Navi Rumors v2.0

Discussion in 'AMD Flavor' started by Boil, May 10, 2019.

Thread Status:
Not open for further replies.
  1. euskalzabe

    euskalzabe Gawd

    Messages:
    969
    Joined:
    May 9, 2009
    A) current DXR accelerating NV cards can do 1080p60 at medium settings - that's hardly "shitting the bed".
    B) I didn't say it needs to be great. I said it needs to be available.
     
  2. XoR_

    XoR_ Gawd

    Messages:
    717
    Joined:
    Jan 18, 2016
    Even with concurrent INT/FPU execution shaders we would see terrible performance.
    AMD still should to add DXR support to drivers for their cards because it is getting more and more ridiculous.

    Navi probably is the same core as they used in Fury with some changes to memory subsystem carried from Vega and support for GDDR6 (if they will have this memory type) and little to none changes to cores themselves. If there will be card with 4096 shaders this will be confirmed just as it was with Fury vs Vega comparison at the same clocks.

    In any case production cost of Navi cards is probably even higher than RTX2070 despite Navi XT being smaller than TU106 because 7nm is new process node and higher TDP increases production cost (RTX 2070 in basic variant is hard limited to 200W) and corporation thinking is that if they want to earn as much per card as competition does at all costs. For them cost will be abysmal sales...
     
  3. Snowdog

    Snowdog [H]ardForum Junkie

    Messages:
    8,601
    Joined:
    Apr 22, 2006
    I disagree. In this case, I think it might be better not do something, than to do it very poorly.

    Right now RT is a tiny niche, and it not having RT benchmark score in their card reviews really doesn't hurt them.

    Include driver support for RT and that helps validate NVidias argument about RT, and the benchmarks will now be there, with a dismal showing.

    I say it's better to ignore RT driver support until they can make a decent showing with some real RT HW.
     
  4. euskalzabe

    euskalzabe Gawd

    Messages:
    969
    Joined:
    May 9, 2009
    A very valid point, but I'll give you the counter point:

    1) AMD releases Navi with no DXR support of any kind.
    Worst headline: A year after Nvidia introduced raytracing, AMD stays behind with barely any advancements in Navi.
    Best headline: AMD releases Navi: lacks raytracing, but good rasterization performance

    2) AMD releases Navi with DXR support.
    Worst headline: A year later, AMD finally catches up to Nvidia in the midrange
    Best headline: AMD releases Navi with limited raytracing support

    Even if RT performance is low, at this point it'll be a statement. It'll tell the world what AMD can do. Any sort of raytracing will announce that they're on top of things. Revealing nothing DXR wise will signal to everyone, that they're painfully behind.
     
  5. Pieter3dnow

    Pieter3dnow [H]ardness Supreme

    Messages:
    6,716
    Joined:
    Jul 29, 2009
    It may be not that fair but isn't it Nvidia's job to get software for their hardware?
    They claimed ray tracing this and that prolly some qualification of being the best ?
    And how is that now suddenly a problem for AMD?

    Nvidia has a promise to keep to their customers and AMD has no obligation at this point whatsoever.

    I will say this that if it is true that both new consoles have dedicated hardware for ray tracing you would be better off on that platform then on PC.
     
  6. Snowdog

    Snowdog [H]ardForum Junkie

    Messages:
    8,601
    Joined:
    Apr 22, 2006
    I think case 2) will create a lot more bad headlines than you think. Things like: farcical RT support. Destroyed in RT. 10X slower than NV (at RT).

    I really don't see any benefit adding driver support for RT like the NV 1600 series.
     
  7. demondrops

    demondrops Limp Gawd

    Messages:
    328
    Joined:
    Jul 7, 2016
    went outside today, all settings maxxed out @ 12k 120 fps mega ultra raytracing. godrays everywhere. not bad, no quest but to walk up this steep 420 m trail :p
     
    N4CR, dvsman, c3k and 2 others like this.
  8. euskalzabe

    euskalzabe Gawd

    Messages:
    969
    Joined:
    May 9, 2009
    AMD has one big obligation: to remain in this industry, if that's what they want. Not keeping up = disappearing. I fear more and more that's what will happen if they don't pick up the pace. I've never, ever been worried about their CPUs. Their GPUs paint a very different picture.

    Yeah I never said it'd be great headlines... but good or bad, DXR support means AMD will be talked about. Plunging deeper into irrelevance means the industry will stop paying attention to AMD. That's how companies disappear. I don't mean to be all doom and gloom here, but AMD's GPU division has me genuinely worried. Even in their bad CPU decade I remained a strong believer in AMD. Radeons are making it quite hard for me to ra-ra. DXR would signal they're still here. Lack of it would signal the opposite.

    But hey, hopefully I'm wrong. We'll see how the industry and consumers react in a few days anyway.
     
  9. Algrim

    Algrim [H]ard|Gawd

    Messages:
    1,407
    Joined:
    Jun 1, 2016
    DXR is now a DX12 feature. nVidia can claim that DXR checkbox for Pascal and Turing and, so far, AMD can't. Think of this as nVidia's async compute. It really didn't fucking matter that nVidia didn't support it since it only mattered in one game at the time of release but many people ridiculed nVidia for not having 'TRUE ASYNC COMPUTE' (an optional DX 12 feature).

    Do I care if AMD supports DXR or not? No. But at the end of the day if you have two graphics cards that cost the same, use about the same power envelope, play games about the same, but one card has support for a standard API (no matter how few games use it) and one card doesn't, guess which card is more likely to be bought?
     
    Auer likes this.
  10. travm

    travm Limp Gawd

    Messages:
    287
    Joined:
    Feb 26, 2016
    History says the green one, irregardless of all other facts.
     
  11. Jandor

    Jandor [H]Lite

    Messages:
    99
    Joined:
    Dec 30, 2018
    Navi is just Polaris updated and optimized for gaming. It will have VRS like Turing and no FP16/FP64 at all. The die will be much smaller than Vega, despite being probably very close to a Radeon VII in gaming. Radeon VII will be much better in compute. Navi will be cheap to produce. AMD will have big margins on the sales at the price spotted by Sapphire while loosing money on Radeon VII. Vega56/64 is also making AMD to loose much money, and production will be terminated. Count on something less than 200mm² for Navi GPU, smaller than Polaris 10/20, compared to 330m² for RadeonVII and 480m² for Vega 56/64. 7nm is not a new process anymore and has improved a lot. I believe the price Sapphire mentioned is unreasonable and will be revised during the sales. We'll see the price drop by some 100$ for sure, if not at launch by AMD. Fact is, if you want big screen gaming with good FPS, you may prefer Navi instead of RTX 2070, if Navi is faster and less expensive and RTX 2060 needs more RAM whatever speed you want. 6GB is insufficient for anything bigger than Full HD. So both Navi cards are okay as long as they drop some of the price Sapphire mentioned.
    Beware ! If the price of Navi is so high for real, expect Zen 2 to be much more expensive than the leaks, like up to 450$ for the 12 cores, 650$ for the 16 cores which will still be better at this point than Intel.
     
  12. Algrim

    Algrim [H]ard|Gawd

    Messages:
    1,407
    Joined:
    Jun 1, 2016
    If you're trying to compete but you can't go toe to toe on features, you have to bring something else to the table; better speed, better power draw, more or faster memory, and/or price. Having everything else equal but not competing on feature set is a losing proposition.

    Again, I want AMD to be competitive. They have to bring something to the table that the green team can't.
     
    euskalzabe likes this.
  13. Ocellaris

    Ocellaris Ginger @le, an alcoholic's best friend.

    Messages:
    18,780
    Joined:
    Jan 1, 2008
    AMD has a better and more forgiving fanbase :)
     
  14. Algrim

    Algrim [H]ard|Gawd

    Messages:
    1,407
    Joined:
    Jun 1, 2016
    As someone who plans on jumping to the red team when I'm done with other priorities (house plans and a new firearm) I personally would pick up a red team card if they are close. If they're not, I'm throwing money away that I can use elsewhere.
     
  15. Jandor

    Jandor [H]Lite

    Messages:
    99
    Joined:
    Dec 30, 2018
    Best way to be competitive now is to drop the price on cards that are doing the same as Nvidia with less garbage (IA and RT) on the chip and put some more Vram on it, like 8GB when Nvidia puts only 6, and 16GB when Nvidia puts 8.
     
  16. euskalzabe

    euskalzabe Gawd

    Messages:
    969
    Joined:
    May 9, 2009
    Ding ding ding! That is very much correct. Marketing is a powerful thing. That's why I say AMD needs to at least match the value, they're in no position to charge the same while offering less.

    Also, "irregardless" does not exist. Regardless is what you mean. SORRY. Literature professor. Can't turn it off. :)
     
    Algrim and KazeoHin like this.
  17. Boil

    Boil [H]ard|Gawd

    Messages:
    1,349
    Joined:
    Sep 19, 2015
    I hope you had a union sanctioned safety meeting after that 420m hike...!
     
  18. XoR_

    XoR_ Gawd

    Messages:
    717
    Joined:
    Jan 18, 2016
    Nvidia added DXR support for GTX cards for two reasons:
    - for developers without RTX hardware to play with DXR effects (UE/Unity/custom)
    - to allow some performance comparisons between GTX and RTX cards, especially
    - to show off Turing shader improvements
    - just to have it and have it before AMD adds software support for non-hardware-DXR chips - any feature, even if kinda useless unless you are game developer/programmer is a feature that can possibly increase sales and improve reputation

    AMD most probably didn't even bother adding DXR support for GCN cards so they do not have properly developed, tested and validated driver yet and simply cannot add it to Navi even if they wanted to. This is my guess though. It is something that I assume is much harder than simply using DX12 fallback driver, something AMD cards can most probably already do to show it can work - hence the news some time ago about this very topic.

    Also it might be that enabling DXR on AMD cards while it works is not running any good... in relative to GTX cards sense, not RTX... and adding feature which would only worsen perceived performance would not be very beneficial to AMD. Shader implementation needs to perform at least as good as on GTX1080/Ti for card around this performance levels.
     
    euskalzabe likes this.
  19. XoR_

    XoR_ Gawd

    Messages:
    717
    Joined:
    Jan 18, 2016
    Simpler core is a plus. Gaming GPU do not need FP64 at all in them and Nvidia locks out performance of this mode anyway. It can be emulated.
    Not having VRS/FP16 is on the other hand pretty bad... not disastrous because game support is not there yet.
     
  20. Pieter3dnow

    Pieter3dnow [H]ardness Supreme

    Messages:
    6,716
    Joined:
    Jul 29, 2009
    Another bad case of apples and oranges.
    Async is a feature which benefits certain aspects of how calls are made on the gpu and as soon as we saw that Nvidia option was to not have that path enabled in one of the titles supporting Async it was clear their architecture was not able to get it working for them it took them a while to fix (hardware). We are not talking about something needed to run certain games it will still play on Nvidia cards, nor was there any sort of craptastic way to slow down games on Nvidia hardware using Async compute which we all know how that went with tessellation.

    It is not about caring about support it is about how it is played out. You are stating that Nvidia cards are bought with a preference because of ray tracing. I can say that is somewhat odd remark to make because not just of the games also how everyone knows it is first generation hardware which in general does not support it in a level which people will see the clear benefits from (say as much as first generation T&L in hardware was slower).

    As far as I know the reason why DXR is there prolly has to do with next generation consoles (Xbox one of course .....)

    There is nothing that says that AMD does not support it but to be honest it takes 3 years to develop a gpu so you can do the math on that one.....
    Correct their obligation is to do the best they can. But can you do shopping with $5 while you need $15 groceries ? Forcing AMD to sacrifice die space because of Nvidia promises is pretty weird way of stating that AMD is not going to be there.
    Most of you people are not realistic.
    Start posting the R&D budget whenever you start to make demands of AMD and then wonder about what they did in comparison with Nvidia.

    Most of you have an extremely short sighted view on these matters.
     
  21. Pieter3dnow

    Pieter3dnow [H]ardness Supreme

    Messages:
    6,716
    Joined:
    Jul 29, 2009
    Nvidia added lip service. Developers were prolly saying/hinting that they can take a hike with their ray tracing features because there is no user base for it. Nvidia thought that they could just change that around the quick and easy way. Not due to selling more dedicated ray tracing hardware but adding a boatload of certified cards that no one knew could do ray tracing but suddenly can.

    The reality of Nvidia piss poor track record on implementing ray tracing is because it does not work with developers or publishers to justify the time it takes to do it. In the super greedy world where Nvidia showed everyone how it works they become a victim of their own inclinations.

    The reality of the situation is that AMD does not do DXR and if they did it still would not improve the software situation since most hardware is not really suited for ray tracing to begin with .......
     
    funkydmunky likes this.
  22. travm

    travm Limp Gawd

    Messages:
    287
    Joined:
    Feb 26, 2016
    :) I forgive you.
    I wasn't writing literature, and have my own dictionary.
    Https://en.m.wikipedia.org/wiki/Irregardless
    Also top hit on Google says otherwise.
     
    Pieter3dnow likes this.
  23. c3k

    c3k 2[H]4U

    Messages:
    2,089
    Joined:
    Sep 8, 2007
    Not to jump on the bandwagon, but if wiki and google say it is acceptable, I will disagree. So, regardless of what those online "sources" say, I must side with the professor on his "irregardless" quest. ;)

    Navi: I am prepared to be disappointed. If it can get V56/64 performance at much lower TDP and less cost, I'll pick one (or two) up. My rigs are getting too far behind. Otherwise, I'll look at a stopgap 1660Ti or two.

    My purchase decision is AMD's to lose.
     
    euskalzabe and Pieter3dnow like this.
  24. Snowdog

    Snowdog [H]ardForum Junkie

    Messages:
    8,601
    Joined:
    Apr 22, 2006
    Oh, I see plenty of reasons for NVidia to add DXR drivers to GTX cards, mainly to promote and up-sell RTX cards.

    It's just that those reasons don't apply to AMD.
     
  25. sabrewolf732

    sabrewolf732 2[H]4U

    Messages:
    3,814
    Joined:
    Dec 6, 2004
    Did you read the source you listed?
    "Most dictionaries list it as non-standard or incorrect usage, and recommend that "regardless" should be used instead.[2][3][4]"
     
    euskalzabe likes this.
  26. Boil

    Boil [H]ard|Gawd

    Messages:
    1,349
    Joined:
    Sep 19, 2015
  27. travm

    travm Limp Gawd

    Messages:
    287
    Joined:
    Feb 26, 2016
    Cambridge dictionary lists it as a word. Also notes it is not considered a word in the USA. I also spell "colour", so we can all agree to disagree.

    On topic I wish they'd release the damn specs and price already. Waiting sucks
     
  28. euskalzabe

    euskalzabe Gawd

    Messages:
    969
    Joined:
    May 9, 2009
    Cambridge lists it because it is used in the UK, however ungrammatical. The US hasn't stooped that low for now. Not that ungrammatical constructions can't become standard words: there's a well-known example in the Spanish equivalent of "with me" = conmigo. It originates in the latin cum me, which could also be prefixed as mecum. Silly romans didn't realize this practice in their many conquests, and started combining cum me + mecum, ending in "cummecum" which evolved into "conmigo". A centuries long grammatical error that became a standard word one millenium later. That doesn't mean that "irregardless" is grammatical. If the word were "regardwith", then it'd make sense to say "irregardwith". The i- exerts oppositve value, but so does the -less. Therefore, by saying i-regard-less, 2 negatives cancel each other out, so instead of saying "irrespective of x" what you're saying is "precisely because of x". As I often remind my students, facts don't change just because you disagree with them. Then again, nobody's forcing you to speak correctly :)

    I'll tie the linguistic discussion with Navi, as I think it still relevant. The big waves about 3080 XT. An arbitrary decision on AMD's part to resemble Nvidia's product numbering, hoping to confuse the average consumer. Not a very confidence inspiring practice. Look how they did the same with their motherboards, forcing Intel to switch to B360. Oh no, what will Nvidia do? Not use 3080 in their next products? I'd like further simplification actually, RTX 38, 37 and 36. Navi should really just use the RX600 series. RX 3080 XT has too many exes for my taste. But I guess bigger numbers and more exes mean MOAR PERFOARMANCE, right?

    I was talking about DXR support via software, which uses the shaders AMD already is including in their GPUs. Therefore, 0 unjustified die space sacrificed. Adding the DXR support would be 1) giving their shaders a new capability via DX12, and 2) a good dry-run before they have hardware acceleration support in the 2020 lineup. Seems like some of us are a bit more realistic than you are...
     
    Last edited: May 23, 2019
    sabrewolf732 likes this.
  29. ManofGod

    ManofGod [H]ardForum Junkie

    Messages:
    10,223
    Joined:
    Oct 4, 2007
    The RX600's are going to be bottom end refreshes of the RX500 series. (Bottom end, just to be clear.) I have no issue with using the name they are doing, if that is what it ends up being.
     
  30. euskalzabe

    euskalzabe Gawd

    Messages:
    969
    Joined:
    May 9, 2009
    You don't think it's precisely intended to confuse the average Joe? I've criticized Nvidia before for their 1060 naming shenanigans, and I hold the same position for AMD. I was fooled into thinking I got a 1060 with 3GB VRAM, but the card I got was a lesser GPU - and I'm an informed consumer, who stupidly bought the card too fast instead of waiting for reviews, assuming that a 1060 was a 1060. I can see people being equally naive with the proposed Navi names.

    This kind of trickery makes me lose respect for both of them, AMD or Nvidia. I've just never seen AMD do this kind of thing (and if I'm missing a historical example, I'm all ears).
     
  31. Algrim

    Algrim [H]ard|Gawd

    Messages:
    1,407
    Joined:
    Jun 1, 2016
    Everything you just stated in regard to async compute can be said about DXR. What is a DXR API call other than an instruction to the GPU to do something? Did you just really type all that out without understanding that you could have just as easily transposed async compute, DXR, AMD, and nVidia and have made the exact same argument?

    Your reading comprehension needs work. I most certainly did not state that. I said

    If everything is the same except for some feature, it again doesn't fucking matter what the feature is, it will be a reason to choose one over the other. A card that supports a feature, no matter how unused, that another card doesn't, is considered to have more features.

    Take, for instance, the RTX 2080 vs. Radeon VII. Both are good cards. The 2080 has RTX and DLSS. Thw Radeon VII doesn't have either of those features but has 16 GB of HBM memory and had enough compute functions to eat up two RTX 2080 cards. The Radeon VII, whilst still costing the same, performing about the same, has power draw a bit higher but nothing catastrophically higher, still brings something to the table that the 2080 can't. Thise differences make the Radeon VII a better buy for some people for whom those features matter (such as myself). For those people who don't care about large, fast memory pools and the gobs of compute power but who do want to experience hardware ray tracing support, the RTX 2080 is the better buy.

    Again, let me restate this yet again as you seem to not understand: If the cards are identical in every other way than one has a feature or two or three above and beyond the other, the card with more features is going to be perceived as the better card.

    Who makes the hardware for the XBOX/Playstation again?

    In my perfect world, AMD's CEOs wouldn't have spent millions of dollars on parties and mismanaged the company almost into irrelevance. You like to bring up the R&D budget all the damn time to absolve them of needing to compete. They had a fantastic R&D budget at one point but R&D meant Recreation and Drugs and not Research and Development. :rolleyes:
     
    N4CR and euskalzabe like this.
  32. Colonel_Blimp

    Colonel_Blimp n00b

    Messages:
    48
    Joined:
    Dec 30, 2017
    My understanding is that irregardless can be used, on it's own.

    So ,

    History says the green one, irregardless.

    Or

    History says the green one, regardless of all other facts.

    But then I'm British, what would I know!
     
    travm likes this.
  33. Pieter3dnow

    Pieter3dnow [H]ardness Supreme

    Messages:
    6,716
    Joined:
    Jul 29, 2009
    Async does not use anything that requires a different render engine ... As a feature both are very different ....
    It is simple features that are useless what use are they?
    Custom hardware made for Sony and MS according to what they want to support. Again takes a couple of years ....
    This does not mean that it is something that is far further along the evolutionary line but it is what the hardware needs to be able to handle according to wishes from company X or Y ....
    I feel silly having to explain this again. No money means hardly any improvement. It is what it is.
     
  34. ManofGod

    ManofGod [H]ardForum Junkie

    Messages:
    10,223
    Joined:
    Oct 4, 2007
    Nope, I do not think it is meant to confuse the customer at all. It is meant to make AMD appear to be current, however, as for what they are selling.
     
  35. Algrim

    Algrim [H]ard|Gawd

    Messages:
    1,407
    Joined:
    Jun 1, 2016
    AMD built Asynchronous Compute Engines (ACEs) and then added the API calls to DirectX 12 as an optional extension so that they could be called using DX12 API calls. I see no real difference than AMD, Microsoft, and nVidia agreeing to having DXR added to the DX12 spec so that API calls could be made to invoke access to hardware built to take advantage of it.

    If you can't see that CARD A and CARD B being identical with the exception of CARD B + SOME FEATURE NOT FOUND IN CARD A would be considered to be a better value no amount of logic is going to get through to you and I'm not going to waste my time.

    Yet my 1070 purchased in 2016 can handle DXR but no AMD card of any kind yet released can. :confused:

    You should feel silly explaining it again. It's a stupid excuse when said once. It's inexcusable now.
     
    euskalzabe likes this.
  36. Digital Viper-X-

    Digital Viper-X- [H]ardForum Junkie

    Messages:
    13,626
    Joined:
    Dec 9, 2000
    your 1070 can handle dxr? in what 0.o
     
  37. euskalzabe

    euskalzabe Gawd

    Messages:
    969
    Joined:
    May 9, 2009
    Last edited: May 23, 2019
  38. XoR_

    XoR_ Gawd

    Messages:
    717
    Joined:
    Jan 18, 2016
    He he
    You are right in that piss poor performing DXR on Radeons might be good motivation to go with RTX instead of new Radeon card so AMD might wait with similar move (shader DXR implementation) until they also have hardware DXR themselves. On the other hand if they do not add any DXR they fall more and more behind and give another reason for a lot of people (like myself XD) to ridicule them. This is situation for AMD in which no matter what they do they loose.

    But again: this is mostly due to ridiculous pricing (which is still a rumor though, hopefully false one)
     
  39. XoR_

    XoR_ Gawd

    Messages:
    717
    Joined:
    Jan 18, 2016
    Well, not everyone need fancy HD 720p
    In my days we played ray-casted games in 320x200 :)
     
    Pieter3dnow likes this.
  40. Algrim

    Algrim [H]ard|Gawd

    Messages:
    1,407
    Joined:
    Jun 1, 2016
    From what I can gather, DXR requires support for the DX12_1 feature set called Conservative Rasterization. Pascal has support for Tier 2; Turing, Vega, and Volta has support for Tier 3. This allowed nVidia to create a driver for the Pascal, Turing, and Volta architectures that can handle DXR. nVidia cards with RTX can hardware accelerate the DXR operations allowing for better (note I did not say stellar) performance. Presumably AMD could use the Tier 3 feature set in Vega to enable DXR support; non-hardware accelerated, of course, just like my 1070 so naturally the performance is piss-poor.

    The point being that my three year old card has support for a feature that was just released without having been designed to do so.
     
Thread Status:
Not open for further replies.