Upcoming Nvidia GTX 1080 pictured, won't use HBM2

It’s no secret that AMD and Nvidia are planning to release new GPUs in the following few months, but there are nonetheless questions about configurations and technology. initially, we anticipated those new playing cards to apply the HBM2 reminiscence wellknown, but the persistent rumor has been that AMD’s upcoming Polaris would be GDDR5 based. That now seems to be proper for Nvidia as properly.

Leaked pix of the GP104 die show the chip with a wellknown memory interface — either GDDR5 or GDDR5X (it’s difficult to tell). computer angle is reporting that there’s an ongoing rumor the GTX 1080 will use GDDR5X, which can be feasible, but could absolutely seem to be a bit early — whilst we spoke to Micron approximately the memory, the employer implied that the manufacturing ramp might be later inside the year. With JEDEC qualification and precise preliminary manufacturing yields, but, it’s feasible that Nvidia may want to deploy GDDR5X on high-end GPUs, and it wouldn’t be the primary time that a GPU producer used a confined reminiscence run. AMD changed into the only business enterprise to make tremendous use of GDDR4 or HBM, in spite of everything.

As for the shortage of HBM2, it’s now not sudden at this juncture. All present day information suggests that HBM2 will arrive with the launch of the excessive-end “massive” GPUs — Nvidia’s GP100 and AMD’s new Vega structure, currently predicted within the q4 2016 / Q1 2017 timeframe. We don’t anticipate AMD to tap HBM for Polaris this day trip, especially not if they’ve focused on scaling 1-2 GPU cores for laptop and cell. recall, HBM calls for an entirely distinctive reminiscence controller architecture than GDDR5, and neither AMD nor Nvidia goes to need to construct two unique memory controllers for the identical GPU middle unless actually vital. It’s a whole lot less complicated to trust Nvidia may do a GDDR5/GDDR5X mixture than an HBM2 / GDDR5 move, and the same is genuine for AMD.

We don’t anticipate Polaris or Pascal to ship with HBM aid in two months (if we’re wrong in this, we’ll have us a bite of crow), and we’re certainly not involved approximately what that means for the GPUs themselves. during the last ten years, each AMD and Nvidia have followed a wide form of memory bus sizes and reminiscence technologies to fit the GPUs they shipped. It’s now not unusual to look both agencies undertake a couple of reminiscence requirements for multiple merchandise — in past years, DDR2, GDDR3, and GDDR5 all shared space in Nvidia’s product stack.

HBM2 is the long-time period future for GPU reminiscence and that’s still going to be genuine if the following-era cards from AMD and Nvidia use GDDR5(X). both companies have many years of experience in matching bandwidth to images processor; the ultimate time we recollect a GPU being terminally memory bandwidth-sure turned into the GeForce 2 generation.