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

Page 98 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Mopetar

Diamond Member
Jan 31, 2011
4,525
458
126
The thing with the large GPU market is with 7nm cost of development, it needs a piggy back market of some sort to make the cost of developing the GPU worth it.

This means adding back the compute. From what I have seen in terms of luxmarks and other programs that test compute, Navi seems to be 2 steps forward for gaming architecture but one or two steps back as far as compute.

This makes sense because Navi seems to be a pure gaming architecture with the consoles in mind and now has IPC similar to Nvidia turing which is very impressive. It seems to be Maxwell + Async in the behavior of it's shaders which makes it no surprise that it IPC is similar to Turing now. However it appears to have taken a step back as far as compute ability goes

https://www.hardwareluxx.de/index.p...0-und-radeon-rx-5700-xt-im-test.html?start=11

To make a A 400mm, 96 rop, 4096 shader navi that pure gaming is possible but it wouldn't have the compute to be competitive with Nvidia next gen 7nm products and would mostly be more inline with Nvidia's volta or less if we look beyond pure tflops(double precision). This would reduce it's TAM in a very fast growing market which is less desirable for AMD in the long run.

I think AMD's next enthusiast card will have the compute added back on and AMD will try to replicate what Nvidia does where they have compute specific cards on one side of the fence and gaming cards one one side of the fence. I didn't think this was possible before with AMD budget, but with the shifting of labor to China for GPU development, some reverse engineering(Navi performs too similarly in regards to IPC in the same games as maxwell and Turing),the piggy backing of R and D development from the PS5 and Xbox next, it looks like AMD managed to do it.

In terms of beyond, AMD will need to look not at the present but to the future for the design of its next gen card. Looking at Nvidia's financials, half of their revenue is now provided by professional and business markets which means Nvidia's professional cards are going to be strong.

I could see coming from Nvidia for their first compute product to succeed Volta being a 600mm-700mm2, 6144 shader card clocks at a conservative(for 7nm standards)1800mhz. This translates into a 22.1tflops card. This is probably Nvidia's target since GV100 already does 14.7 tflops, so roughly a 50% increase is expected with a new node. This will be followed by a 104 series with 0.666 of the specs(This pattern has been repeated since maxwell). So in this case 4096 shaders with a 102 series being the same 6144 shaders but higher clocks. In higher clocks I am guessing 2.2 ghz for the 104 and 2.1ghz for the gv102 since these clocks can be achieved on air/water today on turing and typically new nodes improve on these clocks and become the new default clock. Assuming no changes in architecture and simply a moving of Turing to a new node, AMD will need something with more than 4096 shaders to compete with Nvidia's next gen.

So what specs will AMD need to compete with a 18tflop card(104) or a 25.8tflop card(100/102) using the above specs.

Assuming the clocks don't improve(and the chance of it improving is not good since the wider specs will eat into the clocks, along with compute and the possibility of double precision), this would mean AMD would need something with 5120 shaders to be competitive with Nvidia and perhaps beat the 104.

5120*2*1.8ghz = 18.4 tflops. This becomes tricky for AMD because such a card is a mammoth for AMD in terms of die size consider the space requirements of 2560 shaders at 251mm2. Add in the likely compute ability that needs to come in and AMD is looking at a 500mm2 plus card. With Nvidia, we are likely going to see a growth of die sizes compared to pascal because of how big turing cards are today, I am estimating ga104 being 350mm to 400mm, ga102 being 525 to 600mm and GA100 being600mm- 700mm2.

This playing around with die sizes shows that although RDNA was a good step in the right direction, AMD competitiveness is coming mostly as a result of the move to a new node vs Nvidia. AMD problem with RDNA is scalability and power consumption and this is why Nvidia is not in panic mode.

As is currently, power consumption reveals problems with scalability.

Nvidia's next gen ga106, should be 2036 shader(looking at trends with maxwell, pascal and turing). At 2.25ghz(these cards are clocked a tad higher because of small die and less leakage), we are looking at a 9.16 tflop card. Things look good when the 5700xt produces 9tflops so it initially looks like a competitive race. That is until you look at power consumption. There's a good chance GA106 will consume 120 watts, like the GTX 960, 1060 and 1660 ti before it. It also shows the scalability of the architecture. If 9.1tflops of power can produced for 120watts, there is the possibility of 18tflops at 200-225watts and 25.8tflops at 300watts.

With RDNA there are concerns when 9 tflops already takes 225watts. Where is the scalability to get 18 or 25.8tflops? The 7.6tflops at 180watts isn't much better from the 5700.

If Navi was a 150watt card at 9tflops(don't use a youtube video where someone undervolts, can't guarantee stability across hundreds of thousands of samples across all uses cases), AMD would be in a much more competitive position for scaling but at 225 watts, the outlook does not look good for scaling upward.
I wonder if AMD will use a chiplet approach for compute GPUs since they seem to have made Navi (and RDNA) for gaming.

Interviews with their engineers said that multiple chillers were fine for compute workloads, but it didn’t work well for gaming.

Sure the professional market can justify the high cost of massive compute dies, but the other side of chiplets is better yield and more product to sell.
 

maddie

Platinum Member
Jul 18, 2010
2,734
688
136
I wonder if AMD will use a chiplet approach for compute GPUs since they seem to have made Navi (and RDNA) for gaming.

Interviews with their engineers said that multiple chillers were fine for compute workloads, but it didn’t work well for gaming.

Sure the professional market can justify the high cost of massive compute dies, but the other side of chiplets is better yield and more product to sell.
It's not just better yields, which is also great, but you can circumvent the reticle limit barrier to total silicon for single cards. Bigger and cheaper.
 

krawcmac

Junior Member
Nov 21, 2014
3
1
66
I am surprised no one mentioned latest development regarding Navi overclocking. Look like the card has really nice overclocking/undervolting potential. The overclocking process requires SoftPower play tables tweaking and was discovered by Igor form TomsHardware Germany.

 

Glo.

Diamond Member
Apr 25, 2015
3,110
310
136
Well, maybe RX 5700 XT has power limits.

But 5750 XT, and 5750 might not have them. Who knows, maybe even 5800 will reuse the same chip?

;)
 

Det0x

Senior member
Sep 11, 2014
308
24
116
So apparently the RX 5700XT does ~2150MHz average on 250W by tuning the power tables and placing the vcore at 1.25V (up from 1.2V default):


In the meantime, with undervolt to 0.966V the card consumes slightly above 130W while clocking 40MHz lower.
 

Hitman928

Golden Member
Apr 15, 2012
1,892
334
136
I'd love to see undervolt vs undervolt performance and perf/w with the latest from both camps.
 

Glo.

Diamond Member
Apr 25, 2015
3,110
310
136
In the meantime, with undervolt to 0.966V the card consumes slightly above 130W while clocking 40MHz lower.
Considering the nature of power and core clock control in RDNA GPUs, AMD could've put XT on 6 pin connector only, and call it a day.

The whole situation for me pretty much confirms, that we may see 5750/XT with higher clock speeds, and faster GDDR6 memory, to compete with RTX 2080 and possibly even with 2080 Super in performance.
 

amenx

Platinum Member
Dec 17, 2004
2,446
73
126
I am surprised no one mentioned latest development regarding Navi overclocking. Look like the card has really nice overclocking/undervolting potential. The overclocking process requires SoftPower play tables tweaking and was discovered by Igor form TomsHardware Germany.

So apparently the RX 5700XT does ~2150MHz average on 250W by tuning the power tables and placing the vcore at 1.25V (up from 1.2V default):


In the meantime, with undervolt to 0.966V the card consumes slightly above 130W while clocking 40MHz lower.
Come on guys, why not say it? Navi OC's great............................ when water cooled. To (only slightly) beat the 2070S (also OC'd, but on air), the 5700XT in those vids is using a 125eu EK water block ( + lot more if you dont have the WC kit) which brings its price quite a bit higher than the 2070S. Not to mention the added complexity and hassle. A lot more practical and sensible to just buy the 2070S for cheaper.
 

Hitman928

Golden Member
Apr 15, 2012
1,892
334
136
Come on guys, why not say it? Navi OC's great............................ when water cooled. To (only slightly) beat the 2070S (also OC'd, but on air), the 5700XT in those vids is using a 125eu EK water block ( + lot more if you dont have the WC kit) which brings its price quite a bit higher than the 2070S. Not to mention the added complexity and hassle. A lot more practical and sensible to just buy the 2070S for cheaper.
If the 250W figure while overclocked is correct, I'd imagine a good AIB cooler would also work just fine. . .
 
Feb 6, 2011
1,862
255
136
Come on guys, why not say it? Navi OC's great............................ when water cooled. To (only slightly) beat the 2070S (also OC'd, but on air), the 5700XT in those vids is using a 125eu EK water block ( + lot more if you dont have the WC kit) which brings its price quite a bit higher than the 2070S. Not to mention the added complexity and hassle. A lot more practical and sensible to just buy the 2070S for cheaper.
Why do you ignore the power consumption and focus on the cooling method?
 

Stuka87

Diamond Member
Dec 10, 2010
4,260
191
126
Come on guys, why not say it? Navi OC's great............................ when water cooled. To (only slightly) beat the 2070S (also OC'd, but on air), the 5700XT in those vids is using a 125eu EK water block ( + lot more if you dont have the WC kit) which brings its price quite a bit higher than the 2070S. Not to mention the added complexity and hassle. A lot more practical and sensible to just buy the 2070S for cheaper.
There is zero reason an air cooler would not work for a 250W card. There have been plenty of them. Including 2080Ti and Vega cards, all of which have air coolers.

Water cooling is mostly used to keep noise down.
 
Apr 24, 2019
121
129
76
Come on guys, why not say it? Navi OC's great............................ when water cooled. To (only slightly) beat the 2070S (also OC'd, but on air), the 5700XT in those vids is using a 125eu EK water block ( + lot more if you dont have the WC kit) which brings its price quite a bit higher than the 2070S. Not to mention the added complexity and hassle. A lot more practical and sensible to just buy the 2070S for cheaper.
1) 2070S has more mature drivers.
2) 5700XT 50th edition trades blows with 2070S, on a blower.
3) 5700XT can o/c to 2.1+GHz without water. There are numerous examples.
4) We have no good comparisons on max fan-cooled 5700XT vs 2070S because... there aren't any retail fan-cooled 5700XT.

It's not "a lot more practical and sensible to just buy the 2070S for cheaper" because it's not necessary to put the 5700XT under water to overclock it. Further, with less mature drivers, Navi hardware is hampered and has a much higher potential for better performance. Also, once Navi drivers mature, it's entirely possible that stock AIB 5700XTs will be on-par with the stock 2070S for a 10-20% discount.

I'd say that comparing Navi to Super is difficult because we have a very incomplete assessment of how Navi truly performs. Let's wait til September before claiming a winner.
 
Jan 17, 2013
216
20
91
1) 2070S has more mature drivers.

It's not "a lot more practical and sensible to just buy the 2070S for cheaper" because it's not necessary to put the 5700XT under water to overclock it. Further, with less mature drivers, Navi hardware is hampered and has a much higher potential for better performance. Also, once Navi drivers mature, it's entirely possible that stock AIB 5700XTs will be on-par with the stock 2070S for a 10-20% discount.

I'd say that comparing Navi to Super is difficult because we have a very incomplete assessment of how Navi truly performs. Let's wait til September before claiming a winner.
Ah, the old WaIt FoR tHe DrIvErS argument, where have we heard that before :rolleyes:.

...Conveniently ignoring that Turing is functionally introducing gen 1.0 of Real-time Raytracing in games, meaning that drivers and other software support for that feature are anything but mature now, which will implicitly leave a lot of headroom for future performance improvements.
 

amenx

Platinum Member
Dec 17, 2004
2,446
73
126
1) 2070S has more mature drivers.
2) 5700XT 50th edition trades blows with 2070S, on a blower.
3) 5700XT can o/c to 2.1+GHz without water. There are numerous examples.
4) We have no good comparisons on max fan-cooled 5700XT vs 2070S because... there aren't any retail fan-cooled 5700XT.

It's not "a lot more practical and sensible to just buy the 2070S for cheaper" because it's not necessary to put the 5700XT under water to overclock it. Further, with less mature drivers, Navi hardware is hampered and has a much higher potential for better performance. Also, once Navi drivers mature, it's entirely possible that stock AIB 5700XTs will be on-par with the stock 2070S for a 10-20% discount.

I'd say that comparing Navi to Super is difficult because we have a very incomplete assessment of how Navi truly performs. Let's wait til September before claiming a winner.
This is what I was objecting to:
comment no.1: I am surprised no one mentioned latest development regarding Navi overclocking. Look like the card has really nice overclocking/undervolting potential. The overclocking process requires SoftPower play tables tweaking and was discovered by Igor form TomsHardware Germany.

comment no.2: So apparently the RX 5700XT does ~2150MHz average on 250W by tuning the power tables and placing the vcore at 1.25V (up from 1.2V default)

^ For proper context, it needs to be mentioned that the card was WC'd at considerable expense, then compared to stock air cooling of 2070S OC.
 
Apr 24, 2019
121
129
76
Ah, the old WaIt FoR tHe DrIvErS argument, where have we heard that before :rolleyes:.

...Conveniently ignoring that Turing is functionally introducing gen 1.0 of Real-time Raytracing in games, meaning that drivers and other software support for that feature are anything but mature now, which will implicitly leave a lot of headroom for future performance improvements.
If you'd like, perhaps just wait for an appropriate cooling solution. AMD reference design is behind Nvidia reference design, and third-party cards close the gap. But the drivers will continue to make little gains.

Agree that for RT, Nvidia obviously has a head up. But if you're comparing a $499 card (2070S) to a $399 card (5700XT) perhaps RT isn't worth it. If it is, then great.

I guess my point is that we don't have all the information. We have a subpar cooling solution on 5700XT versus a better one on 2070S. I'll be anxiously awaiting the 3rd party AIBs for both, and hopefully we get some good comparisons.

I just wouldn't throw the towel in yet...
 

Stuka87

Diamond Member
Dec 10, 2010
4,260
191
126
If you'd like, perhaps just wait for an appropriate cooling solution. AMD reference design is behind Nvidia reference design, and third-party cards close the gap. But the drivers will continue to make little gains.

Agree that for RT, Nvidia obviously has a head up. But if you're comparing a $499 card (2070S) to a $399 card (5700XT) perhaps RT isn't worth it. If it is, then great.

I guess my point is that we don't have all the information. We have a subpar cooling solution on 5700XT versus a better one on 2070S. I'll be anxiously awaiting the 3rd party AIBs for both, and hopefully we get some good comparisons.

I just wouldn't throw the towel in yet...
Its behind by choice though. The Radeon VII had a very good reference cooler. AMD knows they will only sell a handful of the reference 5700's. But, I feel it would be better for them to put some money into them just so initial reviews look as a good as possible.
 

Dribble

Golden Member
Aug 9, 2005
1,709
131
126
Its behind by choice though. The Radeon VII had a very good reference cooler. AMD knows they will only sell a handful of the reference 5700's. But, I feel it would be better for them to put some money into them just so initial reviews look as a good as possible.
If it's choice then it's a stupid choice. So what you want is glowing reviews then everyone goes out and buys the cards and they say nice stuff about them to their friends on forums, and you build some sales momentum.

Instead AMD got ok reviews, and no one is buying the cards because of the cooling. So all that hype and build up and instead of building on it they kill their sales. Sure eventually cards will come out with proper cooling but those initial reviews will still be the ones people see, and that sales momentum has to build up all over again.
 
Last edited:

krawcmac

Junior Member
Nov 21, 2014
3
1
66
^ For proper context, it needs to be mentioned that the card was WC'd at considerable expense, then compared to stock air cooling of 2070S OC.[/QUOTE]

Well, WC is nice because it is quiet. But it looks like you do not need water cooling to overclock 5700xt. Take a look. All you need is an aftermarket cooling solution, nothing fancy.


So as someone mentioned AIB cards will trade blows with your 2070 super for 10-25% less.
 

railven

Diamond Member
Mar 25, 2010
6,537
239
126
If it's choice then it's a stupid choice. So what you want is glowing reviews then everyone goes out and buys the cards and they say nice stuff about them to their friends on forums, and you build some sales momentum.

Instead AMD got ok reviews, and no one is buying the cards because of the cooling. So all that hype and build up and instead of building on it they kill their sales. Sure eventually cards will come out with proper cooling but those initial reviews will still be the ones people see, and that sales momentum has to build up all over again.
Hey, at least it's $100 less! Well, the version reviewers are advising not to buy - which by someone's logic is intentional - but the versions everyone will likely buy most certainly won't be $100 less.

AMD needs new people for their optics. This is the second time I'm being led to believe "AMD is just pretending to be stupid."

Rats, Jebaited again!
 

railven

Diamond Member
Mar 25, 2010
6,537
239
126
^ For proper context, it needs to be mentioned that the card was WC'd at considerable expense, then compared to stock air cooling of 2070S OC.
Well, WC is nice because it is quiet. But it looks like you do not need water cooling to overclock 5700xt. Take a look. All you need is an aftermarket cooling solution, nothing fancy.


So as someone mentioned AIB cards will trade blows with your 2070 super for 10-25% less.
And this is exactly what I mean. "Radeon 7 for $400."*

"If you already have a proper cooler, otherwise, +additional cost of better cooler!
 

JDG1980

Golden Member
Jul 18, 2013
1,657
174
136
AMD's graphics division really, really needs to stop overvolting all of their cards by default. All the expertise they've built up in the CPU division about maximizing the efficiency of their products has been completely ignored on the GPU side.
 

Mopetar

Diamond Member
Jan 31, 2011
4,525
458
126
AMD's graphics division really, really needs to stop overvolting all of their cards by default. All the expertise they've built up in the CPU division about maximizing the efficiency of their products has been completely ignored on the GPU side.
I don’t think they’ve done that to the same degree as with Polaris or Vega. The 5700 seems to have good settings, mainly just a poor cooling solution in the reference card due to using a cheap blower again.

We’ve seen people able to push these cards quite a bit harder, which want possible in the past because the cards were already at their limits.

So just wait for the third party cards if you want better cooling and noise levels and aren’t willing to jump through the hoops of putting on a better cooler yourself.
 


ASK THE COMMUNITY

TRENDING THREADS