With GDDR6 Memory Production Scaling Up, New GPUs Won’t Be Far Behind
It’s been almost two years since Nvidia launched its Pascal architecture and the gaming community has been hungry to hear what’s coming next. It has stayed hungry, for the most part, because Nvidia hasn’t had much to say. Rumors from several months ago predicted a launch at GDC or GTC, but we weren’t able to shake anything loose suggesting the rumors had merit. What data we have scrounged up has suggested a launch from Team Green later this year, and new information also indirectly points in that direction.
GamersNexus cornered SK Hynix representatives at GDC last week, and discovered something interesting — plans to ramp up to mass production on GDDR6 within three months. Factor in some additional manufacturing lead time for boards, and we could be looking at new GPUs in the June-August time frame. That would fall neatly in line with our expectations of a launch in later 2018, though we’re leaving some wiggle room — a great deal depends on whether SK Hynix is a major launch partner, what Nvidia’s GPU yields look like, how good VRAM yields are at any given clock rate, and what kind of volume Nvidia wants to bring to market.
It’s not always clear how these issues will shake out. Some of you may recall that AMD and Nvidia launched their 14nm refreshes only after that process node had been used for mobile products and was ostensibly on its second revision. Neither fact prevented abysmal availability from either vendor. Neither AMD nor Nvidia were able to push GPUs into market at anything like acceptable rates, and this was particularly frustrating on Nvidia’s part, since it launched its Pascal GPU family 1-2 months (depending on SKU) ahead of AMD’s RX 480.
This time around, availability should be better. TSMC’s 12nm node is an optimized variant of the 16nm node it already has in production, and shouldn’t have the same problems that plagued 14nm back in 2016. We haven’t heard anything from AMD about any Radeon consumer refreshes this year, but if the company does tweak Polaris or Vega for GlobalFoundries’ new 12nm node, it should have a similarly smooth ramp to launch (again, GF’s 12nm node is an optimized 14nm node).
According to GamersNexus, GDDR6 is expected to be ~20 percent more expensive than GDDR5 at launch, with that to decline to a 10 percent premium over time as GDDR6 ramps up and GDDR5 declines. Hynix expects to ship GDDR6 at 16Gbps signal rates and above, in densities of 8Gb and 16Gb (1GB and 2GB). A GPU equipped with 16GB of GDDR6 in total would target a 256-bit memory bus, and I wouldn’t be surprised to see exactly that kind of sweet spot positioning targeted by Nvidia for its next consumer product.
The larger question, of course, is whether cryptocurrency prices will have declined enough that anyone can afford a card. The jury is out on this one, but so far, things seem to be cooling off from where they were 6-8 weeks ago. GPU prices are still inflated well above normal, but spot-checks this evening suggested that costs are coming down for AMD and NV hardware. Whether that’ll stay the case is anyone’s guess, and we’re far from declaring victory on that front.
Continue reading
Protect Your Online Privacy With the 5 Best VPNs
Investing in a VPN is a smart choice right now, but the options are vast. To help narrow things down a bit, we've rounded up five of our very favorite consumer services.
RISC-V Tiptoes Towards Mainstream With SiFive Dev Board, High-Performance CPU
RISC V continues to make inroads across the market, this time with a cheaper and more fully-featured test motherboard.
The PlayStation 5 Will Only Be Available Online for Launch Day
The PlayStation 5 isn't going to be available in stores on launch day, and if you want to pick up an M.2 SSD to expand its storage, you'll have some time to figure out that purchase.
ARMing for War: New Cortex-A78C Will Challenge x86 in the Laptop Market
ARM took another step towards challenging x86 in its own right with the debut of the Cortex-A78C this week. The new chip packs up to eight "big" CPU cores and up to an 8MB L3 cache.