NVIDIA announces the GTX 980 Ti priced at $650, GTX 980's price slashed to $500
172 replies, posted
[QUOTE=Levelog;47851049]390 and 390x are both HBM cards. Fury will be separate titan esque card.[/QUOTE]
I'm pretty sure Fury is the only HBM card. The 390x is just a beefed up 290x.
[QUOTE=BandClassHAH;47853685]Why do they charge exorbitant amounts for these cards? They're certainly not that expensive to produce. This much of a markup should be illegal. A company can make a profit all they want but there's only two good gpu producers out there and they both want to stuff their pockets so I'm stuck with s 560 Ti that I can't upgrade from on behalf of a single fucking gpu knocking out half my new computer budget. This shit is ridiculous.[/QUOTE]
Because a shitton of money goes into R&D. How do you think they come up with the designs and architecture needed to produce these?
When are the new AMD HBM cards coming out?
I can wait maximum 3 months otherwise I'll buy a 980ti
[QUOTE=Paincake;47853702]I'm getting a new video card for my PC as a high school graduate gift. So when compared performance and price wise of everything mentioned in this thread, should I get two GTX 970s, or a GTX 980 Ti?[/QUOTE]
Best price/performance right now is a 290x I think.
[editline]1st June 2015[/editline]
[QUOTE=maxolina;47853735]When are the new AMD HBM cards coming out?
I can wait maximum 3 months otherwise I'll buy a 980ti[/QUOTE]
Rumour is June 24th but don't count on it.
[QUOTE=Scot;47853713]I'm pretty sure Fury is the only HBM card. The 390x is just a beefed up 290x.[/QUOTE]
Not at all. The 380(x?) and below will be rebrands. The 390, 390x, and Fury will have HBM.
[QUOTE=Levelog;47853776]Not at all. The 380(x?) and below will be rebrands. The 390, 390x, and Fury will have HBM.[/QUOTE]
I've been snooping around wccftech the past few weeks and they claim otherwise.
[QUOTE]The second cards leaked were part of the Radeon R9 390 series that include the Radeon R9 390 and R9 390X. There are no specifications provided except that the Radeon R9 380 comes with 4 GB of GDDR5 VRAM. When specifications are talked about on the Radeon R9 300 series cards, it can be expected that most of the cards will be rebranded from the 200 series lineup. The Radeon R9 380 is the Tonga Pro based Radeon R9 285 which has 1792 stream processors, 128 TMUs, 32 ROPs and a GCN 1.2 revision core. The card has a 256-bit interface that allows a 4 GB GDDR5 VRAM. The Radeon R9 390 and 390X are based on the Grenada / Hawaii core which has the same specifications of the Radeon R9 290 and 290X graphics card except they feature slightly higher clock speeds and 8 GB GDDR5 VRAM.[/QUOTE]
[URL]http://wccftech.com/gigabyte-radeon-r9-380-g1-gaming-radeon-r9-390x-series-pictured-r9-380-features-4-gb-gddr5-memory/[/URL]
Still only rumours but I believe them.
I wouldn't believe a single word that comes from WCCF tbh. They're a known rumor mill and post many clickbait "rumors from anonymous sources"
and if the card has hbm on it then make the card smaller
I do however see some things about the 390/x coming in 2 versions. A 4gb HBM and 8gb GDDR5. (Because current HBM can not get above 4gb for GPU stacking according to Hynix)
[QUOTE=Levelog;47853806]I wouldn't believe a single word that comes from WCCF tbh. They're a known rumor mill and post many clickbait "rumors from anonymous sources"[/QUOTE]
They're usually right in my experience.
[QUOTE=Levelog;47853818]I do however see some things about the 390/x coming in 2 versions. A 4gb HBM and 8gb GDDR5. (Because current HBM can not get above 4gb for GPU stacking according to Hynix)[/QUOTE]
I don't see what the point of that would be when the Fury is apparently releasing soon after if not at the same time.
[QUOTE=mastfire;47853813]and if the card has hbm on it then make the card smaller[/QUOTE]
Why? They're going to have a huge fucking die.
[QUOTE=Levelog;47853595]But my CPU memory runs at 2400mhz :v:
But in all seriousness CPU's don't need bandwidth, they need low latency. The memory that a CPU needs is quite different than the memory that a GPU needs. You're always going to be sacrificing latency for bandwidth so integrated is never really going to be great under current standards.[/QUOTE]
Well, yes. There's a reason for why CPUs use the memory system that they do. It just means that if you have a GPU behind the same memory interface, it's gonna suck. Honestly the tricks they use to make them as decent as they are are pretty impressive.
[QUOTE=Scot;47853819]They're usually right in my experience.[/QUOTE]
They're very very rarely right in mine and many other's.
[editline]1st June 2015[/editline]
Jesus Christ all my merges.
[QUOTE=Levelog;47853820]Why? They're going to have a huge fucking die.[/QUOTE]
the small size is one one the main advantages
[t]http://cdn4.wccftech.com/wp-content/uploads/2015/05/ba377191_ribTDOK.jpeg[/t]
[QUOTE=Scot;47853819]I don't see what the point of that would be when the Fury is apparently releasing soon after if not at the same time.[/QUOTE]
What's the point of the Titan X? It's a slightly more powerful card.
[QUOTE=Levelog;47853820]Why? They're going to have a huge fucking die.[/QUOTE]
HBM uses stacked memory. So instead of 16 memory chips, you get four stacks of memory chips. And it's on an interposer, so they can be packed tighter since the traces are so much smaller.
Also, the next-gen AMD chips are supposed to be at 14nm I think (worst-case 20nm), so the main die should also be smaller, particularly compared to the current late-era 28nm monsters.
[QUOTE=Levelog;47853820]Why? They're going to have a huge fucking die.[/QUOTE]
im talking about the length of the cards
[QUOTE=gman003-main;47853838]HBM uses stacked memory. So instead of 16 memory chips, you get four stacks of memory chips. And it's on an interposer, so they can be packed tighter since the traces are so much smaller.
Also, the next-gen AMD chips are supposed to be at 14nm I think (worst-case 20nm), so the main die should also be smaller, particularly compared to the current late-era 28nm monsters.[/QUOTE]
I realize the memory being stacked will make it smaller, but honestly that's not what takes up the most room on the card. With the massive TDP and iirc it's rumored to have a fucking 550mm^2 die it's not going to be a tiny card.
[editline]1st June 2015[/editline]
I guess I just think the other factors will make up for the space saved via HBM. I could definitely be wrong though. Time will tell.
heres the like about hmb
[url]http://www.setphaserstostun.org/hc26/HC26-11-day1-epub/HC26.11-3-Technology-epub/HC26.11.310-HBM-Bandwidth-Kim-Hynix-Hot%20Chips%20HBM%202014%20v7.pdf[/url]
[QUOTE=mastfire;47853893]heres the like about hmb
[url]http://www.setphaserstostun.org/hc26/HC26-11-day1-epub/HC26.11-3-Technology-epub/HC26.11.310-HBM-Bandwidth-Kim-Hynix-Hot%20Chips%20HBM%202014%20v7.pdf[/url][/QUOTE]
I mean I'm completely aware of HBM and how it works and how it improves things. I was talking about other factors. Sure the memory gets smaller, but that's far from the only thing on a GPU.
At least one of the cards is small. There's that one photo by the paradox dev and the other preview teaser that show it being short.
[QUOTE=Levelog;47853857]I realize the memory being stacked will make it smaller, but honestly that's not what takes up the most room on the card. With the massive TDP and iirc it's rumored to have a fucking 550mm^2 die it's not going to be a tiny card.
[editline]1st June 2015[/editline]
I guess I just think the other factors will make up for the space saved via HBM. I could definitely be wrong though. Time will tell.[/QUOTE]
Is 550mm^2 for the GPU die or for the interposer?
[QUOTE=gman003-main;47854243]Is 550mm^2 for the GPU die or for the interposer?[/QUOTE]
GPU die from what I read. Was a big deal because (I'm completely blanking on AMD's fabricator) has a die limit of 600-650mm^2
[QUOTE=Levelog;47854271]GPU die from what I read. Was a big deal because (I'm completely blanking on AMD's fabricator) has a die limit of 600-650mm^2[/QUOTE]
Global Foundries?
[QUOTE=GoDong-DK;47855034]Global Foundries?[/QUOTE]
TSMC I think.
[editline]1st June 2015[/editline]
But I dunno if TSMC does AMD's stuff. I'm just all mixed up today, had a lot of networking thrown in my face and I'm just done with tech right now.
Damn shit, I was going to buy an EVGA 980 but had a bad experience with my EVGA 660 Ti (constant overheating at relatively low temps) so I got a Gigabyte one, I doubt they'll do the upgrade though.
[QUOTE=Levelog;47855046]TSMC I think.
[editline]1st June 2015[/editline]
But I dunno if TSMC does AMD's stuff. I'm just all mixed up today, had a lot of networking thrown in my face and I'm just done with tech right now.[/QUOTE]
IIRC GlobalFoundries was AMD's old foundries, spun off as a new company because they sucked. They switched to (IIRC) TSMC when they shrunk to 28nm and I seem to recall they're switching to someone else for 20nm, can't remember who though.
Sk hynix makes the hbm and the hbm is on-die so it could be them.
[QUOTE=01271;47855128]Sk hynix makes the hbm and the hbm is on-die so it could be them.[/QUOTE]
Nah, Hynix is pretty much strictly memory.
[QUOTE=Novangel;47851593]It's pretty sad my 780Ti can't max Advanced Warfare or GTA, been needing an upgrade. Probably holding off till August or something though and buy myself a new rig as a birthday present (since my family doesn't do presents)[/QUOTE]
My 780 maxes GTA at 1080p, sans ultra grass and excessive AA.
[editline]2nd June 2015[/editline]
Also not sure whether to sell my 780 and get a 980 Ti, keep it and get a cheaper used 780 for SLi, or just wait for the next generation and sell it then for whatever paltry amount it'll be worth.
Sorry, you need to Log In to post a reply to this thread.