[Rumor, Tweaktown] AMD to launch next-gen Navi graphics cards at E3

[maddie]

Even more damning is the leaked price of the 2060 Super. $429 for a x60 card is downright disgusting (as if $349 wasn't bad enough).

I just hope streaming will be good then I won't need a dGPU anymore.

This got me thinking about Google Stadia and similar tech. If this can begin to take off, then discrete GPUs for even more gamers will become redundant. The marginal buyers that struggle to buy at these inflated prices will begin to abandon the market. No one knows for sure, but something to watch.

 

[Dribble]

This got me thinking about Google Stadia and similar tech. If this can begin to take off, then discrete GPUs for even more gamers will become redundant. The marginal buyers that struggle to buy at these inflated prices will begin to abandon the market. No one knows for sure, but something to watch.

It does give an option of just owning a cheap pc (that plays all old games fine) + stadia to play the new releases. It's also a direct replacement for consoles - you don't need one with stadia in the same way as you no longer need a DVD player now you can stream everything.

 

[Glo.]

It does give an option of just owning a cheap pc (that plays all old games fine) + stadia to play the new releases. It's also a direct replacement for consoles - you don't need one with stadia in the same way as you no longer need a DVD player now you can stream everything.

Stadia's input lag, and connection latency will be terrible for all matters E-Sport.

That's why I have written: Each year consumers use PC's more and more in E-Sport gaming, because it offers best possible responsivity of the platform.

Stadia here is not the solution.

If whay people play that do not requires best possible responsivity, why not just buy a Console with good 4K display?

That is my thinking: I will have 1080p/144 Hz capable computer, for Overwatch and Hearthstone, and for everything else: console(s).

 

[IntelUser2000]

I thought about it, and Navi is actually a good advancement for AMD.

The reason is that all Vega variants use HBM2, and its very power efficient. The 50% perf/watt gain over Vega is with Navi on GDDR6.

If the 5700XT can really rival RTX 2070, it won't be that bad at perf/watt either. RTX 2070 may be 10% lower at power if TBP is comparable. And yes, Navi needs 7nm to do so, but the competition is not there, and if Nvidia brings next gen 7nm say in Spring of next year, that's a difference of 9 months. Especially if the Super cards aren't going to be slightly better at $100 reduction like early rumors said.

Certainly much better than needing the super expensive HBM2 memory and still using a lot more power while performing less as with Vega and Fiji.

 

[Glo.]

I thought about it, and Navi is actually a good advancement for AMD.

The reason is that all Vega variants use HBM2, and its very power efficient. The 50% perf/watt gain over Vega is with Navi on GDDR6.

If the 5700XT can really rival RTX 2070, it won't be that bad at perf/watt either. RTX 2070 may be 10% lower at power if TBP is comparable. And yes, Navi needs 7nm to do so, but the competition is not there, and if Nvidia brings next gen 7nm say in Spring of next year, that's a difference of 9 months. Especially if the Super cards aren't going to be slightly better at $100 reduction like early rumors said.

Certainly much better than needing the super expensive HBM2 memory and still using a lot more power while performing less as with Vega and Fiji.

We haven't seen reviews, yet, and people are claiming AMD already lost in efficiency. Tells more about brand perception than anything else .

2 stacks of HBM2 use at best 10W of power. 8 GDDR6 chips should consume around 40W of power.

 

[DrMrLordX]

If Vega 56/64 disappear from the consumer market then we can reasonably conclude that Navi is a Vega 56/64 replacement.

Or we can reasonably conclude that HBM2 is just too pricey for consumer dGPUs right now, giving AMD an incentive to move their HBM products fully into the datacenter.

This got me thinking about Google Stadia and similar tech. If this can begin to take off, then discrete GPUs for even more gamers will become redundant. The marginal buyers that struggle to buy at these inflated prices will begin to abandon the market. No one knows for sure, but something to watch.

Depends on the cost of services like Stadia. Assuming performance is acceptable, you still have to reckon with PC gamers saving money by just waiting on sales. Smart shoppers get their games 40-50% off or more. I can easily blow $700 on a video card if I am saving $20-$30 on every game I buy compared to a console user that is probably paying full price for everything. Or a Stadia user that is paying monthly no matter what.

If it's only $10/month and I can stream whatever I like whenever I like, then $120/year for that service is cheaper than buying a new dGPU every other year, no matter what the vendor.

 

[swilli89]

Or we can reasonably conclude that HBM2 is just too pricey for consumer dGPUs right now, giving AMD an incentive to move their HBM products fully into the datacenter.



Depends on the cost of services like Stadia. Assuming performance is acceptable, you still have to reckon with PC gamers saving money by just waiting on sales. Smart shoppers get their games 40-50% off or more. I can easily blow $700 on a video card if I am saving $20-$30 on every game I buy compared to a console user that is probably paying full price for everything. Or a Stadia user that is paying monthly no matter what.

If it's only $10/month and I can stream whatever I like whenever I like, then $120/year for that service is cheaper than buying a new dGPU every other year, no matter what the vendor.

GPU every other year isn't needed. A geforce 980 still runs 1080p well and that's what 5 years old?

Don't forget Stadia requires a high internet tier.. Gigabit internet in my area is about $100 while I could get 400mbps service for $50. 100mbps service is $30. I know stadia doesn't require gigabit, but I'm sure the experience is better. That could be a difference of hundreds of dollars a year.

I'm aware Vega is still getting sold for HPC... I was responding to the poster wondering if Navi would replace Vega in consumer which it very much will.

 

[soresu]

The reason is that all Vega variants use HBM2, and its very power efficient. The 50% perf/watt gain over Vega is with Navi on GDDR6.

Certainly much better than needing the super expensive HBM2 memory and still using a lot more power while performing less as with Vega and Fiji.

HBM3 will supposedly be a triple Yahtzee of improvement at least:

Lower cost (for a given stack height)
Lower power
Higher bandwidth per pin/stack

There will likely also be higher stack height limits (16+), and larger dies (32 Gbit) - but the former only increases the cost per stack, and the latter would likely require a new DRAM process node at minimum.

I still don't think these improvements on their own will be enough to surpass price efficiency of GDDR6, for AMD at least - it does as always represent a simplification for GFX card makers, decreasing the total GPU/memory area footprint, as well as reducing necessary area for a cooler to directly contact the thermally constrained areas.

 

[Dribble]

Stadia's input lag, and connection latency will be terrible for all matters E-Sport.

That's why I have written: Each year consumers use PC's more and more in E-Sport gaming, because it offers best possible responsivity of the platform.

Stadia here is not the solution.

If whay people play that do not requires best possible responsivity, why not just buy a Console with good 4K display?

That is my thinking: I will have 1080p/144 Hz capable computer, for Overwatch and Hearthstone, and for everything else: console(s).

There will be a hardcore who are very ping sensitive, but for the unwashed masses I suspect super low ping just isn't that important. To use another streaming analogy how many people are happy with streamed music despite it being a drop in quality over CD's?

 

[Stuka87]

If Vega 56/64 disappear from the consumer market then we can reasonably conclude that Navi is a Vega 56/64 replacement.

The 5700 XT replaces the performance window that Vega occupied, but it does NOT replace the market segment that Vega occupied. That will be filled by big Navi next year.

 

[Stuka87]

There will be a hardcore who are very ping sensitive, but for the unwashed masses I suspect super low ping just isn't that important. To use another streaming analogy how many people are happy with streamed music despite it being a drop in quality over CD's?

Not exactly a proper analogy. The feedback loop for listening to music is not the same as controlling a game. In music, at say 320kbps that Spotify uses on desktop you need a pretty high end audio setup to be able to tell the difference between it and AIFF files on a CD. In a game however, when you are playing a game, and you want your character to move at just the right moment, and that moment ends up being later than you wanted, you are instantly going to know because your character may be dead.

 

[Head1985]

RX5700XT limited edition was rx690
https://www.pcgamesn.com/amd/radeon-rx-5700-xt-690-graphics-card-e3
So they changed names before launch.
RX690-RX5700XT limited ed(rx590 replacement)
RX680-RX5700XT(rx480/580 replacement)
RX670-RX5700(rx470/570 replacement)
Another proof that this is just polaris replacement with doubled price.

 

[PhonakV30]

RX5700XT limited edition was rx690
https://www.pcgamesn.com/amd/radeon-rx-5700-xt-690-graphics-card-e3
So they changed names before launch.
RX690-RX5700XT limited ed(rx590 replacement)
RX680-RX5700XT(rx480/580 replacement)
RX670-RX5700(rx470/570 replacement)
Another proof that this is just polaris replacement with doubled price.

heh ? so the one with 50th is therefor RX590 replacement ? I doubt

 

[Shivansps]

Because you said so?

AMD is placing it as Vega replacement. In every single promotional material it is placed agains Vega 56/64. And yet, you say it is placed as Polaris replacement. What do you have to back that up?

You could consider that it is because i say so, if you decide to ignore the fact that Navi 10 is the middle Navi chip, and there is both bigger and smallers chips coming, I know what AMD is doing, this dosent change the fact that Navi 10 is Polaris 10 replacement, Vega replacement is the big Navi that we are yet to see, if this was Vega replacement it would underwhelming. They can price it wharever they want, it dosent change those facts. They really needed to call these cards RX690, RX680 and RX670 for you to figure out it is Polaris replacement? please.

And we already now the RX5700XT 50th was to be called RX690 limited Edition so there is no much to discuss anymore. So i hope to see people who insisted otherwise to finally accept it now.

 

[Shivansps]

heh ? so the one with 50th is therefor RX590 replacement ? I doubt

Why not? the Navi 10 50th edition is a overclocked Navi 10, very similar to the RX590.

 

[prtskg]

I'm aware Vega is still getting sold for HPC... I was responding to the poster wondering if Navi would replace Vega in consumer which it very much will.

Vega 7nm is being used for HPC. I doubt old Vega will be in demand anymore.

Who here still thinks that Radeon VII was not an opportunistic product for gamers.

Radeon VII exists because there were a lot of 7nm Vega with 60 CUs which neither Apple uses nor it can be utilized comletely in professional market. And of course there's the price of Nvidia gpus. Is this what you meant?

 

[Glo.]

You could consider that it is because i say so, if you decide to ignore the fact that Navi 10 is the middle Navi chip, and there is both bigger and smallers chips coming, I know what AMD is doing, this dosent change the fact that Navi 10 is Polaris 10 replacement, Vega replacement is the big Navi that we are yet to see, if this was Vega replacement it would underwhelming. They can price it wharever they want, it dosent change those facts. They really needed to call these cards RX690, RX680 and RX670 for you to figure out it is Polaris replacement? please.

And we already now the RX5700XT 50th was to be called RX690 limited Edition so there is no much to discuss anymore. So i hope to see people who insisted otherwise to finally accept it now.

So you do not have any reasons about it, only your assumptions.

About that name. Isn't RX 690 SKU suggesting top of the line model?

 

[tajoh111]

Why does everybody read something that I have not written here?

Yes, it will bring power efficiency.

But for those expecting that suddenly 250 mm2, 2560 CUDA core, GTX 1660 Ti replacement on 7nm, clocked at 1.9 GHz will use less than 150W should stop dreaming. It won't happen.

On 7 nm SS EUV? Maybe.

Basically your extremely short selling the improvements of a new node, especially when it comes to keeping things near stock settings of todays cards.

An RTX 2070 can already achieve 1930mhz at roughtly 225 watts or 228watts specifically. Not any sort of underclocking/undervolting involved.

https://www.kitguru.net/components/...z-8gb-review-210mhz-faster-than-reference/15/


https://www.kitguru.net/components/...z-8gb-review-210mhz-faster-than-reference/12/

You know what happens when you clock a card at last gens speed?

You literally cut the power consumption in half. There are AMD slides that say this and TSMC 7nm spec says this(50% reduction at the same performance).

Heck, laptop graphic chips deliver this without the nodal improvements.

Laptop variants of Nvidia GPU's are able to deliver 90% of the performance at 75% the power consumption and 80% the performance at nearly half the power consumption. So 1.9ghz 2560mhz will likely be possible at 150watts. Pascal was able to increase the shader clock by nearly 50% at the same power and 25% more shaders at the same power with between gtx 980 and gtx 1080. A mere 10% increase in clocks, with a modest increase of a 11% in shaders, will consume sustantially less than a RTX 2070. At launch, RTX 2070, with b grade silicon quality were consuming 200watts, and 1750mhz. You don't think a new node designed for high performance is not capable of raising clocks 10% and 11% more shaders while shaving 25% power? Why be so ultra pessimistic. AMD even did this Radeon VII raising stock clocks about 20%, while decreasing power consumption by 17%, while adding 2 stacks of HBM(which would add more power consumption then 4cu) and adding power inefficient double precision hardware at their first and earliest stab at 7nm. You think Nvidia engineers are so incompetent that they can't achieve the performance per watt increase of Radeon VII which had a strong compute focus while having using a better version of the node while having 1 to 1.5 years more to work on the chip? Ridiculous.

If Turing clocks at a very conservative 2.2ghz on 7nm, which some turing chips can nearly achieve on air today, it will likely consume 180 watts or less (25% increase in frequency, 10% reduction in power(B grade oc silicon) or (14% increase with a 20% reduction in power vs A grade silicon today) because it is not using the full capability of the node which turns into power savings(potential for the performance version of node is 40% increase). For the most part, next gen nodal parts are clocked better than last gens air overclocks. This is because much of the big reason for the move to the new node is take take advantage of the performance increase per transistor. That means extending the pipeline to get higher clocks. Leveraging simply the power consumption wouldn't allow companies to charge higher prices per card, hence why most companies move to it. Pascal/maxwell are designed for higher clock rates while having power very controlled which is why pascal came out so good. Considering the ln2 clocks of turing today, this will likely be extended with amphere.

7nm turing/amphere is guaranteed to be more efficient than navi because even with AMD performance slides(which are likely inflated), 12nm turing is already slightly more efficient than 7nm navi.

 

[Glo.]

7nm turing/amphere is guaranteed to be more efficient than navi because even with AMD performance slides(which are likely inflated), 12nm turing is already slightly more efficient than 7nm navi.

You haven't seen any reviews of those GPUs, and yet, you are ready to claim RTX 2070 is more efficient than Navi. What does that tell?

Brand perception, maybe?

 

[Mopetar]

I know what AMD is doing, this dosent change the fact that Navi 10 is Polaris 10 replacement, Vega replacement is the big Navi that we are yet to see, if this was Vega replacement it would underwhelming at best.
There is also a smaller Navi that will take the smaller Polaris place. They can price it wharever they want, it dosent change those facts.

I've really got to disagree with you on this. We already know that there's going to be a bigger Navi, but it doesn't make a lot of sense to claim that Navi 10 (Radeon 5700/XT) is the replacement for the 570/580 when the former has an MSRP of $450/$380 and the later had an MSRP of $220/$170. In what world does a replacement product double the cost of what it replaces?

Those are two different segments of the market entirely. When AMD eventually does release a smaller Navi chip that targets the ~$200 market segment, no one who was in the market for Polaris will buy a 5700/XT unless they've found themselves suddenly flush with cash. If AMD doesn't release a smaller chip, you wouldn't see a massive uptick in 5700/XT sales from people who would otherwise buy a 570/580. If the world actually worked like that, NVidia wouldn't have missed sales expectations with Turing.

I think what you fail to account for is that the bottom of the market is always being eroded by APUs. Often times the bottom of the product stack is really just recycled mid-tier parts from the previous version. For example, the Radeon 530 is really just a rebranded 240 (Orland Pro) which gradually moved down the stack. Eventually something like that becomes pointless to manufacture (hence the recycling of old parts) because incorporating that level of performance in an APU becomes more cost effective than producing a stand-alone GPU.

Replacement occurs from the perspective of the consumer at price points. It doesn't matter what a company wants to claim to the contrary. The amount of performance you can get at a particular price point is what is shifting over time. Typically consumers are seeing solid performance gains for the same price over subsequent generations or simply just on a year-to-year basis as discounts kick in for old hardware. However, the problem with the most recent cards is that we haven't seen this gain from either company and the performance you could get per dollar did not change much if at all. That's perhaps what's causing this illusion that that these more expensive cards are a replacement for some existing product.

 

[PhonakV30]

Why not? the Navi 10 50th edition is a overclocked Navi 10, very similar to the RX590.

RX580 =
Base Clock: 1257MHz
Boost Clock: 1330MHz
RX590 =
Base Clock: 1469MHz
Boost Clock: 1545MHz

 

[Shivansps]

I've really got to disagree with you on this. We already know that there's going to be a bigger Navi, but it doesn't make a lot of sense to claim that Navi 10 (Radeon 5700/XT) is the replacement for the 570/580 when the former has an MSRP of $450/$380 and the later had an MSRP of $220/$170. In what world does a replacement product double the cost of what it replaces?

Im going to stop you right there. In this word, Nvidia did it, and AMD is doing it as well. This was discussed long ago when RTX launched.

You are right about the rest but this happened, you know it.

So you do not have any reasons about it, only your assumptions.

About that name. Isn't RX 690 SKU suggesting top of the line model?

Was the RX590 the top model? Or Vega existed long before?

And what assumptions you are talking about? Is an asumption that Navi 10 is the midrange Navi? when smaller and bigger Navis will exist? That is a fact.
Is assumption that RX590 was not the top model? Is a fact, Vega 64 was the top model.
Is a assumption that X7XX was always mid-range for Radeon? Thats a fact.
Is a assumption that Polaris 10/11 was the midrange cards? Thats a fact.

The only proof that i know of Navi 10 being Vega replacement is price, and that could be very well because they overpriced it like crazy to have big margins. You are assuming it is not because of that. You have any other proof of Navi 10 being Vega replacement?

RX580 =
Base Clock: 1257MHz
Boost Clock: 1330MHz
RX590 =
Base Clock: 1469MHz
Boost Clock: 1545MHz

I dont understand what your point here.

 

[mohit9206]

I've really got to disagree with you on this. We already know that there's going to be a bigger Navi, but it doesn't make a lot of sense to claim that Navi 10 (Radeon 5700/XT) is the replacement for the 570/580 when the former has an MSRP of $450/$380 and the later had an MSRP of $220/$170. In what world does a replacement product double the cost of what it replaces?

Those are two different segments of the market entirely. When AMD eventually does release a smaller Navi chip that targets the ~$200 market segment, no one who was in the market for Polaris will buy a 5700/XT unless they've found themselves suddenly flush with cash. If AMD doesn't release a smaller chip, you wouldn't see a massive uptick in 5700/XT sales from people who would otherwise buy a 570/580. If the world actually worked like that, NVidia wouldn't have missed sales expectations with Turing.

I think what you fail to account for is that the bottom of the market is always being eroded by APUs. Often times the bottom of the product stack is really just recycled mid-tier parts from the previous version. For example, the Radeon 530 is really just a rebranded 240 (Orland Pro) which gradually moved down the stack. Eventually something like that becomes pointless to manufacture (hence the recycling of old parts) because incorporating that level of performance in an APU becomes more cost effective than producing a stand-alone GPU.

Replacement occurs from the perspective of the consumer at price points. It doesn't matter what a company wants to claim to the contrary. The amount of performance you can get at a particular price point is what is shifting over time. Typically consumers are seeing solid performance gains for the same price over subsequent generations or simply just on a year-to-year basis as discounts kick in for old hardware. However, the problem with the most recent cards is that we haven't seen this gain from either company and the performance you could get per dollar did not change much if at all. That's perhaps what's causing this illusion that that these more expensive cards are a replacement for some existing product.

I think its safe to assume that 2070 falls in the same product stack this gen for Nvidia that 970 did back then and x7 series has always been an upper mid range product. So that's near doubling of the price in 2 gens for a mid range product. So since 5700xt is targeting 2070 at similar price/performance it also means that amd too has moved mid range up and almost doubled its price. I don't understand why people would say its a Vega replacement because i would expect a Vega 64 replacement to be about 40 to 60% faster. I don't know how much faster 5700XT is compared to Vega 64 but i doubt its going to be even 40% faster.

 

[Glo.]

And what assumptions you are talking about? Is an asumption that Navi 10 is the midrange Navi? when smaller and bigger Navis will exist? That is a fact.

What are prices of Nvidia midrange GPUs? RTX 2070 and 2080? How much Nvidia wanted for previous gen midrange GPUs: GTX 1070, 1070 Ti and 1080? How much Nvidia charged for even previous gen. midrange GPU: GTX 980?

This is what you guys do not get. It is not mainstream GPU in performance, but Midrange. And priced accordingly to what is on the market.

Vega was Midrange GPU in performance. And priced accordingly. Period.

 

[PhonakV30]

I dont understand what your point here.

You do.RX590 is 200mhz more than RX580.the 50th has 75mhz more than XT version.even RX5700 with OC can reach 2150Mhz on watercooling. RX580 will never able to beat RX590 OC.RX580 chip is 14nm , while RX590 Chip is 12nm