Support NeoGAF

[Shin]

The part about PS5 having 72 CUs is just guess work

It's not gonna have 10CU's as some posters/articles are running with either, that's for sure
10CU (compute units) x 64 SP (stream processors) x 2 memory speed (2 cycles per clock and (PS4 Pro runs at 911MHz, Xbox OX at 1172MHz)) = teraflops (absolute power)
The math doesn't add up then you got the iGPU, it's mind boggling as to why when die space is precious enough as is if not pricey.

 

[ethomaz]

No the FP32 TF number would stay the same no matter how efficient one GPU is over the other but that number is just that "a number" it tells you how many FP32 floating point operations that could be done in a second.


But in the next era of GPUs I'm guessing that Deep Learning is going to be a big deal so you might see fp16 , 10bit , 8bit & so on used for different stuff getting more operations out of the GPU.

I believe he meant the number of units in a CU can change from a arch to another.

Forget the 64SPs per CU if Navi is something new.

BTW GCN is limited to 64ROPS, 64CUs, 4096 SPs.

 

[Imtjnotu]

Link please? I searched "Gonzalo" and didn't find anything. Also scrolled first couple of pages with no results as well :/

EDIT:
OK I see now, guess ignoring people makes you unable to see threads they make.

Hahahaha gotcha

 

[Dabaus]

When I hear numbers like that, I always think; Yeah, but how many have since died? How many are collecting dust in a closet? How many were smashed to bits by some jackass on Youtubue? How many are sitting on a used game store shelf? How many are actually still in use? How many are OG's? How many are Slims? How many are Pro's? How many are new and still sitting on a store shelf? The list goes on.......

I mean, it's great to hear that the console I chose as my main has sold so well, but.....but....I need more detailed stats.

Matching up pretty close to what Osirisblack said

https://www.neogaf.com/threads/possible-ps5-leak-info.1467805/page-7#post-253564893





I really think this will be a lot like what they did with the PS4 going to PS4 Pro with a doubling of the GPU so a 72 CU GPU at 1.121 would be 10.33 TFLOPS

Damn, buddy was right on the money on some parts. What else did he say?

 

[Redneckerz]

I'm just going to say that there will most likely be a wearable VR device in the future but I can't say if it will be a wearable PS4/PS5 or just a portable platform but I would guess that Sony isn't going to waste time on a new portable platform when PS4 already use a mobile CPU & has a large catalog.

This makes even less sense than your previous post. So Sony won't bother with a new handheld... because PS4 has laptop based CPU cores. I can't even the fuck make spaghetti out of that.

The part about PS5 having 72 CUs is just guess work

You mean you said this so months later you can perpetuate you had the truth all this time.

 

[Fake]

Matching up pretty close to what Osirisblack said

https://www.neogaf.com/threads/possible-ps5-leak-info.1467805/page-7#post-253564893





I really think this will be a lot like what they did with the PS4 going to PS4 Pro with a doubling of the GPU so a 72 CU GPU at 1.121 would be 10.33 TFLOPS

A less agressive transition to confirm BC I guess. For me just the remove of jaguar is a big step foward. Someone confirmed if is a 2rd gen Ryzen, or a really custom made (nobody will see in other market) Ryzen?

 

[Dontero]

72CU is not possible. Max CU they can go with is 64.
This is GCN architecture limit. They can't go past 64 CU even if they wanted.

Which is why i think that 10CU part is telling that Navi is not exactly the same GCN people xpect.
Sure it might seem still use some parts of GCN arch for compatibility but under the hood changes could be much more extensive.

 

[bitbydeath]

This is GCN architecture limit.

That isn’t confirmed and is only speculation AFAIK.
Originally they said it would have next-gen memory and I don’t think anything has been said about the memory since.

Feel free to correct me though as I may have missed something.

 

[LordOfChaos]

Largely it's falling in line with expectations, main question for me is what clock they expect it to stay at with all core turbo and a full GPU load. At 1.6GHz it would still beat Jaguar, but not mind blowingly so, if it stays closer to 3.2, well, the range is obviously approximately double.

Navi 10 lite...Hm. If this render config is an accurate stand in, probably towards the lower range of GPU Gflops expectations, though hopefully Navi also substantially lifts instructions per core per clock from Vega.

https://www.techpowerup.com/gpu-specs/amd-navi-10.g861

 

[onQ123]

It's not gonna have 10CU's as some posters/articles are running with either, that's for sure
10CU (compute units) x 64 SP (stream processors) x 2 memory speed (2 cycles per clock and (PS4 Pro runs at 911MHz, Xbox OX at 1172MHz)) = teraflops (absolute power)
The math doesn't add up then you got the iGPU, it's mind boggling as to why when die space is precious enough as is if not pricey.

1 thing for sure is that the new leaks match up with osirisblack numbers for the CPU & GPU clocks so I'm going to assume he is right about that 18GB @ 880GB/s memory also so it's not going to be a weak GPU if it's going to need that much bandwidth ,


RX 590 only has 256 GB/s so even if you account for headroom to use fp16 this 880GB/s is still pretty high so you know it's feeding something.

This makes even less sense than your previous post. So Sony won't bother with a new handheld... because PS4 has laptop based CPU cores. I can't even the fuck make spaghetti out of that.


You mean you said this so months later you can perpetuate you had the truth all this time.

Why are you so worried about my speculations? everything until we get the real specs is speculations so you sound stupid trying to tell someone how to speculate

peace!

 

[Redneckerz]

Why are you so worried about my speculations?

Because:

• Half of the time it is not even true
• The other half you are just vaguely referencing it, yet you present it as definitive proof
• There is a tendency of you dropping these speculative points on purpose so that months later, when nobody remembers, you can pull that post out of your box and be like ''See? I told you so.'' which goes with points 1 and 2

everything until we get the real specs is speculations so you sound stupid trying to tell someone how to speculate

At least some speculations have some basis in reality. Yours are just pie in the sky ideas made out of thin air. If i sound stupid supposely telling people how they should speculate, than you are far deeper in that rabbit hole for making speculations that have no grounds to reality, and done so on purpose so that the one time one of your speculations vaguely resembles something in the final result, you can claim that you knew.

 

[Shin]

AMD Ryzen cpu custom configuration 3.20 , 7nm Navi gpu 2.1 18gddr5 Is what I heard a while ago. As far as the public announcement It will be soon. I sent this info to a mod a while ago I couldn’t post until something else leaked first. I’ll post everything now that something semi accurate is out there.

If what Osiris said is correct or turns out to be close then it's a middle ground of what I wanted and expected (3.5GHz / 12TF / 24Gb / Zen2).
Also what do you know, Samsung GDDR6 went into mass production on their product page (this wasn't the case last month) and more variety.

 

[ethomaz]

Largely it's falling in line with expectations, main question for me is what clock they expect it to stay at with all core turbo and a full GPU load. At 1.6GHz it would still beat Jaguar, but not mind blowingly so, if it stays closer to 3.2, well, the range is obviously approximately double.

Navi 10 lite...Hm. If this render config is an accurate stand in, probably towards the lower range of GPU Gflops expectations, though hopefully Navi also substantially lifts instructions per core per clock from Vega.

https://www.techpowerup.com/gpu-specs/amd-navi-10.g861

If that GPU entry is true and Navi is really only another GCN then what will define the TF is the clock because 64ROP/64CU/4096SP is the limit.

To get anything over 12TFs you will need 1500Mhz on GPU clock.

 

[onQ123]

Because:

• Half of the time it is not even true
• The other half you are just vaguely referencing it, yet you present it as definitive proof
• There is a tendency of you dropping these speculative points on purpose so that months later, when nobody remembers, you can pull that post out of your box and be like ''See? I told you so.'' which goes with points 1 and 2

At least some speculations have some basis in reality. Yours are just pie in the sky ideas made out of thin air. If i sound stupid supposely telling people how they should speculate, than you are far deeper in that rabbit hole for making speculations that have no grounds to reality, and done so on purpose so that the one time one of your speculations vaguely resembles something in the final result, you can claim that you knew.

How about you add me to your ignore list & stop worrying about what I post

 

[LordOfChaos]

If that GPU entry is true and Navi is really only another GCN then what will define the TF is the clock because 64ROP/64CU/4096SP is the limit.

To get anything over 12TFs you will need 1500Mhz on GPU clock.

It's almost undoubtedly another GCN, the "next gen" after it is too conspicuously named. Still, each GCN revision has increased how many instructions per clock each core can actually do, aside from the paper rate (Gflops).

Would Navi efficiency change the teraflops at all compared to going off of a custom Polaris?

Same here - it doesn't change the paper calc, but it does change real performance.

 

[onQ123]

If that GPU entry is true and Navi is really only another GCN then what will define the TF is the clock because 64ROP/64CU/4096SP is the limit.

To get anything over 12TFs you will need 1500Mhz on GPU clock.

https://www.digitaltrends.com/computing/amd-navi-rumors-price-release-date/

"
Early reports suggested that Navi will be the last AMD GPU to be based on its graphics core next (GCN) architecture. However, more recent reports indicate that Navi will debut with a new microarchitecture, which will help it overcome some of the limitations of the GCN framework. This means that we could see faster clock speeds, more shaders, and better efficiency. AMD also confirmed that Navi will be a scalable architecture that will support both HBM2 and GDDR6 memory. AMD said that it won’t use a multi-chip module approach for Navi, according to PC World.

 

[ethomaz]

It's almost undoubtedly another GCN, the "next gen" after it is too conspicuously named. Still, each GCN revision has increased how many instructions per clock each core can actually do, aside from the paper rate (Gflops).

The number of instructions per clock of each core is the same since first GCN... 2.

FLOPS maths = number o cores * clock * 2

Same for nVidia each core can do 2 instructions per clock... that never increased.

Max number of GCN cores are 4096.

 

[Redneckerz]

How about you add me to your ignore list & stop worrying about what I post

I do not ignore people by default just because their rhetoric is flawed.

I won't stop worrying either since you would be influencing lesser knowledgeable users with your speculations.

 

[LordOfChaos]

The number of instructions per clock of each core is the same since first GCN... 2.

FLOPS maths = number o cores * clock * 2

Same for nVidia each core can do 2 instructions per clock... that never increased.

Max number of GCN cores are 4096.

You're talking about the paper calculation I already mentioned. That's not the same as real work done, each revision brings the actual IPC closer to the theoretical two.

You brought up Nvidia and that's the most perfect example - both have two instructions per core per clock as the calculation on paper, but clearly each do different amounts of work per core per clock, or per Gflop.

Trying to find the review that showed this well, they took each GCN generation, found performance, and then normalized for clock speed and core count, showing the work done per core per clock steadily rise per revision. It's still two instructions per core per clock on paper, that's the paper calculation, but none of these is actually filling the full two every clock and every ALU.


(it's the same, by the way, for CPUs - modern ones can issue 6 instructions per clock, but most consumer code isn't crossing two instructions per clock average - though the extra width still matters for those opportunities to fill more)

 

[Esppiral]

Gonzalo Is my project manager's Name lol.

 

[Fake]

Gonzalo Is my project manager's Name lol.

So, you're a leaker.

 

[onQ123]

I do not ignore people by default just because their rhetoric is flawed.

I won't stop worrying either since you would be influencing lesser knowledgeable users with your speculations.

If you don't have any actual input of your own why sit around trying to control what other people post?

 

[CrustyBritches]

I do not ignore people by default just because their rhetoric is flawed.

I won't stop worrying either since you would be influencing lesser knowledgeable users with your speculations.

To be fair, you're arguing with someone who was banned in part for pushing the FP16 narrative with the Xbox One X. As in, he claimed the 6TF was FP16. You can look at the history to confirm this.

 

[ethomaz]

Dammit... two posts? lol

 

[ethomaz]

You're talking about the paper calculation I already mentioned. That's not the same as real work done, each revision brings the actual IPC closer to the theoretical two.

You brought up Nvidia and that's the most perfect example - both have two instructions per core per clock as the calculation on paper, but clearly each do different amounts of work per core per clock, or per Gflop.

Trying to find the review that showed this well, they took each GCN generation, found performance, and then normalized for clock speed and core count, showing the work done per core per clock steadily rise per revision. It's still two instructions per core per clock on paper, that's the paper calculation, but none of these is actually filling the full two every clock and every ALU.


(it's the same, by the way, for CPUs - modern ones can issue 6 instructions per clock, but most consumer code isn't crossing two instructions per clock average - though the extra width still matters for those opportunities to fill more)

I think you need to read that.

https://forums.anandtech.com/threads/two-misconceptions-about-ipc-and-gpus.2476552/

Each GCN has changes that affect performance like ROPs, memory system, etc... that is why difference generations at the same clock have different performance.... it is not IPC die a lower or higher IPC.

GPU cores are actually not like CPU cores.

 

[Redneckerz]

If you don't have any actual input of your own why sit around trying to control what other people post?

I already provided an actual comment, what you do is deflection of your own behavior and inability to understand that outside-of-reality speculation as opposed to speculation with grounds to said reality is simply making a scene and the reasons listed prior.

To be fair, you're arguing with someone who was banned in part for pushing the FP16 narrative with the Xbox One X. As in, he claimed the 6TF was FP16. You can look at the history to confirm this.

Im aware

 

[LordOfChaos]

I think you need to read that.

https://forums.anandtech.com/threads/two-misconceptions-about-ipc-and-gpus.2476552/

Each GCN has changes that affect performance like ROPs, memory system, etc... that is why difference generations at the same clock have different performance.... it is not IPC.

Read that a long time ago. What they're correctly arguing is that GPU IPC goes down with clock rate and this isn't inherently bad - compare a higher clocked GPU to a lower one and it's unfair to complain about it being lower IPC - what I'm saying is that normalized for clock, each generation has also been able to fill more of those two slots per ALU per clock. And yes, increased memory bandwidth and elimination of other bottlenecks plays into that. I have no opposition to that post about lower IPC being a necessity for higher clocks within the same architecture.

This was with the possible exception of Vega iirc.

 

[DeepEnigma]

Goodness it is starting to sound like a bunch of petty children in here, Let It Go.

 

[onQ123]

To be fair, you're arguing with someone who was banned in part for pushing the FP16 narrative with the Xbox One X. As in, he claimed the 6TF was FP16. You can look at the history to confirm this.

No I wasn't

 

[CrustyBritches]

O onQ123 ...You were most definitely the one pushing the Xbox One X 6TF FP16 narrative. Let's not spin. You use the technique of pretty much spray-and-pray, then selectively quote yourself about the stuff you might have been close on, while ignoring your principle claims.

Spoiler: Example: FP16 claims

onQ123 said:


Until I see real Xbox Scorpio specs I'm forced to believe that they did exactly what I think they did..

Remember Microsoft never mentioned Xbox One's 1.3Tflops but now they announce Xbox Scorpio as a 6tflop console.

I'm guessing the APU will also need to have embedded ram for backwards compatibility with Xbox One so what are the chances of them fitting a bigger GPU than what is in the PS4 Neo in a APU with embedded ram?

The fact that Polaris can do half-precision computation gave them the chance to use the 6tflop line & the fact that Neo wasn't announced gave them the chance to say it's the most powerful console ever made.


onQ123 said:


I'm saying that Xbox Scorpio could be just like what the Neo is to the PS4 a APU with 2X the CUs with a higher clock speed & because Polaris is capable of half-precision computation Microsoft was able to say that it is 6tflops because

24 compute units clocked at 1Ghz would be 3tflops single-precision & 6tflops half-precision.

Polaris support native half-precision floating point calculations & because of that Sony could say that PS4 Neo is 8.4tflops because it would be true if they are using Polaris. .


onQ123 said:


I think the more likely scenario for Scorpio is 6TF FP16 but I would rather it be 6TF FP32.





onQ123 said:


PS4 Pro being 4.2tf fp32 / 8.4tf fp16 isn't something being said to make it look more powerful than Scorpio it's the actual specs. the fact that it upset you only says that you don't understand.


onQ123 said:


That's the real specs this isn't secret sauce


djnewwest said:


Why doesn't Sony advertise the pro as 8.4tf to counter the scorpios 6tf?


Syrus said:


Because it is simply not 8.4 no matter what they do, its pure PR nonsense just like ESRAM and Cloud was nonsense

People are upset Scorpio is better then Pro because X1 was a disaster


onQ123 said:


It's 8.4tf fp16

 

[Tripolygon]

No I wasn't

Yea you kinda entertained the idea that when Phil said 6TF he did not specify 32-bit FP or 16-bit FP. I distinctly remember telling you that is not a specification that has been made ever when talking about desktop or console GPUs as FP16 was mainly used at that point for mobile processors.

 

[GlassAwful]

I'll probably get roasted for this but I'd consider anything much less than x10 tflps increase on the gpu side from base Xbox One an abject failure and question the purpose of rushing into a new console cycle. The standard resolution jump is going to eat a shit ton of those apparent flop increases so, in my mind, we need at least as much as we got two gens ago to get that graphical boost.

I feel the CPU jump is being oversold on the impact it will have for games. The possible advancements people keep touting were possible this generation yet most complex AI and physics interactions in games DECREASED from the previous gen.
Why? It wasn't because the jaguar cores were less capable, it's because those advancements require difficult engineering and problem solving that is better off not bothered with over upping the graphics and frame rate to wow people at E3.

By no means am I saying I don't want that CPU upgrade, quite the opposite. I'm just realistic about this shit so if they end up whimping out on the rest of the package all I can think is the next gen jump could end up a big fat Meh.

 

[onQ123]

O onQ123 ...You were most definitely the one pushing the Xbox One X 6TF FP16 narrative. Let's not spin. You use the technique of pretty much spray-and-pray, then selectively quote yourself about the stuff you might have been close on, while ignoring your principle claims.

Spoiler: Example: FP16 claims

onQ123 said:


Until I see real Xbox Scorpio specs I'm forced to believe that they did exactly what I think they did..

Remember Microsoft never mentioned Xbox One's 1.3Tflops but now they announce Xbox Scorpio as a 6tflop console.

I'm guessing the APU will also need to have embedded ram for backwards compatibility with Xbox One so what are the chances of them fitting a bigger GPU than what is in the PS4 Neo in a APU with embedded ram?

The fact that Polaris can do half-precision computation gave them the chance to use the 6tflop line & the fact that Neo wasn't announced gave them the chance to say it's the most powerful console ever made.


onQ123 said:


I'm saying that Xbox Scorpio could be just like what the Neo is to the PS4 a APU with 2X the CUs with a higher clock speed & because Polaris is capable of half-precision computation Microsoft was able to say that it is 6tflops because

24 compute units clocked at 1Ghz would be 3tflops single-precision & 6tflops half-precision.

Polaris support native half-precision floating point calculations & because of that Sony could say that PS4 Neo is 8.4tflops because it would be true if they are using Polaris. .


onQ123 said:


I think the more likely scenario for Scorpio is 6TF FP16 but I would rather it be 6TF FP32.





onQ123 said:


PS4 Pro being 4.2tf fp32 / 8.4tf fp16 isn't something being said to make it look more powerful than Scorpio it's the actual specs. the fact that it upset you only says that you don't understand.


onQ123 said:


That's the real specs this isn't secret sauce


djnewwest said:


Why doesn't Sony advertise the pro as 8.4tf to counter the scorpios 6tf?


Syrus said:


Because it is simply not 8.4 no matter what they do, its pure PR nonsense just like ESRAM and Cloud was nonsense

People are upset Scorpio is better then Pro because X1 was a disaster


onQ123 said:


It's 8.4tf fp16

Like I said I wasn't banned for that also what I said was that if Scorpio's GPU was Vega then the peak TF number would be the fp16 number.

Scorpio GPU isn't Vega so it didn't have double rate fp16 so it's peak came from the fp32 number.


Also I was completely right about PS4 Pro being 8.4TF fp16


Edit: I used the wrong GPU name back then but I knew from the patent that Sony was going to double pack fp16

 

[onQ123]

Yea you kinda entertained the idea that when Phil said 6TF he did not specify 32-bit FP or 16-bit FP. I distinctly remember telling you that is not a specification that has been made ever when talking about desktop or console GPUs as FP16 was mainly used at that point for mobile processors.

I was replying to him saying I was banned for that. I wasn't banned for that.


And in the case of them not using the fp16 number for desktops back in the day it was because it wasn't the peak but now that we have double rate fp16 you will see the fp16 numbers. that's the whole point I was making when everyone got upset

 

[Shin]

Does anyone know or remember the ps4/Xbox apu ID string?
Curious about 1600 or 1400 being Sony, you'd think with a codename like Ariel = Red = Scarlett = Xbox
Also anyone got a survey?
https://twistedvoxel.com/sony-emailing-new-console-gaming-survey-future-playstation/

 

[THE:MILKMAN]

Does anyone know or remember the ps4/Xbox apu ID string?
Curious about 1600 or 1400 being Sony, you'd think with a codename like Ariel = Red = Scarlett = Xbox
Also anyone got a survey?
https://twistedvoxel.com/sony-emailing-new-console-gaming-survey-future-playstation/

I don't remember us getting the equivalent leak for PS4/Xbox One? I do remember that AMD project names were Thebe-J for PS4 and Kryptos for X1.

 

[TLZ]

Need a Hellpoint Dev to make sense of this

Where would we be without them

 

[Toni]

Those specs would be perfect. especially the CPU.

Now Sony needs to make sure the GPU stays on the ballpark of 10 Teraflops, decent and fast Ram, and they'll be profitable during system's first year and they'll have a beastly console that can do 4K, comfortably.

 

[Chozofication_]

O onQ123 my gosh if ps5 were to have 880gb/s that would either leave hbm2 or gddr6 on a 512 bit bus. Both of which I suspect are too expensive unless Sony wants to charge a premium. I dunno.

Perhaps navi will have a yet to be seen memory type.

But my conservative estimation of ps5's bandwidth would be 500-600gb/s. Also dude ignore that fuckwit.

 

[Fuz]

...Gonzalo?

 

[LordOfChaos]

O onQ123 my gosh if ps5 were to have 880gb/s that would either leave hbm2 or gddr6 on a 512 bit bus. Both of which I suspect are too expensive unless Sony wants to charge a premium. I dunno.

Perhaps navi will have a yet to be seen memory type.

But my conservative estimation of ps5's bandwidth would be 500-600gb/s. Also dude ignore that fuckwit.

Fwiw Radeon VII has a 1TB/s VRAM, achieved by 16GB of HBM2 on a 4096 bit bus. So 880GB/s sounds wild with the hardware we're used to, but it would actually be detuned from the high end coming in now, plus we're likely sharing with the CPU on the next gen if it's an APU. On the other hand that HBM2 is also putting a hard lower limit on VII's price which made for its awkward market position. Not sure how much that would be expected to go down in a year.

Or, if they used the old 2048 bit bus, if they could get to 880GB/s by the faster data rate of the new HBM2 chips and clocks? Would likely be cheaper that way.

 

[Chozofication_]

Fwiw Radeon VII has a 1TB/s VRAM, achieved by 16GB of HBM2 on a 4096 bit bus. So 880GB/s sounds wild with the hardware we're used to, but it would actually be detuned from the high end coming in now, plus we're likely sharing with the CPU on the next gen if it's an APU. On the other hand that HBM2 is also putting a hard lower limit on VII's price which made for its awkward market position. Not sure how much that would be expected to go down in a year.

Or, if they used the old 2048 bit bus, if they could get to 880GB/s by the faster data rate of the new HBM2 chips and clocks? Would likely be cheaper that way.

Well yes as i said they could do it with hbm2 or gddr6 512 bit bus. I'm just saying no matter what those solutions are considerably more high end than ps4's 256 bit bus was for its time.

Basically they could do it but i expect sony to be cheap fucks again. I would guess it to be in reason for them to provide a 384 bit gddr6 solution, and thats my personal best case scenario. Happy to be wrong tho.

 

[base]

Already possible, but developers make choices. This is the way of things.

If you want graphics from PS2 era then even 8k is possible.

 

[DemonCleaner]

3.2 is actually pretty high if you ask me.

it's not on Zen2.

that should need around 30W at full load.

 

[IbizaPocholo]

 

[DeepEnigma]

My head hurts.

What hurts even more are people going into full meltdowns with information they have no clue what it completely means, lol.

 

[ANIMAL1975]

Matching up pretty close to what Osirisblack said


https://www.neogaf.com/threads/possible-ps5-leak-info.1467805/page-7#post-253564893





I really think this will be a lot like what they did with the PS4 going to PS4 Pro with a doubling of the GPU so a 72 CU GPU at 1.121 would be 10.33 TFLOPS

Hope that Osiris leak is the closest to the real thing.

Need a Hellpoint Dev to make sense of this

Damn! Beaten like a three legged turtle.

 

[Shin]

[snip

]
Nothing new in either Tweets except 1 unknown person correcting the other about it being a MS SoC.
Pretty much everything in those 2 I already covered and/or corrected in my previous posts so yeah...at least the SoC itself is new info altogether.
On top of all this if it's called Gonzalo that's just something MS would run with as they tend to use US/Mexican naming schemes.

 

[DeepEnigma]

]
Nothing new in either Tweets except 1 unknown person correcting the other about it being a MS SoC.
Pretty much everything in those 2 I already covered and/or corrected in my previous posts so yeah...at least the SoC itself is new info altogether.
On top of all this if it's called Gonzalo that's just something MS would run with as they tend to use US/Mexican naming schemes.

I thought it was Ariel, now it's Gonzalo?

 

[Barakov]

3.2 is actually pretty high if you ask me.

Well, Well. So you can be optimistic.