Read more.Boasts the new chips will offer sevenfold performance increase over current GDDR5.
Read more.Boasts the new chips will offer sevenfold performance increase over current GDDR5.
Place bets - Team Green or Team Red to implement HMB2 first?
fiver on team red!
Old puter - still good enuff till I save some pennies!
Isn't Team Red part of the group that owns HBM? Would be weird if they didn't implement it first
9 layers of TSV is an impressive achievement, although I guess that once you've mastered one layer and aligning that, the rest are easy.
Interesting fact - after the wafers are thinned for use in the stack (compare the top and other dies in the 4Hi stack picture) the wafers are bendy: http://1.bp.blogspot.com/-QQFJ2xq-Q_E/UQwBpc965UI/AAAAAAAAACI/2dZpRLZDWWY/s1600/taiko02.jpg which isn't good, so they need to use special transports (or leave the outside unthinned as a rigid exoskeleton).
Well, if the discussions in http://forums.hexus.net/hexus-news/3...ml#post3597275 are even semi-accurate ( ), nvidia don't even have working Pascal silicon yet, and they have no experience releasing consumer products with HBM or on interposers. Historically ATi/AMD have always been earlier proponents of new memory technologies: I'm pretty sure they were ahead of nvidia with GDDR4, GDDR5, and of course HBM. it would be something of a turn-around for nvidia to beat them to the release of HBM2 GPUs; it'd be the first time in more than a decade that they'd be first to exploit a new memory technology...
However HBM2 isn't even needed for NV to win, as shown with HBM1. Just put out faster GDDR5 (8ghz etc), up the bus, and your done for the next gen chips. AMD made a mistake going HBM1 IMHO and forced costs too high, shortages, and as a result no gpu profits, leading to losses again and again. NV chips get far more from increasing gpu speed than mem speed in overclocking. There is more then enough bandwidth for their chips as shown with 980 vs. 780ti going to a smaller bus, with better tech, still beat the older model handily. OCing mem already hits 8.2ghz in reviews and 980ti ships with 7ghz. Next rev coming up we could see gddr5 hit near 9ghz making 8ghz the new norm, so plenty of room to run, and NV could even go 512bus for the top end if forced.
http://www.anandtech.com/show/9883/gddr5x-standard-jedec-new-gpu-memory-14-gbps
Lastly, GDDR5x. Blows out options I already mentioned. Note low costs to switch also from GDDR5. 14GBps is just 12% shy of HBM1 which is already overkill. 448 is a massive increase over 336 of 980ti, so plenty of room to grow without HBM2. NV could wait until it matures before switching (which AMD should have done).
"While internally a GDDR5X chip is different from a GDDR5 one, the transition of the industry to GDDR5X is a less radical step than the upcoming transition to the HBM (high-bandwidth memory) DRAM. Moreover, even the transition from the GDDR3/GDDR4 to the GDDR5 years ago was considerably harder than transition to the GDDR5X is going to be in the coming years."
AS they say we could see cards in 6 months with this, and I'd rather be the guy with cheaper/easy to make GDDR5x (with plenty of bandwidth), than the guy trying HBM2 which will surely jack up the price and kill profits (see hbm1, and how well that went for AMD bottom line). Hype and bandwidth didn't sell furyx or HBM when 980ti sat next to it. Margins were not helped for AMD either (nor shortages). HBM2+new process+new chip design=shortages again and high prices netting few profits. Go with easy, mass production and higher margins, cheaper cost to produce etc. That's what got NV to 82% discrete share. If that massive bandwidth had been needed, it would have been different, but that wasn't the case. Better compression tech, higher speed mem etc killed the need for more bandwidth and it looks like we'll see it yet again. Also, NV wisely concentrated on 97% of the market (2560x1600 and under), while AMD forgot it. Win where most of your audience actually plays, or people buy the other guy. 4K matters to so few, there is no point making it your centerpiece in your slides like AMD had to. Maybe next year that will be a point of contention, but not likely much this year and most 4K people buy multi-gpu setups (same with multi monitor people, as steam shows for both). AMD workers unfortunately have been plagued by bad management decisions for a years. Paying 3x more than ATI was worth (then writing down that exact amount...LOL), Consoles, HBM, chasing low dollar APU instead of cpu kings etc all wrong decisions. Those and other decisions have killed the great work the employees can put out. Bummer. Let's hope ZEN's die size is 1.5-2x INtel's cpu side of the die! That is a KING that can sell for $50 above all Intel chips at every level. We pay for 1st place, but 90% of people want a massively discounted 2nd place offering. Go for the KING or go home. AMD needs to stop trying to be a cheaper 2nd that's "good enough". Good enough=losses yearly. Having the king makes NV/Intel profitable yearly.
with reports elsewhere showing pascal wont be here till H2 / Q3 - with AMD being to market in H1/Q2 they can lead the way this time
As memory increases in clockspeed so does the complexity of PCB routing.
HBM has little or none of this disadvantage so will ultimatly be much much easier to implement. GDDR5/6/7/8/X/29.3/etc will be for low end parts only.
There are currently 1 users browsing this thread. (0 members and 1 guests)