• 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.

Albert Penello - Impossible for changes to the die to be made 12 months before release.

A post was made on another forum, whos name escapes me, asking Albert Penello an interesting question.

The question was
"
Could you answer this for me.

When you were working on the one X, and it was 12 months before release, if Bill Gates walked in and said " Hey guys, I want you to wack on 4 more compute units", just how much time and cost are we talking about to just chuck on some more Cu's?

Would you have plenty of time to make the launch date? What sort of things would you need to crash test the GPU for by adding said Cu's?
Thanks muchly."

Alberts reply was interesting.

"
It would be impossible. You'd either delay the program, try and see what you could get done within the chip you had (enable redundant CU's or upclock) or ship what you have.

I'm not a silicon engineer. Someone who is may be able to speak in more depth. My experience would say that "adding CU's" to a chip that doesn't have them is no longer that chip, it's a new program.

What I can tell you is that 12 months before release, you're actually manufacturing consoles. Not final consoles, but you have production lines going and are manufacturing units for validation and testing.

Phil tweeted in December that the team had takeome units. Those are pre-production units but "pre-production" is still made on a production line. It may not be using all the final parts, but the process is designed to work out manufacturing kinks.

To ship for the end of the year, manufacturing is likely starting in late summer which at this point is only 4 - 5 months away."
 

Hobbygaming

has been asked to post in 'Grounded' mode.
A post was made on another forum, whos name escapes me, asking Albert Penello an interesting question.

The question was
"
Could you answer this for me.

When you were working on the one X, and it was 12 months before release, if Bill Gates walked in and said " Hey guys, I want you to wack on 4 more compute units", just how much time and cost are we talking about to just chuck on some more Cu's?

Would you have plenty of time to make the launch date? What sort of things would you need to crash test the GPU for by adding said Cu's?
Thanks muchly."

Alberts reply was interesting.

"
It would be impossible. You'd either delay the program, try and see what you could get done within the chip you had (enable redundant CU's or upclock) or ship what you have.

I'm not a silicon engineer. Someone who is may be able to speak in more depth. My experience would say that "adding CU's" to a chip that doesn't have them is no longer that chip, it's a new program.

What I can tell you is that 12 months before release, you're actually manufacturing consoles. Not final consoles, but you have production lines going and are manufacturing units for validation and testing.

Phil tweeted in December that the team had takeome units. Those are pre-production units but "pre-production" is still made on a production line. It may not be using all the final parts, but the process is designed to work out manufacturing kinks.

To ship for the end of the year, manufacturing is likely starting in late summer which at this point is only 4 - 5 months away."
The rumor though is that PS5 had its plans changed years ago, possibly in 2018
 
The rumor though is that PS5 had its plans changed years ago, possibly in 2018
Considering AMD were still running benchmarks on the 36 Cu GPU in June, that doesn't ring true.
Most likely it is a 36/40CU GPU. Interesting that Albert said that they could enable redundant Cu's, as well as up clocks. I said that I think they could do that and people jumped up and down about how that would be impossible due to failure rate. Well, glad to see Albert says the same thing. And thats what I think we will end up seeing. A 40 CU GPU clocked at as close to 2ghz as they can get. That would give the equivalent of a 13tflop GCN GPU. That might be why it is said that the early GCN dev kits Sony sent out were 13tflop, and the new ones using Navi GPUs was 10.x.
 
Last edited:

Hobbygaming

has been asked to post in 'Grounded' mode.
Considering AMD were still running benchmarks on the 36 Cu GPU in June, that doesn't ring true.
Most likely it is a 36/40CU GPU. Interesting that Albert said that they could enable redundant Cu's, as well as up clocks. I said that I think they could do that and people jumped up and down about how that would be impossible due to failure rate. Well, glad to see Albert says the same thing. And thats what I think we will end up seeing. A 40 CU GPU clocked at as close to 2ghz as they can get. That would give the equivalent of a 13tflop GCN GPU. That might be why it is said that the early GCN dev kits Sony sent out were 13tflop, and the new ones using Navi GPUs was 10.x.
I don't think they would enable the other CU's because of yields but I also don't believe they will come in at 36 CU's. Jason Schreier mentioned that each console has strengths in different areas and both are very powerful. He also said it might take some time before we'll know which console is the most powerful, this leads me to believe that they are very close in power just like most insiders have said

I think it'll be a 15% or less difference and that's nothing when it comes to real-world performance
 
Last edited:

sinnergy

Member
Yes , this is how it works. Better is to start producing real unit is around this time ...

I bet they will starting to roll of the line this month or the next.
 
Last edited:

Panajev2001a

GAF's Pleasant Genius
A post was made on another forum, whos name escapes me, asking Albert Penello an interesting question.

The question was
"
Could you answer this for me.

When you were working on the one X, and it was 12 months before release, if Bill Gates walked in and said " Hey guys, I want you to wack on 4 more compute units", just how much time and cost are we talking about to just chuck on some more Cu's?

Would you have plenty of time to make the launch date? What sort of things would you need to crash test the GPU for by adding said Cu's?
Thanks muchly."

Alberts reply was interesting.

"
It would be impossible. You'd either delay the program, try and see what you could get done within the chip you had (enable redundant CU's or upclock) or ship what you have.

I'm not a silicon engineer. Someone who is may be able to speak in more depth. My experience would say that "adding CU's" to a chip that doesn't have them is no longer that chip, it's a new program.

What I can tell you is that 12 months before release, you're actually manufacturing consoles. Not final consoles, but you have production lines going and are manufacturing units for validation and testing.

Phil tweeted in December that the team had takeome units. Those are pre-production units but "pre-production" is still made on a production line. It may not be using all the final parts, but the process is designed to work out manufacturing kinks.

To ship for the end of the year, manufacturing is likely starting in late summer which at this point is only 4 - 5 months away."

It makes sense, with enough money thrown at it you may still be able to make good changes within that window (iPhone’s A series chips get one or two new designs per year and ship in waaay higher volume than Xbox or PS dare to dream of), but you also risk RROD’s/failures to reach customers.

I was perplexed about the first post target since it felt a bit out of the blue... thankfully the PS5 reference and the GitHub leak as gospel came shortly after ;).
 

MoreJRPG

Suffers from extreme PDS
You’re gonna have consoles bricking left, right and centre if you start changing the silicon at this stage. These units need to start hitting mass production.

All the speculation though about Sony reworking Playstation at this time leads me to be believe that yes, the leaks were correct. This thing will be underpowered and overpriced and looking at Godfall and how unimpressive that looks graphically, I would say the hardware isn’t worth showing off which is why they haven’t.
 

Panajev2001a

GAF's Pleasant Genius
You’re gonna have consoles bricking left, right and centre if you start changing the silicon at this stage. These units need to start hitting mass production.

All the speculation though about Sony reworking Playstation at this time leads me to be believe that yes, the leaks were correct. This thing will be underpowered and overpriced and looking at Godfall and how unimpressive that looks graphically, I would say the hardware isn’t worth showing off which is why they haven’t.

Believing the best rumours about a console and only the worst ones about another? no way... :rolleyes:
 

Thaedolus

Gold Member
This should be obvious to anyone who’s worked in R&D and manufacturing, and GPUs/CPUs are almost certainly the most complicated and intricate things humankind have ever made. You can’t just juice them up on a whim.
 

Woo-Fu

Banned
People who think you simply make your console more powerful when you find out your competition is stronger shortly before introduction---without incurring a significant delay to that introduction---should ask themselves why that has never actually happened.

You don't have to know anything about design or manufacturing to figure out it isn't practical simply by reviewing history and noticing that it hasn't been done. You don't even have to look past the current generation to figure it out.
 
Last edited:
I don't think they would enable the other CU's because of yields but I also don't believe they will come in at 36 CU's. Jason Schreier mentioned that each console has strengths in different areas and both are very powerful. He also said it might take some time before we'll know which console is the most powerful, this leads me to believe that they are very close in power just like most insiders have said

I think it'll be a 15% or less difference and that's nothing when it comes to real-world performance
It could be that there is in fact a 2 or 3 tflop advantage to the XSX, but there may be other tech that makes up for it? I'm not talking the SSD, but maybe the RT solution Sony has doesn't hit the GPU as hard, while the XSX RT does, therefore the PS5 puts out similar results to XSX.

Meh who knows. They just need to announce the specs already.
 

Hobbygaming

has been asked to post in 'Grounded' mode.
You’re gonna have consoles bricking left, right and centre if you start changing the silicon at this stage. These units need to start hitting mass production.

All the speculation though about Sony reworking Playstation at this time leads me to be believe that yes, the leaks were correct. This thing will be underpowered and overpriced and looking at Godfall and how unimpressive that looks graphically, I would say the hardware isn’t worth showing off which is why they haven’t.
Imagine thinking an internal demo from a build that's 2 years away from being a finished product is how all PS5 games will look 🤦‍♂️
 

MadAnon

Member
It makes sense, with enough money thrown at it you may still be able to make good changes within that window (iPhone’s A series chips get one or two new designs per year and ship in waaay higher volume than Xbox or PS dare to dream of), but you also risk RROD’s/failures to reach customers.

I was perplexed about the first post target since it felt a bit out of the blue... thankfully the PS5 reference and the GitHub leak as gospel came shortly after ;).
That's because the projects overlap and are worked on in different capacities. It's not like there's one team just doing all the chip designs in consecutive fashion.
 
Last edited:

Panajev2001a

GAF's Pleasant Genius
That's because the projects overlap and are worked on in different capacities. It's not like there's one team just doing all the chip designs in consecutive fashion.

Of course (and having multiple teams with overlapping projects and sharing tech helps but it costs a lot of money to setup, scale, and coordinate nonetheless), but then you do have clear sign of iteration of previous teams work and other major major elements such as node transitions to cover. I am not saying it would be cheap, but modifications to final silicon can be done within a 12 months window depending on the impact (adding self contained replicable blocks like SRAM or CU’s has a different cost than implementing SMT on a design that was not built with it in mind, despite the latter having a relatively tiny die area impact).
 
Last edited:

cireza

Banned
How strange, some smart people have been repeating that Sega added the second CPU to the Saturn two weeks before release. Did they lie ?
 

joe_zazen

Member
Considering AMD were still running benchmarks on the 36 Cu GPU in June, that doesn't ring true.
Most likely it is a 36/40CU GPU. Interesting that Albert said that they could enable redundant Cu's, as well as up clocks. I said that I think they could do that and people jumped up and down about how that would be impossible due to failure rate. Well, glad to see Albert says the same thing. And thats what I think we will end up seeing. A 40 CU GPU clocked at as close to 2ghz as they can get. That would give the equivalent of a 13tflop GCN GPU. That might be why it is said that the early GCN dev kits Sony sent out were 13tflop, and the new ones using Navi GPUs was 10.x.

here is what happened:

summer of 2018, Charlie Demerjian leaks ps5 info on Semiaccurate. This is the guy who broke the story of nintendo going with nvidia, ps4 specs in 2012, and sonys plan for pspro in 2013. Guy isnt 100%, but he is connected in silicon world. He stated a large number of sony dev kits went out in 2018, indicating a 2019 launch.. there were so many kits out that he thought they could have done a 2018 launch if they had to, he DID NOT predict a 2018, 2019 was the year.

Oberon was the 2019 chip. Since they did not release a 2019 console, Something changed between summer of 2018 and 2019, and sony shifted to a 2020 date.

Most likely they saw MS’ plan for a huge chip based on superior tech on a smaller node with hugely marketable things like RT (2019 Oberon would not have had RT), as well as Microsoft’s plans for a cheap box.

2019 PS5 would look like crap come holiday 2020, and for the next six years, and they couldn’t just lower the price because of lockhart, which would also have things like RT. So they scrambled, and paid AMD to fast track their own competitive 2020 chip. Very expensive, but necessary.

this would explain sony’s silence in 2019 and yesterday’s story about sony having to shift third party games around 2 years ago.

So whats up with github oberon still being tested and worked on as late as summer 2019? It is sony’s version of lockhart. Maybe Oberon is dead or maybe it is coming as ps5 weaksauce, but github does not mean a 8-9 TF ps5.

all is good.

people hoping for a shit tier ps5 because of git hub will be disappointed.
 

DESTROYA

Member
Considering AMD were still running benchmarks on the 36 Cu GPU in June, that doesn't ring true.
Most likely it is a 36/40CU GPU. Interesting that Albert said that they could enable redundant Cu's, as well as up clocks. I said that I think they could do that and people jumped up and down about how that would be impossible due to failure rate. Well, glad to see Albert says the same thing. And thats what I think we will end up seeing. A 40 CU GPU clocked at as close to 2ghz as they can get. That would give the equivalent of a 13tflop GCN GPU. That might be why it is said that the early GCN dev kits Sony sent out were 13tflop, and the new ones using Navi GPUs was 10.x.
Or they could of been testing different hardware to see what fits and works better.
Prototypes are not final hardware.
 
Last edited:

Thaedolus

Gold Member
How strange, some smart people have been repeating that Sega added the second CPU to the Saturn two weeks before release. Did they lie ?

SEGA added a VDP late in development based on what Sony was planning on PSX, but there’s zero chance it was done two weeks prior to launch, and adding a chip to a PCB is like orders of magnitude simpler than changing the silicon of a processor. Even in today’s world there are huge lead times in acquiring components and developing processes on manufacturing lines to get reliable product out the door. A two week turnaround on adding a chip to a mass produced product which ostensibly has software already created and tested on it sounds completely impossible.

Then again, 90s SEGA was completely insane.
 

wolffy71

Banned
Everyone is looking so deep for anything that points to the power or lack thereof for both consoles. But wouldn't a really big difference between consoles have been whispered a little louder if they were really that far apart?
 

Hobbygaming

has been asked to post in 'Grounded' mode.
here is what happened:



people hoping for a shit tier ps5 because of git hub will be disappointed.
This makes a lot of sense and it goes along with other rumors that there was a PS5 prototype sent out for devs before summer 2019 when Sony sent the new Prospero SDKs to devs
 

Hobbygaming

has been asked to post in 'Grounded' mode.
Everyone is looking so deep for anything that points to the power or lack thereof for both consoles. But wouldn't a really big difference between consoles have been whispered a little louder if they were really that far apart?
That's a great point. No one has came out and said there's a huge gap in capabilities. Everyone seems to be saying both consoles are close in power
 

The Alien

Banned
Is this post in response to "9 tf Gate" and recent "reports" that Sony reworked the PS5 and delayed it from launching on 2019?

If not...the just a completely random thread, I suppose.
 

DeepEnigma

Gold Member
You’re gonna have consoles bricking left, right and centre if you start changing the silicon at this stage. These units need to start hitting mass production.

All the speculation though about Sony reworking Playstation at this time leads me to be believe that yes, the leaks were correct. This thing will be underpowered and overpriced and looking at Godfall and how unimpressive that looks graphically, I would say the hardware isn’t worth showing off which is why they haven’t.

Fanboy hopes & dreams.
 
I said that I think they could do that and people jumped up and down about how that would be impossible due to failure rate.
Sometimes a very high failure rate is your best option, they can always do the first "few" millions of consoles like that and ask for the creation of a new batch of chips with more redundancy, or wait for the production to matures.

Either way we saw how it could work, MS was in that situation before the xbone launched, they "overclocked" their chip in the hope that it would compete a bit better (just image if they had not done this, games like Ryse would have run like Ocarina Of time on the N64!).
 

I'm nobody

Member
Dunno why people say more power is better the xbx is most powerful console now and we note no difference.

Tbh it also done Xbox ms no favors an all

I work in game and I've heard nothing but people wanting to pre order ps5 over xbox for months now
 
Last edited:

diffusionx

Gold Member
It might be obvious to many, but I’ve seen several posts on GAF where users acted like Sony could just increase the CUs or increase the CPU frequency at the drop of a hat.

MS over clocked the Xbox one right after the 2013 E3. It was probably done at high expense and was an act of desperation and obviously didn’t matter much. But it could be done.

Increasing CUs or changing the architecture, no way. The decisions have been made and everyone will have to live with them.
 

Journey

Banned
You’re gonna have consoles bricking left, right and centre if you start changing the silicon at this stage. These units need to start hitting mass production.

All the speculation though about Sony reworking Playstation at this time leads me to be believe that yes, the leaks were correct. This thing will be underpowered and overpriced and looking at Godfall and how unimpressive that looks graphically, I would say the hardware isn’t worth showing off which is why they haven’t.

I agree that it's too late for any changes, but I disagree that Godfall is a good example of what the PS5 will be capable of, that's just silly. Sure XSX will likely be more powerful, but PS5 will be no slouch.
 

Jigsaah

Gold Member
My prediction is that each console will tout their differences as a supposed power or speed advantage of some type but when it really comes down to it, it's what the developers can do with it that will matter the most. The games will tell all.
 

cireza

Banned
SEGA added a VDP late in development based on what Sony was planning on PSX
I totally don't believe that they added the second VDP late considering how each VDP is dedicated to very specific graphic tasks.

I was joking when saying "two weeks" by the way.
 
Last edited:
Top Bottom