Microsoft internal emails show Xbox executives reacting to the strengths and weaknesses of the PS5 tech specs

Bernoulli

M2 slut
Microsoft internal emails show Xbox executives reacting to the strengths and weaknesses of the PS5 tech specs

The emails show Liz Hamren, former head of platform engineering and hardware at Xbox, pointing out the strengths and weaknesses of Sony's PS5 specs compared to Microsoft's Xbox Series X. PlayStation hardware lead Mark Cerny detailed the PS5 specs just two days before Microsoft publicly announced its own Xbox Series X specs.

Hamren lists the variable GPU and CPU clock rates of the PS5, "versus us running at higher sustained rates." Hamren admits Sony has a clear advantage on SSD performance with the PS5:

Cerny talked at length about the move to SSDs and the advantages for game developers and consumers. They have optimized for raw higher raw throughput (2x ours with slightly better hardware compression and associated performance improvements) as opposed to a more integrated streaming architecture enabled by Sampler Feedback Streaming.


Elsewhere, the 12 teraflops of performance versus Sony's 10 teraflops was also briefly discussed. "[Cerny] emphasized that GPU teraflops and CU is not a good measurement of performance. We made this same point with Digital Foundry, but we do have a clear performance advantage (12 v 10)," wrote Hamren.

Ptqnfvp.jpg


The former Xbox exec also said that Cerny "spent what seemed like a disproportionate amount of time on audio innovations," but the rest of this discussion is redacted from the email chain. Hamren also claims Sony's expandable storage solution for the PS5 is "similar to us in reality due to minimum size and speed requirements." But in reality, Sony's approach means consumers can use any regular NVMe SSDs, instead of the proprietary and often more expensive Xbox drives that are only available from Seagate and Western Digital.
 
Last edited:
If they really did interally equate 10 v 12 TFLOPs = "clear performance advantage" then I'm actually not that surprised the difference between the two consoles doesn't show up to more than a negligble extent

Edit: By this I don't mean the compute is neglible, but equating TFLOPs to performance is kind of wierd when there are other factors involved such as API/tools(lol)/hardware customizations/etc
 
Last edited:
no that's Mark Cerny
MS say we have a clear performance advantage 12 vs 10
I read it as them saying something along the same lines to DF, but in the email it immediately turns back to that ever holy TFLOP number so I really don't know where the author was going with that
 
Last edited:
Quite interesting that they appear to have found out what the specs were at the same time as everyone else. I assumed that Sony, Microsoft and Nintendo would have all been whispered the info by shared partners.
 
Quite interesting that they appear to have found out what the specs were at the same time as everyone else. I assumed that Sony, Microsoft and Nintendo would have all been whispered the info by shared partners.

Of course they knew because the specs for the PS5 SoC were more or less known public information by first half of 2019 thanks to benchmark/debugging leaks.
 
no that's Mark Cerny
MS say we have a clear performance advantage 12 vs 10
No...MS says "we made the same point with DF," in regards to GPU TFLOPs and CUs not being a good measurement for performance but they still said that they have a clear TFLOPs advantage. ie, yes, Cerny is correct that it isn't the be-all end-all of performance and we agree with him but we're still superior there.
 
Last edited:
Why do you ignore the start of that very same sentence, where they say the same thing as Cerny?

It's right there.
So they knew he's right, but immediately continued their 12V10 Tflop narrative.
No...MS says "we made the same point with DF," in regards to GPU TFLOPs and CUs not being a good measurement for performance but they still said that they have a clear TFLOPs advantage. ie, yes, Cerny is correct that it isn't the be-all end-all of performance and we agree with him but we're still superior there.
they are the ones ignoring it, after all that they say :

BUT we have a clear performance advantage 10 vs 12

not teraflops but Performance
 
it's pretty lazy honestly.
nothing deeper than your average forum post around that time. (probably where the person parsed the "summary" from, and that explains why they fell for the "clear advantage" bullshit narrative from same time)
i would have fired this person by now.
 
they are the ones ignoring it, after all that they say :

BUT we have a clear performance advantage 10 vs 12

not teraflops but Performance
Yes, we have a clear performance advantage when it comes to TFLOPs..."10 vs 12" is what? TFLOPs. It doesn't just stop at performance. They specify where they have an advantage.
 
If they really did interally equate 10 v 12 TFLOPs = "clear performance advantage" then I'm actually not that surprised the difference between the two consoles doesn't show up to more than a negligble extent

Edit: By this I don't mean the compute is neglible, but equating TFLOPs to performance is kind of wierd when there are other factors involved such as API/tools(lol)/hardware customizations/etc
Big part of it is that devs aren't going to spend a lot of extra resources on bells and whistles for a comparably tiny userbase.
 
they are the ones ignoring it, after all that they say :

BUT we have a clear performance advantage 10 vs 12

not teraflops but Performance
This is what they are talking about. Microsoft has made the same argument before that TF is not a good measurement of performance which is what the internal document says even though on paper, they have a clear performance advantage in terms of TFLOP 12 v 10.
qMxGE3O.jpg
 
This is what they are talking about. Microsoft has made the same argument before that TF is not a good measurement of performance which is what the internal document says even though on paper, they have a clear performance advantage in terms of TFLOP 12 v 10.
qMxGE3O.jpg
and after that they say that they have a clear performance advantage what's your point?
 
and after that they say that they have a clear performance advantage what's your point?
Do you even read? There is a (12 vs 10) in parentheses at the end of the sentence. This is where they have a clear performance advantage.
 
Last edited:
"We made this same point with Digital Foundry"

What I don't get, is why they then proceeded to imply a clear advantage anyway.
I mean they had no other talking point tbh, Xbox has always pushed 'Power' just about every gen. And they have small dick syndrome so bad that Im sure they see it as a victory.
 
This is what they are talking about. Microsoft has made the same argument before that TF is not a good measurement of performance which is what the internal document says even though on paper, they have a clear performance advantage in terms of TFLOP 12 v 10.
qMxGE3O.jpg
Then why did DF still act like "deer in the headlights" for almost a year or more with the face-off's, and kept scratching their heads when the PS5 was coming out on top or inline? Did they forget this memo? Or was it all an act early on?
 
Last edited:
Do you even read? There is a (12 vs 10) in parentheses at the end of the sentence. This is where they have a clear performance advantage.
Does your brain stop reading after performance advantage? Did you miss the 12 v 10 in parenthesis? Which is the teraflops advantage they are talking about.

Bernoulli Bernoulli 's point is that performance =/= TFLOPs and this MS staffer should be informed enough to know that.
 
they are the ones ignoring it, after all that they say :

BUT we have a clear performance advantage 10 vs 12

not teraflops but Performance

It's written immediatly after that they're talking about teraflops:
We made this same point with Digital Foundry, but we do have a clear performance advantage (12 v 10)

In any case the power difference was a total nothing burger and they knew that teraflops weren't a reliable metric.
 
Last edited:
Bernoulli Bernoulli 's point is that performance =/= TFLOPs and this MS staffer should be informed enough to know that.
This "MS staffer" was the head engineer. They are saying in that email that Cerny is correct that TFLOPs and CUs aren't accurate measures of performance but that they still hold the TFLOPs advantage. Or do you seriously think the head engineer is a moron who admits Cerny is correct and even MS themselves told the same thing to DF, only to contradict themselves in the very next sentence?
 
Last edited:
This "MS staffer" was the head engineer. He's saying in that email that Cerny is correct that TFLOPs and CUs aren't accurate measures of performance but that they still hold the TFLOPs advantage. Or do you seriously think the head engineer is a moron who admits Cerny is correct and even MS themselves told the same thing to DF, only to contradict himself in the very next sentence?

I don't think he's a moron. But I do think she gaffed with the wording, yes.

You might think engineers are infallible speakers (lol), but they're clearly not.
 
Last edited:
"more integrated streaming architecture" my ass.

SFS is hardly a revolutionary feature in streamlining the asset/data streaming pipeline, it just allows for more efficient texture streaming into the main memory. Not too different to what Epic are doing with Virtual Texturing in UE4 and UE5.
 
I don't think he's a moron. But I do think he gaffed with his wording, yes.

You might think engineers are infallible speakers (lol), but they're clearly not.
But they didn't...and that's not being "not infallible" that's not knowing something even a layman would be aware of. There is a 12vs10 in parentheses because that's where they advantage lies. There is even a "but" after the part where they said they told the same thing to DF. They have a clear performance advantage in TFLOPs is what it's saying.
 
"Yes Cerny is right but we have a clear (marketing) advantage (12 vs 10)"
Seems pretty clear to me.
Also lmao @ the SFS.
 
Last edited:
This "MS staffer" was the head engineer. They are saying in that email that Cerny is correct that TFLOPs and CUs aren't accurate measures of performance but that they still hold the TFLOPs advantage. Or do you seriously think the head engineer is a moron who admits Cerny is correct and even MS themselves told the same thing to DF, only to contradict themselves in the very next sentence?
in that case why not write we have a clear teraflops advantage?

or performance = teraflops now?
 
I think at this point it's pretty obvious that the overall throughput, compression and memory considerations are the main sticklers.
The CPUs and GPUs are not really what's really making up the difference.
 
in that case why not write we have a clear teraflops advantage?

or performance = teraflops now?
We've got the real-world results.

What he said could only mean Tflops and nothing else.
Actual performances of both console are basically on par.
 
Last edited:
Actual performances of both console are basically on par.
pretty-cool-cerny.gif


But man oh man, the marketing on forums were off the chain prior to launch. First it was the "tools." Then it shifted to "cross-gen games, you wait until next gen games" which was always a head scratcher to me, then it shifted to "VRS2.0 and other buzzword AMD features..."

This train never stops! Choo choo!
train GIF
 
so liz hamren started the whole variable clocks vs sustained clocks debacle

1693853677960

cant believe i fell for her..............................statement
 
Last edited:
Top Bottom