For many years now, high-end stand-alone graphics processors have used a type of memory known as GDDR5. However, to drive a substantial increase in memory bandwidth to feed increasingly powerful graphics processors, graphics chipmakers are looking to transition to a new memory standard known as high-bandwidth memory, or HBM for short.
Advanced Micro Devices (AMD 3.46%) recently launched its first graphics processor family using the first version of HBM. The use of HBM has allowed AMD to gain a solid memory bandwidth advantage over the highest-end graphics chips from rival graphics specialist NVIDIA (NVDA 1.55%), but with a major caveat: The first version of HBM was limited to just 4 gigabytes of memory.
The next version of HBM, known as HBM2, is expected to enable significantly larger capacities as well as greater speeds. Both Advanced Micro Devices and NVIDIA are expected to adopt this technology in their next-generation graphics architectures, slated to launch in 2016.
Interestingly enough, though, there was a report out of website WCCFTech not too long ago claiming that AMD would have "priority access" to the HBM2 supply out of memory maker SK Hynix. The implication here would be that AMD would be able to get the jump on NVIDIA in releasing next-generation graphics cards.
Is there any truth to this report? Let's take a closer look.
First things first: reaching out to AMD and NVIDIA
The first thing I did was to reach out to both AMD and NVIDIA for comment on this story. AMD's Drew Praire responded, "We haven't said this and I wouldn't expect we would talk about our supplier arrangement on future products publicly."
NVIDIA's Ken Brown offered this: "We can't comment on rumors or speculation."
Well, I had to ask!
Let's see what SK Hynix has to say
At MemCon, a well-known memory industry event, SK Hynix -- which WCCFTech claims is giving AMD "priority access" to HBM -- gave a presentation on HBM. In the presentation, Hynix boasted that it would begin volume production on HBM1 in early 2015 and that "design wins" were in progress "with major SoCs in multiple markets."
In the presentation, SK Hynix further went on to explain the virtues of HBM as a next-generation memory technology for high-performance processors such as GPUs, citing lower power consumption, a smaller physical footprint, and lower latency compared to current DDR solutions.
Nothing conclusive, but here's why I doubt WCCFTech's report
From SK Hynix's presentation, it would seem that the memory maker is eager to sell HBM2 to as many customers as possible. I doubt that SK Hynix's other customers -- many of which are probably larger than AMD -- would be all too thrilled to know that they will have to wait in line behind AMD, especially with memory giant Samsung (NASDAQOTH: SSNLF) having said that it will go into production on HBM2 in the first half of 2016.
If SK Hynix tells potential customers to take a hike because AMD has dibs on HBM2, then I suspect they will take their business on over to Samsung. Unless AMD is writing huge checks to secure "priority access" (which, given AMD's relatively fragile balance sheet position, seems unlikely), I doubt that AMD will have the "priority access" that WCCFTech claims it has.
