Soldato
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
** The AMD Navi Thread **
- Thread starter subbytna
- Start date
Tbh with you, mid-late 2020 might as well be next century. Hard to care about 2080ti performance 2 years later, unless it has at least some great redeeming feature for other purposes/extreme use cases ala Radeon VII with its 16gb hbm 2 vram. Even there, it's nice but hardly exciting.
WCCFTech, of course it's nonsense
To be fair to WCCFTech (no I cannot believe I just wrote that either
) the parent article is from RedGamingTech.TLDR:
What they say about Raja is for the most part true.
'Detail' about Navi 20 is mostly just speculation / fabrications, as are their projected dates.
What is true however is that there is definitely a larger Navi coming, and likely significantly faster than the RVII for gaming. Its release date depends almost entirely on how much foundry capacity at TSMC that AMD can secure on the 7nm node, and potentially Samsung.
If they can get enough wafers, it'll probably launch before Thanksgiving in the US. If they can't get enough wafers and Zen 2 (across Ryzen / TR / EPYC) is as popular as it could be, it could well get delayed into Q2 2020 ... but that would be far from ideal as RVII is expensive, still eats 7nm wafers, and is almost certainly lower yield than NAVI will be due to be being the pipe-cleaner shrink.
If it gets pushed out far enough, it could end up being their mid tier card, due to proximity to 'Next Gen' launch, as AMD will be virtually certain to want at least some GPU and CPU on 7nm EUV in 2020.
What they say about Raja is for the most part true.
'Detail' about Navi 20 is mostly just speculation / fabrications, as are their projected dates.
What is true however is that there is definitely a larger Navi coming, and likely significantly faster than the RVII for gaming. Its release date depends almost entirely on how much foundry capacity at TSMC that AMD can secure on the 7nm node, and potentially Samsung.
If they can get enough wafers, it'll probably launch before Thanksgiving in the US. If they can't get enough wafers and Zen 2 (across Ryzen / TR / EPYC) is as popular as it could be, it could well get delayed into Q2 2020 ... but that would be far from ideal as RVII is expensive, still eats 7nm wafers, and is almost certainly lower yield than NAVI will be due to be being the pipe-cleaner shrink.
If it gets pushed out far enough, it could end up being their mid tier card, due to proximity to 'Next Gen' launch, as AMD will be virtually certain to want at least some GPU and CPU on 7nm EUV in 2020.
Soldato
So RedGamingTech is talking about the 7nm+ 2020 "next gen" arch then? But then if GCN has been "fixed" then there's no rush (I guess) to push the new arch out.
7nm EUV will be a bit more expensive, but it'll also be quite a bit more dense and higher clock / lower power, everything else being equal. I'm virtually certain that Zen 3 will be on it in a year or slightly over. Whether we see 'Next Gen' consumer cards on it in 2020 probably depends mainly on wafer supply - they're not going to be too bothered about competing with what ever NVIDIA are doing if they're biting off huge chunks of Intel's share, and wafer supply is finite. Also, of course, I don't think almost anyone outside of AMD knows when 'Next Gen' is actually going to be ready to go. Maybe it won't be ready for a while, or maybe it's virtually good to go and waiting on 7nm EUV.
We know NAVI core architecture has been done for a while, and the smaller chips are due in early summer. Bigger NAVI is probably very similar to smaller NAVI, design wise, as it's not intended for enterprise or pro stuff. So it would stand to reason it should be ready when smaller NAVI is, or shortly thereafter.
But like I say ... wafers. AMD have already warned twice about supply constraints for 7nm this year, during investor calls. Whether TSMC are able to meet demand for the PC market is highly suspect, in 2019. Samsung's EUV will be ready to go before TSMC, but it'll also be more expensive than TSMC's non-EUV, and being EUV, it would probably require some significant tweaks. But, almost certainly, at least some of XBOX / PS Next GPUs will be on Samsung EUV, if not to begin with, then shortly thereafter. But, yet again, we still don't know if they'll actually be NAVI, or 'Next Gen'.
Soldato
Personally I don't think we'll see a high-end Navi, simply because I don't think Koduri "fixed" GCN. Architectures are planned years in advance, and the timeline suggested in the RedGamigTech video doesn't line up. Koduri returned to AMD in 2013 yet Vega was supposedly finished in design so nothing could be done to tweak it. We know Vega was late, but it didn't show up until 2017. 4 years from a final design to product? Navi showed up in roadmaps in 2016 and was slated for 2018, then it got pushed to 2019 and now it's still unconfirmed if we'll see it until late this year and there's still "next Gen" on 7nm+ on the roadmap for 2020. And Koduri was put on gardening leave almost 18 months ago, and allegedly a contributing factor was being at odds with Lisa Su given his engineers were shipped over to Ryzen.
So what has he done then? I guess we'll see with Navi, but I suspect Navi will just be a 7nm exercise. It will be the first arch designed specifically for 7nm, not a shrink, and a purely gaming-focussed design. Small, lean, efficient, cheap, make your profit from high volume, low margin sales, get some steady hands at the GPU tiller and go all-in on "next gen" which has been suggested as moving away from GCN finally.
Based on the AdoredTV spec leaks, I've suggested that we could see a higher-end Navi in the form of a full-fat Navi 10, the logic being RX3070 was a full Navi 12, RX3060 a cut-down Navi 12 but only a single Navi 10 product in the RX3080 and we know there's very little chance all of the Navi 10 dies will be perfect. So there's scope for the RX3080 to be a cut-down Navi 10 with a full-fat Navi 10 available after let's say RX3090 for ease of naming, and replace Radeon VII.
In my mind it comes to this: if Navi is scalable beyond Radeon VII performance, then chances are good for a Navi 20, GCN sticks around for a while long because it was "fixed" and "next gen" is pushed back. Or Navi 20 is actually "next gen" with the extra bells and whistles like ray tracing on 7nm+. If Navi isn't scalable then it's the last hurrah for GCN, "next gen" comes along in 2020 and there never will be a high-end Navi.
ALL conjecture given we haven't even seen the small Navi cards yet.
So what has he done then? I guess we'll see with Navi, but I suspect Navi will just be a 7nm exercise. It will be the first arch designed specifically for 7nm, not a shrink, and a purely gaming-focussed design. Small, lean, efficient, cheap, make your profit from high volume, low margin sales, get some steady hands at the GPU tiller and go all-in on "next gen" which has been suggested as moving away from GCN finally.
Based on the AdoredTV spec leaks, I've suggested that we could see a higher-end Navi in the form of a full-fat Navi 10, the logic being RX3070 was a full Navi 12, RX3060 a cut-down Navi 12 but only a single Navi 10 product in the RX3080 and we know there's very little chance all of the Navi 10 dies will be perfect. So there's scope for the RX3080 to be a cut-down Navi 10 with a full-fat Navi 10 available after let's say RX3090 for ease of naming, and replace Radeon VII.
In my mind it comes to this: if Navi is scalable beyond Radeon VII performance, then chances are good for a Navi 20, GCN sticks around for a while long because it was "fixed" and "next gen" is pushed back. Or Navi 20 is actually "next gen" with the extra bells and whistles like ray tracing on 7nm+. If Navi isn't scalable then it's the last hurrah for GCN, "next gen" comes along in 2020 and there never will be a high-end Navi.
ALL conjecture given we haven't even seen the small Navi cards yet.
That AdoredTV 'spec' leak was complete fabrication / conjecture. 100%
As I said before, any big NAVI would almost certainly be very, very similar design wise to smaller iterations. Navi is purely consumer / gaming focused. They should be able to make as many iterations as they feel suits their market. Wafers permitting ... and I'm not sure they will be very permitting at all.
Soldato
Based on what? We literally know nothing about Navi, other than the long-standing belief that it's going to be a mid-range part that is gaming-focussed and cost-effective. Those leaked specs support that narrative. But we've seen nothing whatsoever to confirm nor dismiss those leaked specs, so to say they're "100% fabrication" is baseless, unless you have some source you'd like to share with the class?
Saying that a "big" Navi would be similar to smaller iterations is fine in principle, but it's assuming that Navi can scale. How scalable is GCN? Is that what Koduri "fixed", if he actually fixed anything? And you also say AMD could make as many iterations as they'd like to suit the market, again assuming Navi can scale.
And to be honest your talk about wafers is fud. Do you really think AMD would have gone all-in on 7nm across their entire product stack if they didn't think TSMC could handle it? In fact, I'd be pretty confident in saying that AMD have negotiated a priority clause with them to guarantee wafer throughput; TSMC won't care who their customers are because a wafer is a wafer is a wafer to them so it'll be no skin off their nose who maxes out capacity, and if AMD guaranteed the lion's share then TSMC would snap it up.
I know AMD have been seriously cash-strapped for the longest time, and with a lack of money comes a lack of R&D, but I have to say that if GCN-based architectures were properly scalable to saturate all market tiers we would've actually seen it by now. If Navi is based on GCN then it'll take a miracle for it to surpass Radeon VII performance, but I just don't think it's actually supposed to. Vega is too expensive to produce, Radeon VII is not actually a real product, Polaris is long in the tooth and the RX 590 is just as much of a short-term filler and PR stunt as Radeon VII. AMD need a cost effective GPU that can bring stability back to their graphics division and unified under Lisa Su, and push to reclaim market share and make some money. That's Navi, and that's Navi in 2019. Not some mythical ray-tracing "big" Navi that is showing up in 2020 or later. Unless, as I said from the beginning, Koduri really did work miracles and GCN is suddenly a massively scalable architecture.
Soldato
Isnt Arcturus 2020, which is the non Gcn architecture card? All to compete with nvidia 7nm and Intel Arctic sound (lmao at the name) or is it Xe
Soldato
This is what I keep saying. Based on this latest article, unless Koduri really did work miracles and breathe life into GCN, Navi will only last for about a year. AMD can't keep milking GCN otherwise it won't even have competitive low and mid-range graphics cards any more. As I've said before, there is next to no information about Navi so any discussion we have is all subjective and baseless, but after languishing in the doldrums for so long, it seems logical for AMD to use stick with something they know as a basis to get their house in order. OK, that's GCN but it's as good a starting point as any. Design something explicitly for 7nm, rather than just shrink something, get the production costs down, get the yields up, stabilise the graphics unit and properly integrate under Lisa Su's single leadership. High volume, low margin sales to cement the midrange and get some much-needed stability under their feet, just as Zen did for CPUs.
Then it's a much healthier foundation to push the new arch and move away from GCN. Nobody ever said Navi would return AMD to bleeding-edge competitiveness so yes we are waiting for "next gen" now in 2020, or whenever that will be.
But ultimately we need to see Navi before anything else can be discussed rationally and sensibly.
Navi will definitely last more than a year. Should be obvious by now that new gens are going to take longer and longer to come out and will be milked ever more. At most you'll get a re-hash ala Polaris where they re-release it after a year for ~10% more performance.
Just look at what Nvidia's doing. Unless AMD has some genius breakthroughs they won't be able to R&D faster than the green team, so I wouldn't expect shorter intervals between gens. And once the new crypto wave comes, which it most assuredly will, they will try to milk that further again, as has happened before.
I put absolutely no hope in a timely release of Arcturus, but frankly am happy with AMD so long as it keeps offering a superior value proposition to Nvidia. Truth is, the hardware is plenty strong already from both vendors, the bottleneck is still software (on game devs' side; which is frankly due to economic realities, so not gonna change any time soon). If we are to see more frequent new GPU releases it will be once ray tracing adoption accelerates but that's also "far" (>2020) into the future, so meh.
Just look at what Nvidia's doing. Unless AMD has some genius breakthroughs they won't be able to R&D faster than the green team, so I wouldn't expect shorter intervals between gens. And once the new crypto wave comes, which it most assuredly will, they will try to milk that further again, as has happened before.
I put absolutely no hope in a timely release of Arcturus, but frankly am happy with AMD so long as it keeps offering a superior value proposition to Nvidia. Truth is, the hardware is plenty strong already from both vendors, the bottleneck is still software (on game devs' side; which is frankly due to economic realities, so not gonna change any time soon). If we are to see more frequent new GPU releases it will be once ray tracing adoption accelerates but that's also "far" (>2020) into the future, so meh.
Soldato
Nvidia milk the gens because they can, AMD aren't milking anything because they haven't had the resources to produce anything, i.e. they're stuck. There is a big difference. And how much more can AMD milk GCN? If Navi is even remotely close to what has been talked about, it isn't offering a new performance tier, merely shifting the goalposts on what the cheap midrange market should be in 2019. There is no way AMD can run Vega 64 + 15% performance as their principle offering for a few years, and sure as hell can't maintain Radeon VII as an actual, real product (unless we do get another crypto wave, but even then no amount of Radeon VII sales will ever make up for sacrificing MI50 cards).
AMD aren't going to bang out a new arch every year but I'm fairly sure that slowdown will happen after Navi. It's a stop gap arch to stabilise the GPU division and generate revenue. It can't be anything more otherwise AMD will end up too far behind Nvidia and lose the midrange segment too i.e. that superior value proposition. And then there's Intel joining the fray.
AMD aren't going to bang out a new arch every year but I'm fairly sure that slowdown will happen after Navi. It's a stop gap arch to stabilise the GPU division and generate revenue. It can't be anything more otherwise AMD will end up too far behind Nvidia and lose the midrange segment too i.e. that superior value proposition. And then there's Intel joining the fray.
It is all getting a bit boring now, for me at least. Used to love the anticipation of a new GPU coming out, especially an AMD GPU as they would almost always improve price for performance in a big way. These days we get relatively small performance bump and hardly any improvement on price for performance. Just all a bit boring.
Soldato
True, but as I said that's because AMD haven't had the money to push anything, and without that push from AMD Nvidia has zero incentive to release something significant. I think this is why I want the AdoredTV Navi leaks to be (more or less) true, not because of the performance (because that's nothing new) but because of the price, or the ballpark thereabouts. That'll shake things up a bit I reckon.
RTX 2070-esque raster performance for $250-ish? Nvidia would either have the slash prices of RTX (doubtful), really push devs to utilise RTX tech (to actually make RTX appealing and therefore worth the premium), produce a Turing GTX that performs at RTX 2070 levels (which would be hilarious), accelerate their 7nm plans, or any combination of the above.
Of course this won't happen, but it's nice to dream. But in any event, the excitement in the GPU sector that you miss hinges on AMD getting a decent R&D budget and producing something different (and that surely is post-GCN). Don't expect Intel to get involved in the fray in any meaningful way until their 2nd iteration of product.
Associate
- Joined
- 17 Sep 2018
- Posts
- 1,474
AdoredTVs lineup doesn't make sense from a product portfolio perspective. So Navi replaces Vega and 500 series cards right? Yet the top tier of cards is going to be $250 or £250. So are we to believe AMD won't have any GPU between a £250 '3080' and a £650 Radeon 7? Nothing in the Vega price range? I just can't see that happening, even if it is a Ryzen type strategy, you'd think the cheapest would be at least £350 if it does Vega 64 +15%. Really AMD need a Navi that can compete with the 2080 because Radeon 7 probably won't have the parts to last very long.
Soldato
This is where my crazy idea of a full-fat Navi 10 comes from
And I freely admit it's conjecture based on rumour and not to be taken as anything more than a discussion point.The AdoredTV leak had 2 products built on Navi 12, yet only 1 on Navi 10. We all know that you never get perfect silicon and vendors will turn bits off to feed lower-tiered products. So...
If RX 3060 is a cut-down Navi 12 and RX 3070 is a full-fat Navi 12, logic would dictate that RX 3080 is a cut-down Navi 10. There's no way that Navi 10 die is coming off the wafer in a perfect condition. Which means then there could be a 4th product potentially waiting in the wings as full-fat Navi 10 dies get binned and stocked, let's call it RX 3090. And following this crazy, baseless logic through, if RX 3080 is Vega 64 +15% then RX 3090 is Vega 64 +25% or so, which is where Radeon VII sits. Ergo, the expensive PR stunt that is Radeon VII gets replaced (although how AMD will justify breaking the 16GB VRAM precedent is an issue, unless 16GB GDDR6 gets slapped on RX 3090).
So now then AMD have everything from RX 580 to RTX 2080 (ish) performance covered with Navi, which is where most of the consumer dollar is going anyway, and keep things ticking over for Arcturus (or whatever the hell it's called) and is where the real test begins to see if AMD can fight at the top (if AMD ever decide they want the hassle of halo epeen bragging rights).
As for pricing, I've said from the start the leak isn't going to be realistic as the actual real price, but I do believe it's a show of intent as AMD's tentative MSRP (which was set before anybody knew how much gouging RTX was going to do): high volume, low margin sales to bring some semblance to GPU pricing and gain market share in the face of Nvidia's greed and consumer dissatisfaction. And if Navi is cheap enough to produce all in, they can still make money. Throw 25-40% on those AdoredTV prices; RX 3080 comes in at about $329 which would give RX 3090 a bit of space to be something up to $399. Add a wee bit more if it has 16GB GDDR6.
AMD Navi Preliminary Support Added To HWINFO As AMD Gears Up For Launch https://wccftech.com/amd-navi-preliminary-support-added-to-hwinfo-as-amd-gears-up-for-launch/
- The AMD Partner’s Summit will happen somewhere around the 23rd of this month.
- This will be a regional event (probably North America only), not a global event, and the company plans to have more of these throughout the year.
- Ryzen 7nm CPUs and Navi GPU are on the agenda.
- There is a high likelihood that the embargo for Ryzen 7nm and Radeon (Navi?) will be revealed here (read: launch date).
Caporegime
