The Intel Atom Thread

[Shivansps]

Just tested the GRID 2 on my E-350, on the first race (what i belive is what Anandtech gus tested) at 1366x768 with Ultra low details it gets 13 to 20 fps, after that connected an HDMI monitor to check CPU usage and it spikes to 100% way too often.

Not sure what level of details was used by Anadtech, not sure what "igp settings" even means.

 

[piesquared]

People play games for the Hardware they have. You dont expect people to play Crysis 3 on Tablets do you ?? But if you have a faster GPU you will play more 3D games with higher IQ characteristics. And here comes Temash, you will be able to play more games or with higher Image Quality than BayTrail.






From AT preview, its clear that BayTrail has half the Performance of HD3000 at 15W in real Games. That is, it will almost have the performance of HD2000 at 15W.
3D Mark 2013 (IceStorm) must be the best case in GPU performance for Baytrail. I believe the driver is 3D Mark optimized. BayTrail scores 13850 in IceStorm, when two year old AMD Z-60 scores close to 12000. Im expecting Temash to be close to ~20000 in the same Benchmark.
In real games, that drivers play more important role in performance, im expecting a 3.9W A4-1200 Temash to be on average close to ~70-80% faster than BayTrail Z3770. Ahh and thats for DX-9, i dont believe BayTrail can even run DX-11 games(drivers).

They can't and it's undoubtedly the reason Futuremark included a DX10 version of the benchmark. The Radeon architecture is much more advanced. Nothing against power vr, their products are intended for a different market.

 

[Abwx]

I gave the numbers before but perhaps you didnt see them,

AMD Z-60 is a dual core 1GHz 4.5W TDP (+ 1-2W from the PCH) 80 Radeon cores(VLIW5) at 276MHz and scores 12199 on Ice Storm.

Temash A4-1200 is a dual core 1GHz 3.9W TDP (SoC) 128 GCN Radeon Cores at 225MHz, it will score higher than BayTrail Z3770(13850).

But, real games is a different thing than Ice Storm.

FWIW a 4C temash at 1Ghz has 35% better 3Dmark06 perf
than a 1.7ghz 2C bobcat , so it s likely that a 1GHz dual core
Temash has still better GPU perf than a 1.6-1.7g bobcat.

http://www.notebookcheck.biz/Courte-critique-du-APU-AMD-A4-5000-Kabini.95199.0.html

 

[blackened23]

So we're still at the point of taking "guesstimates" of A4-1200's performance. Extrapolating. Good stuff. Where are the real tablets and where are the real benchmarks? I'll pay attention when someone has them. I think it's a fair guess that the Bay Trail T will have less power consumption while still having better CPU and GPU performance.

 

[Sweepr]

Just tested the GRID 2 on my E-350, on the first race (what i belive is what Anandtech gus tested) at 1366x768 with Ultra low details it gets 13 to 20 fps, after that connected an HDMI monitor to check CPU usage and it spikes to 100% way too often.

Not sure what level of details was used by Anadtech, but sure what "igp settings" even means.

Yup, seems like you are completely CPU limited here. It would just get worse with a slow 1GHz dual-core Jaguar.

 

[Sweepr]

To be quite honest i'm waiting for better tablets as well. I think the chip has the potential to be used to great effect, so if there's a tablet with a 1920x1080 screen with full Windows 8.1, i'm game for 300-350$. I'd buy it.

Dell Venue will have a 1080p screen, not sure about its price though:

~$350-400 is my guess.

 

[piesquared]

Yup, seems like you are completely CPU limited here. It would just get worse with a slow 1GHz dual-core Jaguar.

What's that got to do with it? It's the 1GHz quad-core[/q] Jaguar's with GCN cores at 3W SDP you should be comparing to.

 

[AtenRa]

Of yourse it can. Gen7 is DX11.

The hardware may be DX-11, but as always with Intel, there is no DX-11 driver yet. Thats why you only see Ice Storm which is DX-9, why Anand or anyone else didnt measured Cloud Gate (DX-10) and Fire Strike (DX-11) on BayTrail ??? :whiste:

 

[VengenceIsMine]

Please tell me how a sky high margins company like Intel is going to battle these almost zero ones. Not to talk about bigger players that are already on razor thin margins on their CPUs. And all that not having in mind the sick amount of cash all of them have while Intel is dealing with one bad quarter to one worse.

Intel is the desperate one here, not the other way around.

EDIT: And chinese makers like Rockchip have been dumping 28nm into the market for a while now.

I don't know about Rockchip but I know people at Allwinner and they aren't close to profitable, they are just burning through low interest loans from state banks at this point to build market share. Which is fine, it's a normal way to build a business but the problem with that business plan is that is that some point you have to turn the corner and they don't really have any advantages over any other bottom of the barrel ARM provider, same IP, same fab as everyone else, it is very hard to build a sustainable business that way. MediaTek would be a better example of a company that is building a sustainable business and moving up the food chain somewhat.

Intel's situation is very, very different. They are moving into a market they haven't taken seriously and have been failing at, no denying that. However, pretending that Bay Trail doesn't put them in the game with class leading CPU performance & acceptable GPU performance and a very aggressive looking & achievable roadmap does change the game from it's present ARM domination is willfully ignoring reality. Firstly, the PC OEM's have an acceptable x86 tablet platform which will move significant volume, check out the comments on AnandTech and The Verge for the T1000 if you need to gauge interest in a baytrail Windows device & there is huge pent up demand from Enterprise IT for a tablet they can live with and Android hasn't done much to address manageability and Apple is too expensive and has all the same issues. That won't move ipad like volume but no Android device is moving close to ipad like volume either, not Samsung or even Nexus 7. That will give Intel something to build on in the Android market as well since most of the PC OEM's are producing Android tablets as well, they will certainly look at BayTrail as a possible Android platform. 2ndly Intel's perf is at the bottom of where it will be with Android right now, expect it to significantly improve in 3-4 months as the software optimization takes place. 3rd, Intel's roadmap for the next 12 months is pretty achievable: die shrink (which intel is by far the best at moving to on time), moving to the better Haswell GPU cores, & MHz opened up by die shrink. Intel's going to get more of a MHz gain out of their move to 14nm than TSMC will get out of 20nm planar, the leakage issues are just too nasty at 20nm planar, they will either have to sacrifice power efficiency or not gain as much MHz which is an ugly choice in mobile. The only real iffy thing on Intel's roadmap for their next 12 months is 450mm wafers (& it is even more iffy for TSMC since they started 12 months later), if they do pull that off, Atom's are going to be very cheap or Intel to make. LTE is the other big question mark but obviously that is a minor consideration in tablets for right now. There isn't an equivalent roadmap coming out of ARM right now, sure 64 bit but that adds significant die area & storage cost for a minimal return on performance in the mobile world. 20nm TSMC planar will be a smaller improvement than Intel's die shrink and TSMC has a long history of slow ramp ups of new process tech. big.LITTLE not only increases ARM's cut but it seriously increases die area for what are at this point very unproven performance gains though obviously it gives you a great way to market to idiots. When you evaluate this market you have to look at the future from the manufacturer's point of view for the next year or two, not just the consumer's today point of view to see the trends.

 

[Sweepr]

What's that got to do with it? It's the 1GHz quad-core[/q] Jaguar's with GCN cores at 3W SDP you should be comparing to.

We have already seem Z3770 outperform 15W TDP A4 5000 1.5GHz quad with <2.5W power comsumption under CPU load (@ Intel reference platform).

The chip you are talking about took a hit from 1.5GHz to 1GHz (its much slower than the A4 5000) to reach the 8W TDP mark and until someone properly tests it (with a reference platform or final product) the 3W number is pure marketing just talk like Intel's SDP numbers pre Bay Trail launch. Also, wake me up when someone launches an 8'' fanless <9mm <400g tablet based on this 8W TDP chip.

 

[mikk]

I gave the numbers before but perhaps you didnt see them,

AMD Z-60 is a dual core 1GHz 4.5W TDP (+ 1-2W from the PCH) 80 Radeon cores(VLIW5) at 276MHz and scores 12199 on Ice Storm.

Temash A4-1200 is a dual core 1GHz 3.9W TDP (SoC) 128 GCN Radeon Cores at 225MHz, it will score higher than BayTrail Z3770(13850).

But, real games is a different thing than Ice Storm.

A4-1200 scores ~12K in Ice Storm. Z-60 isn't a relevant model anymore.

They can't

Nonsense. Gen7 is fully dx11.

The hardware may be DX-11, but as always with Intel, there is no DX-11 driver yet. Thats why you only see Ice Storm which is DX-9, why Anand or anyone else didnt measured Cloud Gate (DX-10) and Fire Strike (DX-11) on BayTrail ??? :whiste:

Once again nonsense. There is no problem with dx11. You are clueless. Drivers for Gen7 including DX11 are available since the launch of Ivy Bridge.

 

[piesquared]

We have already seem Z3770 outperform 15W TDP A4 5000 1.5GHz quad with <2.5W power comsumption under CPU load (@ Intel reference platform).

The chip you are talking about took a hit from 1.5GHz to 1GHz (its much slower than the A4 5000) to reach the 8W TDP mark and until someone properly tests it (with a reference platform or final product) the 3W number is pure marketing just talk like Intel's SDP numbers pre Bay Trail launch. Also, wake me up when someone launches an 8'' fanless <9mm <400g tablet based on this 8W TDP chip.

Software plays a major part in a devices usability. Wake me up when intel's GPU is anywhere near as capable as GCN. I wonder how BT will do on Adobe's new products. There are a lot of people out there who will benefit from being able to experience creative work on their mobile devices with Temash. Did you miss that the industry is embracing heterogeneous computing?

 

[Shivansps]

Ivy Bridge also has OpenCL support, and it performs really well for what they are.

 

[Imouto]

@VengenceIsMine

Well, two different points of view. I don't think that this chip is going anywhere while Intel is holding hands with Microsoft. Bay Trail based Android tablets will have to wait since they're not even close to be ready.

I don't think that Intel will do anything meaningful in the mobile space, it's too late for that. Just thinking about how Qualcomm has to update its old Krait cores while they're still kicking ass doesn't bode well for Intel... or anyone else for that matter except Samsung.

It's going to be fun checking quarterly reports from everyone.

 

[Herr Kutz]

Dell Venue will have a 1080p screen, not sure about its price though:

~$350-400 is my guess.

Is that aero glass? Is MSFT reintroducing it in v8.1?

 

[krumme]

@VengenceIsMine

Well, two different points of view. I don't think that this chip is going anywhere while Intel is holding hands with Microsoft. Bay Trail based Android tablets will have to wait since they're not even close to be ready.

I don't think that Intel will do anything meaningful in the mobile space, it's too late for that. Just thinking about how Qualcomm has to update its old Krait cores while they're still kicking ass doesn't bode well for Intel... or anyone else for that matter except Samsung.

It's going to be fun checking quarterly reports from everyone.

Its going to take some time - if at all - for Intel to have profit on tablets and phones. Fair enough thats how business always is. As windows is kind of dead for that market what becomes intereting is how Intel can devellop the entire soc and adapt to android software. Proof will be there when the top players use it for the highend android or ios devices. Intel needs the margins from that market to continue because of its cost structure. With Anands words it needs to be "classleading" product all over and that means connectivity, dsp, multimedia...everything. And even then its going to one of the toughest sell ever to Apple and Samsung.

All the same time Qualcomm is executing its roadmap. And that probably means some bad ass eg. dsp, camera and connectivity solutions.

Intel needs to move now. When 16nm tscm process comes online, a lot of their huge process lead for performance will end and then much of their leverage disapear.

I dont know why Intel enters the phone market. I cant see the profit. But for us consumers more compettition is always great. So we win anyway while shareholders lose. Fine for me

 

[mrmt]

Intel needs to move now. When 16nm tscm process comes online, a lot of their huge process lead for performance will end and then much of their leverage disapear.

Don't expect much from TSMC 16nm process. Not only it will have the same die sizes of 20nm nodes, it will spend a significant part of its useful life competing against intel 10nm.

 

[Broheim]

it's funny to think not so long ago people were claiming that intel would never be able to take the fight to arm, yet here we are...

 

[liahos1]

it's funny to think not so long ago people were claiming that intel would never be able to take the fight to arm, yet here we are...

totally agree man.:thumbsup:

 

[NostaSeronx]

Don't expect much from TSMC 16nm process. Not only it will have the same die sizes of 20nm nodes, it will spend a significant part of its useful life competing against intel 10nm.

TSMC 16nm FinFET isn't expected to get any customers. TSMC is skipping right to 10nm FinFETs for 2015.

The big three(TSMC, GlobalFoundries, Intel) will all be 10nm FinFET with Quad Patterning.

 

[mikk]

TSMC 16nm FinFET isn't expected to get any customers. TSMC is skipping right to 10nm FinFETs for 2015.

The big three(TSMC, GlobalFoundries, Intel) will all be 10nm FinFET with Quad Patterning.

Interesting. Do you have a source?

 

[Exophase]

http://techon.nikkeibp.co.jp/english/NEWS_EN/20130702/290691/

First products using 20nm will be in 2014. Trial production of 16nm starts in 2014. Trial production of 10nm starts in 2015 and continues into 2016. Does not sound like them skipping 16nm and getting no customers, and certainly doesn't sound like it'll be available in products in 2015. I doubt there will be 16nm TSMC products before 2015 or 10nm products before 2017.

 

[Khato]

Given the discussion regarding whether or not Baytrail's graphics are adequate I figured that I'd quote Sweepr's post from earlier in the thread as it would certainly seem that this subjective observation had been missed.

We tried the Android version of Grand Theft Auto III at 2560 x 1440 resolution with maxed out settings, and it ran smooth as butter on the new Atom chip — smoother than the Tegra 4 inside Nvidia's Shield gaming handheld. Bay Trail won't be your first choice for Windows gaming, but it's already impressive in mobile.

http://www.theverge.com/2013/9/11/4...tablet-processors-heres-what-you-need-to-know

Sure, AMD's Kabini would likely perform just as well... But keep in mind that according to the Tech Report's coverage - http://techreport.com/review/25329/intel-atom-z3000-bay-trail-soc-revealed/4 - the Baytrail SoC was only using ~1.4W while gaming.

 

[Imouto]

it's funny to think not so long ago people were claiming that intel would never be able to take the fight to arm, yet here we are...

They're not fighting a thing. They took the wrong road several years ago and they're still going through the wild to reach the runners that took the right one on time. Once they're on track they'll still be tired and scratched having to outrun fresh and loaded runners.

Qualcomm has a huge lead here don't let it fool you. The've been just managing the huge success with the old Krait 400 cores till now. Qualcomm doesn't need to update its desings for Bay Trail since it will be a low volume chip tied to Microsoft. Once Intel releases anything threatening that runs on Android it will just smash everything else again. Just like that.

Meanwhile Samsung is confident about its fab prowess to beat Qualcomm (and anything else in its way) in a few years making buckets of money out of anything that dares to move meanwhile.

I just don't see how AMD, Nvidia or Intel have any chance here with MS in the same bag. They're locked out.

 

[liahos1]

They're not fighting a thing. They took the wrong road several years ago and they're still going through the wild to reach the runners that took the right one on time. Once they're on track they'll still be tired and scratched having to outrun fresh and loaded runners.

Qualcomm has a huge lead here don't let it fool you. The've been just managing the huge success with the old Krait 400 cores till now. Qualcomm doesn't need to update its desings for Bay Trail since it will be a low volume chip tied to Microsoft. Once Intel releases anything threatening that runs on Android it will just smash everything else again. Just like that.

Meanwhile Samsung is confident about its fab prowess to beat Qualcomm (and anything else in its way) in a few years making buckets of money out of anything that dares to move meanwhile.

I just don't see how AMD, Nvidia or Intel have any chance here with MS in the same bag. They're locked out.

D: i like your analogy but it lacks anything remotely resembling business acumen. you act as if the world is set in stone (based on your views) and there is no capacity for change or disruption.

What if Tizen takes off? Let me guess in your mind it never will. but in 2008 would you have predicted how powerful a force android would be? probably not.