Windows tablets

[Deleted member 13524]

Brazos can use LV DDR3, which runs at 1.35V. LPDDR2 runs at 1.2V so the power savings will be minimal i guess.

Not that you wouldn't be right, but power draw isn't dictated by voltage alone.
I don't know what is the current needed for either tech.

And im sure Brazos could use the extra b/w provided by DDR3.

Yes. I also think AMD needs to overcome the b/w shortage on all its APUs somehow and I hope they don't just wait for higher clocked DDR3 memory to become cheaper and consume less power.

When are Cortex A15 SoC's slated to be available? From what i've read so far, OMAP 5 is the first and we wont see that until H2 2012 (unspecified if Q3 or Q4). Even if its Q3, Brazos-T will be available in Q2 and that will give them time to get design wins for the crucial back to school period.

Are you sure OMAP 5 is coming before Nova A9600?
And isn't Kal-El more likely to get design wins for tablets than Brazos-T, at least until Windows 8 is launched?

Why would "next gen" graphics neccesarily mean low power?

Erm, no.. Next gen graphics means it'll be based on the new SIMD architecture, or GCN as AMD calls it. I don't think they'd call it "next gen" if it was VLIW4, as that's current gen already.
Being a low-power version of that architecture is implicit, since we're talking about the lowest power-consuming APUs from AMD.
You're not thinking AMD would put a 50W iGPU for a ~4W APU, right?

And are AMD going to be able to target smartphones at all?

They're definitely headed that way in terms of power consumption, while launching APUs with reduced TDP every 9 months or so.
Either they'll stop at the tablet form factor and increase performance or keep going down until they reach a smartphone form factor depends on a lot of things, so maybe not even AMD knows that yet.

I dont think they've publicly stated their intent of going after the smartphone market. Besides the SoC market is crowded enough as it is

x86 Smartphone SoC market (as in, smartphones capable of running Windows 8 with legacy software support) is completely vacant right now.

Did anybody also notice that they compared power consumption at different temperatures? 5.9W @ 75 C for Desna while for Hondo they mention < 4.5W @ 60 C

Yes. It probably has to do with the fact that the Brazos-T is "optimized for fanless designs", unlike the original Brazos. AFAIK, even C-50 tablets and netbooks have fans.

 

[Lazy8s]

OMAP5 is scheduled to come a few months before A9600, yes, and may well be the first A15 implementation to hit.

 

[Erinyes]

Are you sure OMAP 5 is coming before Nova A9600?
And isn't Kal-El more likely to get design wins for tablets than Brazos-T, at least until Windows 8 is launched?

Yep im fairly certain it is. Though MSM 8960 might beat it to market by a bit.

Oh yes im sure Kal El will get a lot more design wins than Brazos-T for Android tablets. Its already entered production so Kal El tablets should start shipping by September, its got at least a half a year lead on Brazos-T. But IMO for Windows 8, Brazos-T and Cloverview are a big threat as they are x86 (it remains to be seen if legacy apps will work on ARM). Kal El will probably get a majority of the Windows 8 ARM tablets though, coz there is literally no competition. Apart from Kal El I dont know of any other quad core ARM CPU which is slated to be available in that timeframe. Hell even Wayne has a chance of being available by then (its apparently sampling in Dec 2011). APQ8064 is set to sample in Q4 2011. Though i remember reading something about Freescale coming out with a quad core A9.

Erm, no.. Next gen graphics means it'll be based on the new SIMD architecture, or GCN as AMD calls it. I don't think they'd call it "next gen" if it was VLIW4, as that's current gen already.
Being a low-power version of that architecture is implicit, since we're talking about the lowest power-consuming APUs from AMD.
You're not thinking AMD would put a 50W iGPU for a ~4W APU, right?

Not neccesarily. Do you think Krishna/Witchita (on which the "Samara" part is presumably based) is going to sport a GCN Class GPU? Just like Trinity i expect its going to be VLIW4. I never mentioned anything about using a high power GPU so dont know what ur last sentence is about

They're definitely headed that way in terms of power consumption, while launching APUs with reduced TDP every 9 months or so.
Either they'll stop at the tablet form factor and increase performance or keep going down until they reach a smartphone form factor depends on a lot of things, so maybe not even AMD knows that yet.

x86 Smartphone SoC market (as in, smartphones capable of running Windows 8 with legacy software support) is completely vacant right now.

They're still a long way off from reaching smartphone power levels. And given their limited resources i dont expect them to be able to go for the smartphone market even if they wanted to. I think they would be wise to stick to the tablet market where they can compete well. Like i said already, the SoC market is cut-throat enough as it is.

Mighty Intel already tried and failed, what makes you think AMD has a better chance? They're still trying and probably wont be able to catch ARM for another generation. Im not saying AMD cant, just that its unlikely

Yes. It probably has to do with the fact that the Brazos-T is "optimized for fanless designs", unlike the original Brazos. AFAIK, even C-50 tablets and netbooks have fans.

Fanless designs would presumably run at a higher temperature so i dont know how that would result in a lower temperature rating

 

[Rys]

Kal El will probably get a majority of the Windows 8 ARM tablets though, coz there is literally no competition. Apart from Kal El I dont know of any other quad core ARM CPU which is slated to be available in that timeframe. Hell even Wayne has a chance of being available by then (its apparently sampling in Dec 2011). APQ8064 is set to sample in Q4 2011. Though i remember reading something about Freescale coming out with a quad core A9.

Is quad-core a hard requirement now, and does GPU performance not matter?

 

[Arun]

Yep im fairly certain it is. Though MSM 8960 might beat it to market by a bit.

OMAP5 was supposed to sample in October 2011, A9600 was supposed to sample in Q4 2011. Even if that means December, then that's only a two months difference if both hit their schedule. Of course A9600 is more disruptive (Rogue, heterogeneous CPU cores, etc.) so they've got more reasons to be delayed (both for sampling and end-devices). Here's hoping they execute - they really need it given ST-Ericsson's financial problems.

Kal El will probably get a majority of the Windows 8 ARM tablets though, coz there is literally no competition. Apart from Kal El I dont know of any other quad core ARM CPU which is slated to be available in that timeframe. Hell even Wayne has a chance of being available by then (its apparently sampling in Dec 2011). APQ8064 is set to sample in Q4 2011. Though i remember reading something about Freescale coming out with a quad core A9.

Kal-El does have a very good positioning for Windows 8, but as Rys said it's insane to look only at CPU performance. And even then, it's insane to look at only the number of cores. OMAP5 and A9600 are going to thrash Kal-El for real-world workloads (and even Wayne to a slightly lesser extent).

I'm willing to bet Wayne still only has a 32-bit memory controller and 4xA9 at up to 2GHz. It's more of a high-end part than an ultra-high-end one IMO. Logan is the real monster (4xA15 and next-gen GPU arch) but who knows when they'll be sampling that. I don't think that's a bad thing though: with the Icera acquisition, they've got an opportunity to penetrate the wider market, and Wayne is the perfect chip for that. They'll temporarily lose the performance crown but I think it's probably worth it.

 

[rpg.314]

OMAP5 was supposed to sample in October 2011, A9600 was supposed to sample in Q4 2011. Even if that means December, then that's only a two months difference if both hit their schedule. Of course A9600 is more disruptive (Rogue, heterogeneous CPU cores, etc.) so they've got more reasons to be delayed (both for sampling and end-devices). Here's hoping they execute - they really need it given ST-Ericsson's financial problems.

Kal-El does have a very good positioning for Windows 8, but as Rys said it's insane to look only at CPU performance. And even then, it's insane to look at only the number of cores. OMAP5 and A9600 are going to thrash Kal-El for real-world workloads (and even Wayne to a slightly lesser extent).

I'm willing to bet Wayne still only has a 32-bit memory controller and 4xA9 at up to 2GHz. It's more of a high-end part than an ultra-high-end one IMO. Logan is the real monster (4xA15 and next-gen GPU arch) but who knows when they'll be sampling that. I don't think that's a bad thing though: with the Icera acquisition, they've got an opportunity to penetrate the wider market, and Wayne is the perfect chip for that. They'll temporarily lose the performance crown but I think it's probably worth it.

So when does Denver land in Tegra? Stark?

 

[Arun]

So when does Denver land in Tegra? Stark?

No, Stark is the equivalent of Wayne: a shrink of the 28nm Logan to 20nm with a few incremental improvements. Project Denver will be used on the chip right after Stark. But as always NVIDIA's roadmap changes quite a bit over time so who knows - for example, Stark might lose its head (sorry, couldn't resist! )

Remember that Denver uses a fair bit of custom logic and they'll implement it in on the desktop first, so they need to port it from 20HP to 20SoC (yes that really is the name of TSMC's low-power 20nm process, or at least it was the last time I checked).

 

[Rys]

Remember that Denver uses a fair bit of custom logic and they'll implement it in on the desktop first, so they need to port it from 20HP to 20SoC (yes that really is the name of TSMC's low-power 20nm process, or at least it was the last time I checked).

OT, but I just Googled 20SoC to check, and your post less than 10 mins ago is in the first page of results

 

[Arun]

OT, but I just Googled 20SoC to check, and your post less than 10 mins ago is in the first page of results

haha, who needs real time Twitter results when Google can archive the entire internet in real time! And the same quick Google tells me they've renamed 20HP to 20G. So it looks like they are randomly changing names again for no reason whatsoever, yay!

 

[metafor]

No, Stark is the equivalent of Wayne: a shrink of the 28nm Logan to 20nm with a few incremental improvements. Project Denver will be used on the chip right after Stark. But as always NVIDIA's roadmap changes quite a bit over time so who knows - for example, Stark might lose its head (sorry, couldn't resist! )

Remember that Denver uses a fair bit of custom logic and they'll implement it in on the desktop first, so they need to port it from 20HP to 20SoC (yes that really is the name of TSMC's low-power 20nm process, or at least it was the last time I checked).

I have a difficult time imagining a design scaling all the way from HPC to mobile SoC's. It's possible they designed it modularly enough such that one can trim down various features to get it within a power envelope, but it wouldn't be very optimized.

IMO, they'd be better off sticking to the Cortex cores until they can get sufficient momentum to split into two CPU teams.

 

[Arun]

I have a difficult time imagining a design scaling all the way from HPC to mobile SoC's. It's possible they designed it modularly enough such that one can trim down various features to get it within a power envelope, but it wouldn't be very optimized.

IMO, they'd be better off sticking to the Cortex cores until they can get sufficient momentum to split into two CPU teams.

Two points:
1) Hopefully it's heterogeneous for handhelds (next to lower-end ARMv8 cores, maybe Kingfisher?)
2) NVIDIA likely wants most of the flops to be on the GPU side, so I wouldn't expect mind-blowing NEON performance.

Is that enough? We'll see.

 

[metafor]

Two points:
1) Hopefully it's heterogeneous for handhelds (next to lower-end ARMv8 cores, maybe Kingfisher?)

What would be the point though? I don't think it'd be better in area and if you scale down performance of an HPC processor to the point where it'd be within the thermal/current envelope of a mobile chip, I'd wager the performance will be nowhere near as good as Athena or whatever the leading edge ARMv8 Cortex will be.

2) NVIDIA likely wants most of the flops to be on the GPU side, so I wouldn't expect mind-blowing NEON performance.

I was more concerned with performance on the integer/branch/ldst side. Granted, SIMD is generally the worse case part of the chip when it comes to power but the rest is a close second.

Is that enough? We'll see.

It's still 2 years out. I doubt even nVidia has solidified their plans; but I would suspect that they may release two CPU designs perhaps reusing parts of Denver.

 

[Erinyes]

Is quad-core a hard requirement now, and does GPU performance not matter?

Sorry i should have been more clear, i was talking about quad core tablets. Of course the GPU matters but the average consumer can hardly differentiate between different GPU's can they? Im thinking the majority of tablet sales will be quad cores anyway due to the marketing advantages of having four cores over two. And the fact that Kal El will be on the market early will help too (just like for Tegra 2)

Kal-El does have a very good positioning for Windows 8, but as Rys said it's insane to look only at CPU performance. And even then, it's insane to look at only the number of cores. OMAP5 and A9600 are going to thrash Kal-El for real-world workloads (and even Wayne to a slightly lesser extent).

I'm willing to bet Wayne still only has a 32-bit memory controller and 4xA9 at up to 2GHz. It's more of a high-end part than an ultra-high-end one IMO. Logan is the real monster (4xA15 and next-gen GPU arch) but who knows when they'll be sampling that. I don't think that's a bad thing though: with the Icera acquisition, they've got an opportunity to penetrate the wider market, and Wayne is the perfect chip for that. They'll temporarily lose the performance crown but I think it's probably worth it.

True, i agree. And quad A9's on 1.5 ghz arent gonna be a whole lot faster than dual A15's at say 1.5 ghz anyway right? But as i was saying earlier, the marketability of having four cores is still a factor. And timing should also be taken into account. If Win 8 ships in Oct 2012, will OMAP5 and A9600 be available by then?

Im no expert but I'd expect them to move to 64 bit for Wayne. From what ive read so far they're pushing it by going for just 32 bit for Kal El already. With clocks going up by say 33%, and GPU performance expected to double they'd really need it wont they? And what about the possibility of them increasing the L2 cache to 2 MB from the current 1 MB in Kal El?

 

[Lazy8s]

Users feel a difference in graphics acceleration more than in CPU performance when part of a UI or page has a jarring loss of smoothness while scrolling, when it isn't responsive during the delay while it draws/scales the webpage, or when they can't play the latest game well nor take advantage of rendering smoothly on video out to their HDTV.

 

[wco81]

Rumors that Samsung tablets with early build of W8 will be handed out to developers at a conference next week.

Of course it would be what, almost a year before actual products would ship?

But maybe they feel the train is leaving the station and they have to get on now, even if the product is a long ways from being ready?

 

[Ike Turner]

Rumors that Samsung tablets with early build of W8 will be handed out to developers at a conference next week.

It's true http://forum.beyond3d.com/showpost.php?p=1578305&postcount=4

 

[Laurent06]

It's true http://forum.beyond3d.com/showpost.php?p=1578305&postcount=4

Hmm I guess that won't fit the OP need for an x86 tablet

 

[Ike Turner]

Hmm I guess that won't fit the OP need for an x86 tablet

Well he can still get the X86 version in a few weeks (the Samsung Slate 7)

 

[Ike Turner]

No Kal-El .. MS is giving away the Samsung Slate 7 loaded with Win8:

http://www.mobiletechworld.com/2011/09/13/microsofts-first-windows-8-tablet-from-samsung/

 

[Deleted member 13524]

A device with sandybridge Core i5, active cooling and low-ish battery life is what Microsoft decided to hand out to developers in order to test Windows 8 (-> which is supposedly going to counter Android tablets and iPads)?

Whose brainfart was that?
There are C60 tablets out there now, why go with inadequate hardware?