Qualcomm shows working MSM8x60 at CES.

  • Thread starter Deleted member 13524
  • Start date
A bit late but I'd imagine they'll be showing it off at CES next year, dont think there'll be an opportunity before that.

Btw does MSM8960(or APQ8064) retain the ascynchronous CPU architecture introduced in MSM8260/8660?

All multi-core CPU's are said to be asynchronous.
 
All multi-core CPU's are said to be asynchronous.

Depends, there are various levels of "asynchronous". Some dual-core implementations can only scale both cores to the same frequency. Some can scale using a multiplier -- since they only have one PLL. Some have separate PLL's for each core, thus allowing fine-grain control over the frequency for each core.

And then there's various levels of voltage scaling as well. Some can only turn the voltage on and off, while other implementations use separate regulators for each core, thus allowing fine-grain control over voltage.
 
Depends, there are various levels of "asynchronous". Some dual-core implementations can only scale both cores to the same frequency. Some can scale using a multiplier -- since they only have one PLL. Some have separate PLL's for each core, thus allowing fine-grain control over the frequency for each core.

And then there's various levels of voltage scaling as well. Some can only turn the voltage on and off, while other implementations use separate regulators for each core, thus allowing fine-grain control over voltage.

Yep and in the case of MSM8260/8660 each core had a separate voltage plane and could run at independent clocks and voltages. Qualcomm made a big deal about it in all their press material as well. But in the case of the Krait CPU's, i havent come across it so far.
 
Yep and in the case of MSM8260/8660 each core had a separate voltage plane and could run at independent clocks and voltages. Qualcomm made a big deal about it in all their press material as well. But in the case of the Krait CPU's, i havent come across it so far.

It would depend on the SoC. I don't see any reason why MSN8960 wouldn't implement such a scaling method but perhaps the tablet oriented apq series and the eventual move to quad core would change that. What with the need for 4 PLL's and 4 voltage regulators...
 
Depends, there are various levels of "asynchronous". Some dual-core implementations can only scale both cores to the same frequency. Some can scale using a multiplier -- since they only have one PLL. Some have separate PLL's for each core, thus allowing fine-grain control over the frequency for each core.

And then there's various levels of voltage scaling as well. Some can only turn the voltage on and off, while other implementations use separate regulators for each core, thus allowing fine-grain control over voltage.

I should have said all Qualcomm multi-core CPU's.
See here in official press release about Krait.
 
I should have said all Qualcomm multi-core CPU's.
See here in official press release about Krait.

Yep the press release does mention asynchronous but makes no mention of voltage control. If it is only different clocks then thats standard, different voltages are a different thing altogether.

Just saw the Qualcomm's rebrand. Is all windows phone still based on S1 chipsets?.

I thought so but just saw this on AT - http://www.anandtech.com/show/4570/qualcomms-march-into-the-gaming-market

Text on the first image says "Xbox Live already running on Snapdragon S2 devices today with Windows Phone 7"
 
screenshot20110803at121.jpg


So Qualcomm thinks quad-core Krait + Adreno 320 is superior to PS3 & X360?
Isn't this too much of a wishful thinking?


Apparently, this would happen with the high-performance, baseband-less version, the APQ8064, when the device is plugged in:
screenshot20110803at122.jpg


Maybe they'll allow the GPU to achieve very high clocks when the device is plugged in?
 
Why does it say "playing time matches that of consoles (if plugged in)"? Err, obviously?
 
Great, so tablets out in a year or two from now can exceed consoles that are currently 6 years old. This might do a good job in replacing consoles if all three console companies decide to never update their platforms again. Exceeding Wii U on < 1W seems highly dubious to me as well; does Qualcomm actually even know what Wii U's specs will be?

Apples to apples it's just too disadvantaged. The mobile devices will always be limited to a much smaller TDP because of its small/thin footprint and lack of active cooling. With the same thermal budget, at the same process, and with a similar level of design competence the dedicated console will be able to do a lot more.

Moreover, the real problem with total consolidation will be the incompatible business models, with gaming consoles being sold entirely subsidized by games and subscriptions. Good luck trying that with a tablet or other mobile device that is simultaneously trying to sell to a market that's not the least bit interested in buying console games. People won't want to need subsidy from a data plan in order to get a console-replacement tablet at a similar price.

It seems to me that Qualcomm is beating the hype drum hardest of any of the SoC manufacturers today, and least believably.. I'm going to take it all with some salt until we see real hardware. Still waiting to see how much of an A15-killer Krait is supposed to be.
 
Great, so tablets out in a year or two from now can exceed consoles that are currently 6 years old. This might do a good job in replacing consoles if all three console companies decide to never update their platforms again. Exceeding Wii U on < 1W seems highly dubious to me as well; does Qualcomm actually even know what Wii U's specs will be?

It's marketing. Devices that will compete against consoles aren't going to be < 1W. Krait and Adreno on HPM will scale to sub-10W when the frequency is increased. Similar to ARM's plans for A15, this will start encroaching onto console and laptops.

Apples to apples it's just too disadvantaged. The mobile devices will always be limited to a much smaller TDP because of its small/thin footprint and lack of active cooling. With the same thermal budget, at the same process, and with a similar level of design competence the dedicated console will be able to do a lot more.

While true, the Wii kinda taught us that you can be an order of magnitude less powerful processing wise and still come out ahead in sales.

Moreover, the real problem with total consolidation will be the incompatible business models, with gaming consoles being sold entirely subsidized by games and subscriptions. Good luck trying that with a tablet or other mobile device that is simultaneously trying to sell to a market that's not the least bit interested in buying console games. People won't want to need subsidy from a data plan in order to get a console-replacement tablet at a similar price.

As I look at it, it's the console model that's losing out right now. In fact, I see more and more titles being released as ~$5-$10 games that are bought and downloaded from XBox Live or PSN. Yes, there will always be a market for games like LA Noire but the vast majority of the market just wants Wii Sports.

It seems to me that Qualcomm is beating the hype drum hardest of any of the SoC manufacturers today, and least believably.. I'm going to take it all with some salt until we see real hardware. Still waiting to see how much of an A15-killer Krait is supposed to be.

It'll be at least another year before any SoC manufacturer in the mobile space can realistically move their base designs up to the console/laptop space. But with Windows 8, this is a better time than ever for them to try. However, I expect nVidia will be the one leading the pack with Denver.
 
Great, so tablets out in a year or two from now can exceed consoles that are currently 6 years old. This might do a good job in replacing consoles if all three console companies decide to never update their platforms again. Exceeding Wii U on < 1W seems highly dubious to me as well; does Qualcomm actually even know what Wii U's specs will be?

Apples to apples it's just too disadvantaged. The mobile devices will always be limited to a much smaller TDP because of its small/thin footprint and lack of active cooling. With the same thermal budget, at the same process, and with a similar level of design competence the dedicated console will be able to do a lot more.

Moreover, the real problem with total consolidation will be the incompatible business models, with gaming consoles being sold entirely subsidized by games and subscriptions. Good luck trying that with a tablet or other mobile device that is simultaneously trying to sell to a market that's not the least bit interested in buying console games. People won't want to need subsidy from a data plan in order to get a console-replacement tablet at a similar price.

It seems to me that Qualcomm is beating the hype drum hardest of any of the SoC manufacturers today, and least believably.. I'm going to take it all with some salt until we see real hardware. Still waiting to see how much of an A15-killer Krait is supposed to be.

While it's true that a <1W SoC will never be able to match contemporary console hw, mobile devices have much less resolution, iterate much faster, are often subsidized and have much cheaper games - which indirectly implies simpler assets.

While a lot of what Qualcomm is saying is obviously hype, I think there is an element of truth to the proposition that consoles will become increasingly marginalized at the hands of mobile devices. And for that to happen, there is absolutely no need for mobile devices to beat consoles in the raw muscle dept.
 
It's marketing. Devices that will compete against consoles aren't going to be < 1W. Krait and Adreno on HPM will scale to sub-10W when the frequency is increased. Similar to ARM's plans for A15, this will start encroaching onto console and laptops.

While true, the Wii kinda taught us that you can be an order of magnitude less powerful processing wise and still come out ahead in sales.

10W is nowhere close to what a serious higher end console will use. I'm not really questioning the CPU designs here - those have the potential to scale quite a bit - but these devices will always be limited by the one that has the cooling surface area, fan(s), and heavier duty power supply.

Yes, you can make sales and even be the market leader with something quite underpowered. This does not equate to replacing everything else, which is Qualcomm's language. I think the current climate of consoles has reinforced the notion that there are different gamer demographics and neither XBox 360/PS3 nor Wii is coming close to addressing them all.

As I look at it, it's the console model that's losing out right now. In fact, I see more and more titles being released as ~$5-$10 games that are bought and downloaded from XBox Live or PSN. Yes, there will always be a market for games like LA Noire but the vast majority of the market just wants Wii Sports.

Again, there's room for both, and just because one is growing doesn't mean it's going to eventually take over the other completely. I strongly disagree that the "vast majority" just want Wii Sports, or else the sales ratios between the three systems wouldn't be what they are. And if there were only one higher end console instead of being split between Sony and MS the story could be even different; combined, the sales of the two exceed Wii's.

Of course, even Wii Sports level of development isn't really sustainable on $5-10 a game, which is why games costing that much are on a MUCH smaller budget or are ports. The XBLA/PSN libraries are actually pretty small compared to the disc games, a better comparison would be games on iOS and Android.

While it's true that a <1W SoC will never be able to match contemporary console hw, mobile devices have much less resolution, iterate much faster, are often subsidized and have much cheaper games - which indirectly implies simpler assets.

The resolution argument is moot when talking about an otherwise mobile device that's supposed to out-perform consoles when being connected to a TV. The quick iteration as well as much greater market segmentation is itself an impediment towards development actually pushing into the higher end of these device's capabilities. The cheaper expected cost and much simpler assets is also hardly an argument for how games running on Snapdragon-powered tablets are supposed to look better than console games.

Re-subsidy, are tablets even being subsidized at all right now? When it comes to data plans buyers will get one subsidy buy-in against it and that's it. In most cases that'll probably be their phone, and the phone will ostensibly be less powerful than the high-end tablets in the GPU department. If you can't subsidize the tablets they stay $500+ for the high end which will hopefully not be the running cost for new consoles (which instead are subsidized).

I really don't think that phones and tablets are threatening home consoles (not gaming handhelds) as much as some say they are. Yes, they'll probably continue to take some share, but I see that plateauing. Most of their sales have been due to growing a new market more than destroying that particular one.
 
From Anandtech's review of the Adreno 220-equipped Evo 3D:
The Adreno 220 is Qualcomm's highest end GPU (for now). Unfortunately we can't talk about its architecture as Qualcomm isn't publicly disclosing much but it is the fastest in the shipping Adreno lineup...
The huge advantage really comes from Basemark ES2 which seems to prefer deferred rendering architectures like the Adreno 220 and SGX 540 to NVIDIA's immediate mode renderer...
(bolded for emphasis)
I know deferred rendering is thrown around loosely by marketing departments and is often times confused by the press, but a little clarification here and in the case of Mali would be helpful again. Obviously Adreno 220 is mostly a scaled version of the architecture in the earlier 2xx parts.
 
Last edited by a moderator:
I saw a presentation from AMD (when they owned Adreno), and they said quite plainly that they were a TBDR.

Is adreno tile based earlyZ renderer?
 
I saw a presentation from AMD (when they owned Adreno), and they said quite plainly that they were a TBDR.

Is adreno tile based earlyZ renderer?

Pretty certain that it isn't TBDR or at least not what PowerVR calls TBDR (and has done for the last 17 years!), it is really an eary Z TBR. This is probably just ongoing confusion as originaly started by ARM who also claim to be TBDR when they aren't. I think we've had the debate here a couple of times on the naming and don't want to really be drawn into it again as it's fairly pintless, but fudametally Qualcom and ARM are saying that the D in TBDR refers to deferring scene rasterisation unitl they have collected all the scene data, in our case it refers to deferred texturing/shading, the later is definitely only something PowerVR do automatically (i.e. without app implementing deferred shading).

John.
 
Back
Top