This site may earn chapter commissions from the links on this page. Terms of use.

GTX980M feature

Earlier this year, AMD became the first GPU vendor to coil out graphics cards that utilized HBM, or Loftier Bandwidth Memory. We've discussed this and other new retention technologies extensively over the past 10 months, and there's broad agreement that HBM is the GPU RAM of Tomorrow for both AMD and Nvidia. What's less clear, withal, is how quickly HBM and its successor, HBM2, volition waterfall downwardly the product lines. Now there'southward a rumor that Nvidia could adopt a new type of GDDR5, GDDR5X, for some of its Pascal GPUs in 2016.

Back on September 1, Micron made a pair of announcements apropos the future of its GDDR5 evolution. Showtime, the company stated that it had adult new 8Gbps GDDR5 modules, which would let companies to double RAM densities without calculation more RAM chips. That's important relative to scaling concerns, since one of the problems with GDDR5 scaling is that you can't just keep embedding more memory around the GPU. At that place'south a minimal amount of room betwixt traces to go along the electrical signaling make clean.

HBM-1

This slide shows the problem of stacking 2D RAM configurations indefinitely. 8Gb GDDR5 modules does help with that — for a little while.

Second, and more than interesting for our purposes, was Micron's declaration that it had created a new type of GDDR5, dubbed GDDR5X, that would offer 10-14Gbps signaling rates. Existing transfer rates for GDDR5 are 6-8Gbps, which is why this is being spun in some circles as a "doubling" of GDDR5 performance. According to micron, it can do this past doubling the amount of data that GDDR5X prefetches, thereby allowing for devices with every bit much as 16Gbps worth of throughput.

GDDR5X

Could Micron use GDDR5X to slow HBM adoption — and will Nvidia adopt it as a solution for next-generation hardware? Nosotros're dubious in both cases.

The GDDR5 Dilemma

Micron's month-onetime announcement is getting press-fourth dimension because there's a kernel of truth to all the GDDR vs. HBM rumors. For all of HBM / HBM2's benefits, these are new retentiveness technologies, and new technologies typically command a premium compared to long-established alternatives. AMD confined its first-generation of HBM products to the upper end of the market, where it could command a commensurate price for the technology. Next twelvemonth, HBM2 is expected to boost capacities and cut costs, but it'southward not realistic to expect the technology to roll out at every toll point.

Looking dorsum across the years, in that location'southward e'er been a transition phase betwixt old and new memory standards. Dorsum in the HD 3000 and 4000 flow, AMD regularly rolled out its newest retention standards on upper-range cards, while midrange and budget models relied on older applied science. GDDR5 debuted on the Hard disk 4870 in 2008, but wasn't deployed on anything simply the flagship until 2009. Equally late as 2012, AMD was all the same using GDDR3 for the lowest-stop GCN ane.0 GPUs.

Nosotros tin can assume, therefore, that AMD and Nvidia will go along to use GDDR5 in mod GPUs, even as those chips curlicue out on 14/16nm. This is partly because the benefits of HBM, like increased RAM densities, smaller PCBs, and decreased power consumption grow every bit the corporeality of RAM on the GPU grows. If you merely demand a 2GB frame buffer, HBM doesn't offer much of an reward compared to a conventional solution. Lower-end GPUs often don't pack enough firepower to saturate a huge corporeality of RAM bandwidth — a two-broad HBM configuration would withal offer 256GB/southward of RAM bandwidth. If your $150 GPU doesn't demand that much capability, paying a premium for an HBM-based design makes no sense.

Thus far, nosotros've only discussed GDDR5. What nearly GDDR5X? In that location are two problems with this technology. First, it'southward proprietary to Micron and not expected earlier the terminate of 2016, when both AMD and Nvidia volition hopefully have shipped next-generation chips from top to bottom. Second, nonetheless, information technology just addresses the bandwidth part of GDDR5's problem. At that place'due south always a power toll associated with increasing retentivity bandwidth, and simply doubling prefetch isn't going to magically reduce GDDR5'south power consumption. One of the advantages of HBM that nosotros covered long before AMD began talking it up this yr is that HBM significantly improves GPU power efficiency. Micron will be pushing its technology in the contrary direction.

If rumors are true, AMD and Nvidia take both already taped out their next-generation designs — and that means they won't be retrofitting them for GDDR5X at the last second. The change from GDDR5 to HBM requires a completely different memory controller, and it'due south a not-trivial switch to move from i to the other.

I'm not going to say that GDDR5X won't come to market, but it would surprise me to encounter AMD or NV use it. HBM and HBM2 have as well many advantages in likewise many categories over GDDR5 as a whole, and bandwidth is only one part of that equation. HBM allows for products similar the Radeon R9 Nano, which packs far more performance into a smaller space than was possible before. We look to see that trend spread to lower-finish cards as well, and when every millimeter counts, HBM/HBM2 is always going to have an advantage over GDDR5/5X.