General Next Generation Rumors and Discussions [Post GDC 2020]

I don’t understand. Why would potential PS5 yield issues create heat issues?

The SOC has a defined power limit that the cooling system is built around. The system is designed to mitigate stress caused by pulling too much power.

Any yield issues should show up as frequency clock issues not increased heat production.

It makes me take any heat related rumors with a grain of salt.
 
Sony absolutely had a design for a ~ 9.x TF SOC based on Ariel.

And we've been hearing about 13 TF PS5 targets since early 2020, like around the Series X reveal.

Exactly. That leak for Ariel seems exactly right except clock speeds. And yup sony doesn't act on rumors , once MS made their announcement and the line in the sand was drawn so to speak Sony would then know how their machine stacked up. Spencer in the past has said that they know what type of tech is out there and from what companies and so they can reasonably assume what their competitors can release.

It could have been that Sony was caught off guard by MS willing to go with a unique case design in the console world and such a large apu and then differently clocked ram.

I have no idea. I have no idea other than what some personal friends have told me about the ps5 and yields and cost. So they could be a 100% correct or 0% correct. I trust the sources because they have been right before. However I freely admit that my MS sources are much stronger than the sony ones and that of course any source I have can be wrong.


But its intriguing at least. Sony planned for a 9tflop console with super fast ssd. Were caught off guard higher tflop xbox that microsoft announced. Chip redesign is out of the question so what are the options ? redesign cooling and clock higher ? of course both things can negatively effect pricing and availability. But perhaps for sony that was better than having a much less powerful machine ?

To be clear my sources have only told me that yield is bad on the sony product and that costs are higher than for MS because of that and other reasons I listed in my original post. I don't know anything about a console redesign or any delay.

I don't even think a delay of a few months would impact sony all that much in the long term.
 
in his presentation Cerny said they capped the clock at 2,23Ghz, implying they tested even higher clocks.
just looking at the dev kit design shows that they were planning to go high.
 
No one is talking latest benchmarks or latest designs.

I thought someone was disputing Sony having a 9.x TF design at some point.

In retrospect, I think the 9TF design shows that their strategy was always very high clocks. I remember how many people here (myself included IIRC) were in disbelief at 1.8GHz when Vega couldn't dream of those sorts of clockspeeds. What was it, about a year before we knew anything about RDNA?
 
Just because Sony is using MS for their future PSNow doesn't mean they are using the azure blades.
Could azure emulate the PS5 even, with everything that that entails. Its more likely that they use the data centre infrastructure with PS5 hosted hardware.
The blades are only part of the end to end infrastructure required.
Why didn't MS run xcloud in emulation on azure, or why would they host xsx in the cloud, but Sony use PC emulation.
 
"We remain absolutely on track. We're going to launch this holiday, we will have a global launch... We're not seeing anything saying to us there's going to be problems with quantities or we won't be able to launch the way we want." - Jim Ryan to CNET 7 days ago
 
just offering an explanation, I don't necessarily believe everything I write. But I also don't believe that makes experience infallible either. Sometimes you make a design for X, and someone realizes it needs changing because of Y and a whole design change is made. A lot of last minute changes can lead to problems.
100%. You absolutely don't want to have to be revising a design at the last minute. Equally plausible explanations for unneeded revisions is you hit a design that works just fine, yields are good and cost/budget are within tolerance but your chip designer (AMD) and/or fabber (TSMC) tell you there is scope to improve things if you try further refinements, so if you're willing to bite those cost and that pans out, you run with them. If not, you revert to the earlier design that you know works.

But who knows, maybe Sony/AMD/TSMC screwed up. We'll know soon enough. :yep2: I give AMD and TSMC more credit than some weird unexpected last minute issue rather than a bad run. They happen, TSMC just eat the cost, and re-run. Or if the wafers are defective the wafer supplier eats the cost.
 
Sort of a variable target though. If you are looking for the cream of the crop chip and not getting as many, you're bound to have poorer yields than a lower binned chip.
Aren't console chips some of the lowest binned chips there are? Even "cream of the crop" means bottom of the barrel when compared to desktop chips.
 
Aren't console chips some of the lowest binned chips there are? Even "cream of the crop" means bottom of the barrel when compared to desktop chips.

They used to be, when not aiming for stratospheric speeds.
 
in his presentation Cerny said they capped the clock at 2,23Ghz, implying they tested even higher clocks.
just looking at the dev kit design shows that they were planning to go high.
I mean they were already going to use adjusting frequencies to deal with BC so why wouldn't Sony just find another use that kind of tech for another valuable purpose ?
Speeding things up is pretty much always a good thing if you can do it at the right price point.
 
Aren't console chips some of the lowest binned chips there are? Even "cream of the crop" means bottom of the barrel when compared to desktop chips.
If 2230 Mhz is lowest binned, it's going to be an exciting year for Big Navi and AMD.

The 5700XT 50th Anniversary edition: 1980 Mhz boost with 40 CUs
5700XT is 1905 Mhz boost with 40CUs
5700 is 1725 Mhz boost with 36CU (4 disabled)

All of them on the 7nm process.

I would be shocked if the 7nm+ node would move the 'bottom of the barrel' clock rate to 2230Mhz, 230 Mhz over the cream of the crop PC currently.

With liquid cooling they can get the 5700XT to 2.1/2.2GHz. @ 210 W

With the “EvenMorePower5700XT” registry modification installed, we found 2.1 GHz to be the limit of both our Asus and PowerColor reference cards. We tried a few other tweak profiles but the same 2.1 GHz hard limit on our hardware was observed.

https://www.tomshardware.com/reviews/amd-radeon-rx_5700-rx_5700_xt,6216-6.html
AMD says Navi has a maximum junction temperature of 110°C, so we must imagine that throttling would happen many degrees below that point.
...
So when we observe measurements in the 80° range, we’re typically looking at junction temperatures in the 90s.
Honestly, not sure how many chips can operate at that temperature for 6-7 years for long gaming sessions. So the cooling will need to be really good to keep it cool.... except that those frequencies are of course boost clocks.

And boost clocks are going to drop more as the load gets heavier:
You can see this with division and tomb raider taking the clock downwards to 1760 Mhz.

https://www.anandtech.com/show/14618/the-amd-radeon-rx-5700-xt-rx-5700-review/15
I1KfTeb.png


So we got this situation where Cerny has declared boost to be 2230Mhz _almost_ all the time. Standard reference Navi's can't get above 2.1GHz. They require liquid cooling. And then its eating quite a bit of power; Cerny cites that if the CPU needs more power, it's only going to lose a little bit on the GPU side.

Honestly perplexing to me. They have flipped everything I have understood about clocks and voltages with PS5.

When the rumours started coming out about heat and yield issues; I mean, this sort of lines up with some of the challenges they would face when I look at the above behaviour for the 5700XT series.

And then on this line of thinking the 5700XT is marketed as the ultimate 1440p video card.
https://www.amd.com/en/products/graphics/amd-radeon-rx-5700-xt
Designed from the ground up for exceptional 1440p performance and high-fidelity gaming.
Then I guess coincidentally PS5 hits about 1440p on UE5. The only benchmark I have of the console.

So to me, it sounds like it's pretty darn close to a super heavily overclocked 5700. (4CUs disabled) And it would be no doubt confirmation bias for me to agree with rumours about heating issues, yield issues, or expensive cooling solution issues.
 
Last edited:
With the “EvenMorePower5700XT” registry modification installed, we found 2.1 GHz to be the limit of both our Asus and PowerColor reference cards. We tried a few other tweak profiles but the same 2.1 GHz hard limit on our hardware was observed.

What's the point of comparing to RDNA 1 ?
 
What if it's not only Him desinging the console? I highly doubt He's the sole man that can do what he wants with a console that's going to sell over 80 million units.
It's not only him, but he's the chief architect so he signs off on every final choice and tells the underlings what console to make. The only way someone else could be involved would be his boss reacting to MS's TF figure and demanding PS5 is faster and Cerny saying, "you canna break the laws o' physics!" That'll be John Kodera or Jim Ryan, depending on timeline.

It's a chance, of course, but an outside one.
 
Will someone leak if only a tiny bit of Horizon sequel for me! I have a feeling Horizon 2 would feature a lot of aerial battles thanks to the SSD's extreme data streaming rate. Avatar's final battle is the closest comparison that comes to mind and I'm tipping the graphics would come close to it too.
 
Back
Top