>According to team red, Don, the Radeon RX Vega performance compared to the likes of NVIDIA’s GeForce GTX 1080 Ti and the Titan Xp looks really nice >looks really nice
Why he did not say it faster but ''looks really nice''? What did he mean by this?
What else can it mean besides vague non answers? He can't tell you until NDA
Evan Martinez
As long as the GPU gives me at least 100+ fps in games like Witcher 3, GTA V and other demanding games at a 1440p, I'll buy the one who gives me the most FPS per Dollar.
Camden Anderson
It's going to be underwhelming piece of shit.
Aiden Fisher
He saying vega had RGB led on it
Jaxson Price
Even Fury had a tachmeter
Daniel Martinez
So is Vega going to be 8GB of VRAM? Wish they would include at least 10, or 11 like the 1080ti. Not that I need that much today, but it might give the card some more longevity.
Isaac Bailey
Stick more VRAM is a pretty shitty way of going at it, I'd rather engines use VRAM smarter instead of allocating 6GB while only using 3GB
Keepa cost, size and power down
Robert Hernandez
If Vega doesn't have 16GB of HBM2 VRAM, then it's trash.
Julian Cook
16gb why
I would be happy if they releases 4gb with cheap price
Ryder Ramirez
>novidia now care about VRAM after 3.5GB meme
...my almonds...wew
Jeremiah Price
16gb is probably stupidly much
But 4gb is not enough today, assuming you play games. Lots of games use more than 4gb, 8gb is a minimum.
Grayson Martin
>3.5 maymay still going on >after NVIDIA settled it in a class action suit and made sure to never repeat that same mistake again AMDrones sure love holding onto the past, huh?
4GB isn't enough for even ultra settings on a 1080p monitor. The 1080 Ti has 11GB of memory. If Vega wants to be competitive, it should at the very least top it in terms of memory. That's why 16GB of HBM2 should be the minimum amount of memory it has.
Ryan Barnes
There are rumors that the small Vega will use perhaps only 4GB of VRAM, but I think also 16 GB may be possible (in the big variant).
Cameron Roberts
What's even the point of having 8GB+ in a video card when no games will use it and the card will be obsolete in a year?
Jace Perez
>MOAR EVERYTHING you have to go back >>Sup Forums
Ryan Robinson
4gb is enough, since amd hyping that new cache controller
Parker Ramirez
You mean like how some games even at 1080p try to use 5+GB of ram? Many games are unoptimized garbage whose bloat is only beaten by windows itself. 6GB vram should be a minimum by now. If AMD doesnt have a card with minimum 10GB ram, I'd be disappointed.
Josiah Diaz
The only video game I play is League of Legends and that's only because I've been in the top 100 of players in North America since it was released.
Justin Flores
i doubt it will be much faster than a 1070 if we're being honest with ourselves.
Evan Evans
>Wait™
Robert Carter
I'll take an AMD card with 1070 performance. I'm on a 980ti and while it still performs, their forced Nvidia account required to use shadowplay and other features is fucking bullshit.
Austin Evans
Apparently AMD is throwing a decent amount of die space at logic for a new "cache controller" that just makes some or all of the HBM just a fuckhuge L3 to the host's DDR and even NVMe.
It's not clear how much software support it will take, but the promise is that it will be relatively easy to support obscene texture and mesh asset sets and only keep render targets and currently used assets on the HBM. They're throwing around a 2x factor difference in recent games between what's allocated on the GPU vs. what's actually used in any given scene.
There scheme will probably actually work, but you better get used to 300 GB game installs with unique art assets for every piece of debris or navel lint.
Jason Brown
I would like to see a high memory 4k model and one specifically designed for high refresh gaming with lower memory and perhaps other optimizations. I feel like going for 4k 60fps is so much easier than shooting for 144fps at lower resolutions as you don't have to fight with engines. i had to buy fucking $160 3600 cl16 ram because my most played game needed it to go from ~110 to ~140fps
Elijah King
The rumor mill is so Goddamn worthless
I hope it will be good
Carter Baker
ITS FAKE NOOOO WE NEED MORE VRAAAMMMMM
Logan Sanders
I actually wouldn't be upset if game installs shot up that large but there needs to be a better solution to using flash storage as a cache for spinning rust. A 500gb SSD should be enough to cache a couple of terabytes of HDD storage and have it all be seamless and feel as though it's all "fast" storage. I have a decent raid array in my machine but I would love to drop down to 2 4tb drives in raid0 cached to a 500gb nvme.
Nicholas Jackson
it should easy to make it work, only few game using more then 4 gb anyway
if not, what the fuck
Parker Mitchell
So you would rather spend money to sidegrade than install an older version of GFE?
Kayden Thomas
>Why he did not say it faster but ''looks really nice''? What did he mean by this?
My bet is that it's gonna be beating the Titan Xp but ONLY in 4k due to having HBM memory.
Which is great because i'm getting my 4k IPS panel next week.
Ethan Cruz
Titan Xp has more memory bandwidth than Vega though.
Adrian Turner
>side grade
Yes. I had an Intel processor and side graded to a Ryzen setup to get away from Intel. AMD has their relive (I.e. shadowplay clone). Plus then I can get a freesync monitor and take advantage of it.
Lucas Stewart
It wont beat Titan Xp overall, maybe in vulcan and some dx12 that about it
Cooper Russell
>it should easy to make it work, only few game using more then 4 gb anyway
Rendering the frames in 4k requires fuckton of VRAM and bandwidth. that's why even 1080ti barely scratches the 4k with it's GDDR5 overclocked till it catches fire and still not cutting it.
Vega will be the first actual 4K gaming gpu.
Hudson Perry
>Titan Xp has more memory bandwidth than Vega though. >12GB of GDDR5X memory running at 11.4 Gbps No, just no.
Parker Lewis
>That shitty English
Fuck off pajeet
John Evans
SSD read caches for HDD only tend to work well with small parts of the data frequently accessed and the rest accessed infrequently and at low bandwidth.
Caching things like FS structures, text documents, image thumbnails, or the first few seconds of an audio file makes sense and works well. Caching entire video files is wasteful and does not.
Game data is in a weird spot where it takes a ton of space, doesn't necessarily get read out very linearly within and between scenes or sessions, and it needs to be gulped down as fast as possible for level loads etc.
I think we'd be better off if games just had better holistic compression or perhaps procedural generation/reconstruction of detail that ran at level load time. For example, Google made some proof of concept image compression scheme that downsampled pics by 2x and then used machine learning to reconstruct detail on the 2x upscale, and it has pretty impressive objective and subjective results.
Hudson Baker
Vega is going to have 2 stacks of HBM2, that's 512GB/s. Titan Xp has 548GB/s.
1080Ti's bandwidth is 480 GB/s vs. Vega's presumed 512 GB/s.
Vega's not exactly gonna be head and shoulders above 1080Ti at 4k unless a shit-ton of games get shader patches that rely heavily on double-rate fp16 and are more limited by shader throughput than memory.
Brandon Thompson
>Vega is going to have 2 stacks of HBM2, that's 512GB/s. >Vega's presumed 512 GB/s.
That's same exact value as Fury X And that is assuming it's going to run at same exact 500MHz. For starters Fury X could reach higher 570 MHz on memory, and dog only knows how fast the Vega's memory will be. I'm pretty sure it's not gonna be slower.
John Ross
Ewww that's fucking disgusting
Christian Ward
>That's same exact value as Fury X Yes, Fury X has 4 stacks. They downgraded to 2 stacks of HBM2 this time >For starters Fury X could reach higher 570 MHz on memory, and dog only knows how fast the Vega's memory will be. The spec for HBM2 is already published, it's not some arcane mystery. The spec says 256GB/s per stack, Vega has 2 stacks. Maybe they'll OC it, but you can OC Titan Xp as well and it has a lead in any case at stock, according to the spec. There's not going to be any big difference in memory bandwidth between Vega and Titan Xp either way.
Jacob Sanchez
lmao you are just prepping yourself to be disappointed
Leo Nelson
>AMDrones will have Just Waited™ for Vega for longer than the gap between Vega and Volta >three to six months after Vega comes out, Volta will come out on a 12nm process and with proper async compute
Andrew Lopez
Well considering Vega provides at least 60 solid FPS in Battlefront and Doom it must be way above 1080 performance on the level of Titan Xp.
If it's priced competitively, and it just fucking must be, it's fucking great.
Wyatt Watson
>>three to six months after Vega comes out, Volta will come out on a 12nm process and with proper async compute
Citation needed. Sounds more like Nvidia is spreading rumors in hopeless panic hoping to discourage people from getting Vega when it comes out.
Soon you'll be the one waiting, lol.
Matthew Lee
> Volta 12nm
you know that's just a TSMC marketing name for a tweaked variant of their 16nm process, right?
Julian Brooks
Volta isn't coming before 2018. mid-late 2018 if they are actually waiting for 12nm.
Levi Davis
Based on the DOOM demo it's like 10% above a 1080 at best and that's in Vulkan. 1080 does 60FPS in Battlefront too and the leaked AotS benchmark results puts it right at the same level. Literally nothing points to Vega 10 being Titan Xp level.
I hope you're right, might make NVIDIA reduce their high-end pricing a bit, it's been slipping out of control since AMD decided not to compete with them.
Angel Wood
Isn't 14nm just 20nm with finfet or something? I was watching a lecture on YouTube about Moore's law from before 14nm launched, and the lecturer said 14nm is just going to be 20nm with finfet and some other "boosters".
Dylan Hughes
If it comes out in early 2018 like originally planned, then the gap between Vega and Volta will still be smaller than the gap between Polaris and Vega in terms of launch times.
Juan Mitchell
which will cost a shitton and age like milk. no thanks, I'll take that value finewine.
Mason Johnson
>cut down engineering sample on GCN 1.2 drivers >already faster than 1080 Titan XP is only 30% faster than a 1080. Vega can easily reach that.
Christopher Anderson
Die shrinks make things cheaper. Pascal is 16nm (except the 1050 and 1050ti, which are 14nm). Going down to 12nm would make cards with the same level of performance as Pascal cheaper.
Kayden Reed
>with proper async compute lmao
Nicholas Hall
They're all bullshit marketing terms at this point, but 12nm is even more so than usual.
They literally just renamed in November a previously announced lower-leakage 16nm FinFET variant as 12nm. Magically, Volta and a couple other customers' chips will now be "12nm" instead of "14nm" from the competitor...
Lincoln Wilson
>Die shrinks make things cheaper. Unless the yields are 1.7-tier. Fermi will be back. Someday.
Robert Gray
>the fire rises
Isaiah Thomas
The GTX 1080 was the first actualy 4K gaming GPU.
Brody Powell
>40 fps in Witcher Hardly
Lucas Gonzalez
That's pure baseless conjecture at this point, I'll judge it when I see a real benchmark.
Luis Perez
>You merely adopted the fire. I was born in it, molded by it.
Camden Torres
I'm wondering how much of a range vega will cover in terms of performance... it's semi-confirmed theres going to be 3 vega cards right?
Charles Bell
Nvidia has been working on Volta for a long time. They planned for it to come after Maxwell, but decided to put out a die shrunk Maxwell (Pascal) first to give them more time on Volta. I wouldn't judge Volta based on the fact that Pascal doesn't bring much new to the table architecturally.
40fps at max settings is good. You can easily get 60fps by turning a few settings down to medium.
Andrew Rogers
Probably two. But only one, full Vega10-based is confirmed.
Lucas Price
>if i pull that fan off, will you die?
Easton Price
>I wouldn't judge Volta based on the fact that Pascal doesn't bring much new to the table architecturally They can slap hardware scheduling back and give us our lovely housefires. That's what, it?
Jason Brooks
Two chips (Vega 10, Vega 11), which means probably 4 or 5 cards
Dominic Jenkins
It would be extremely hot.
Dylan Edwards
>for you
Carson Turner
>Vega11 We know nothing about it though.
Brandon Phillips
There's only a hard confirmation that there will be two Vega chips, the Vega 10 with 2 HBM2 stacks and 4096 ALUs, and a smaller Vega 11 that nobody has any clue about that may or may not even ship at the same time.
Between laser-cut binnings and potential different cooling solutions, there will likely be at least 3 variants of Vega 10 shipped eventually. (Fury X, Fury, and Nano style)
David Reed
>not 'you're a big card' Fag.
Nicholas Johnson
> :|
Carson Anderson
For you
Christian Allen
>Nvidia has a proper hardware scheduler >LOL HOUSEFIRES!!!
>Nvidia uses software for more power efficiency >LOL NO ASYNC COMPUTE!!!
Connor Gray
>Die shrinks make things cheaper nYidia isn't a charity. they'll continue to ask a thousand shekels for their flagship consumer cards because there are enough dumb gaymer kids who will spend their parents money on it. AMD isn't a charity either, but they can't pull this level of jewry simply because that doesn't work with their "good value" public image.
Jayden Reed
if you believe this youre a fucking retard "looks really nice" means its a good price to performance ration compared to the 1080 ti and titan Xp. It could be comparable to a 1050 but be cheaper to balance it out. name the last time amd ever made anything comparable to nvidia in performace, you can't.
David Carter
>>not 'your a big card' even bigger faggot
Nolan Gomez
>the performance they have seen is very good when compared to the competition.
This literally says nothing
Charles Green
I'm just wondering if there will be a "low end" vega thats somewhere in the 1070 ballpark of performance, or if it'll start at the 1080's range and go beyond that, which will be pretty insane.
Anthony Reyes
It's up to novideo to properly balance their architecture to implement hardware scheduling without core temps reaching 100 degrees celsius.
Charles Ortiz
Vega11 is supposed to compete with 1070.
Sebastian Ward
>novideo Volta isn't even out yet, how do you know their architecture isn't balanced and that it's a housefire?
Adrian Roberts
>AMD isn't a charity either, but they can't pull this level of jewry simply because that doesn't work with their "good value" public image. They never did that even when they had superior cards. Because Titan Xp already eats 250watts without fucking hardware scheduling.
Zachary Edwards
AMD needs something smaller than Vega 10 (500-550 mm^2) they can shove on an enterprise MCM APU, and a half-Vega 10 with 1 HBM stack would fit the bill pretty well and probably come in around 1070 perf levels as well.
Jaxon Wright
You've been samefagging this in threads all day It's not even remotely funny
Bentley Scott
t. butthurt nvidiacuck
Colton Rivera
...
Aaron Jones
The highest end cards have never been the most power efficient. Nvidia cards are still leaps and bounds ahead of AMD in the efficiency department. And the 1080ti is a different architectures than Volta.
Dylan Myers
...
Jonathan Foster
...
Jeremiah Davis
>with proper async compute They're too terrified to try that again since they aren't capable of making an efficient GPU with hardware scheduling like AMD are
Robert Lewis
>The highest end cards have never been the most power efficient. No fucking shit, Sherlock. >Nvidia cards are still leaps and bounds ahead of AMD in the efficiency department. Considering each GCN design is FAR denser, nah.
Nathan Thompson
...
Levi Stewart
Compare the TDP of AMD and Nvidia cards in the same price/performance class. The 1060 consumes less power than the 580. The 1050ti consumes less power than the 570.
Michael Peterson
>The highest end cards have never been the most power efficient Lol no they aren't, they use software scheduling and have lower shader counts in general.