For those who have not heard about the thing, HSA is a design for processing data that AMD is convinced will generate much better performance for a whole new generation of electronic gear.
AMD is preparing the ground before it releases chips based around the standard. Word on the street is that it wants to avoid a cock up like the launch of its first Opteron chip.
In 2003, AMD launched the Opteron with the AMD64 x86 64-bit instruction set. Four months later AMD launched the consumer version of the instruction set in Athlon 64 branded chips. While the chip was the best on the market until Intel matched it in 2003 there was little software to run then. Microsoft did produce a late Windows XP AMD64 edition but it was too little and too late.
AMD wants desperately to convert journalists and the IT industry to support the technology a year before its first HSA chip arrives. This should be the end of the year, meaning there should be software that takes advantage of AMD’s latest chips around launch time.
This time AMD has prepared the ground well by forming the HSA Foundation. HSA has the backing of chip and system makers, with only AMD’s main rivals Nvidia and Intel been sitting back and watching.
Going through AMD’s bumph, you can’t help but wonder why everyone is not rushing to sign up for what appears to be a cure for cancer.
AMD claims HSA means faster and more power-efficient personal computers, tablets, smartphones and cloud servers. It works with hUMA, which is the latest way for processors to access the memory inside an Accelerated Processing Unit.
HSA allows developers to take control of the GPU and makes graphics an equal partner with the CPU (central processing unit) and other processors.
HSA uses “heterogenous queuing,” which allows software to communicate with the GPU, treating it as an equal partner along side a CPU when it comes to accessing data quickly. The software does not have to wait for the CPU when what it really needs to is to access the GPU.
So far AMD has converted Imagination Technologies, ARM, Samsung, Mediatek, Qualcomm, and Texas Instruments to the technology. The reason they are interested might be simply that it gives them access to APU technology that they would not have for years.
But why didn’t Intel sign up? Officially Chipzilla’s view is that HSA will not provide enough the flexibility for its customers. This is code for “AMD is involved and we do not want to touch it.” In many ways this is pure complacency. It has control of the market and even if AMD and all its chums get their technology under the bonnet of low priced servers it is not going to harm Intel that much. Taking part in a project that will help AMD on the other hand will probably do longer term damage so it is better to sit back and wait.
Nvidia is using a similar philosophy – we don’t want what our rival wants. However, the feeling is that while Intel can afford to be stroppy, Nvidia is being dumb not supporting HSA, particularly if it wants developers to run more GPU cores in Tegra.
Nvidia has been supporting programming languages outside its own CUDA, but a lack of HSA could leave it isolated. The only logic for this is that HSA’s advantages are in applications and servers, both areas where AMD and ARM vendors can pitch its upcoming system on chip APUs. Nvidia wants to push into the smartphone and tablet markets where it already has a presence and control the games market.
AMD’s evangelists will not be losing much sleep in not recruiting Intel or Nvidia. With the likes of ARM on board it means that Intel’s biggest enemies will have signed up and the standard will be out there when it makes its first chips.
What will be interesting is what happens next year when the HSA chips start appearing. If they really shake up the industry the way AMD hopes, then Intel and Nvidia could end up kicking themselves as they lose whole chunks of their industry to an innovation steamroller from their rivals.