hUMA has come with a weapon to slay the memory latency dragon

Subject: General Tech | April 30, 2013 - 01:23 PM |
Tagged: Steamroller, piledriver, Kaveri, Kabini, hUMA, hsa, GCN, bulldozer, APU, amd

AMD may have united GPU and CPU into the APU but one hurdle had remained until now, the the non-uniformity of memory access between the two processors.  Today we learned about one of the first successful HAS projects called Heterogeneous Uniform Memory Access, aka hUMA, which will appear in the upcoming Kaveri chip family.   The use of this new technology will allow the on-die CPU and GPU to access the same memory pool, both physical and virtual and any data passed between the two processors will remain coherent.  As The Tech Report mentions in their overview hUMA will not provide as much of a benefit to discrete GPUs, while they will be able to share address space the widely differing clock speeds between GDDR5 and DDR3 prevent unification to the level of an APU.

Make sure to read Josh's take as well so you can keep up with him on the Podcast.

huma_02.jpg

"At the Fusion Developer Summit last June, AMD CTO Mark Papermaster teased Kaveri, AMD's next-generation APU due later this year. Among other things, Papermaster revealed that Kaveri will be based on the Steamroller architecture and that it will be the first AMD APU with fully shared memory.

Last week, AMD shed some more light on Kaveri's uniform memory architecture, which now has a snazzy marketing name: heterogeneous uniform memory access, or hUMA for short."

Here is some more Tech News from around the web:

Tech Talk

Author:
Subject: Processors
Manufacturer: AMD

heterogeneous Uniform Memory Access

 

Several years back we first heard AMD’s plans on creating a uniform memory architecture which will allow the CPU to share address spaces with the GPU.  The promise here is to create a very efficient architecture that will provide excellent performance in a mixed environment of serial and parallel programming loads.  When GPU computing came on the scene it was full of great promise.  The idea of a heavily parallel processing unit that will accelerate both integer and floating point workloads could be a potential gold mine in wide variety of applications.  Alas, the promise of the technology did not meet expectations when we have viewed the results so far.  There are many problems with combining serial and parallel workloads between CPUs and GPUs, and a lot of this has to do with very basic programming and the communication of data between two separate memory pools.

huma_01.jpg

CPUs and GPUs do not share common memory pools.  Instead of using pointers in programming to tell each individual unit where data is stored in memory, the current implementation of GPU computing requires the CPU to write the contents of that address to the standalone memory pool of the GPU.  This is time consuming and wastes cycles.  It also increases programming complexity to be able to adjust to such situations.  Typically only very advanced programmers with a lot of expertise in this subject could program effective operations to take these limitations into consideration.  The lack of unified memory between CPU and GPU has hindered the adoption of the technology for a lot of applications which could potentially use the massively parallel processing capabilities of a GPU.

The idea for GPU compute has been around for a long time (comparatively).  I still remember getting very excited about the idea of using a high end video card along with a card like the old GeForce 6600 GT to be a coprocessor which would handle heavy math operations and PhysX.  That particular plan never quite came to fruition, but the idea was planted years before the actual introduction of modern DX9/10/11 hardware.  It seems as if this step with hUMA could actually provide a great amount of impetus to implement a wide range of applications which can actively utilize the GPU portion of an APU.

Click here to continue reading about AMD's hUMA architecture.

Qualcomm Joins HSA Foundation to Further Heterogeneous Computing Standards

Subject: General Tech | October 4, 2012 - 05:28 PM |
Tagged: qualcomm, hsa, APU, amd, AFDS

The HSA Foundation announced today that Qualcomm would be joining as its newest Founder-level member. The mobile ARM System on a Chip company joins AMD, ARM, Imagination Technologies (the company who licenses out PowerVR graphics), MediaTek, Samsung, and Texas Instruments. Reportedly, the HSA Foundation has doubled its total members since its inception in June where it was announced at the AMD Fusion Developer Summit (AFDS 2012).

500px-QualcommLogo.png

Senior Vice President of Engineering at Qualcomm Jim Thompson has stated that the company is joining the HSA Foundation in an effort to standardize aspects of heterogenous computing. Those programming and hardware standards will then be incorporated into devices running future Snapdragon ARM processors.

HSA Foundation President Phil Rogers welcomed the mobile communications giant to the organization by stating the following.

“It’s great to see an innovative company like Qualcomm, which has revolutionized the wireless communications market, placing their support behind HSA.”

It is unclear from the press release where Qualcomm and the HSA Foundation will go from here, but it is promising to see additional companies lending their expertise to further heterogeneous computing standards. Here's hoping that the HSA Foundation is the opposite of the PC Gaming Alliance and actually gets things done to further the technology. After all, AMD is betting the company on APUs and could likely benefit from a big HSA programming standard push and the low power computing prowess of the ARM chip designers in its ranks.

Podcast #206 - Corsair 550D Chassis, AMD licensing ARM, AMD Tahiti 2 GPUs and more!

Subject: General Tech | June 14, 2012 - 02:53 PM |
Tagged: tahiti 2, podcast, nvidia, Intel, hsa, corsair, arm, amd, 550d

PC Perspective Podcast #206 - 06/14/2012

Join us this week as we talk about the Corsair 550D Chassis, AMD licensing ARM, AMD Tahiti 2 GPUs and more!

You can subscribe to us through iTunes and you can still access it directly through the RSS page HERE.

The URL for the podcast is: http://pcper.com/podcast - Share with your friends!

  • iTunes - Subscribe to the podcast directly through the iTunes Store
  • RSS - Subscribe through your regular RSS reader
  • MP3 - Direct download link to the MP3 file

Hosts: Jeremy Hellstrom, Josh Walrath, Allyn Malvantano and Scott Michaud

This Podcast is brought to you by MSI!

Program length: 1:22:58

Program Schedule:

  1. 0:00:20 Introduction
  2. 1-888-38-PCPER or podcast@pcper.com
  3. http://pcper.com/podcast
  4. http://twitter.com/ryanshrout and http://twitter.com/pcper
  5. 0:02:00 ioSafe SoloPro and Synology DiskStation 212+ Review
  6. 0:13:05 Origin EOS17 Gaming Notebook Review
  7. 0:18:00 Corsair Obsidian 550D Case Review
  8. 0:22:00 This Podcast is brought to you by MSI!
  9. 0:24:10 AMD, ARM, Ti, Imagination and MediaTek for HSA Foundation
  10. 0:34:30 AMD licenses ARM Cortex-A5 for APUs
  11. 0:39:45 Sapphire passive Radeon HD 7770
  12. 0:42:50 ASUS ROG laptop first with 802.11ac
  13. 0:47:50 AMD could be releasing Tahiti 2 GPU next week
  14. 0:49:16 Unreal Engine 4 looks pretty awesome...
  15. 0:55:05 AMD Wireless Display standard coming soon
  16. 0:56:45 Apple does indeed release high-res 15" laptop
  17. 1:02:00 New MacBooks Sporting 6Gb/s Samsung 830 Series SSD Controllers
  18. 1:04:18 AMD Kevari 3rd gen APU to hit 1 TFLOPS performance
  19. 1:06:45 Link_A_Media controller explored
  20. 1:09:45 AMD FirePro W600 launched
  21. 1:13:55 Hardware / Software Pick of the Week
    1. Ryan: That Doctor he was getting drunk with
    2. Jeremy: It's heeere and on the Leaderboard
    3. Josh:  Not for the faint of heart. Or wallet.
    4. Allyn: Windows 8 Release Preview is out
    5. Scott: Mount and Blade: Warband: Napoleonic Wars (because you can never have too many subtitles)
    6. Tim: Corsair Obsidian 550D I've been drooling over this since CES! )
  22. 1-888-38-PCPER or podcast@pcper.com
  23. http://pcper.com/podcast   
  24. http://twitter.com/ryanshrout and http://twitter.com/pcper
  25. 1:22:00 Closing

 

AMD Licenses ARM Technology: AMD Leans on ARM for Security

Subject: Processors | June 13, 2012 - 10:00 AM |
Tagged: TrustZone, hsa, Cortex-A5, cortex, arm, APU, amd, AFDS

Last year after that particular AFDS, there was much speculation that AMD and ARM would get a whole lot closer.  Today we have confirmed that in two ways.  The first is that AMD and ARM are founding members of the HSA Foundation.  This endeavor is a rather ambitious project that looks to make it much easier for programmers to access the full computer power of a CPU/GPU combo, or as AMD likes to call them, the APU.  The second confirmation is one that has been theorized for quite some time, but few people have actually hit upon the actual implementation.  This second confirmation is that AMD is licensing ARM cores and actually integrating them into their x86 based APUs.

HSAFoundation-FINAL-Desktop.png
 
AMD and ARM are serious about working with each other.  This is understandable as both of them are competing tooth and nail with Intel.
 
ARM has a security functionality that they have been working with for several years now.  This is called ARM TrustZone.  It is a set of hardware and software products that provide a greater amount of security in data transfer and transactions.  The hardware basis is built into the ARM licensed designs and is implemented in literally billions of devices (not all of them enabled).  The biggest needs that this technology addresses are that of secure transactions and password enabled logins.  Money is obviously quite important, but with identity theft and fraud on the rise, secure logins to personal information or even social sites are reaching the same level of importance as large monetary transactions.
 
AMD will actually be implementing a Cortex-A5 processor into AMD APUs that will handle the security aspects of ARM TrustZone.  The A5 is the smallest Cortex processor available, and that would make sense to use it in a full APU so it will not take up an extreme amount of die space.  When made on what I would assume to be a 28 nm process, a single A5 processor would likely take up as little as 10 to 15 mm squared of space on the die.
 
This is not exactly the licensing agreement that many analysts had expected from AMD.  It is a start though.  I would generally expect AMD to be more aggressive in the future with offerings based on ARM technologies.  If we remember some time ago Rory Read of AMD pronounced their GPU technology as “the crown jewel” of their IP lineup, it makes little sense for AMD to limit this technology just to standalone GPUs and x86 based APUs.  If AMD is serious about heterogeneous computing, I would expect them to eventually move into perhaps not the handheld ARM market initially, but certainly with more server level products based on 64 bit ARM technology.
 
cor_a5.jpg
 
Cortex-A5: coming to an AMD APU near you in 2013/2014.  Though probably not in quad core fashion as shown above.
 
AMD made a mistake once by selling off their ultra-mobile graphics group, Imageon.  This was sold off to Qualcomm, who is now a major player in the ARM ecosystem with their Snapdragon products based on Adreno graphics (“Adreno” is an anagram of “Radeon”).  With the release of low powered processors in both the Brazos and Trinity line, AMD is again poised to deliver next generation graphics to the low power market.  Now the question is, what will that graphics unit be attached to?
 
 
Source: AMD

AFDS 2012: HSA Foundation Joins AMD, ARM, Ti, Imagination and MediaTek with Open Architecture

Subject: Graphics Cards, Processors | June 12, 2012 - 01:31 PM |
Tagged: texas instruments, mediatek, imagination, hsa foundation, hsa, arm, amd, AFDS

Today is a big day for AMD as they, along with four other major players in the world of processors and SoCs, announced the formation of the HSA Foundation.  The HSA Foundation is a non-profit consortium created to define and promote an open approach to heterogeneous computing.  The primary goal is to make it easier for software developers to write and program for the parallel power of GPUs.  This encompasses both integrated and discrete of which the HSA (heterogeneous systems architecture) Foundation wants to enable users to take full advantage of all the processing resources available to them.

On stage at the AMD Fusion Developer Summit in Bellevue, WA, AMD announced the formation of the consortium in partnership with ARM, Imagination Technologies, MediaTek, and Texas Instruments; some of the biggest names in computing. 

The companies will work together to drive a single architecture specification and simplify the programming model to help software developers take greater advantage of the capabilities found in modern central processing units (CPUs) and graphics processing units (GPUs), and unlock the performance and power efficiency of the parallel computing engines found in heterogeneous processors.

HSA Foundation_Logo.png

There are a lot of implications in this simple statement and there are many questions that are left open ended to which we hope to get answered this week while at AFDS.  The idea of a "single architecture specification" set a lot of things in motion and makes us question the direction of both AMD and the traditionally ARM-based companies of the HSA Foundation will be moving in.  AMD has had the APU, and the eventual complete fusion of the CPU and GPU, on its roadmap for quite a few years and has publicly stated that in 2014 they will have their first fully HSA-capable part.  We are still assuming that this is an x86 + Radeon based part, but that may or may not be the long term goal; ideas of ARM-based AMD processors with Radeon graphics technology AND of Radeon based ARM-processors built by other companies still swirl amongst the show.  There are even rumors of Frankenstein-like combinations of x86 and ARM based products for niche applications.

hsa01.jpg

Looks like there is room for a few more founding partners...

Obviously ARM and others have their own graphics IP (ARM has Mali, Imagination Technology has Power VR) and those GPUs can be used for parallel processing in much the same way that we think of GPU processing on discrete GPUs and APUs today.  ARM processor designers are well aware of the power and efficiency benefits of utilizing all of the available transistors and processing power correctly and the emphasis on an HSA-style system design makes a lot of sense moving forward.  

My main question for the HSA Foundation is its goals: obviously they want to promote the simplistic approach for programmers, but what does that actually translate to on the hardware side?  It is possible that both x86 and ARM-based ISAs can continue to exist with libraries and compilers built to correctly handle applications for each architecture, but that would seem to me to be against the goals of such a partnership of technology leaders.

amdarmcombo.jpg

In a meeting with AMD personnel, the most powerful and inspiring idea from the HSA Foundation is summed up with this:

"This is bigger than AMD.  This is bigger than the PC ecosystem."

The end game is to make sure that all software developers can EASILY take advantage of both traditional and parallel processing cores without ever having to know what is going on under the hood.  AMD and the other HSA Foundation members continue to tell us that this optimization can be completely ISA-agnostic – though the technical blockages for that to take place are severe. 

AMD will benefit from the success of the HSA Foundation by finally getting more partners involved in promoting the idea of heterogeneous computing, and powerful ones at that.  ARM is the biggest player in the low power processor market responsible for the Cortex and Mali architectures found in the vast majority of mobile processors.  As those partners trumpet the same cause as AMD, more software will be developed to take advantage of parallel computing and AMD believes their GPU architecture gives them a definite performance advantage once that takes hold.  

What I find most interesting is the unknown – how will this affect the roadmaps for all the hardware companies involved?  Are we going to see the AMD APU roadmap shift to an ARM-IP system?  Will we see companies like Texas Instruments fully integrate the OMAP and Power VR cores into a single memory space (or ARM with Cortex and Mali)?  Will we eventually see NVIDIA jump onboard and lend their weight towards true heterogenous computing?

hsa02.jpg

We have much more the learn about the HSA Foundation and its direction for the industry but we can easily say that this is probably the most important processor company collaboration announcement in many years – and it does so without the 800 pound gorilla that is Intel in attendance.  By going after the ARM-based markets where Intel is already struggling to compete in, AMD can hope to create a foothold with technological and partnership advantages and return to a seat of prominence.  This harkens back to the late 1990s when AMD famously put together the "virtual gorilla" with many partners to take on Intel.

Check out the full press release after the break!

AMD Fusion Developer Summit 2012 - What to expect

Subject: General Tech, Processors, Shows and Expos | June 7, 2012 - 06:49 PM |
Tagged: hsa, fusion, amd, AFDS

One of the best show experiences I had last year was a surprise to me - AMD's first annual Fusion Developer Summit (AFDS) was hosted in the Seattle / Bellevue area.  I say that it was a surprise only because the inaugural year for vendor-specific shows like this tend to be pretty bland and lack interesting information, but that wasn't the case in 2011.  We saw ARM get on stage with AMD to talk about the idea of "dark silicon" and how to prevent it, we saw the first AMD Trinity notebook and even got details of the Tahiti GPU architecture well ahead of release.  

We expect even better things in 2012.

afds.jpg

While I don't know exactly what surprises will be on display this year I am looking forward to seeing the improvement from software developers after having another 12 months to work on APU-accelerated applications.  HSA (heterogeneous system architecture) has been getting a lot of buzz from AMD and the industry as we push towards a combined memory address space and the ultimate acceleration of programs across both serialized and parallel processors on the same die.

If you are in the Seattle / Bellevue area and you have the ability to attend AFDS, I would highly encourage you to do so.  You'll have access to:

  • Keynotes
  • Never before seen demos
  • Technical tracks and sessions to learn about HSA and programming for it

If you can't make it though, you should definitely follow the whole event right here at PC Perspective - the easiest way is to keep track of our AFDS tag to make sure you don't miss any of the potentially industry shifting news! 

You can also expect us to have a live blog from the event as well!

AMD shows 18mm thin reference ultrathin notebook based on Trinity

Subject: Graphics Cards, Processors, Mobile | February 2, 2012 - 02:02 PM |
Tagged: amd, trinity, hsa, ultrabook, ultrathin

Today at the AMD Financial Analyst day in Sunnyvale, Lisa Su, Senior Vice President and General Manager, Global Business Units, showed off a reference design from Compal of an 18mm think ultrathin notebook that they are obviously hoping to compete with Intel's Ultrabook push.

ultrathin1.jpg

The notebook is based on AMD's upcoming Trinity APU that improves on the CPU and GPU performance of the currently available Llano APU.  There weren't many details though Su did state they were hoping for prices in the $600-800 range would could but a lot of pressure on Intel. 

ultrathin2.jpg