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.

Author:
Subject: Processors
Manufacturer: AMD

Jaguar Hits the Embedded Space

 

It has long been known that AMD has simply not had a lot of luck going head to head against Intel in the processor market.  Some years back they worked on differentiating themselves, and in so doing have been able to stay afloat through hard times.  The acquisitions that AMD has made in the past decade are starting to make a difference in the company, especially now that the PC market that they have relied upon for revenue and growth opportunities is suddenly contracting.  This of course puts a cramp in AMD’s style, but with better than expected results in their previous quarter, things are not nearly as dim as some would expect.

Q1 was still pretty harsh for AMD, but they maintained their marketshare in both processors and graphics chips.  One area that looks to get a boost is that of embedded processors.  AMD has offered embedded processors for some time, but with the way the market is heading they look to really ramp up their offerings to fit in a variety of applications and SKUs.  The last generation of G-series processors were based upon the Bobcat/Brazos platform.  This two chip design (APU and media hub) came in a variety of wattages with good performance from both the CPU and GPU portion.  While the setup looked pretty good on paper, it was not widely implemented because of the added complexity of a two chip design plus thermal concerns vs. performance.

soc_arch.jpg

AMD looks to address these problems with one of their first, true SOC designs.  The latest G-series SOC’s are based upon the brand new Jaguar core from AMD.  Jaguar is the successor to the successful Bobcat core which is a low power, dual core processor with integrated DX11/VLIW5 based graphics.  Jaguar improves performance vs. Bobcat in CPU operations between 6% to 13% when clocked identically, but because it is manufactured on a smaller process node it is able to do so without using as much power.  Jaguar can come in both dual core and quad core packages.  The graphics portion is based on the latest GCN architecture.

Read the rest of the AMD G-Series release by clicking here!

XFX Announces Malta Dual-GPU Radeon HD 7990

Subject: Graphics Cards | April 24, 2013 - 10:14 PM |
Tagged: xfx, malta, hd 7990, GCN, dual gpu, amd

Now that AMD’s dual-gpu Malta graphics card is official, cards from Add-In Board (AIB) partners are starting to roll in. One such recently announced card is the XFX Radeon HD 7990 card. The XFX card is based on the reference AMD design, which includes two Radeon HD 7970 GPUs in a Crossfire configuration.

The two GPUs can boost up to 1GHz clock speeds and feature a total of 4096 stream processors, 256 texture units, 64 ROPs, and 8.6 billion transistors. The card also includes 3GB of GDDR5 memory per GPU running off a 384-bit bus. It supports AMD’s Eyefinity technology and offers up one DL-DVI and four mini-DisplayPort video outputs.

XFX Radeon HD 7990.jpg

The XFX HD 7990 uses the reference AMD heatsink as well, which includes a massive aluminum fin stack with five copper heatpipes that run the length of the heasink and directly touch the two 7970 GPUs. Three shrouded fans, in turn, keep the heatsink cool.

The dual-GPU monster is eligible for AMD’s Never Settle bundle which includes eight free games. With purchase of the HD 7990 (from any eligible AIB), you get free key codes for the following games:

  • Bioshock Infinite
  • Crysis 3
  • Deus Ex: Human Revolution
  • Far Cry 3
  • Far Cry 3: Blood Dragon
  • Hitman: Absolution
  • Sleeping Dogs
  • Tomb Raider

The XFX press release further assures gamers that the card can, in fact, play Crysis 3 at maximum settings at a resolution of 3840 x 2160. The company did not mention pricing, however.

For those interested in AMD’s new Malta GPU, check out our review as well as how the card performs when paired with a prototype AMD driver that seeks to address some of the frame rating issues exhibited by AMD's Crossfire multi-GPU solution.

Source: XFX

PowerColor Launches Revised Factory Overclocked Radeon HD 7790 OC V2 Graphics Card

Subject: Graphics Cards | April 13, 2013 - 10:07 PM |
Tagged: radeon hd7790, powercolor, GCN, amd, 7790

PowerColor launched a new factory overclocked graphics card recently that is a revision of a previous model. The PowerColor HD7790 OC V2 is based on AMD’s Graphics Core Next (GCN) architecture and measures a mere 180 x 150 x 38mm.

PowerColor Radeon HD7790 1GB GDDR5 OC V2 Graphics Card.jpg

The AMD Radeon HD 7790 GPU features 896 stream processors, 56 texture units, and 80 ROP units. The GPU is clocked at 1000 MHz base and 1030 MHz boost while the 1GB of GDDR5 memory is clocked at the 6Gbps reference speed. PowerColor has fitted the overclocked card with an aluminum heatsink cooled by a single 8mm copper heatpipe and 70mm fan.

The new card features two DL-DVI, one HDMI, and one DisplayPort video outputs. Its model number is AX7790-1GBD5-DHV2/OC. According to Guru3D, the new/revised card is priced at 120 pounds sterling. However, considering the currently available OC (non-V2) card is $150, the revised card is likely to come in around that price when it hits US retailers.

Also: If you have not already, read our latest Frame Rating article to see how the Radeon HD 7790 graphics card stacks up against the competition!

Source:

AMD Hosting an Event for Fans In San Francisco on April 6th

Subject: General Tech | April 3, 2013 - 05:57 AM |
Tagged: prizes, GCN, fan day, APU, amd

AMD has announced that is will be hosting an event for fans in San Francisco this weekend. The AMD Fan Day is free with registration (register here), and fans will give enthusiasts a chance to go hands-on with the company's 2013 hardware lineup, play several newly released (and some not-yet-released) games, talk with industry experts, check out modded PCs, and have a chance to win free hardware and swag from AMD, Corsair, and Gigabyte.

AMD Fan Day.png

Gamers will get a chance to speak with the developers for Bioshock Infinite, Far Cry 3, Crysis 3, Devil May Cry (DMC), and Tomb Raider as well as AMD representatives. VIZIO, IGN, Ubisoft, Sapphire, and Logitech will also be attending the AMD fan day to show off their latest products.

The event will held at City View at Metreon (address below) at 5:30pm on Saturday, April 6th. Best of all, the first 1,000 registered attendees in the door will get a free AMD A8 5600K APU. The first 120 attendees will win both an A8 5600K APU and an A85X motherboard.

AMD Modded PC.jpg

One of the modded PCs that will be on the event floor.

If you're going to be in the area this weekend and are interested in going, be sure to head over to the AMD site and register. It sounds like it should be a fun time, and the free hardware doesn't hurt!
The AMD Fan Day will be held at the following address:

City View at Metreon
135 4th Street
San Francisco, CA 94013

Will you be checking out the AMD fan day to enjoy some gaming and PC hardware?

Source: AMD

GDC 2013: AMD Reveals Radeon Sky Specifications

Subject: Graphics Cards | March 31, 2013 - 03:06 AM |
Tagged: GDC 13, sky 900, sky 700, sky 500, RapidFire, radeon sky, GCN, cloud gaming, amd

Earlier this week, AMD announced a new series of Radeon-branded cards–called Radeon Sky–aimed at the cloud gaming market. At the time, details on the cards was scarce apart from the fact that the cards would use latency-reduction "secret sauce" tech called RapidFire, and the highest-end model would be the Radeon Sky 900. Thankfully, gamers will not have to wait until AFDS after all, as AMD has posted additional information and specifications to its website. At this point, pricing and the underlying details of RapidFire are the only aspects still unknown.

AMD Radeon Sky Lineup_AMD Slide.jpg

According to the AMD site, the company will release three Radeon Sky cards later this year, called Sky 500, Sky 700, and Sky 900. All three cards are passively cooled with aluminum fin heatsinks and are based on AMD's Graphics Core Next (GCN) architecture. At the high end is the Sky 900, which is a dual Tahiti graphics card clocked at 825 MHz. The Sky 900 features 1,792 stream processors per GPU for a total of 3,584. The card further features 3GB of GDDR5 RAM per GPU on a 384-bit interface for a total GPU bandwidth of 480GB/s. AMD claims this dual slot card draws up to 300W while under load. In many respects the Sky 900 is the Radeon-equivalent to the company's professional FirePro S10,000 graphics card. It has similar hardware specifications (including the 5.91TFLOPS of single precision performance potential), but a higher TDP. It is also $3,599, though whether AMD will price the gaming-oriented Sky 900 similarly is unknown.

The Sky 700 steps down to a single-GPU graphics card. This card features a single Tahiti GPU clocked at 900 MHz with 1792 stream processors and 6GB of GDDR5. The graphics card memory uses a 384-bit memory interface for a total memory bandwidth of 264GB/s. Although also a dual slot card like the Sky 900, the cooler is smaller and it draws only 225W under load.

Finally, the Sky 500 represents the low end of the company's cloud gaming hardware lineup. It is the Radeon Sky equivalent to the company's consumer-grade Radeon HD 7870. The Sky 500 features a single Pitcairn GPU clocked at 950 MHz with 1280 stream processors, 4GB of GDDR5 on a 256-bit memory bus, and a rated 150W power draw under load. It further features 154GB/s of memory bandwidth and is a single slot graphics card.

  Sky 900 Sky 700 Sky 500
GPU(s) Dual Tahiti Single Tahiti Single Pitcairn
GPU Clockspeed 825 MHz 900 MHz 950 MHz
Stream Processors 3584 (1792 per GPU) 1792 1280
Memory 6GB GDDR5 (3GB per GPU) 6GB GDDR5 4GB GDDR5
Memory Bus 384-bit 384-bit 256-bit
Memory Bandwidth 480GB/s 264GB/s 154GB/s
TDP 300W 225W 150W
Card Profile dual-slot dual-slot single-slot

Additionally, the Radeon Sky cards all employ a technology called RapidFire that allegedly reduces latency immensely. As Ryan mentioned on the latest PC Perspective Podcast, the Radeon Sky cards are able to stream up to six games. RapidFire is still a mystery, but the company has indicated that one aspect of RapidFire is the use of AMD's Video Encoding Engine (VCE) to encode the video stream on the GPU itself to reduce game latency. The Sky cards will output at 720p resolutions, and the Sky 700 can support either three games at 60 FPS or six games at 30 FPS.

In addition to working with cloud gaming companies Ubitus, G-Cluster, CiiNow, and Otoy, AMD has announced a partnership with VMWare and Citrix. AMD is reportedly working to allow VMWare ESX/ESXi and Citrix XenServer virtual machines to access the GPU hardware directly, which opens up the possibility of using Sky cards to run workstation applications or remote desktops with 3D support much like NVIDIA's VCA and GRID technology (which the company showed off at GTC last week). Personally, I think the Sky cards may be late to the party but is a step in the right direction. Even if cloud gaming doesn't take off, the cards could still be used to great success by enterprise customers if they are able to allow direct access to the full graphics card hardware from within virtual machines!

More information on the Radeon Sky cards can be found on the AMD website.

Source: AMD

Podcast #244 - Frame Rating Launch, HD 7790 vs. GTX 650Ti BOOST, and news from GDC

Subject: General Tech | March 28, 2013 - 03:47 PM |
Tagged: sli, podcast, pcper, nvidia, kepler, HD7790, GTX 560Ti BOOST, GCN, frame rating, crossfire, amd

PC Perspective Podcast #244 - 03/28/2013

Join us this week as we discuss the launch of Frame Rating, HD 7790 vs. GTX 650Ti BOOST, and news from GDC

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: Ryan Shrout, Jeremy Hellstrom, Josh Walrath, and Allyn Malventano

This Podcast is brought to you by MSI!

Program length: 1:19:22

Podcast topics of discussion:
  1. Week in Review:
  2. News items of interest:
  3. 1:12:00 Hardware/Software Picks of the Week:
  4. 1-888-38-PCPER or podcast@pcper.com
  5. Closing/outro

 

Gaming for $150 with the Radeon HD 7790

Subject: Graphics Cards | March 22, 2013 - 01:56 PM |
Tagged: hd 7790, graphics core next, GCN, ea Islands, bonaire, amd

AMD is trying to fill a gap in their product line between the less than $200 HD 7850 and the ~$120 HD 7770 with a $150 card, the HD 7790.  The naming scheme implies two GPUs but this is not the case, it is a single Bonaire GCN chip with 896 stream processors, 56 texture units and an impressive fill rate of up to 1.79 TFLOPS thanks to some optimization of the GCN architecture.  It has 1GB of GDDR5 at 6GHz effective and a CPU speed dependent on the model, in [H]ard|OCP's case the ASUS Radeon HD 7790 DirectCU II OC runs at 1.075GHz.  [H] passed it a Silver Award for being a vast improvement over the 7770 and good competition for the GTX 650 Ti but feel the card does need to be faster.

This card also makes an appearance on our front page, with a lot of Frame Rating charts so you can see not only the raw FPS data you are used to, but also an indept look at how the game is going to 'feel' while you play.

H_7790s.gif

"AMD is launching the Radeon HD 7790 today. This new video card should give the sub-$200 video card segment a kick in the pants. Will it provide enough performance for today's latest games at $149? We will find out, testing the new ASUS Radeon HD 7790 DirectCU II OC with no less than six of today's hottest games."

Here are some more Graphics Card articles from around the web:

Graphics Cards

Source: [H]ard|OCP
Author:
Manufacturer: AMD

A New GPU with the Same DNA

When we talked with AMD recently about its leaked roadmap that insinuated that we would not see any new GPUs in 2013, they were adamant that other options would be made available to gamers but were coy about about saying when and to what degree.  As it turns out, today marks the release of the Radeon HD 7790, a completely new piece of silicon under the Sea Islands designation, that uses the same GCN (Graphics Core Next) architecture as the HD 7000-series / Southern Islands GPUs with a handful of tweaks and advantages from improved clock boosting with PowerTune to faster default memory clocks.

slide02.png

To be clear, the Radeon HD 7790 is a completely new ASIC, not a rebranding of a currently available part, though the differences between the options are mostly in power routing and a reorganization of the GCN design found in Cape Verde and Pitcairn designs.  The code name for this particular GPU is Bonaire and it is one of several upcoming updates to the HD 7000 cards. 

Bonaire is built on the same 28nm TSMC process technology that all Southern Islands parts are built on and consists of 2.08 billion transistors in a 160 mm2 die.  Compared to the HD 7800 (Pitcairn) GPU at 212 mm2 and HD 7700 (Cape Verde) at 120 mm2, the chip for the HD 7790 falls right in between.  And while the die images above are likely not completely accurate, it definitely appears that AMD's engineers have reorganized the internals.

slide03.png

Bonaire is built with 14 CUs (compute units) for a total stream processor count of 896, which places it closer to the performance level of the HD 7850 (1024 SPs) than it does the HD 7770 (640 SPs).  The new Sea Islands GPU includes the same dual tessellation engines of the higher end HD 7000s as well and a solid 128-bit memory bus that runs at 6.0 Gbps out the gate on the 1GB frame buffer.  The new memory controller is completely reworked in Bonaire and allows for a total memory bandwidth of 96 GB/s in comparison to the 72 GB/s of the HD 7770 and peaking theoretical compute performance at 1.79 TFLOPS.

The GPU clock rate is set at 1.0 GHz, but there is more on that later.

Continue reading our review of the Sapphire AMD Radeon HD 7790 1GB Bonaire GPU!!