QQ Snapdragon 810 overheating issues: delay?

[jdubs03]

I imagine their 20nm A57 APU for skybridge will be at TSMC. But I'm pretty sure AMD's skybridge is a server platform, so I'm not sure it'll be well suited for tablets and phones

I doubt it would be TSMC, AMD still has their WSA agreement with GlobalFoundries, who is licensing 20nm and 14nmFF technology.

 

[geoxile]

I doubt it would be TSMC, AMD still has their WSA agreement with GlobalFoundries, who is licensing 20nm and 14nmFF technology.

Didn't GF cancel their 20nm nodes?

 

[NTMBK]

I imagine their 20nm A57 APU for skybridge will be at TSMC. But I'm pretty sure AMD's skybridge is a server platform, so I'm not sure it'll be well suited for tablets and phones

Skybridge is targeting Android: http://www.anandtech.com/show/7989/...bridge-pincompatible-arm-and-x86-socs-in-2015 I suspect that the idea is to let OEMs create a single tablet design and ship Windows and Android variants.

 

[mrmt]

Skybridge is targeting Android: http://www.anandtech.com/show/7989/...bridge-pincompatible-arm-and-x86-socs-in-2015 I suspect that the idea is to let OEMs create a single tablet design and ship Windows and Android variants.

I don't think AMD is targeting Android. This is a face-saving move, just something to enable management to write ">>technically¹²³<< this APU can be used in Android products". AMD doesn't have Mediatek lean cost structure and high volume, Intel financial muscle and foundry expertise, much less Qualcomm experience in developing custom cores plus its connectivity IP.

If they have nothing of that why would an OEM pick AMD to build Android products? That's right, they wouldn't. Maybe in some embedded product they might have a small chance, but I really doubt we'll ever see something meaningful from their part in the consumer market.

 

[NTMBK]

I would look at it as a way to add value to their x86 tablet designs. For the same cost of designing a tablet for their x86 design, the OEM gets an ARM design thrown in for free.

Certainly an uphill struggle though. Nvidia already tried to differentiate themselves with only GPU IP, and we all know how well that is going for them...

 

[witeken]

Intel just has to be better than Qualcomm per unit of money. They're already doing that in tablets and it works well.

 

[mrmt]

Intel just has to be better than Qualcomm per unit of money. They're already doing that in tablets and it works well.

I also want to see the performance of their LTE modem, both in terms of costs and performance.

 

[mrmt]

Look at contrarevenue for non samsung and apple oem. 4B a year. Its way off. How many B to get to apple or ss. 20B a year? Naa probably way more. Ss and apple want 100% control of their extremely profitable market. They dont earn anything worth compared to mobile phones market. And they dont want to share a dime with anyone.

Contra-revenue is 1 billion, not 4 billion.

 

[Arachnotronic]

Contra-revenue is 1 billion, not 4 billion.

No, mrmt, you've got it all wrong. Contra revenue is $4 billion because that's how big the operating loss is for the year. R&D and SG&A costs are obviously minimal for complex mobile SoCs and IPs.

/sarcasm

 

[Arachnotronic]

I would look at it as a way to add value to their x86 tablet designs. For the same cost of designing a tablet for their x86 design, the OEM gets an ARM design thrown in for free.

Certainly an uphill struggle though. Nvidia already tried to differentiate themselves with only GPU IP, and we all know how well that is going for them...

In the mobile market, "differentiating on GPU" or even on CPU is not going to work, as we have seen with Tegra K1 (which arguably has the fastest merchant solutions on both counts, depending of course on what benchmark(s) you care about). ARM/Imagination will license you great IP.

The differentiation comes from delivering on every key vector, including CPU, GPU, image signal processor(s), video encode/decode, connectivity, comms, and so on.

AMD has no connectivity story and no cellular story to speak of, so its place in the ultramobile market is unlikely to be meaningful -- no matter how many times people point to Mullins benchmarks run in a bulky test system with no power consumption/battery life measured

 

[III-V]

No, mrmt, you've got it all wrong. Contra revenue is $4 billion because that's how big the operating loss is for the year. R&D and SG&A costs are obviously minimal for complex mobile SoCs and IPs.

/sarcasm

It's astounding to me that people are still so ignorant about contra revenue.

 

[krumme]

Contra-revenue is 1 billion, not 4 billion.

You know that i know that. But whatever. Anything to pick a fight. Even writing in red. Are we in a little troll mood today?

 

[krumme]

It's astounding to me that people are still so ignorant about contra revenue.

Yeaa. All people is so stupid.

 

[III-V]

Yeaa. All people is so stupid.

It's been explained to death. People like to conveniently "forget" things around here.

In the mobile market, "differentiating on GPU" or even on CPU is not going to work, as we have seen with Tegra K1 (which arguably has the fastest merchant solutions on both counts, depending of course on what benchmark(s) you care about). ARM/Imagination will license you great IP.

The differentiation comes from delivering on every key vector, including CPU, GPU, image signal processor(s), video encode/decode, connectivity, comms, and so on.

AMD has no connectivity story and no cellular story to speak of, so its place in the ultramobile market is unlikely to be meaningful -- no matter how many times people point to Mullins benchmarks run in a bulky test system with no power consumption/battery life measured

Yep. I don't see AMD ever becoming a significant player in the tablet space. I think they'd do quite well doing custom, embedded stuff, though.

 

[Thala]

In the mobile market, "differentiating on GPU" or even on CPU is not going to work, as we have seen with Tegra K1 (which arguably has the fastest merchant solutions on both counts, depending of course on what benchmark(s) you care about). ARM/Imagination will license you great IP.

I disagree. Tegra K1 had some design wins, most notably Nexus 9. CPU and GPU are most likely the single most important differentiation factors in contrary to all the commodity and check-box features as the likes of connectivity etc.

Of course you can sell just about anything if the price is right, as the Baytrail conter-revenue example shows. In this regard Intel has bigger up-hill battle to fight than NVidia. At least they earn some money with a superior product.

 

[SirCanealot]

2+4 with aggressive scaling on the big 2 would probably be good.

you know, that's probably what apple will do next, but it'll be 2+2

You may as well have the 2 extra 'big' cores though to make it 4+4. 2+4 will actually work great (and it works well in the Galaxy K Zoom which is 2+4), but quad is excellent for multitasking (eg: a classic usage pattern for me is browsing a few websites and opening 10-20 tabs as I browse down the news page... kotaku.com is terrible for this if I haven't checked it in a few days ). Without 4 fast cores, you'd still get lags and stutters, imo. The above usage pattern will basically max out my Note 3 as it loads all the tabs up in the background into RAM.

Try and remember you could have a SOC with 200 cores, but it will use the same power as a SOC with 4 cores since the other 196 cores could be power gated

2+4 would still be fantastic, and in my experience A7 cores are solid as hell (browsing with my girlfriend's Tab S in power save mode locks to A7 cores and it was surprisingly usable), but I'd still be looking at 4+4 for a flagship device for myself.

For any kinda mid-range device though, I'm hoping 2+4 does have some design wins though, as it is a massive performance boost over just A7/A53 cores

 

[III-V]

I disagree. Tegra K1 had some design wins, most notably Nexus 9. CPU and GPU are most likely the single most important differentiation factors in contrary to all the commodity and check-box features as the likes of connectivity etc.

Of course you can sell just about anything if the price is right, as the Baytrail conter-revenue example shows. In this regard Intel has bigger up-hill battle to fight than NVidia. At least they earn some money with a superior product.

Nvidia's edge has always been time to market (for premium devices). In this case, Nvidia has a "next gen" product before anyone else. That's how they've gotten their design wins historically. In the case of Tegra 4, the product was late, and what do you know, it wasn't that successful either.

 

[NTMBK]

Don't read too much into Nexus 9. Google wanted an ARM V8 device for developers, and K1 was the only thing ready in time.

 

[jhu]

Would there be any difficulty getting it running on Android? Trying to get a proper Linux distro running on all these platforms would be a pain.

Actually, it's not. For example, I have a full Debian 7 distribution on my phone. It runs as a chroot. Just need to be able to root the phone. It's kind of neat being able to run gcc on my phone (at least that's what the girls at cocktail parties tell me...).

 

[sontin]

Nvidia's edge has always been time to market (for premium devices). In this case, Nvidia has a "next gen" product before anyone else. That's how they've gotten their design wins historically. In the case of Tegra 4, the product was late, and what do you know, it wasn't that successful either.

Tegra 4 was successful even after the delay.

The Android market is a race to the bottom. Last year a lot of companies have released high end Android tablets. This year there is nothing.

nVidia won these design because companies went after the high-end.

 

[mrmt]

Tegra 4 was successful even after a the delay.

Yeah, a success to the point Nvidia is still gunning after the same markets Tegra 4 was targeted.

 

[III-V]

Tegra 4 was successful even after the delay.

The Android market is a race to the bottom. Last year a lot of companies have released high end Android tablets. This year there is nothing.

nVidia won these design because companies went after the high-end.

That's more of a timing issue -- everything has gotten delayed past the holiday season.

 

[Exophase]

Actually, it's not. For example, I have a full Debian 7 distribution on my phone. It runs as a chroot. Just need to be able to root the phone. It's kind of neat being able to run gcc on my phone (at least that's what the girls at cocktail parties tell me...).

Oh, I hadn't even thought about that. Are there any real limitations in what you can do with the OS chrooted? Nothing that would inhibit SPEC I'm sure.

I did hear that sometimes you get performance degradation this way, is there any truth to that?

 

[jhu]

Oh, I hadn't even thought about that. Are there any real limitations in what you can do with the OS chrooted? Nothing that would inhibit SPEC I'm sure.

I did hear that sometimes you get performance degradation this way, is there any truth to that?

Not that I know of. There is the issue of not knowing what speed the processor is actually running at, but that has nothing to do with being in a chroot.

If there's anything you'd like run, let me know. The only caveat is that phones throttle a lot. I have a Snapdragon 801 (HTC One M8) and an OMAP 4470 (Nook HD+).

 

[jdubs03]

found this searching online via gsmarena:

http://browser.primatelabs.com/geekbench3/1466744

vs.

exynos 5433 (i know sweepr has benches but they got a bit lost)
http://www.androidauthority.com/tegra-k1-exynos-5433-snap-805-541582/

results are as expected, similar as least per geekbench.

edit: here is that thread:

http://forums.anandtech.com/showthread.php?t=2407781

1282 single-core, that's a 154 point difference from the androidauthority benches.