?time_continue=21&v=ZVIhtbfLfpk

Clover Trail IGP does not work well under W10, it never got a proper driver W7, never got a driver updated for W8 and it does just dont work under W10, Windows Update warms you NOT to update to Windows 10 if you have a Atom CloverTrail, and they never even released a X64 driver...

The olders Atom PineTrail with GMA 3150 and DiamondVille with GMA950 can run W10 32/64B no problem.

 

[IntelUser2000]

Second, Apple is selling you similarly priced ARM based products with the best and most expensive PC equivalents. The iPad Pro is asking for the same $$$ as the Surface Pro 4. If Intel's perf/dollar ratio is bad, who has the better one?

Mind you Intel is a chip maker. A dedicated one at that.

The fact that iOS/Android doesn't have productivity apps is just a cover that doesn't allow Intel's full vulnerabilities to be shown. They've got no technological advantage(process, architecture, wireless, power management) which is why they can't compete at all in mobile.

The regular iPad does the same thing as the Pro at prices that significantly undercut both. If you want to compare using products that show one side at the worst light, perhaps you should look at the product that's the focus of this thread.

 

[jhu]

Security, for one thing.



No, but maybe Intel will have the right SoC to make this happen soon.

You're implying Ubuntu phone is going to be basically a desktop computer in phone form-factor. I highly doubt that, especially with so many binary blobs that need to be taken care of. I think it will be closer to current Android x86 phones instead.

Current Ubuntu phones are on ARM, and they have the same limitations that Android phones have.

 

[coercitiv]

Mind you Intel is a chip maker. A dedicated one at that.

And still we can only evaluate their work by looking at specific implementations, even if we choose the best ones. Moreover, it was this chip maker who dictated the ultrabook form factor, through clever market mechanics.

The fact that iOS/Android doesn't have productivity apps is just a cover that doesn't allow Intel's full vulnerabilities to be shown.

Agreed. See bellow.

The regular iPad does the same thing as the Pro at prices that significantly undercut both. If you want to compare using products that show one side at the worst light, perhaps you should look at the product that's the focus of this thread.

I don't think I relayed my point of view clearly enough: the Atom as it presents to us today is done for, I don't see it as a viable product for the future. Last year we were talking about x86 trying to get into mobile, next year we'll be talking about ARM trying to get into traditional PC market.

The only alternative I see today for Intel when it comes to passively cooled devices is a new breed of Atom built from high performance arch DNA: modified cores in a true SOC (chipset included) with just enough I/O capabilities to make it work in tablets / hybrids / thin products.

I won't insist on this any further, since I understand it's borderline OT, but I believe we'll discuss the subject again in 1-2 years time, if only to prove me wrong

 

[Sweepr]

Broxton Pro (ES?) SiSoftware Scores

- Processor Arithmetic
Broxton: 29.88GOPS
Atom x7-Z8700 (average): 22.79GOPS
Core m3-6Y30 (average): 29.31GOPS

- Processor Multi-Media
Broxton: 54.84Mpix/s
Atom x7-Z8700 (average): 31.47Mpix/s
Core m3-6Y30 (average): 81.65Mpix/s

- Processor Cryptography (High Security)
Broxton: 1.83GB/s
Atom x7-Z8700 (average): 0.80GB/s
Core m3-6Y30 (average): 2.07GB/s

http://ranker.sisoftware.net/show_s...e3deeec8a09dad8bf3cefed8bdd8e5d5f380bd85&l=en

Being recognized as a Pentium II, just like Geekbench. 4C/4T 1.19GHz 2x 1MB L2.

 

[mikk]

Processor Multi-Media is quite a big difference.

 

[Exophase]

This is why the contra-revenue was needed with Bay Trail in tablets...lots of corners that they needed to cut in the platform cost that they hadn't done. The mobile guys had it figured out a while ago, out of necessity

It's interesting to see all of the places where the contra-revenue was needed. I only really knew about the need for a discrete BIOS chip, which was an embarrassment. But AFAIK most or all of these are features ARM SoCs have had for years. I know they don't USB hubs or level shifters for 3.3V I/O anymore.

 

[Exophase]

Well it's not bad for sure, against one of the best x7-Z8700 results:
http://browser.primatelabs.com/geekbench3/compare/2517688?baseline=6034406

For fun, against a Kabylake at similar clockspeed:
https://browser.primatelabs.com/geekbench3/compare/6044389?baseline=6034406

Looks like SHA is finally implemented in hardware (7x score...), AES and many other floating point subtest are up 1.5-2x too.

Memory is awful but that is 100% single channel and low speed, LP-DDR4 should help there

The "core" jump of the atom line?

Eh. Yes, they've got a huge jump in SHA and AES. Who cares. This is why the GB3 aggregate scores are useless and you should always look at the individual ones on their individual merit (although none of them are really that good)

Comparison with x7-Z8700:

https://browser.primatelabs.com/geekbench3/compare/5964399?baseline=6034406

There's a respectable ~25% IPC improvement in the best integer cases, but in a lot of other cases it's nowhere near that.

The FP scores are less interesting, both because they're not really great tests and because FP performance in general isn't that critical on these platforms. But I would say they could have gotten these improvements by moving to proper OoOE for the SIMD/FP side of the core. On some of them it's pretty flat again.

 

[mikk]

Eh. Yes, they've got a huge jump in SHA and AES. Who cares. This is why the GB3 aggregate scores are useless and you should always look at the individual ones on their individual merit (although none of them are really that good)

Comparison with x7-Z8700:

https://browser.primatelabs.com/geekbench3/compare/5964399?baseline=6034406

There's a respectable ~25% IPC improvement in the best integer cases, but in a lot of other cases it's nowhere near that.

Another 32 bit vs 64 bit comparison.

 

[Exophase]

Another 32 bit vs 64 bit comparison.

Pretty sure they're both 64-bit...

 

[mikk]

You are wrong.

https://browser.primatelabs.com/geekbench3/5964399
https://browser.primatelabs.com/geekbench3/6034406

Geekbench 3.4.1 Tryout for Windows x86 (32-bit)
Geekbench 3.4.1 for Windows x86 (64-bit)

 

[Shivansps]

Remember that Z8700 its the top sku, or at least it was until CT refresh, and we cant be sure what this ES sample is representing, 2.3Ghz is kinda low, Z8700 can turbo up to 2.4, and even Z8550 is now 2.4Ghz, its really bad if they regress freq on a 14nm refresh compared to the first 14nm product.

As a Z8300 replacement is very good actually, the awful single channel ram kinda support this, compare it to this one https://browser.primatelabs.com/geekbench3/6104069

 

[Exophase]

You are wrong.

https://browser.primatelabs.com/geekbench3/5964399
https://browser.primatelabs.com/geekbench3/6034406

Geekbench 3.4.1 Tryout for Windows x86 (32-bit)
Geekbench 3.4.1 for Windows x86 (64-bit)

Oh I see, I didn't even see that part, sorry >_> I just looked at the OS version and assumed that referred to binary type. When you have it in compare mode it doesn't even list the binary type, it seems...

Here's a proper 64-bit one, I hope:

https://browser.primatelabs.com/geekbench3/5964263

It's pretty close.

As a Z8300 replacement is very good actually, the awful single channel ram kinda support this, compare it to this one https://browser.primatelabs.com/geekbench3/6104069

But that one is max 1.84GHz.. I really don't think single channel RAM makes much of a difference here outside of the memory scores and maybe some of the FP scores.

 

[VengenceIsMine]

Remember that Z8700 its the top sku, or at least it was until CT refresh, and we cant be sure what this ES sample is representing, 2.3Ghz is kinda low, Z8700 can turbo up to 2.4, and even Z8550 is now 2.4Ghz, its really bad if they regress freq on a 14nm refresh compared to the first 14nm product.

As a Z8300 replacement is very good actually, the awful single channel ram kinda support this, compare it to this one https://browser.primatelabs.com/geekbench3/6104069

A lot depends on which end of the product range this is and how fully baked that ES & platform is. If it's the high end 8700 equivalent, it's a disaster for Intel and the PC business in general. If it's the low end or a not fully cooked mid range part with crappy ram then it's pretty respectable, it can compete with ARM big.LITTLE parts based on a57/a72 reasonably well and presumably a high end part with decent ram can at least be remotely close to lower end Core-M parts CPU performance for a much cheaper platform cost and power envelope.

 

[Shivansps]

But that one is max 1.84GHz.. I really don't think single channel RAM makes much of a difference here outside of the memory scores and maybe some of the FP scores.

But even the mid range X5-Z8550 is right now 2.4Ghz turbo, if we are talking about a 2.3Ghz turbo sku here (no way in hell for a 2.3Ghz base), its more likely a Z8300/Z8350 replacement, and the single channel ram seems to back that up. If Intel regress frecuency on a 14nm refresh compared to its first 14nm product this is really, really bad, so im hoping that 2.3Ghz its not the top end sku.

Btw, Z8750 is 2.56Ghz turbo, thats the top sku right now.

And if we are talking about Braswell its even worse, unless we are talking about Celeron N3160 replacement.

 

[Exophase]

But even the mid range X5-Z8550 is right now 2.4Ghz turbo, if we are talking about a 2.3Ghz turbo sku here (no way in hell for a 2.3Ghz base), its more likely a Z8300/Z8350 replacement, and the single channel ram seems to back that up. If Intel regress frecuency on a 14nm refresh compared to its first 14nm product this is really, really bad, so im hoping that 2.3Ghz its not the top end sku.

Btw, Z8750 is 2.56Ghz turbo, thats the top sku right now.

And if we are talking about Braswell its even worse, unless we are talking about Celeron N3160 replacement.

We're looking at an engineering sample here. They're not supposed to represent some particular SKU. Its clock speed could very well be in between what real products will offer.

 

[Nothingness]

Here's a proper 64-bit one, I hope:

https://browser.primatelabs.com/geekbench3/5964263

What about this proper 32-bit vs 32-bit? https://browser.primatelabs.com/geekbench3/compare/6034406?baseline=5964335

 

[Arachnotronic]

What about this proper 32-bit vs 32-bit? https://browser.primatelabs.com/geekbench3/compare/6034406?baseline=5964335

Goldmont seems to have brought a very good increase in integer performance, but FP performance is still garbage.

Intel has yet to design even an A72 class mobile core.

 

[Exophase]

What about this proper 32-bit vs 32-bit? https://browser.primatelabs.com/geekbench3/compare/6034406?baseline=5964335

Ugh sorry, I'm having the hardest time dealing with these GB results.. I really loathe their page.. okay yeah, I see now that it was the Broxton one that was 32-bits all along :| Bleh, sorry again...

I think FP takes a dive on some 32-bit tests because of being so register constrained, but the integer ones are pretty close?

 

[Nothingness]

Isn't that back to 64-bit vs 32-bit again?

No

EDIT: you edited. No there's no big difference.

@Arachnotronic: it's unimpressive on all fronts, most of the integer increase comes from cryptographic benches, isn't it ironic?