AMD Delays Navi to Q4 2019

[RejZoR]

How Navi performs compared to Radeon 7 is irrelevant. Radeon 7 is just a filler to salvage Instinct GPU's from generating loss. And people buying them are those who prefer AMD and want fastest thing right now and don't care if Navi already beats it in late 2019. They'll just go with Navi then if it proves to be even better.

[leadeater]

4 minutes ago, RejZoR said:

How Navi performs compared to Radeon 7 is irrelevant. Radeon 7 is just a filler to salvage Instinct GPU's from generating loss. And people buying them are those who prefer AMD and want fastest thing right now and don't care if Navi already beats it in late 2019. They'll just go with Navi then if it proves to be even better.

It's not irrelevant, all GPUs have a performance bar they need to reach. That could be another of product, a competing product or something else more subtle like improvement over last generation which is something Nvidia contends with on their high end. 

 

With respect to Navi it actually has to compete, hence if midrange isn't up to scratch which would by that time be Radeon 7 like performance. If it can't do that then it's another high end write off like Vega is/was.

 

Navi can still deliver competitive products at certain price points yet at the same time be an architectural and commercial failure in the same way Vega was. Conciliatory praise only goes so far.

[RejZoR]

4 minutes ago, leadeater said:

It's not irrelevant, all GPUs have a performance bar they need to reach. That could be another of product, a competing product or something else more subtle like improvement over last generation which is something Nvidia contends with on their high end. 

 

With respect to Navi it actually has to compete, hence if midrange isn't up to scratch which would by that time be Radeon 7 like performance. If it can't do that then it's another high end write off like Vega is/was.

 

Navi can still deliver competitive products at certain price points yet at the same time be an architectural and commercial failure in the same way Vega was. Conciliatory praise only goes so far.

What I meant with that was that even if Navi has performance of Radeon 7 or higher at half the price just few months later, that's not important because Radeon 7 is the filler for enthusiasts. Not the other way around. Maybe I worded it weirdly.

[Humbug]

1 hour ago, leadeater said:

If Navi is as good as it's supposed to be, or needs to be, mid range should performance match the Radeon 7.

How? 

 

The fastest Polaris mid-range cards like the rx 580 had 36 CUs. If mid-range Navi has the same or maybe a bit more CUs it will crush the RX 580 thanks to the newer architecture and higher clock speeds. But how can it compete with a Radeon 7 which has 60 CUs?  I think it's reasonable to expect it to compete with the RTX 2060 and 2070.

 

RTX 2080 / radeon 7 performance looks like it's going to remain high end this year at around $700.

 

Unless mid-range Navi is a much beefier chip than we all expected and lands with 50+ CUs.

[RejZoR]

People need to drop all the CU expectations because I don't think Navi will be based on CU's as we know them now. It's just not feasible anymore, especially not for a new GPU design. Or it'll be a very radical design like Polaris was. Same old, but radically redesigned.

[cj09beira]

1 minute ago, RejZoR said:

People need to drop all the CU expectations because I don't think Navi will be based on CU's as we know them now. It's just not feasible anymore, especially not for a new GPU design. Or it'll be a very radical design like Polaris was. Same old, but radically redesigned.

pretty sure it will be very similar, after all its still gcn

[Humbug]

9 minutes ago, RejZoR said:

People need to drop all the CU expectations because I don't think Navi will be based on CU's as we know them now. It's just not feasible anymore, especially not for a new GPU design. Or it'll be a very radical design like Polaris was. Same old, but radically redesigned.

Well what they can do with a given number of ROPs and CUs gets better right? For example an R9 290x has way more resources than an RX 580, but the newer Polaris GPU can get more work done with fewer CUs, apart from the clock speed differences... I guess we can expect a similar trend going from Polaris to Navi.

[MyName13]

I thought navi is coming in a year or two.

[cj09beira]

2 minutes ago, MyName13 said:

I thought navi is coming in a year or two.

no navi is this year

[RejZoR]

35 minutes ago, Humbug said:

Well what they can do with a given number of ROPs and CUs gets better right? For example an R9 290x has way more resources than an RX 580, but the newer Polaris GPU can get more work done with fewer CUs, apart from the clock speed differences... I guess we can expect a similar trend going from Polaris to Navi.

If it was that simple, Vega 64 would just annihilate everything. But it doesn't. It's not that simple.

[leadeater]

1 hour ago, RejZoR said:

What I meant with that was that even if Navi has performance of Radeon 7 or higher at half the price just few months later, that's not important because Radeon 7 is the filler for enthusiasts. Not the other way around. Maybe I worded it weirdly.

Radeon 7 status doesn't matter, what matters is Navi, because that is the performance being speculated. If it's not high enough it's another dead end which nobody actually wants. Navi could be great, Navi could be bad, Navi in some way could effect Radeon 7 but that is the thing that actually doesn't matter. Radeon 7 is only being used as a measuring stick.

[RejZoR]

4 minutes ago, leadeater said:

Radeon 7 status doesn't matter, what matters is Navi, because that is the performance being speculated. If it's not high enough it's another dead end which nobody actually wants. Navi could be great, Navi could be bad, Navi in some way could effect Radeon 7 but that is the thing that actually doesn't matter. Radeon 7 is only being used as a measuring stick.

It'll be good because they don't need king of the hill, but another Polaris like success (especially now that crypto mining is finally gone). They nailed it with RX480 which made my GTX 980 basically worth half the price over night. I don't see why Navi couldn't do the same. I think if they hit roughly GTX 1080 performance at around 250-300 $/€ for the mid range it would be a sweet spot. Just what RX480 did back then.

[leadeater]

1 hour ago, Humbug said:

The fastest Polaris mid-range cards like the rx 580 had 36 CUs. If mid-range Navi has the same or maybe a bit more CUs it will crush the RX 580 thanks to the newer architecture and higher clock speeds. But how can it compete with a Radeon 7 which has 60 CUs?  I think it's reasonable to expect it to compete with the RTX 2060 and 2070.

Just like it's reasonable that RTG is only able to performance match a competitor product released on May 2016 in Feburary 2019? A competitor product that has already been superseded. There is no nice way to put it but that is market failure.

 

As much as I've always liked AMD GPUs, recommended them where it makes sense, they have for a while now been a competitive technical failure. It does take skill to deliver a nice well put together rump steak meal but sometimes you do actually want the eye fillet steak, yea I'm hungry right now . AMD at best has been able to serve up scotch fillet, not quite the best grade scotch filet but scotch fillet none the less. Nvidia however has eye fillet along with scotch fillet, ribeye on bone, t-bone, porterhouse etc all with many combinations of seasoning and sauces, some would say more options than we need but if you want a steak Nvidia will have an option you'd like to eat. Customers over at restaurant RTG have to make do with more limited options.

 

Did I mention I'm hungry?

 

Anyway more back on topic, Navi needs to be competitive in relation to when it's released and that will be a time after the competitor has already released it's current generation of products and is working on the next, that next being on a major node shrink which is an important factor.

 

So what does a forward looking future competitive GPU that isn't doomed to irreverence look like in the mid range, something near Radeon 7 performance. Why? because the high end is going to need to scale much higher in performance than that, higher than RTX 2080 Ti.

[leadeater]

13 minutes ago, RejZoR said:

It'll be good because they don't need king of the hill, but another Polaris like success (especially now that crypto mining is finally gone). They nailed it with RX480 which made my GTX 980 basically worth half the price over night. I don't see why Navi couldn't do the same. I think if they hit roughly GTX 1080 performance at around 250-300 $/€ for the mid range it would be a sweet spot. Just what RX480 did back then.

Yea GTX 1080 would be fine, not excellent but not quite as much as I suspect they need. The problem I see is that mid range Navi sets the tone for the higher end, it lets us know the potential to be. Unless of course big Navi will have some extra significant improvements that would mean a simple scale up performance estimate wouldn't be accurate.

 

I really don't want to see another round of being 2-3 years behind, I don't mind a bit behind but honestly the Pascal generation was far and away better architecturally and enjoyed much better implementation and support from the market due to that.

 

Performance king certainly is nice, they don't really need to get that but they need to be climbing that hill at the same time as Nvidia not scaling it after they have finished and moved on to the next hill.

[Mihle]

13 minutes ago, Jack_of_all_Trades said:

 

What I am thinking is that they don't have a choice to fail with Navi since Microsoft and Sony are expecting that shit to bring consoles to the next level. It has to be good or it will hurt too much to be just another mediocrity.

I bet console manufacturers don't really want to go to Nvidia unless something drastic like Navi not being a thing for another 2 years or something would happen. With AMD they can have full APUs or at least have one partner supply both CPU and GPU. And that AMD is probably cheaper/more willing to work with them for then to get what they want. But that's speculation tho.

[Trik'Stari]

Seems like a bad decision to me. I would think more people would be interested in the rumored "2070 performance at half the price" than the VII.

 

I know I am. The 2080/VII is far too much for me to even contemplate getting one, but half of that price? Now that's a bit more possible.

[ZestyJalapeno]

Give me the gun, I'll do it myself!

 

[Agosto]

7 hours ago, Humbug said:

How? 

 

The fastest Polaris mid-range cards like the rx 580 had 36 CUs. If mid-range Navi has the same or maybe a bit more CUs it will crush the RX 580 thanks to the newer architecture and higher clock speeds. But how can it compete with a Radeon 7 which has 60 CUs?  I think it's reasonable to expect it to compete with the RTX 2060 and 2070.

 

RTX 2080 / radeon 7 performance looks like it's going to remain high end this year at around $700.

 

Unless mid-range Navi is a much beefier chip than we all expected and lands with 50+ CUs.

As far as we know, Navi was the architecture which "stole" resources from Vega during their developement and it's almost sure it will go inside new consoles.

Moreover, IIRC, RTG staff has specified that Vega didn't get wider because of time/resources constraint, thus the limit to 4 geometry engines/64 ROPs. GCN has always been front end limited, if "big Navi" can get to 8 geometry engines it could probably match or surpass Radeon VII with less CUs.

Hopefully all those Vega features that didn't actually get activated/used will make their way inside Navi (in a functional way): this way the last iteration of GCN could have the largest jump in performance since the architecture was launched, paving the way to the supposed SuperSIMD stuff coming after Navi.

[Taf the Ghost]

5 hours ago, leadeater said:

Yea GTX 1080 would be fine, not excellent but not quite as much as I suspect they need. The problem I see is that mid range Navi sets the tone for the higher end, it lets us know the potential to be. Unless of course big Navi will have some extra significant improvements that would mean a simple scale up performance estimate wouldn't be accurate.

 

I really don't want to see another round of being 2-3 years behind, I don't mind a bit behind but honestly the Pascal generation was far and away better architecturally and enjoyed much better implementation and support from the market due to that.

 

Performance king certainly is nice, they don't really need to get that but they need to be climbing that hill at the same time as Nvidia not scaling it after they have finished and moved on to the next hill.

AMD needs to hit performance in price categories. That's been the name of the gaming for a long time. If they hit one out of the ballpark, they'll be able to put Vega 64/1080/2070 performance in the 250-300 USD price range, and then good performance in the categories below that. That's where dGPUs actually sell in Volume. There's a reason Jensen was so pissed off when AMD flooded the channel (and forced prices down) during the Holiday 2018 season. AMD's production costs for Polaris, at this point, is wholly about the BOM cost and R&D is all paid off. Good time for a volume play right into Nvidia trying to clear stock without dropping margins. (Frankly, it was one of the shrewder business moves AMD has made in a long while.)

 

GDC is in March. We're going to know a lot about Navi & Zen after that conference. The leaks from credible sources have put a lot of emphasis on the Console stuff getting displayed there, in a ramp for the 2020 release of the next console generation. We should have a better idea where Navi will land. AMD could really use about a ~60% performance uplift per price category. 

[Doobeedoo]

2 minutes ago, valdyrgramr said:

Nobody knows to be honest.  The only confirmation about the high end is that Radeon VII is part of it as stated by Lisa at CES.  We probably won't know anything official until at least Computex.

True. I know they sometimes want to make a surprise so trying to read between the lines. Same how for Radeon 7 it was thought to be only for Instinct cards with first 7nm product and also no new architectures for GPU side yet. But yeah, wait.

[Doobeedoo]

55 minutes ago, valdyrgramr said:

I'm waiting until August before deciding on a Radeon VII, or maybe Navi.  Might go 2080/2080 Ti still, but I need a card great for Maya that games at least at the 2080 level hence why I want the Radeon VII.  I just don't wanna pay Jen's ridiculous 2080 Ti price.  Not that I can't afford it, but there's no justification to the cost of the one card other than, "Because we can ask that much!  No competition!"  Right now, though, the Radeon VII is the ideal card for what I want/need, but I don't need it at this exact moment either.  So, I can wait until August to see how things play out.  XD  As for Maya the benches used were synth benches by a 3D party company.  Someone who bought a Radeon VII is going to do workload tests aka real world application for me in Blender since Maya and Blender are somewhat similar.

Cool. I've seen workloads that can use the 16GB VRAM rather well. For like video work. I'll buy a new PC this year will also see what Navi brings, for now R9 290 works rather well.

[Taf the Ghost]

1 hour ago, valdyrgramr said:

Has that been confirmed?  Cuz everyone else has been stating Computex in June.

Motherboards for Zen are coming around Computex. GDC has a lot of technical information about the Consoles coming, which means Navi & Zen 1 or 2 information.

[Taf the Ghost]

1 hour ago, Doobeedoo said:

True. I know they sometimes want to make a surprise so trying to read between the lines. Same how for Radeon 7 it was thought to be only for Instinct cards with first 7nm product and also no new architectures for GPU side yet. But yeah, wait.

Radeon VII was on & off the schedule at least twice in the rumor mill during 2018. They were mostly correct. It was off, then put back on when the schedule got strung out for Navi. This was always a card they could release, it was just a matter if they felt the need to. (People forget this is exactly what the 2080 and TI class cards are like as well.)

[THraShArD]

The Summer is a loooong wait.  I'm expecting AMD to release another video card before then

[leadeater]

2 hours ago, Taf the Ghost said:

AMD needs to hit performance in price categories. That's been the name of the gaming for a long time. If they hit one out of the ballpark, they'll be able to put Vega 64/1080/2070 performance in the 250-300 USD price range, and then good performance in the categories below that. That's where dGPUs actually sell in Volume. There's a reason Jensen was so pissed off when AMD flooded the channel (and forced prices down) during the Holiday 2018 season. AMD's production costs for Polaris, at this point, is wholly about the BOM cost and R&D is all paid off. Good time for a volume play right into Nvidia trying to clear stock without dropping margins. (Frankly, it was one of the shrewder business moves AMD has made in a long while.)

 

GDC is in March. We're going to know a lot about Navi & Zen after that conference. The leaks from credible sources have put a lot of emphasis on the Console stuff getting displayed there, in a ramp for the 2020 release of the next console generation. We should have a better idea where Navi will land. AMD could really use about a ~60% performance uplift per price category. 

They still need something like Nvidia has been able to do with their archs, even if they don't hit the highest performance mark by 10%-15% something that can scale down easily would be more successful and cheaper for AMD. Being cheap to make alone isn't enough, products need to be desirable. Stressing archs outside of where they are well balanced just to compete doesn't make them desirable, they can still be well priced and meet the performance for the price point.

 

Getting general business aspects done very well unfortunately doesn't counteract being on a technical level behind your market competitors.

 

Polaris while good took too long to get there, binning on binning on binning is not as good as "Here's GP106 and it's good now, oh and here's a bunch of cut downs".

Fiji wasn't any good and hasn't aged well.

Vega is doing what Fiji didn't but similarly too late to matter.

Vega 20 is literally an entire generation too late. Seriously think how good and profitable it would have been competing in the same market as the 1080 Ti without having to do ultra cut rate pricing. Yes I know 7nm wasn't ready for that to happen but requiring 7nm to compete with 16nm is not a good look.

 

AMD may have been able to make Nvidia release reactionary products but it was in no way difficult for Nvidia to do so and utilized existing dies and cutting them down which is vastly cheaper and easier to do than what AMD had to do to drive Nvidia to do that, e.g. 1070 Ti.

 

Navi needs to be good, not good for AMD and compared to it's products but actually good for the products it will compete against. Navi won't just be competing with RTX 20 series because due to coming later to market than would be best it's in the timeline of Nvidia's potential 7nm generation. If Navi is going to be anything like Zen the performance jump required is significant. Timing just isn't on AMD's side right now.