• Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.

|OT| Next-Gen PS5 & XSX |OT| Speculation/Analysis/Leaks Thread

Braldryr

Banned
Apr 9, 2019
197
248
275
Tricked my ass. I'd say he never had any source and made it all up for the clicks. Just like many of his childish baity videos.
Lol I don't know how dumb or naive you have to be to even make a video like that in the first place. Sony came out and said its AMD and his sitting here talking Nvidia nonsense. 😂😂😂
 
  • Like
Reactions: TLZ

SonGoku

Gold Member
Aug 16, 2018
4,754
6,572
595
So, summing up:
  • 105 Watts for faster than Vega 64/RTX 2060 performance (840mV, 1500Mhz target).
  • 120 Watts for faster than RTX 2070 performance (900mV, 1600 Mhz target).
  • 160 Watts for RTX 2080/stock VII performance.

Why doesnt AMD do this?
 
Last edited:
  • Like
Reactions: thelastword and TLZ

SonGoku

Gold Member
Aug 16, 2018
4,754
6,572
595
Because most chips won’t run at that profiles... they need to be safe that all Vega 64 chips will run fine.

That happens in NVidia, Intel, etc... it is called binned lottery chip.
Ah so yields then?
As 7nm matures is it reasonable to expect much better perf/watt next year?
 

SonGoku

Gold Member
Aug 16, 2018
4,754
6,572
595
If AMDs 1.5X (or more) perf perf/watt claim compares to Radeon VII that means a 150W 12TF GPU (maybe even 13TF) should be perfectly doable on consoles APU.
 

ethomaz

is mad because DF didn't do a video on a video of a video of a video on PS5
Mar 19, 2013
35,291
28,349
1,205
38
Brazil
Ah so yields then?
As 7nm matures is it reasonable to expect much better perf/watt next year?
Remember Polaris 14nm? They launched what? RX 480, RX 580 and RX 590... all due the process got better and more and more chips were safe for high frequencies.

That happens with any process.

Overclock exists because that... some chip get high clocks with same or lower vcore but because it is so few they choose to set they product at safe margin... so overclocked get the hands on these few chips by lottery and break records over records.
 
Last edited:
  • Like
Reactions: SonGoku

SonGoku

Gold Member
Aug 16, 2018
4,754
6,572
595
Remember Polaris 14nm? They launched what? RX 480, RX 580 and RX 590... all due the process got better and more and more chips were safe for high frequencies.

That happens with any process.

Overclock exists because that... some chip get high clocks with same or lower vcore but because it is so few they choose to set they product at safe margin... so overclocked get the hands on these few chips by lottery and break records over records.
Damn im feeling very optimistic now, 13TF consoles might not be a pipe dream after all.
 

shark sandwich

tenuously links anime, pedophile and incels
Jun 9, 2012
3,866
1,632
760
If AMDs 1.5X (or more) perf perf/watt claim compares to Radeon VII that means a 150W 12TF GPU (maybe even 13TF) should be perfectly doable on consoles APU.
?? According to my math a 12TF GPU would be 180W.

Radeon VII is 13.4TF at 300 W. X 12/13.4 = 268W for 12 TF. Divide by 1.5 and you’re at 180W
 

SonGoku

Gold Member
Aug 16, 2018
4,754
6,572
595
?? According to my math a 12TF GPU would be 180W.

Radeon VII is 13.4TF at 300 W. X 12/13.4 = 268W for 12 TF. Divide by 1.5 and you’re at 180W
Uh?
Radeon VII 13.8TF would be 200W and that's hitting diminishing returns (300/1.5). By downclocking to hit 12.5TF you can easily shave 50W from 200W.
It can comfortably hit 12TF at 150W, maybe even 12.9TF!

At this moment my base expectation is 12.5TF. All in all even if Navi underwhelms compared to Nvidia it turned out to be a huge MEGATON for consoles.
 
Last edited:

LordOfChaos

Member
Mar 31, 2014
11,305
6,337
915
Pretty sure 1.5x perf/watt was counting the node gains, so compared to Vega and not VII. Power also doesn't scale linearly with performance like this


?? According to my math a 12TF GPU would be 180W.

Radeon VII is 13.4TF at 300 W. X 12/13.4 = 268W for 12 TF. Divide by 1.5 and you’re at 180W
 

SonGoku

Gold Member
Aug 16, 2018
4,754
6,572
595
Pretty sure 1.5x perf/watt was counting the node gains, so compared to Vega and not VII.
Are you sure? I think it was comparing arch improvements, from anandtech
50% better power efficiency than AMD’s current-generation Vega architecture.
1.5x perf/watt improvement over
Vega10: 12TF consoles
Vega20: 13TF consoles
 
Last edited:

ethomaz

is mad because DF didn't do a video on a video of a video of a video on PS5
Mar 19, 2013
35,291
28,349
1,205
38
Brazil
Are you sure? I think it was comparing arch improvements, from anandtech


1.5x perf/watt improvement over
Vega10: 12TF consoles
Vega20: 13TF consoles
The issue is what is current Vega architecture? Both Vega 16nm and Vega 7nm are GCN 1.5, no?
 
Last edited:

ethomaz

is mad because DF didn't do a video on a video of a video of a video on PS5
Mar 19, 2013
35,291
28,349
1,205
38
Brazil
No idea i though Vega20 was "new"
Well it is the same Arch shrink to 7nm.

But I believe they numbers are comparing the Arch alone so they are using Vega 7nm.
 
Last edited:
  • Thoughtful
Reactions: SonGoku

ethomaz

is mad because DF didn't do a video on a video of a video of a video on PS5
Mar 19, 2013
35,291
28,349
1,205
38
Brazil
AMD said that about the results.

1.25x perf/clock.
Testing done by AMD performance labs 5/23/19, showing a geomean of 1.25x per/clock across 30 different games @ 4K Ultra, 4xAA settings. Performance may vary based on use of latest drivers. RX-327

1.5 perf./watt
Testing done by AMD performance labs 5/23/19, using the Division 2 @ 25x14 Ultra settings. Performance may vary based on use of latest drivers. RX-325

The 1.5x perf./watt test they did with one game... they did not say which cards where used too.
 
Last edited:
  • Like
Reactions: llien

LordOfChaos

Member
Mar 31, 2014
11,305
6,337
915
Are you sure? I think it was comparing arch improvements, from anandtech


1.5x perf/watt improvement over
Vega10: 12TF consoles
Vega20: 13TF consoles
But I believe they numbers are comparing the Arch alone so they are using Vega 7nm.

Last but certainly not least of course is overall power efficacy. Thanks to the combination of AMD’s architectural improvements and TSMC’s 7nm process, AMD is promoting a 50% increase in performance per watt for Navi.

Architecture + new node, so the comparison was definitely to 16nm Vega


Edit: They actually had the same question:


Update: Since there have been some questions about what the efficiency number is in reference to – if it's against the 7nm Vega 20 GPU or the 14nm Vega 10 GPU – I went back over AMD's keynote presentation and transcribed it. Here is what Dr. Lisa Su specifically said on the subject (emphasis mine).

And then, when you put that together, both the architecture – the design capability – as well as the process technology, we're seeing 1.5x or higher performance per watt capability on the new Navi products
 
Last edited:

ethomaz

is mad because DF didn't do a video on a video of a video of a video on PS5
Mar 19, 2013
35,291
28,349
1,205
38
Brazil
Last but certainly not least of course is overall power efficacy. Thanks to the combination of AMD’s architectural improvements and TSMC’s 7nm process, AMD is promoting a 50% increase in performance per watt for Navi.

Architecture + new node, so the comparison was definitely to 16nm Vega


Edit: They actually had the same question:


Update: Since there have been some questions about what the efficiency number is in reference to – if it's against the 7nm Vega 20 GPU or the 14nm Vega 10 GPU – I went back over AMD's keynote presentation and transcribed it. Here is what Dr. Lisa Su specifically said on the subject (emphasis mine).
Thanks for the work but that still looks not clear to me :D

Are they using Vega 20 or Vega 10 in the comparison?
 

bitbydeath

Member
Nov 25, 2015
13,045
20,721
1,200
What I find interesting is her mentioning Cerny revolutionising gaming for the next decade, and then later mentioning RDNA being the gaming engine for the next decade. Coincidence?
Or maybe Sony just bought them and they’re obligated to say nice things. :messenger_grinning_squinting:
 
  • LOL
Reactions: Tarkus98

DemonCleaner

Member
Jun 11, 2004
1,968
885
1,685

Update: Thanks to our own Andrei Frumusanu for doing the leg work, we now have a die size estimate based on these and some unpublished photos. Our working guess right now is 275mm2, give or take a few percent. Though that's going to be less precise than what a proper micro-caliper measurement will turn up, so this is very much still just an estimate.

well that's confirmation. AMD just does not give a shit anymore. just go with the flow and sell a 275mm² GPU for 500 bucks. it's not like the customers have much of an alternative.


i'd never thought i'd say this: but now there's just intel to save our sorry asses :poop:
 

Ovech-King

Member
Feb 27, 2017
227
219
285
Anyone care to explain why estimated performance around here seems to expect RTX 2080 levels? With 12tf and such a high memory bandwith to me it would go beyond even the 2080ti ... so what did I missed ?
 

ethomaz

is mad because DF didn't do a video on a video of a video of a video on PS5
Mar 19, 2013
35,291
28,349
1,205
38
Brazil
Anyone care to explain why estimated performance around here seems to expect RTX 2080 levels? With 12tf and such a high memory bandwith to me it would go beyond even the 2080ti ... so what did I missed ?
Vega couldn’t beat even RTX 2080 with 13.7TFs... what made you think Navi with 12TFs will beat RTX 2080ti with 25% improvident in IPC???
 
Last edited:

bitbydeath

Member
Nov 25, 2015
13,045
20,721
1,200
Anyone care to explain why estimated performance around here seems to expect RTX 2080 levels? With 12tf and such a high memory bandwith to me it would go beyond even the 2080ti ... so what did I missed ?
It’s a case of setting expectations to not be disappointed.
 
  • Thoughtful
Reactions: SonGoku

DemonCleaner

Member
Jun 11, 2004
1,968
885
1,685
Vega couldn’t beat even RTX 2080 with 13.7TFs... what made you think Navi with 12TFs will beat RTX 2080ti with 25% improvident in IPC???
well to be fair radeon VII + 25% IPC would come pretty close to a 2080ti in a lot of cases
 

DemonCleaner

Member
Jun 11, 2004
1,968
885
1,685




well that's confirmation. AMD just does not give a shit anymore. just go with the flow and sell a 275mm² GPU for 500 bucks. it's not like the customers have much of an alternative.


i'd never thought i'd say this: but now there's just intel to save our sorry asses :poop:
only hope is, if nvidia really anouces the 2070ti they might go down with the prices $50-100 before launch (compared to the sapphire "leak"). that might be the reason they haven't released more info yet.

so GO NVIDIA!
 

ethomaz

is mad because DF didn't do a video on a video of a video of a video on PS5
Mar 19, 2013
35,291
28,349
1,205
38
Brazil
well to be fair radeon VII + 25% IPC would come pretty close to a 2080ti in a lot of cases
It needs avg. 35% to match RTX 2080ti.

In simple terms you need to match RTX 2080TI:
- Vega @ 18.6TFs
- Navi @ 14.8TFs

12TFs Navi doing beyond RTX 2080TI is just wishful thinking... it will probably around RTX 2080 perf.
 
Last edited:
  • Thoughtful
Reactions: SonGoku

DemonCleaner

Member
Jun 11, 2004
1,968
885
1,685
It needs avg. 35% to match RTX 2080ti.

In simple terms you need to match RTX 2080TI:
- Vega @ 18.6TFs
- Navi @ 14.8TFs

12TFs Navi doing beyond RTX 2080TI is just wishful thinking... it will probably around RTX 2080 perf.
well i think your 35% are wrong it is more like 25% in relevant cases (makes no sense to compare an average that is heavly skewed by outliers to get the technical potential of a card. you only should average technical competent games to do such a comparison). but i really don't want to argue about a few percentage points.

i think your absolutely right in your messaging, because we won't see a 14TF (Radeon VIIs raw perf) Navi in the PS4. that would be whats technical doable. the very best case so to speak. not that what makes sense from a manufacturing standpoint. so i never go by best possible cases, because they never come true. i also think a PS4 that reaches 2080 levels of performance would be an archivement and we should be happy if that came true.
 
  • Like
Reactions: SonGoku
Jul 29, 2013
2,684
5,038
985
US
Asrock showing off RX 5700 models. The 2x8pin pci-e power connectors did end up on Navi 10 retail config. Says in the article they showed 3 models, but I see a 4th with a single 8-pin pci-e connector. I see 2 Phantoms and a Taichi, so maybe that's a different model. Says 150W(180W total board power) and 180W(200W/225W? total board power) respectively.


That's some beast cooling. Reminds me of my 390PCS+. Wouldn't be surprised if it's a little more on the hotter and hungrier side of 150W/180W.
 
Last edited:
  • Like
Reactions: ethomaz

ethomaz

is mad because DF didn't do a video on a video of a video of a video on PS5
Mar 19, 2013
35,291
28,349
1,205
38
Brazil
well i think your 35% are wrong it is more like 25% in relevant cases (makes no sense to compare an average that is heavly skewed by outliers to get the technical potential of a card. you only should average technical competent games to do such a comparison). but i really don't want to argue about a few percentage points.

i think your absolutely right in your messaging, because we won't see a 14TF (Radeon VIIs raw perf) Navi in the PS4. that would be whats technical doable. the very best case so to speak. not that what makes sense from a manufacturing standpoint. so i never go by best possible cases, because they never come true. i also think a PS4 that reaches 2080 levels of performance would be an archivement and we should be happy if that came true.
I was being generous to Radeon VII.

It is 35% only in 1080p... over that is 39%.

 

Gamernyc78

Banned
Jun 11, 2018
5,815
15,322
655
Or maybe Sony just bought them and they’re obligated to say nice things. :messenger_grinning_squinting:
Yeah, they are just paid for escorts saying nice things about everyone they work with because that's good business. Not to say good things aren't coming out of these relationships but again it was the same prior to Ps4/Xbox One release. Im definitely interested to see Sony's role in the new arch Navi.
 

shark sandwich

tenuously links anime, pedophile and incels
Jun 9, 2012
3,866
1,632
760
Uh?
Radeon VII 13.8TF would be 200W and that's hitting diminishing returns (300/1.5). By downclocking to hit 12.5TF you can easily shave 50W from 200W.
It can comfortably hit 12TF at 150W, maybe even 12.9TF!

At this moment my base expectation is 12.5TF. All in all even if Navi underwhelms compared to Nvidia it turned out to be a huge MEGATON for consoles.
Radeon VII is 13.4 TF. And I am taking AMD at their word that Navi is a 50% perf/watt improvement over that. Not a 50% improvement over a Radeon VII that has first been downclocked to achieve better perf/watt first. That makes zero sense.
 
Last edited:

LordOfChaos

Member
Mar 31, 2014
11,305
6,337
915
Thanks for the work but that still looks not clear to me :D

Are they using Vega 20 or Vega 10 in the comparison?

And then, when you put that together, both the architecture – the design capability – as well as the process technology, we're seeing 1.5x or higher performance per watt capability on the new Navi products
-SuBae direct

I don't see that there's any question remaining, it was compared to Vega on 16nm
 
Last edited:

Munki

Member
Apr 2, 2013
570
69
490
What I find interesting is her mentioning Cerny revolutionising gaming for the next decade, and then later mentioning RDNA being the gaming engine for the next decade. Coincidence?
I think that's a bit of a reach. She mentions Cerny wanting to revolutionize gaming for the next decade, not that his input was revolutionary. In the end only 42 seconds of the Keynote was devoted to their partnership with Sony, compare that to a spokesperson from MS on stage with Lisa to speak about their relationship/involvement is a much more telling (totalling roughly 5 mins).

And let's not forget this glorious moment.

 
Last edited:

thelastword

Member
Apr 7, 2006
11,447
10,610
1,850
A year of hyping complete fanfiction "leaks" about Ryzen 3000 is what I gather

I vote they go on the banned sources list, why even bother with him now
No need for that. Adored TV does very good videos, well researched ones, I think that's his Forte.... However, this whole I'm a leaker thing is like the new rage, many pastebin accounts later, many uncles and janitors working at major corporations and gaming studios later, this is what we get....An influx of horse manure sources, all searching for a bit of likes and boosts........

I don't think Adored Tv's Ryzen leak is way off base though, things can change during development and there is no way AMD should give their products away for peanuts when they outperform Intel at much lower prices already..... I don't think his Navi source or information is as great, but it's still in the ball park save the prices.......

He jumped the gun on gcn and Navi having major issues, but quite a few sources and journalists were ready and waiting to pile on this anyway, they wanted Navi to have issues.... They were so convinced Navi was gcn and could not wait to air their disappointment, when in truth no one knew, except that amd told us it was based on some newness, but people scoffed regardless, believe them over AMD, "My sources dawg".......
 

ethomaz

is mad because DF didn't do a video on a video of a video of a video on PS5
Mar 19, 2013
35,291
28,349
1,205
38
Brazil
And then, when you put that together, both the architecture – the design capability – as well as the process technology, we're seeing 1.5x or higher performance per watt capability on the new Navi products
-SuBae direct

I don't see that there's any question remaining, it was compared to Vega on 16nm
Well that is not I call clear :D

But if true makes the claim no that impressive because just the shrink should give you near that 1.5x performance per watt increase... people were talking about Maxwell moment (Maxwell never relied on die shirking to increase the perf. per watt).

Vega 64 to Radeon VII has a 40% perf. per watt increase in Uk for example.

 
Last edited:

shark sandwich

tenuously links anime, pedophile and incels
Jun 9, 2012
3,866
1,632
760
And then, when you put that together, both the architecture – the design capability – as well as the process technology, we're seeing 1.5x or higher performance per watt capability on the new Navi products
-SuBae direct

I don't see that there's any question remaining, it was compared to Vega on 16nm
It’s marketing. I’m assuming they picked whichever one sounds the most impressive. Which is why all these posts dissecting every word and interpreting them in the most optimistic possible way is kind of ridiculous.
 

LordOfChaos

Member
Mar 31, 2014
11,305
6,337
915
I think that's a bit of a reach. She mentions Cerny wanting to revolutionize gaming for the next decade, not that his input was revolutionary. In the end only 42 seconds of the Keynote was devoted to their partnership with Sony, compare that to a spokesperson from MS on stage with Lisa to speak about their relationship/involvement is a much more telling (totalling roughly 5 mins).
Yeah I had to go back and listen to that part again because of what people were saying.

It wasn't "Mark Cerny was a key revolutionary for Navi", it was "We helped Mark Cerny with their semicustom effort for his vision for the next decade of console gaming", or something more like that.

Still don't know if the ray tracing was a custom Sony addition or inherent to Navi, but this non-answer almost seems like an answer, "that custom silicon"



It’s marketing. I’m assuming they picked whichever one sounds the most impressive. Which is why all these posts dissecting every word and interpreting them in the most optimistic possible way is kind of ridiculous.
The architecture alone achieving 1.5x perf/watt would sound a lot more impressive, and this is to Computex, not normie central, where people know the difference. I think it's a stretch to read it any other way than being a combination of both.
 
Last edited:
  • Like
Reactions: SonGoku

TLZ

Member
Oct 20, 2015
10,318
14,683
930
Sydney
I think that's a bit of a reach. She mentions Cerny wanting to revolutionize gaming for the next decade, not that his input was revolutionary. In the end only 42 seconds off the Keynote were devoted to their partnership with Sony, compare that to a spokesperson from MS on stage with Lisa to speak about their relationship/involvement is a much more telling.
Nah. Not a reach at all. She mentions Navi and RDNA and Cerny wanting to revolutionize gaming for the next decade in the same breath. Not a coincidence.

What she had with MS there imo was different. It looked like more a big business agreement between them, like powering MS hardware; Azure's servers, Xbox consoles, Surfaces etc, and also probably working closer for better W10 drivers. This is all also massive and a huge deal, on a different scale.

In the end, AMD is the big winner here making all these great business deals with everyone, supplying their needs with good hardware at good cost.
 

ethomaz

is mad because DF didn't do a video on a video of a video of a video on PS5
Mar 19, 2013
35,291
28,349
1,205
38
Brazil
Yeah I had to go back and listen to that part again because of what people were saying.

It wasn't "Mark Cerny was a key revolutionary for Navi", it was "We helped Mark Cerny with their semicustom effort for his vision for the next decade of console gaming", or something more like that.

Still don't know if the ray tracing was a custom Sony addition or inherent to Navi, but this non-answer almost seems like an answer, "that custom silicon"

AMD already said Navi PC won't have ray-tracking dedicated hardware.

They said ray-tracing hardware only in "others places".

The architecture alone achieving 1.5x perf/watt would sound a lot more impressive. I think it's a stretch to read it any other way than being a combination of both.
That is what Maxwell did in the past... if you look at Radeon VII bench you will see a 40% perf. per watt increase due the die shrink... so maybe there is a 10% due the Architecture in Navi that doesn't look that impressive.
 
Last edited:

xool

Member
May 29, 2018
1,780
2,154
415
Quoted TSMC figures for energy efficiency are ~35% and ~40% reduction (16 to 10nm and 10 to 7nm) - cumulatively that's ~55% power reduction just from the die shrink.

If the 50% reduction quoted by AMD is from 14/16nm then their figure would look kind of meh, as they should be getting that ballpark anyway without changing anything ..

My antenna are twitching and wondering if this is a "new architecture" in the same way Skyrim had a "completely new game engine" .. ie barely at all.

[edit - I just realized a ton of people just said the exact same thing]
 
Last edited:

ethomaz

is mad because DF didn't do a video on a video of a video of a video on PS5
Mar 19, 2013
35,291
28,349
1,205
38
Brazil
Asrock showing off RX 5700 models. The 2x8pin pci-e power connectors did end up on Navi 10 retail config. Says in the article they showed 3 models, but I see a 4th with a single 8-pin pci-e connector. I see 2 Phantoms and a Taichi, so maybe that's a different model. Says 150W(180W total board power) and 180W(200W/225W? total board power) respectively.


That's some beast cooling. Reminds me of my 390PCS+. Wouldn't be surprised if it's a little more on the hotter and hungrier side of 150W/180W.
There are pics from cards coming from Computex... all of them shows two SKUs:

225W TBP
180W TBP

Basically the Sapphire leasks.
 
  • Like
Reactions: shark sandwich

ethomaz

is mad because DF didn't do a video on a video of a video of a video on PS5
Mar 19, 2013
35,291
28,349
1,205
38
Brazil
Quoted TSMC figures for energy efficiency are ~35% and ~40% reduction (16 to 10nm and 10 to 7nm) - cumulatively that's ~55% power reduction just from the die shrink.

If the 50% reduction quoted by AMD is from 14/16nm then their figure would look kind of meh, as they should be getting that ballpark anyway without changing anything ..

My antenna are twitching and wondering if this is a "new architecture" in the same way Skyrim had a "completely new game engine" .. ie barely at all.
TSMC data is in theory... AMD reached around 40% increase avg. in real benchs with Vega 20 going from 16nm to 7nm.
There is obvious some Architecture improvement with Navi but it is way lower than the marketing tries to show.... 10% improvement in perf./watt with a new Arch is good it is just not Maxwell level.
 
Last edited: