[Digitimes] AMD updates product roadmap for 2014 and 2015

[frozentundra123456]

I am glad to read that they are making progress but I was really expecting them to demo some of that progress...instead all they had in the video was two dudes talking anecdotally about how great the product was (so I guess we'll just take their word for it?)

Do you know of any videos that actually demonstrate this feature in a functioning capacity? I'd really like to see that.

None of the APUs on the market are HSA ready anyway are they?

 

[USER8000]

I am glad to read that they are making progress but I was really expecting them to demo some of that progress...instead all they had in the video was two dudes talking anecdotally about how great the product was (so I guess we'll just take their word for it?)

Do you know of any videos that actually demonstrate this feature in a functioning capacity? I'd really like to see that.

Here is a video made a while back:

https://www.youtube.com/watch?v=J6MroxhsLT0

 

[ShintaiDK]

None of the APUs on the market are HSA ready anyway are they?

Nope. Not even the consoles.

 

[Idontcare]

Here is a video made a while back:

https://www.youtube.com/watch?v=J6MroxhsLT0

Thanks :thumbsup:

 

[AtenRa]

Not even the consoles.

And you know this how exactly ??

 

[ShintaiDK]

And you know this how exactly ??

Because Jaguar doesnt support HSA.

Even the Xbox One APU doesnt support hUMA either.

 

[Dresdenboy]

Because Jaguar doesnt support HSA.

Even the Xbox One APU doesnt support hUMA either.

It's not the CPU, which needs to be changed, since it already is cache coherent and supports task switching. The important changes should happen in the iMC, IOMMU, and - of course - the GPU.

 

[zlatan]

Because Jaguar doesnt support HSA.

Even the Xbox One APU doesnt support hUMA either.

Every CPU can support HSA. The actual implementation is need an HSA-MMU, a hUMA-like architecture, and an iGPU that use the same memory management as the CPU.

hUMA is an AMD thing. The HSA only specify what you need to implement in the hardware. Unified virtual memory and full coherency for example. The actual hardware implementations may be very different. The Xbox One and the PS4 are hUMA-like architectures, but neither implement hUMA itself.

I don't know Xbox One very much, because I'm not work for that system, but the PS4 is very close to hUMA. Sony will release final HSA tools in 2014 so the developers can program the system very different than today. The first tier games won't use too much power of the new consoles. As far as I know only EA Sports use some of the brand new features in the Ignite engine. This is the reason why they can't port the next-gen Fifa to PC.

 

[SocketF]

Every CPU can support HSA. The actual implementation is need an HSA-MMU, a hUMA-like architecture, and an iGPU that use the same memory management as the CPU.

How can something be "huma"-like??
Either the unified memory has common adresses for any hosts (CPU/GPu etc.), or not. It's not a greyscale thing, its simply black or white.

The XBOX cant do it, because it has 32MB EDRAM which is *NOT* working as cache, therefore these 32MB need their own addresses, and if these addresses are not usable by the CPU, then it is simply not a huma setup.

In case of the PS4 it is crystal clear to me:

Fully unified address space -> huma.

 

[USER8000]

http://techreport.com/news/24725/ps4-architect-discusses-console-custom-amd-processor

Sony has also taken steps to make it easier for developers to use the graphics component for general-purpose computing tasks. Cerny identifies three custom features dedicated to that mission:

• An additional bus has been grafted to the GPU, providing a direct link to system memory that bypasses the GPU's caches. This dedicated bus offers "almost 20GB/s" of bandwidth, according to Cerny.
• The GPU's L2 cache has been enhanced to better support simultaneous use by graphics and compute workloads. Compute-related cache lines are marked as "volatile" and can be written or invalidated selectively.
• The number of "sources" for GPU compute commands has been increased dramatically. The GCN architecture supports one graphics source and two compute sources, according to Cerny, but the PS4 boosts the number of compute command sources to 64.

The GPU does seem modified in the PS4,whether this means HSA,I don't know.

 

[NTMBK]

Even if the PS4 doesn't support HSA, it doesn't really matter. HSA is just another abstraction layer- you wouldn't want to use it on a console game, even if the silicon supported it. The PS4 has the useful bit, the fully unified address space, and that will do it just fine.

 

[piesquared]

Nice, more from IBC 2013.

AMD Directly Contributes to 4K Video Production

AMD announced at the 2013 IBC (International Broadcasting Convention) that it was collaborating with Fusion-io, Supermicro, Dell, Blackmagic, Autodesk and Adobe.

Even though it doesn't specialize in video creation and editing, Advanced Micro Devices has a significant stake in that field simply because it is one of just two corporations that make video cards capable of meeting customer needs.


The company has decided that simple role isn't enough, however. So it figured it may as well start to directly involve itself in video creation, especially now that the 4K resolution is gaining steam (3840 x 2560 pixels).

It's easy to see FirePro and professional APU's gaining major share in the professional market, developers are flocking to GCN and heterogeneous computing.

 

[monstercameron]

all that sh!te about amd kabini but look at this
http://www.notebookcheck.net/Review-Acer-Aspire-E1-522-45004G50Mnkk-Notebook.101677.0.html

Power consumption
Off / Standby    0.2 / 0.5 Watt
Idle    5.2 / 9.5 / 9.6 Watt
Load    16.8 / 20.6 Watt
     
Key: min: , med: , max:         Voltcraft VC 940

We measure the battery runtime via the BatteryEater tool. The maximum possible runtime is ascertained using minimum screen brightness, disabled communication modules (Wi-Fi, Bluetooth) and energy-saving mode while the screen is permanently active and without any user interaction. Acer's E1-522 achieves nine hours and 37 minutes with its 37 Wh battery and is thus on a par with Toshiba's equally sized Satellite C50D-A-10E based on the same hardware. In return, Samsung's ATIV Book 2 (40 Wh) is drained after approximately seven hours while the Lenovo (41 Wh) only manages six hours.

they completely borked the temash a6-1450 in the v5-122p
http://www.notebookcheck.net/Review-Acer-Aspire-V5-122P-61454G50NSS-Notebook.93321.0.html

Power consumption
Off / Standby    0.1 / 1.0 Watt
Idle    6.9 / 9.4 / 12.4 Watt
Load    18.3 / 23.3 Watt
     
Key: min: , med: , max:         Voltcraft VC 940

The battery runtimes of the Acer Aspire V5-122P are not one the strong suits of the subnotebook. Despite the low energy intake, the 30 Wh lithium-ion battery proves to be insufficient for long trips far away from a socket.
While the battery lasts an acceptable 6 hours and 19 minutes in idle mode (which is lacking practical relevance, Battery Eater Reader's Test, minimum brightness), it shrinks to a little over 3 hours in our practically oriented WLAN test. Under load the device needs to be connected again to a power point in less than 2 hours.

 

[Idontcare]

Who disables their communication modules when using their mobile devices as a mobile device? That is precisely when you want to have mobile connectivity.

 

[USER8000]

Wait,an A4-5000 based device with a 15.6" screen is consuming less power than an A6-1450 based device with an 11.6" screen?? WTF,what the heck was Acer doing?? Did they plonk an 8 phase VRM for the A6?? It reminds of a one Gigabyte E350 motherboard which had a very inefficient VRM section,and ended up consuming over 50% more power than any other E350 based motherboard.

It just shows how much the rest of the components outside the CPU used contribute to system power consumption. This is why I am always wary of manufacturer reference systems used during reviews,as they might not be fully indicative of what you might or might not get.

 

[krumme]

Haswell have proven where the vrm needs to be
Intel at its finest using their far superior analogue like knowledge

 

[Enigmoid]

Who disables their communication modules when using their mobile devices as a mobile device? That is precisely when you want to have mobile connectivity.

Word processing or watching videos on the device (from the HDD)

all that sh!te about amd kabini but look at this
http://www.notebookcheck.net/Review-Acer-Aspire-E1-522-45004G50Mnkk-Notebook.101677.0.html

Power consumption
Off / Standby    0.2 / 0.5 Watt
Idle    5.2 / 9.5 / 9.6 Watt
Load    16.8 / 20.6 Watt
     
Key: min: , med: , max:         Voltcraft VC 940

they completely borked the temash a6-1450 in the v5-122p
http://www.notebookcheck.net/Review-Acer-Aspire-V5-122P-61454G50NSS-Notebook.93321.0.html

Power consumption
Off / Standby    0.1 / 1.0 Watt
Idle    6.9 / 9.4 / 12.4 Watt
Load    18.3 / 23.3 Watt
     
Key: min: , med: , max:         Voltcraft VC 940

What on earth are you talking about?

The E1 uses a 37 Whr battery and lasts 9:37 idle (useless measurement) and 3:47 browsing. The v5 uses a 30 whr battery and lasts 6:19 idle and 3:07 browsing. Normalized battery life under typical use conditions is pretty much similar. Screen size is different but then there are multiple other differences between the two.

 

[monstercameron]

Word processing or watching videos on the device (from the HDD)



What on earth are you talking about?

The E1 uses a 37 Whr battery and lasts 9:37 idle (useless measurement) and 3:47 browsing. The v5 uses a 30 whr battery and lasts 6:19 idle and 3:07 browsing. Normalized battery life under typical use conditions is pretty much similar. Screen size is different but then there are multiple other differences between the two.

that's what I am getting at, the e1-522 uses ~3.8W SDP idling, while the v5-122p uses ~4.7W.

the e1-522 uses ~9.78W SDP browsing, while the v5-122p uses ~9.6W.

the e1-522 uses ~17W SDP stresses, while the v5-122p uses ~15.93W.

the power use seems very high for the v5-122p on an 8W tdp chip. Even in turbo mode 14W tdp in a smaller formfactor should use a bit less power thaan a 15" laptop.

 

[Enigmoid]

that's what I am getting at, the e1-522 uses ~3.8W SDP idling, while the v5-122p uses ~4.7W.

the e1-522 uses ~9.78W SDP browsing, while the v5-122p uses ~9.6W.

the e1-522 uses ~17W SDP stresses, while the v5-122p uses ~15.93W.

the power use seems very high for the v5-122p on an 8W tdp chip. Even in turbo mode 14W tdp in a smaller formfactor should use a bit less power thaan a 15" laptop.

AMD's tdp ratings for kabini are sketchy (double the cores and keep the same tdp for some skus) same as intels i5 and i7s having the same tdp while the i7 consumes significantly more power.. Then there is the inherent problem of different notebook configurations (expect +/- 15% as perfectly normal, perhaps more at the ULV level where RAM and HDD can have a fairly large effect on power).

We need more evidence at this point in time to say its kabini rather than acer. I agree that acer screwed it up but again more data is needed to see if the v5 is particularly bad or the E1 particularly good.

Ironically, it seems as mobile efficiency increases battery size decreases. 10 years ago you could expect a 50whr + battery in budget, now you are lucky with 40whr+. 30/37 watts is simply too small for a notebook.

 

[USER8000]

Haswell have proven where the vrm needs to be
Intel at its finest using their far superior analogue like knowledge

Which didn't help in all parts did it as it introduced its own set of problems,as a number of posters have elaborated on in other threads. AFAIK,Bay Trail does not use the same technology last time I checked:

http://images.anandtech.com/doci/7314/Screen%20Shot%202013-09-11%20at%208.27.39%20AM_575px.png

Moreover,CPU infrastructure makes a big change to power consumption anyway,as you can see realworld differences on desktop for example,via just using a different motherboard. This is even seen with socket 1150:

http://www.bit-tech.net/hardware/2013/08/02/asrock-z87-extreme3-review/6

 

[krumme]

Which didn't help in all parts did it as it introduced its own set of problems,as a number of posters have elaborated on in other threads. AFAIK,Bay Trail does not use the same technology last time I checked:

http://images.anandtech.com/doci/7314/Screen Shot 2013-09-11 at 8.27.39 AM_575px.png

Moreover,CPU infrastructure makes a big change to power consumption anyway,as you can see realworld differences on desktop for example,via just using a different motherboard. This is even seen with socket 1150:

http://www.bit-tech.net/hardware/2013/08/02/asrock-z87-extreme3-review/6

Yes it doesnt work well for desktop high perf parts. And yes the entire system is more important. But I have no doubt that in 5 years every cpu/soc will have vrm integrated. Obviously its very hard to do right, and perhaps Intel was 1 year to early on this. But for the ulv i guess it made a huge improvement.

The others will catch up but its the kind of thech that is extremely difficult to design and make into production. I guess Intel can really use their integration of design and production for this kind of challenge.

 

[USER8000]

Another Jaguar based touch enabled netbook has been previewed:

http://www.cnet.com/laptops/dell-inspiron-11-intel/4505-3121_7-35827688.html

It seems to be quite decent for the price.

 

[monstercameron]

Another Jaguar based touch enabled netbook has been previewed:

http://www.cnet.com/laptops/dell-inspiron-11-intel/4505-3121_7-35827688.html

It seems to be quite decent for the price.

madness!!!!!!
under $400 with a 50Wh battery?

 

[Shivansps]

Another Jaguar based touch enabled netbook has been previewed:

http://www.cnet.com/laptops/dell-inspiron-11-intel/4505-3121_7-35827688.html

It seems to be quite decent for the price.

Sorry but $30 for the Haswell version? whats the point of buying the Kabini one?

 

[USER8000]

Sorry but $30 for the Haswell version? whats the point of buying the Kabini one?

????

Intel-powered versions of the Inspiron 11 start at $379, and CPU options include Intel's latest Haswell fourth-generation chips, although it'll be interesting to see how different CPU choices affect the price.

Its most likely an Atom derivative for the lower priced Intel models.