• 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.

Next-Gen PS5 & XSX |OT| Console tEch threaD

Status
Not open for further replies.

bitbydeath

Member
Yes, Bob in street will put high performing SOC at 7nm for manufacturing and validation to TSMC and AMD will design this mythical chip without anyone know.

Chips, especially complex monolithic dies, have ES (engineering samples) well before system is out. So yea, 1.5yr is nothing out of the ordinary. There are 2 more revisions after QS and they need to lock the chip half a year before going retail.

You didn’t answer my question.
Everything your backing your theories on are PS4 and PS4 Pro being possibly detected.

How far in advance were they detected?
 

LordOfChaos

Member
Yeah but leaks and physical hardware detection are two very different things as the former implies what they’re aiming for.

Ah, gochya


Yes, Bob in street will put high performing SOC at 7nm for manufacturing and validation to TSMC and AMD will design this mythical chip without anyone know.

Chips, especially complex monolithic dies, have ES (engineering samples) well before system is out. So yea, 1.5yr is nothing out of the ordinary. There are 2 more revisions after QS and they need to lock the chip half a year before going retail.

Or they altered the driver strings and ran a benchmark on existing hardware, it often is what happens on "leaked" geekbench scores, which they later figure out and mark as fake. Firestrike runs on Windows, Android, and iOS, how would a PS5 SoC be running it? The scores are very near Vegas.
 
Last edited:

R600

Banned
You didn’t answer my question.
Everything your backing your theories on are PS4 and PS4 Pro being possibly detected.

How far in advance were they detected?
Tbh dont remember, APISAK did find Thebe in 3DMark by 2013, but I am not sure such a scene was this rabid back in the day :) Back then we had sweetvar26 here who leaked everything from his friend who worked on Thebe chip (PS4) and Xbox One. Later they got sued IIRC.


Or they altered the driver strings and ran a benchmark on existing hardware, it often is what happens on "leaked" geekbench scores, which they later figure out and mark as fake. Firestrike runs on Windows, Android, and iOS, how would a PS5 SoC be running it? The scores are very near Vegas.
Benchmarking in 3DMark is a bit weird, but as I posted they benchmarked PS4 and Pro back in the day as well. Codenames where clearly visible.
 
Last edited:

LordOfChaos

Member
Benchmarking in 3DMark is a bit weird, but as I posted they benchmarked PS4 and Pro back in the day as well. Codenames where clearly visible.

Again, what verifies that it actually happened? Did they ever share a picture of the setup they were running? How did they run it on an OS it doesn't support, or are they saying they got it running Windows or Android well before everyone else had other operating systems on PS4?

Online benchmark results are notoriously fakeable with a few string modifications, happens all the time, it's awesome if this is real but just approach all of this with a healthy dose of skepticism until good evidence is provided. Primate Labs marks benchmarks fake after the fact all the time, plenty of time between them checking in on it and it fooling enough people


 
Last edited:

R600

Banned
Again, what verifies that it actually happened? Did they ever share a picture of the setup they were running? How did they run it on an OS it doesn't support, or are they saying they got it running Windows or Android well before everyone else had other operating systems on PS4?

Online benchmark results are notoriously fakeable with a few string modifications, happens all the time, it's awesome if this is real but just approach all of this with a healthy dose of skepticism until good evidence is provided. Primate Labs marks benchmarks fake after the fact all the time, plenty of time between them checking in on it and it fooling enough people


No one verifies, doubt Sony/AMD would officialy reply to mail if we asked. This is why its a leak coming from guy who is decoding and pulling benches from AMD chips for over decade, so I guess he would know a thing or two about fake benches.

We also have precedent of PS4 and Pro being in database with, what appears, very likely benchmark scores. Its not like bench itself says "this is PS5". Someone had to dig up PCI IDs database and find which codename could be for which product etc.

It might all be ruse, but too many things are tying up IMO.
 

R600

Banned
It's above 2070, not below.
There are alot od Vega56 scores higher then 20K, but its still not close to 2070 in regular gaming, so I assume this will be again AMD scoring more on syntetic benchmarks compared to Nvidia but losing out in games. For example, 1080 scores more then 2070 yet it regulary trails it in games by 15-20%

We know Navi XT is ~3-5% faster then 2070 so we will 100% know on 7th of July, when someone can make benchmark of Navi card with Ryzen2 and compare it.
 
Last edited:

pawel86ck

Banned
What in the name... we did it! Team 16.8 beibi!


They are meat for the cannons in the Great floating point operations per second Wars... Dust in the wind, join the winning side my friend 😊
Wars like that are fun for sure😉, but there is no way we will hear any real details soon (15 months before launch). The problem is current estimations are based on current knowledge and we dont know to what extend Navi will be modified for Xbox Scarlett and PS5. We dont even know how REAL TDP will look like for Navi cards (Igor calvulations suggest better TDP than people expect) while people here already concluded how fast xbox scarlett and PS5 will be.

Just before PS5 was announced people were laughing when I have suggested PS5 or next Xbox may use HW RW, in fact some people were even attacking me for suggestions like that saying only a fool will expect HW RT on consoles. The same with regards to Navi, I was expecting better gaming performance compared to Vega while people were saying it will be a disaster and even worse gaming performance compared to Vega was expected based of their current knowledge (leaks). I think Sony and MS can still surprise people with PS5 and Scarlett specification, and especially when both companies want to deliver a premium product.

PaNaMa, based on Mark Cerny quote I also thought PS4P GPU is indeed dual GPU, and if you run PS4 pro games both chips are activated.

We doubled the GPU size by essentially placing it next to a mirrored version of itself. That gives us an extremely clean way to support the existing 700 [PS4] titles. We just turn off half the GPU and run it at something quite close to the original GPU." stick a Pro game, in and both chips are activated, doubling the GPU’s power while the CPU runs at the same rate - something Cerny says is important to ensure compatibility with older games.

4chan leak suggest dual GPU as well in PS5 at 14TF, so one chip must be 7TF. I wonder how much more power efficient RX5700 would be at 7TF, maybe it's within reasonable limits.
 

Gamernyc78

Banned
Can someone who speaks English tell me what's going on here today?

Basically arm chair analysts throwing numbers around with a peppering of fanboyism. Depending on what console they are a fan of, the competitors guesstimated tera flops go down lol 😭😭😭 Trolling at its best lol 😘 e.g. If the poster is a Microsoft fanatic the guesstimate is 11/12tf Anaconda versus 9/10 TF PS5 and vice versa

Got to love some of them 😁

BTW not to knock some posters who are actually neutrally taking educated guesses.
 
Last edited:
Basically arm chair analysts throwing numbers around with a peppering of fanboyism. Depending on what console they are a fan of the competitors guesstimated tera flops go down lol 😭😭😭 Trolling at its best lol 😘 e.g. If the poster is a Microsoft fanatic the guesstimate is 11/12tf Anaconda versus 9/10 TF PS5 and vice versa

Got to love some of them 😁

BTW not to knock some posters who are actually neutrally taking educated guesses.
I'm thinking both consoles will range from 8-10 without a single inkling of a difference in visuals between the systems.

Considering that the RTX Titan @ $2,499 sits around 16.3 TFLOPS with 24GB of GDDR6, expecting more than previously stated would be a fool's gambit. Also consider the 2080 TI @ $1,300 sits at 13.4 TFLOPS with 11GB GDDR6. 8-10 TFLOPS seems fairly reasonable for a console.
 
Last edited:

R600

Banned
What's with all the talk about a dual soc like the PS4 Pro? where is this coming from?
From people who didnt get the memo that weaker Navi (7.9TF version) sucks entire console worth of watts - 180.

All that without CPU, RT and with another 8GB missing. So missing another 50ish watts.

What does 20,000 mean in Laymens terms?
We dont know 100%, as this score includes entire system performance, but in AMD terms something between Vega56 and Vega64 in syntetic benchmark.

We will have to wait for Zen2 and Navi to be benchmarked by 7th of July, but should be be between 8 and 9TF on GPU size (as AMD always has higher synthetic benchmarks then Nvidia, but trails in gaming).
 
Last edited:

Gamernyc78

Banned
I'm thinking both consoles will range from 9-10.5 without a single incling of a difference in visuals between the systems.

Considering that the RTX Titan @ $2,499 sits around 16.3 TFLOPS with 24GB of GDDR6, expecting more than previously stated would be a fool's gambit.

Yeah I think thy will be super close in tflops to the point in being negligible. The real difference will come either in the RT implementation and how they implement the SSD solution. I think both will be hardware RT I just think thy will be slightly different.
 

xool

Member
PS4 Pro and Xbox One X APUs.

No Dual-GPU nonsense.

jpg


This is why people say "dual GPU" :
 
Last edited:
Yeah I think thy will be super close in tflops to the point in being negligible. The real difference will come either in the RT implementation and how they implement the SSD solution. I think both will be hardware RT I just think thy will be slightly different.
I edited my comment pretty heavily. Just a heads up. It's not much different however.
 
Last edited:

CrustyBritches

Gold Member
People seem confused. Let's clarify:

1. The guy(Apisak) who found PS4, PS4 Pro, and SuborZ early benchmarks has been tracking a chip called "Gonzalo" that is designated "G" for gaming console APU.

2. Digital Foundry believes that it's plausible that it's PS5

3. Last night the guy(Apisak) who found those chips in 3dMark benchmarks found the score for Gonzalo, the plausible PS5 chip.

4. It scores in the range of a Ryzen 1700 paired with Vega 56

5. If Ryzen 3000 8-core @ 3.2GHz is faster than the Ryzen 1700, then an even weaker card could be used to get that overall score.
fscomp.jpg


tldr: PS5 might be in between a Vega 56 and a Vega 64 in performance.
 

vpance

Member
Seems like the only way 10TF+ is likely is if Sony goes 7nm EUV. With 6nm and 5nm right around the corner going with a tiny 7nm chip would be a mistake.
 

R600

Banned
What does a benchmark mean?
CB posted good wrap in post above.

In layman terms, it will probably be betwen Vega56 and Vega64. If we know Navi XT is 1.14x Vega64, we could be looking at chip right between Navi10 (7.9TF) and Navi10 XT (9.75TF)

Is there a reason why Scarlett is not connected to this chip?
Scarlett chip is codename "Arden", while Gonzalo's ID is referencing Ariel, which fits under Sony's custom chips in PCI ID base.
 
Last edited:

Munki

Member
CB posted good wrap in post above.

In layman terms, it will probably be betwen Vega56 and Vega64. If we know Navi XT is 1.14x Vega64, we could be looking at chip right between Navi10 (7.9TF) and Navi10 XT (9.75TF)


Scarlett chip is codename "Arden", while Gonzalo's ID is referencing Ariel, which fits under Sony's custom chips in PCI ID base.

Ahh.. any leaked benchmarks for Arden?
 

Munki

Member
CB posted good wrap in post above.

In layman terms, it will probably be betwen Vega56 and Vega64. If we know Navi XT is 1.14x Vega64, we could be looking at chip right between Navi10 (7.9TF) and Navi10 XT (9.75TF)


Scarlett chip is codename "Arden", while Gonzalo's ID is referencing Ariel, which fits under Sony's custom chips in PCI ID base.

Nice, thank you for the information bro!
 

LordOfChaos

Member
No one verifies, doubt Sony/AMD would officialy reply to mail if we asked. This is why its a leak coming from guy who is decoding and pulling benches from AMD chips for over decade, so I guess he would know a thing or two about fake benches.

We also have precedent of PS4 and Pro being in database with, what appears, very likely benchmark scores. Its not like bench itself says "this is PS5". Someone had to dig up PCI IDs database and find which codename could be for which product etc.

It might all be ruse, but too many things are tying up IMO.


If he's just reading results off the online browser he has even less means of determining a fake.

No one seems to have explained why such a benchmark would run on a FreeBSD based OS. Do Sony develop working drivers for bespoke hardware just to test it? Because it takes years after a consoles release to get another OS tepidly running on it unless it's officially supported, and the PS4 still runs Linux poorly, the PS4 results are being brought up to add validity to this but what ever said those were real either?

Occam's razor here, it's much much easier to get parts that are close to the expected performance and alter driver strings...


Isn't 3dmark directx based? If these chips are being tested under windows then the real performance results under ps5 api will most likely be different.

It can use Vulkan, but even then, no one is asking the simple question of why it would run on PS5 hardware at all...
 
Last edited:

pawel86ck

Banned
From people who didnt get the memo that weaker Navi (7.9TF version) sucks entire console worth of watts - 180.
And how do you know Navi RX5700 TDP will be 180W, have you tested it? If Igor calculatuins are correct, RX5700 TDP should be lower.
 

bitbydeath

Member
Everything we actually do know flies in the face of a sub 10TF machine.

1. Sony demoing GT Sport in 8K at 120FPS on a supposed PS5 dev kit.


2. Mark Cerny confirming the PS5 will support 8K graphics.


3. Mark Cerny stating the pricing will be appealing despite its advanced feature set.


4. Ubisoft stating a major leap with graphics ranging from 4K to 8K.


Along with the numerous rumours of devs being really impressed with the specs.

Edit: And you know Sony will want PSVR2 to run in 4K which won’t come cheap.
 
Last edited:

Ar¢tos

Member
Isn't 3dmark directx based? If these chips are being tested under windows then the real performance results under ps5 api will most likely be different.
 

Fake

Member
Isn't 3dmark directx based? If these chips are being tested under windows then the real performance results under ps5 api will most likely be different.
Yeap. The chief of Digital Foundry already alert for not making comparison between PC and console. We can just speculate at this point.
 
Last edited:
Everything we actually do know flies in the face of a sub 10TF machine.

1. Sony announcing 8K and having demoed GT Sport in 8K at 120FPS on a supposed PS5 dev kit.


2. Mark Cerny stating the pricing will be appealing despite its advanced feature set.


3. Ubisoft stating a major leap with graphics ranging from 4K to 8K.


Along with the numerous rumours of devs being really impressed with the specs.

Edit: And you know Sony will want PSVR2 to run in 4K which won’t come cheap.
To be fair, they were pretty impressed with a 1.8TF machine also.
 

TeamGhobad

Banned
what i got from that ps5 benchmark was that under the worse conditions it performs just under a vega 64. not bad.
 
Last edited:
4.2 x 4 = 16.8....


16.8 GCN TF / 1.25 IPC = 13.44 Navi TF

Wait, so one of the theories is that PS5 is dual GPU, like PS4 pro? Like dual 5700s? Won't that be a lot of TDP for a console, like even if it's underclocked a little? Would a dual GPU design be cost effective for Sony as its mainstream console? (wouldn't something like this be more logical mid generation refresh type thing, aimed at the same audience as Pro?). Hrmm
RSX was 2 x GeForce 6800 and PS4 Pro was OG PS4 x 2 (butterfly setup for easy BC).

Sony has a precedent with this. They just need the proper lithography tech to pull this off.

It could literally be Bob down the street.
If the hardware were completed already then it’d release in 2019.

Was PS4 / PS4 Pro detected 1.5-2 years in advance of release?
The final PS4 APU was completed for dev kits in January 2013 (10 months before launch/November 2013):


It's too early for a final PS5 APU, so I wouldn't bet on Gonzalo being the real thing (assuming it is for Sony).

Early PS3 dev kits had Cell running at 2.4 GHz, not 3.2 GHz. They also used off-the-shelf GeForce 6800/7800 cards before RSX was finally completed and PCIe instead of FlexIO:

 
Last edited:

R600

Banned
16.8 GCN TF / 1.25 IPC = 13.44 Navi TF


RSX was 2 x GeForce 6800 and PS4 Pro was OG PS4 x 2 (butterfly setup for easy BC).

Sony has a precedent with this. They just need the proper lithography tech to pull this off.


The final PS4 APU was completed for dev kits in January 2013 (10 months before launch/November 2013):


It's too early for a final PS5 APU, so I wouldn't bet on Gonzalo being the real thing (assuming it is for Sony).

Early PS3 dev kits had Cell running at 2.4 GHz, not 3.2 GHz.
Amm....there is nothing dual GPU about RSX and Pro. Not a thing.

Pro is using "butterfly" (whatever that PR speak is) as a way to say "We doubled CUs between the two so that we can have rudimental BC support by disabling half a chip when running PS4 games".

As for PS4 dev kits, this is for 3rd party as I understand, because 1st party gets it months in advance.

PS3 dev kits with official SOC came more then 1 year before release, 2005.
 
Last edited:
Status
Not open for further replies.
Top Bottom