| Page 276 of 782 |
Werelds
Special Little Man
Posts: 15098
Location: 0100111001001100
|
 Posted: Wed, 3rd Jun 2015 12:07 Post subject: |
|
 |
| russ80 wrote: | "won" console contracts. Good one     |
Here we go again, I'll entertain you for a minute. What are you trying to imply? That Nvidia gave them up because "it's not worth it"?
AMD have made $350m net income off those contracts already. That's just 1 year of these consoles' lifespan. For comparison's sake, Nvidia's net income for FY'15 (2014) was $630m. If it wasn't for AMD's weak CPU division, that one contract on its own would've put them at 55% of Nvidia's income.
No, clearly not worth it. |
|
| Back to top |
|
|
|
|
| Posted: Wed, 3rd Jun 2015 13:27 Post subject: |
|
|
Last edited by AmpegV4 on Mon, 23rd Feb 2026 09:38; edited 1 time in total
|
|
| Back to top |
|
|
|
|
| Posted: Wed, 3rd Jun 2015 15:53 Post subject: |
|
|
| Werelds wrote: | @sabin1981 the "rank" in that table doesn't say that much. Even though Intel IGP technically do have some 11_1 features, they still won't be fast enough to benefit. Also keep in mind that AMD's VLIW4/VLIW5 isn't on the list at all purely because VLIW isn't efficient at these kind of tasks, so AMD opted to not support it from a driver perspective rather than offering it with crap performance. Technically Nvidia's support goes back further than AMD's 
You'll still get the performance benefits. The limits are in fact visual limitations; taking Star Swarm as an example, the difference will not be in how quick stuff renders. It'll be in how much can be rendered; i.e. for Fermi there'll be fewer independent lighting sources for example. It'll still be higher than D3D11 though, because even though it's "limited" compared to GCN or even Kepler, there's just that much more headroom. You're still gonna get all the benefits, they'll just not be as extreme as they are for GCN or Maxwell.
These feature levels in fact are a good thing for Fermi, because without them Fermi would've had 11.0 support and nothing more. It was never 11.1 compliant (nor is Kepler), but does support some features from it. Same for 12.0 - it's not compliant, but some features do exist and thanks to the feature levels that means it can still claim compatibility. To give you an idea of how they work: level 11_1 includes a load of features that are mandatory for hardware conforming to that level. Some of those same features however are optional ones on levels 10_0, 10_1 and 11_0; developers can still check whether those features are supported even if the hardware doesn't conform to 11_1. They're confusing as fuck but without them you wouldn't get 12.0 at all. The way MS do things now is better than it was with 6-9 (or even 9.0a through c). |
Thanks Paul, much appreciated as always 
| Quote: | | It's 100% Nvidia's fault for being so lazy to implement things though. For Kepler to not even be compliant with 11.1 is just ridiculous. That's just how the company works these days though, most of the stuff you see in D3D11, 12 and Mantle come from AMD. Gaming would be much better if Nvidia participated in the D3D programmes and the Khronos group. They're members now, but they're far from the biggest contributors. They like their proprietary stuff. |
I don't even think it's laziness, it's calculated spite and greed. They deliberately eschew older cards and compliance with newer technology (the 7xx Keplers are a YEAR newer than the Intel HD4200, yet has less functionality  ) in order to push their next big thing. They're a company, I get it, I just don't like it.
|
|
| Back to top |
|
|
Nui
VIP Member
Posts: 5720
Location: in a place with fluffy towels
|
|
| Back to top |
|
|
Werelds
Special Little Man
Posts: 15098
Location: 0100111001001100
|
| Posted: Wed, 3rd Jun 2015 16:32 Post subject: |
|
|
Well it all boils down to where these companies come from, Nvidia is still a baby in the industry and when they started, they had to compete against some top dogs straight away.
Intel and AMD are both approaching their 50th birthday. ATI would've been 30 years old this year. Matrox is turning 40 next year. S3 was founded in 1989. All of these companies come from times where there was more competition, but at the same time more collaboration. Intel wouldn't be what it is now without AMD and AMD wouldn't have been what it used to be (hate myself for saying it, please let Zen be awesome  ) without Intel. ATI and Matrox had the same kind of synergy. None of these would've been able to push their proprietary shit, because no consumer or OEM was interested in having to deal with those differences, they all knew they had to get their shit supported by others in order to sell stuff and actually make money. Same shit happened with soundcards for example. And the companies that didn't realise that all went under. It's these collaborations where so many things that we now take for granted come from. Not because one company tried to push them.
Nvidia was founded in 1993. It took them 5 years to become even remotely competitive with ATI, S3 and Matrox (who still mattered back then). S3 and Matrox had already started falling behind a bit, but ATI hadn't. The bigger problem by then was that other baby in the industry though: 3DFX. Founded in 1994 and by 1998 no one had anything on them. And Nvidia's 2D performance was incredibly poor; the best performance came from an ATI + Voodoo. It took Nvidia and ATI another 18-24 months to really hit back.
And that's where the difference comes from: Nvidia started out having to fight against the superpower that was ATI, the incredibly clever people at 3DFX who were wrecking everyone else - and then still had 2 other minor players to deal with. And they did so by marketing. Starting with the first GeForce (late 1999) their marketing machine started spinning. That's when they figured out the acronym wars. A couple of years later there was crap like TwinView, T&L2, NSR, HDVP, DVC. None of it new or really proprietary yet; Matrox did multi monitor shit long before then, T&L came from 3DFX originally (this is before Nvidia bought them), and so on.
Fast forward a few years and you're looking at a time where all but ATI have vanished, D3D has won the desktop API war and they have to deal with that highly standardised API, leaving little room for their own specialties. Enter G80 and CUDA and presto, they've found their calling. They can make money off that shit and so they did (at least for 1 year, until the underfill costs kicked in and cost them $800m...and then another year with mobile underfill + Fermi delays  ). New player trying to change the landscape? Fuck it, buy their shit, lock it down and then shelve it for the better part of a decade (if I have to explain what acquisition I'm talking about here, shame on you  ).
Meanwhile, ATI (and a few years later, AMD) are still doing what they did 20 years earlier. Develop new shit, (try to) get it into common APIs.
The reasons behind both philosophies is the same: sell products.
|
|
| Back to top |
|
|
Nui
VIP Member
Posts: 5720
Location: in a place with fluffy towels
|
| Posted: Wed, 3rd Jun 2015 16:55 Post subject: |
|
|
|
|
|
| Back to top |
|
|
|
|
|
| Back to top |
|
|
Nui
VIP Member
Posts: 5720
Location: in a place with fluffy towels
|
|
| Back to top |
|
|
Werelds
Special Little Man
Posts: 15098
Location: 0100111001001100
|
| Posted: Wed, 3rd Jun 2015 17:34 Post subject: |
|
|
| Nui wrote: | | Please shame me then |
Tolanri got it
| Quote: | | Thanks for the read! Very appreciated. Just a little biased towards nvidia, don't you think? |
I don't like Nvidia as a company and I certainly don't agree with how they do business either, doesn't prevent me from buying their stuff if it's good value. For every ATI or AMD GPU I've had, I've had 2 Nvidia GPUs I think. Mach 64, Rage Pro or 128 (can't remember), Radeon 9700, 5870, 6950. That's about it I think. As opposed to GF2MX400, 6800 GT, 8800 GTS-640, another 8800 GTS-640 (), a 9800 GTX+ as replacement for that last one ( ), GTX 460 (in between the 5870 and 6950), GTX 680, GTX 770 and now a GTX 970.
If it's a mediocre chip, I don't care for it whoever makes it. Also note the absence of the first Radeon and GeForce. My Voodoo was just that good and it served me better than any GPU after it. Same reason I skipped the GF3/4/5/FX or ATI's X300 and successors - they were just not very interesting. |
|
| Back to top |
|
|
Nui
VIP Member
Posts: 5720
Location: in a place with fluffy towels
|
|
| Back to top |
|
|
|
|
|
| Back to top |
|
|
DimasL
Posts: 1766
Location: Spain
|
|
| Back to top |
|
|
|
|
| Posted: Wed, 3rd Jun 2015 22:38 Post subject: |
|
|
DX 12 will come very fast and its very good. Firstly the hardware is in the consumers hands right now. Secondly Win 10 will spread fast due to being free.
So I think we will see DX12 games very soon.
|
|
| Back to top |
|
|
|
|
|
| Back to top |
|
|
|
|
| Posted: Thu, 4th Jun 2015 09:54 Post subject: |
|
|
cards appeared here, 792eur
| Lutzifer wrote: | | and yes, mine is only average |
|
|
| Back to top |
|
|
|
|
|
| Back to top |
|
|
couleur
[Moderator] Janitor
Posts: 15073
|
| Posted: Thu, 4th Jun 2015 11:12 Post subject: |
|
|
That screams for a shrink.
"Enlightenment is man's emergence from his self-imposed nonage. Nonage is the inability to use one's own understanding without another's guidance. This nonage is self-imposed if its cause lies not in lack of understanding but in indecision and lack of courage to use one's own mind without another's guidance. Dare to know! (Sapere aude.) "Have the courage to use your own understanding," is therefore the motto of the enlightenment."
|
|
| Back to top |
|
|
Frant
King's Bounty
Posts: 24975
Location: Your Mom
|
|
| Back to top |
|
|
|
|
| Posted: Thu, 4th Jun 2015 12:05 Post subject: |
|
|
| Lutzifer wrote: | | and yes, mine is only average |
|
|
| Back to top |
|
|
Frant
King's Bounty
Posts: 24975
Location: Your Mom
|
| Posted: Thu, 4th Jun 2015 12:09 Post subject: |
|
|
As usual when it comes to AMD the performance of the R9 300/Fiji series will improve over ~6-12 months through serious driver optimization (last official big update was the Omega driver). When Win10 and DX12 hits the market end of July it will get another boost from DX12 (but so will nVidia if they've managed to write new low-level drivers). The good thing with DX12 is that ALL GCN-based AMD cards will get a boost.
We'll hopefully see patches for a bunch of popular games from the last ~12 months that adds a D3D12-renderer path (here's hoping the GTA V PC-team is actually allowed to do just that).
Ph'nglui mglw'nafh Cthulhu R'lyeh wgah'nagl fhtagn!
|
|
| Back to top |
|
|
russ80
Posts: 4679
Location: Romania
|
| Posted: Thu, 4th Jun 2015 12:36 Post subject: |
|
|
| Frant wrote: | | As usual when it comes to AMD the performance of the R9 300/Fiji series will improve over ~6-12 months through serious driver optimization |
Yeah i guess that makes sense considering they release 1 driver each 6 months
Also
| Quote: | AMD's global head of technical marketing Robert Hallock confirmed to Computerbase GCN-based graphics cards will support DX12 and at least some will also support DX12 feature level 12_0. Computerbase mentions Tonga, Hawaii and the Xbox One GPU will surely support 12_0, however question marks remain for older chipsets like Bonaire, Tahiti and Pitcairn.
Meanwhile, Nvidia confirmed earlier their chipsets GM200, GM204 and GM206 will already support DX12 feature level 12_1. At GDC this year, MS gave an overview of feature level 12_0 and 12_1:
Hallock said it shouldn't be an issue that AMD cards only support feature level 12_0 since all the features relevant for games are included in 11_1 and 12_0. Also, most games will stick closely to what the consoles use and will therefore mostly use 12_0. Keep in mind, though, he's the global head of technical marketing, so his words shouldn't just be taken as gospel. Computerbase also pointed out Nvidia said the same thing when their Kepler cards didn't support feature level 11_1.
tl;dr: If you want a fully compatible DX12 card you should look into getting a Nvidia GTX 960, 970, 980, 980 Ti or Titan X. |
http://www.computerbase.de/2015-06/directx-12-amd-radeon-feature-level-12-0-gcn/
|
|
| Back to top |
|
|
Werelds
Special Little Man
Posts: 15098
Location: 0100111001001100
|
| Posted: Thu, 4th Jun 2015 13:37 Post subject: |
|
|
Maxwell isn't fully DX12 compatible either. Maxwell only supports 11_1, not 11_2 so the resource binding features in 12_0 and 12_1 are limited on on it. The 7790, R7 260(X) and R9 290(X) are the only cards that can run 12_0 without limits.
If you want a fully DX12 compatible card you need to wait. Fury will likely be 100% compatible, for Nvidia it'll be next year.
|
|
| Back to top |
|
|
|
|
| Posted: Thu, 4th Jun 2015 14:01 Post subject: |
|
|
this whole DX12 talk is stupid. it will be years before games are out that require 100% DX12 feature set and refuse to work on current cards.
|
|
| Back to top |
|
|
tonizito
VIP Member
Posts: 52331
Location: Portugal, the shithole of Europe.
|
| Posted: Thu, 4th Jun 2015 14:04 Post subject: |
|
|
| Breezer_ wrote: | Dat die size (Radeon Fury).
 |
Godamn it's almost as big as GT200 was 
| boundle (thoughts on cracking AITD) wrote: | | i guess thouth if without a legit key the installation was rolling back we are all fucking then |
|
|
| Back to top |
|
|
|
|
| Posted: Thu, 4th Jun 2015 14:10 Post subject: |
|
|
Maybe the one holding it has tiny hands? No? 
|
|
| Back to top |
|
|
Werelds
Special Little Man
Posts: 15098
Location: 0100111001001100
|
| Posted: Thu, 4th Jun 2015 14:40 Post subject: |
|
|
She does, actually, it's Lisa Su
But yes, the package will be big. There's no way to avoid that with HBM, because the memory has to go on there too. The bus width it gives would be near impossible on a PCB though. They'd have to add lots of layers and half the PCB would consist of just the etched lanes.
The actual GPU die (the middle shiny bit) looks like it'll sit somewhere around 550 mm, bit hard to judge (Google "GF100 die size" or "GM200 die size" to get an idea what a 550-600 mm chip looks like - this looks to be closer to GF100 than to GM200). That does include the 4 1024-bit busses though, so in those 4 corners where the HBM stacks are, is a lot of extra space just for those lanes. Without HBM that thing would've been 20% smaller.
Definitely gonna be the biggest thing AMD have ever baked
|
|
| Back to top |
|
|
DimasL
Posts: 1766
Location: Spain
|
| Posted: Thu, 4th Jun 2015 14:46 Post subject: |
|
|
| Werelds wrote: | Maxwell isn't fully DX12 compatible either. Maxwell only supports 11_1, not 11_2 so the resource binding features in 12_0 and 12_1 are limited on on it. The 7790, R7 260(X) and R9 290(X) are the only cards that can run 12_0 without limits.
If you want a fully DX12 compatible card you need to wait. Fury will likely be 100% compatible, for Nvidia it'll be next year. |
Wait... so a 980ti wont be full compatible with DX12 and my 7970 will be???
MSI MEG X570 UNIFY / A.D2 (AGESA 1.2.0.7) / AMD Ryzen 9 3950X @ 4.70 GHz / Noctua NH-D15 / EVGA GeForce RTX 3080 FTW3 ULTRA / 32GB Corsair Vengeance RGB Pro 3200Mhz CL16 / Corsair AX750w / SilverStone Fortress FT02B-W (Black)
Samsung 970 EVO Plus 1TB SSD NVMe M.2 / Samsung 850 Evo 500gb / Samsung F3 1TB 7200.11rpm / Samsung F4 2TB 5400rpm / Seagate 3TB 7200.14rpm / Seagate 8 TB Expansion Edición Especial USB 3.0
Logitech G502 Lightspeed / Logitech G640 / Logitech G915 Lightspeed / DROP Sennheiser + EPOS PC38X / Xbox360 Pad / Logitech G27 / Logitech X56 HOTAS / Logitech StreamCam
ASUS ROG SWIFT OLED PF27AQDM / BenQ XL2730Z 1ms 144Hz 1440p / HP 27q 1440p 60Hz
Internet Connection: 1Gb/1Gb Fiber
|
|
| Back to top |
|
|
|
|
|
| Back to top |
|
|
Werelds
Special Little Man
Posts: 15098
Location: 0100111001001100
|
| Posted: Thu, 4th Jun 2015 14:55 Post subject: |
|
|
Neither is fully compatible.
See http://www.nfohump.com/forum/viewtopic.php?p=2731246#2731246 and http://en.wikipedia.org/wiki/Direct3D#Direct3D_12_levels
Basically, AMD don't support the 12_1 feature level on the current versions of GCN (because 12_1 is a very recent addition, so GCN is just too old), but they have full support for everything else, without limits.
Nvidia do support 12_1, but because their hardware does not support 11_2, some of the features in 12_0 are limited.
I'll try to explain what the limits mean. If you take the UAV feature as an example, that's a buffer that allows reading/writing from multiple threads. So it's a bit of data that multiple threads can use. On Maxwell, you can only have 64 of those in total; on GCN, there's no limit, you can do whatever the fuck you want. What that means is that that GCN has more freedom and can do more things in parallel.
Does it matter? Not really, because even Maxwell's "limits" are way, way, WAY higher than anything in D3D11. But neither company can claim full compatibility at the moment.
|
|
| Back to top |
|
|
russ80
Posts: 4679
Location: Romania
|
| Posted: Thu, 4th Jun 2015 14:59 Post subject: |
|
|
| Werelds wrote: | Neither is fully compatible.
See http://www.nfohump.com/forum/viewtopic.php?p=2731246#2731246 and http://en.wikipedia.org/wiki/Direct3D#Direct3D_12_levels
Basically, AMD don't support the 12_1 feature level on the current versions of GCN (because 12_1 is a very recent addition, so GCN is just too old), but they have full support for everything else, without limits.
Nvidia do support 12_1, but because their hardware does not support 11_2, some of the features in 12_0 are limited.
I'll try to explain what the limits mean. If you take the UAV feature as an example, that's a buffer that allows reading/writing from multiple threads. So it's a bit of data that multiple threads can use. On Maxwell, you can only have 64 of those in total; on GCN, there's no limit, you can do whatever the fuck you want. What that means is that that GCN has more freedom and can do more things in parallel.
Does it matter? Not really, because even Maxwell's "limits" are way, way, WAY higher than anything in D3D11. But neither company can claim full compatibility at the moment. |
Except they can and have
Main PC : I7 12700, MSI Ventus RTX 4090 24gb, Alienware AW3423DW QD-OLED
Laptop : I5 4200H @ 3400mhz boost, GTX 850m 2gb Vram DDR3, 4gb RAM DDR3
Derpsole : Playstation 5 disc edition, Ninty Switcherino
TV+audio: LG CX 65" / Sonos ARC + SL ones + Sonos sub 3
VR Headset: Meta quest 2 airlinked
|
|
| Back to top |
|
|
| Page 276 of 782 |
All times are GMT + 1 Hour |
