AMD "Navi" GPU by Q3-2018: Report

[btarunr]

AMD is reportedly accelerating launch of its first GPU architecture built on the 7 nanometer process, codenamed "Navi." Graphics cards based on the first implementation of "Navi" could launch as early as by Q3-2018 (between July and September). Besides IPC increments with its core number-crunching machinery, "Navi" will introduce a slew of memory and GPU virtualization technologies.

AMD will take its multi-chip module (MCM) approach of building high-performance GPUs a step further, by placing multiple GPU dies with their HBM stacks on a single package. The company could leverage its InfinityFabric as a high-bandwidth interconnect between the GPU dies (dubbed "GPU module"), with an I/O controller die interfacing the MCM with the host machine. With multi-GPU on the decline for games, it remains to be seen how those multiple GPU modules are visible to the operating system. In the run up to "Navi," AMD could give its current "Vega" architecture a refresh on a refined 14 nm+ process, to increase clock speeds.



View at TechPowerUp Main Site

 

[cucker tarlson]

No way this is coming out by Q3 2018. Maybe a working ES by Q3 2018. Maybe. AMD pulling the wool over people's eyes again, that's been their strategy for quite some time. But I guess they need something new to get their fanbase hyped again after Vega failed hard. Infinityfabric,HBM2 and all their fancy features will always fall short until they do something to come up with a power efficient architecture that can achieve higher clocks without pulling ridiculous amounts of power.

btw this is a nvidia slide

 

[CyberBuddha]

No way this is coming out by Q3 2018. Maybe a working ES by Q3 2018. Maybe. AMD pulling wool over people's eyes again.

btw this is a nvidia slide

Is it better now?

 

[cucker tarlson]

Is it better now?
View attachment 92866

now that's a credible source.

 

[Xzibit]

AMD is reportedly accelerating launch of its first GPU architecture built on the 7 nanometer process, codenamed "Navi." Graphics cards based on the first implementation of "Navi" could launch as early as by Q3-2018 (between July and September). Besides IPC increments with its core number-crunching machinery, "Navi" will introduce a slew of memory and GPU virtualization technologies.

AMD will take its multi-chip module (MCM) approach of building high-performance GPUs a step further, by placing multiple GPU dies with their HBM stacks on a single package. The company could leverage its InfinityFabric as a high-bandwidth interconnect between the GPU dies (dubbed "GPU module"), with an I/O controller die interfacing the MCM with the host machine. With multi-GPU on the decline for games, it remains to be seen how those multiple GPU modules are visible to the operating system. In the run up to "Navi," AMD could give its current "Vega" architecture a refresh on a refined 14 nm+ process, to increase clock speeds.

So they took a look back at the Capsaicin slides and came up with a headline for you to click.

AMD did have a wide frame time to guess at all this.

 

[evernessince]

No way this is coming out by Q3 2018. Maybe a working ES by Q3 2018. Maybe. AMD pulling the wool over people's eyes again, that's been their strategy for quite some time. But I guess they need something new to get their fanbase hyped again after Vega failed hard. Infinityfabric,HBM2 and all their fancy features will always fall short until they do something to come up with a power efficient architecture that can achieve higher clocks without pulling ridiculous amounts of power.

btw this is a nvidia slide

Actually Vega is quite power efficient. If you look at GamersNexus, they did a video where they compared Vega to the Fury X and Clock for Clock Vega is significantly more power efficient. In fact I managed to get my hands on a Vega 56 to test this. I flashed the Vega 64 bios onto the card and lowered the voltage to 950mv and left everything else alone. I was getting Vega 64 performance for 80 watts less. The card doesn't consume much more than a GTX 1080 and is on par performance wise. 70c max on the stock blower cooler. There are plenty of other reports of Vega undervolting as well. AMD simply shipped the card a bit too early with too high stock voltage which made them look like they were power suckers.

Also, how can infinity fabric fall short when it's already very well proven in Ryzen? Everyone already knows it's good shit.

 

[RejZoR]

No way this is coming out by Q3 2018. Maybe a working ES by Q3 2018. Maybe. AMD pulling the wool over people's eyes again, that's been their strategy for quite some time. But I guess they need something new to get their fanbase hyped again after Vega failed hard. Infinityfabric,HBM2 and all their fancy features will always fall short until they do something to come up with a power efficient architecture that can achieve higher clocks without pulling ridiculous amounts of power.

btw this is a nvidia slide

You're not getting this. Navi is FAR more feasible than Vega. Vega is a MASSIVE chip. Massive chips are expensive, hard to produce and with low yields.
Why do you think there are no shortages of Ryzen CPU's? It's because they are "glued" together using smaller cores that are cheaper, easier to make and with excellent yields. They have experience with Infinity Fabric already on Ryzen, so if they say Navi comes in Q3 2018, I find that perfectly reasonable timeframe. Hopefully, HBM2 shortages will be fixed by then as well. And this is where HBM2 will come into play. Surface area. For Navi cores, they'll need larger surface for the GPU itself. Which is in a way also a benefit. It's easier to cool larger surface area of GPU than one tiny super hot core. Which is why I think Navi is a great idea if they can make it function.

 

[CyberBuddha]

You're not getting this. Navi is FAR more feasible than Vega. Vega is a MASSIVE chip. Masive chips are expensive, had to produce and with low yields.
Why do you think there are no shortages of Ryzen CPU's? It's because they are "glued" together using smaller cores that are cheaper, easier to make and with excellent yields. They have experience with Infinity Fabric already on Ryzen, so if they say Navi comes in Q3 2018, I find that perfectly reasonable timeframe. Hopefully, HBM2 shortages will be fixed by then as well. And this is where HBM2 will come into play. Surface area. For Navi cores, they'll need larger surface for the GPU itself. Which is in a way also a benefit. It's easier to cool larger surface area of GPU than one tiny super hot core. Which is why I think Navi is a great idea if they can make it function.

I somehow believe that by utilising infinity fabric on GPUs ( I mean larger heat dissipation area) we may see smaller GPU coolers or at least they won't be as noisy as it use to be nowadays. What do you think?

 

[the54thvoid]

So they took a look back at the Capsaicin slides and came up with a headline for you to click.

AMD did have a wide frame time to guess at all this.

This looks like another 'fabricated' AMD story, as if AMD have just said this when TT are just making news? We could nip it in the bud and ignore it, otherwise, cue a 100 page thread running into 2019.

 

[Hood]

Also, how can infinity fabric fall short when it's already very well proven in Ryzen? Everyone already knows it's good shit.

I know, it puzzles me why all Ryzen CPUs are being heavily discounted at retail - their latest and greatest ever - with no successor in sight? Doesn't make a lot of sense. Please enlighten me as to why Ryzen is losing popularity so quickly. I offered to build a Ryzen system for my nephew for his birthday this November. A week later he tells me he wants an Intel system, apparently because his friends laughed at him when he told them he was getting a Ryzen rig. Has AMD lost their mojo this quickly? What gives? What do teenage gamers know about it that we don't?

 

[Boosnie]

Do not expect this to be available to the general public.
If the cores will be accessible as indipendent CUs, those cards will be immediately loaded on a 747 parked in the lot outside the fab.

 

[GorbazTheDragon]

You're not getting this. Navi is FAR more feasible than Vega. Vega is a MASSIVE chip. Masive chips are expensive, had to produce and with low yields.
Why do you think there are no shortages of Ryzen CPU's? It's because they are "glued" together using smaller cores that are cheaper, easier to make and with excellent yields. They have experience with Infinity Fabric already on Ryzen, so if they say Navi comes in Q3 2018, I find that perfectly reasonable timeframe. Hopefully, HBM2 shortages will be fixed by then as well. And this is where HBM2 will come into play. Surface area. For Navi cores, they'll need larger surface for the GPU itself. Which is in a way also a benefit. It's easier to cool larger surface area of GPU than one tiny super hot core. Which is why I think Navi is a great idea if they can make it function.

I hope whatever they make actually performs closer to NVidias chips of similar core count... Vega is massive and still has trouble beating a chip 2/3rd of its size, thats not something they should be looking to repeat.

 

[Boosnie]

I know, it puzzles me why all Ryzen CPUs are being heavily discounted at retail - their latest and greatest ever - with no successor in sight? Doesn't make a lot of sense. Please enlighten me as to why Ryzen is losing popularity so quickly. I offered to build a Ryzen system for my nephew for his birthday this November. A week later he tells me he wants an Intel system, apparently because his friends laughed at him when he told them he was getting a Ryzen rig. Has AMD lost their mojo this quickly? What gives? What do teenage gamers know about it that we don't?

Well, yes, children want the newest toys don't they?
Nothing new really.

 

[ShurikN]

Well, yes, children want the newest toys don't they?
Nothing new really.

Ryzen is new as well...
Children want a bragging right. A name. Like a $1000 iPhone even tho they will only use it for social media and taking selfies, which can be done equally well on a $200 android phone.

 

[R0H1T]

So a full year of back & forth trash talking with incendiary news flashes

 

[prtskg]

I know, it puzzles me why all Ryzen CPUs are being heavily discounted at retail - their latest and greatest ever - with no successor in sight? Doesn't make a lot of sense. Please enlighten me as to why Ryzen is losing popularity so quickly. I offered to build a Ryzen system for my nephew for his birthday this November. A week later he tells me he wants an Intel system, apparently because his friends laughed at him when he told them he was getting a Ryzen rig. Has AMD lost their mojo this quickly? What gives? What do teenage gamers know about it that we don't?

Because it is matured, thick, long and it makes people come money when there is 'Intel inside'!

 

[RejZoR]

I somehow believe that by utilising infinity fabric on GPUs ( I mean larger heat dissipation area) we may see smaller GPU coolers or at least they won't be as noisy as it use to be nowadays. What do you think?

Not really. The heat output will be the same or possibly even tiny bit higher. The thing is, if you have a tiny hot GPU, you have very little surface area to take heat away and send it to the fins. Because of this everyone heavily relies on heatpipes because they can achieve that well. With massive GPU area, it's very likely that even blower style coolers will again become very much functional and also normal cards with big heatsinks could potentially go back to simple extruded aluminium heatsinks without any heatpipes. Which can mean aftermarket cards that cost the same as reference ones. I mean, if you don't have a single hot spot, any cooler will do. Where with single hot spot, you need sophisticated heat transfer to get rid of it as fast as possible from that tiny spot. So, I think AMD is going to win on multiple fronts. Manufacturing cost of GPU, manufacturing costs of coolers thanks to larger heat production surface.

I just think main problem will still be down to HBM2 availability which is still a turd. But now they'll need HBM2 more than ever, because they'll need larger GPU surface area and they can gain that with HBM2 easily. Second problem here along with availability, HBM2 will be the new hot spot, so now that I think about it, this might be a problem for coolers...

 

[Xzibit]

I know, it puzzles me why all Ryzen CPUs are being heavily discounted at retail - their latest and greatest ever - with no successor in sight? Doesn't make a lot of sense. Please enlighten me as to why Ryzen is losing popularity so quickly. I offered to build a Ryzen system for my nephew for his birthday this November. A week later he tells me he wants an Intel system, apparently because his friends laughed at him when he told them he was getting a Ryzen rig. Has AMD lost their mojo this quickly? What gives? What do teenage gamers know about it that we don't?

Well, yes, children want the newest toys don't they?
Nothing new really.

I'd be more concerned that all he got out of it was polling children on their CPU preferences

Nephew being laughed at. He jumps to it being AMDs fault and asks questions. Never mind that he was the one who offered to build him a AMD system in the first place putting his nephew in this situation.

I'd teach him not to give in to peer pressure and not cave in to children making fun of him.

 

[evernessince]

I know, it puzzles me why all Ryzen CPUs are being heavily discounted at retail - their latest and greatest ever - with no successor in sight? Doesn't make a lot of sense. Please enlighten me as to why Ryzen is losing popularity so quickly. I offered to build a Ryzen system for my nephew for his birthday this November. A week later he tells me he wants an Intel system, apparently because his friends laughed at him when he told them he was getting a Ryzen rig. Has AMD lost their mojo this quickly? What gives? What do teenage gamers know about it that we don't?

Oh yes, the anti-AMD shill offered to build his imaginary nephew an AMD system. That's not even the funny part, the funny part is that you are making YOUR rig building decisions based on the bulling of school children.

"I know, it puzzles me why all Ryzen CPUs are being heavily discounted at retail "

?? Define heavily discounted. Or perhaps I should show you a list of Intel CPU prices drops recently to show you the meaning of heavily discounted. I can guarantee that Intel's processors have been discounted much more.

 

[evernessince]

I hope whatever they make actually performs closer to NVidias chips of similar core couny... Vega is massive and still has trouble beating a chip 2/3rd of its size, thats not something they should be looking to repeat.

Well yes, that's because Vega has HBM on the GPU. Not really a fair comparison since Nvidia GPUs do not integrate memory onto the die.

 

[GorbazTheDragon]

Well yes, that's because Vega has HBM on the GPU. Not really a fair comparison since Nvidia GPUs do not integrate memory onto the die.

What?

HBM is not on the GPU die for one, and my point really is that VEGA 64 has 4k CUs while the 1080 which it competes with has only 2.5k.

 

[evernessince]

Not really. The heat output will be the same or possibly even tiny bit higher. The thing is, if you have a tiny hot GPU, you have very little surface area to take heat away and send it to the fins. Because of this everyone heavily relies on heatpipes because they can achieve that well. With massive GPU area, it's very likely that even blower style coolers will again become very much functional and also normal cards with big heatsinks could potentially go back to simple extruded aluminium heatsinks without any heatpipes. Which can mean aftermarket cards that cost the same as reference ones. I mean, if you don't have a single hot spot, any cooler will do. Where with single hot spot, you need sophisticated heat transfer to get rid of it as fast as possible from that tiny spot. So, I think AMD is going to win on multiple fronts. Manufacturing cost of GPU, manufacturing costs of coolers thanks to larger heat production surface.

I just think main problem will still be down to HBM2 availability which is still a turd. But now they'll need HBM2 more than ever, because they'll need larger GPU surface area and they can gain that with HBM2 easily. Second problem here along with availability, HBM2 will be the new hot spot, so now that I think about it, this might be a problem for coolers...

HBM2 may or may not be a problem depending on how well AMD has their high bandwidth cache controller working at the time. It's made to allow the GPU to have a much larger virtual address space than actual VRAM it has. So long as your system still has space available in RAM, your GPU should be able to stream data between the vram and system RAM. That's the main reason AMD chose HBM2 for Vega as regular GDDR cannot have the cells ready quick enough for the constant data access required of streaming data between RAM and VRAM. AMD should be able to reduce the amount of RAM they have to put on their cards. The HBCC is actually the foundation for AMD's SSG line of products that have an SSD integrated into the card that acts as VRAM, which allows you to utilize massive assets or datasets.

 

[evernessince]

What?

HBM is not on the GPU die for one, and my point really is that VEGA 64 has 4k CUs while the 1080 which it competes with has only 2.5k.

Comparing CUs between architectures is like comparing GHz between architectures, worthless.

 

[Prima.Vera]

So basically we are moving to a multi-core GPU, and AMD is preparing the first 4 Cores GPU. Can we have H.T. with that too??

 

[RejZoR]

@evernessince
I'm well aware of all that. The fact remains, HBM2 availability is garbage at the moment. It wasn't Vega core that was not available, it was the HBM2 that set back the Vega launch. It has somewhat stabilized now, although Vega isn't exactly a selling hit anymore which explains why availability has stabilized now. Or they solved the availability. One or another. Or mix of both...