wccftech.com: Why The PS5’s 10.3 TFLOPs Figure Is Misleading

Status
Not open for further replies.

takriel

The Fallen
Oct 25, 2017
5,376
User Banned (1 Day): Platform warring
Some people really can't stand the thought that the PS5 might be inferior to the Xbox SeX this time around huh?
 

Gemüsepizza

Member
Oct 26, 2017
1,449
Meh, with that SSD it's basically 30 TFLOPS anyway so it doesn't matter.
Show me please one post where somebody claims that the SSD is doing any graphics processing. What is being discussed is, that a faster SSD allows loading more high-quality or more varied assets, for example. If this is the new level of discourse here, then this is kind of sad.
 

modiz

Member
Oct 8, 2018
9,289
I generally trust Usman, but i dont believe that we know enough about RDNA2 to deduce what clocks can and cant it hit. Only things we know is dropping the clock by 2% drops the power consumption by 10% on the PS5 and that RDNA2 brings with it a 50% performance per watt increase which is a major improvement and actually falls closely in line with the clocks.
 

gofreak

Member
Oct 26, 2017
3,836
if the gpu is indeed 9.2 tflops then the difference in resolution will be 30%. we will find out come launch.

sony wont be able to hide real world differences just like MS couldnt last gen.

i dont see why cerny would be lying but this is definitely new tech. we need to see it in action.
I respect Usman on his insider reports on AMD, but the rationale for why 2Ghz would be the 'sustained' clock is flimsy to say the least. He seems to have picked that number because it showed up in the leaked github test. That was one benchmark test.

The 'sustained' clock-rate will depend on the workload, no more, no less. It may be 2, 2.1, 2.2. Maybe it can even go lower! But there isn't 'a' sustained clockrate applicable to everything. It will vary by game. It will even vary within games.

I'd suggest one side point, and speaking of misleading - teraflops alone may be a misleading indicator of real world performance. Of course there'll be a gap, but I'd be surprised if the real world performance delta lines up neatly with the teraflops delta (16% or 30% or whatever is in between) - or if it is, it won't simply be because of the compute difference. More likely to be bandwidth. If you look at benchmarks of rdna1 chips at higher resolutions and settings, you rarely get real world performance gains between the lower and higher end of the family that are in line with the tflops gaps. It's often half as large as the paper gap. Bandwidth, however, is better on XSX, but narrow/fast vs slower/wide throws another complication into the comparison with e,g, rdna1 benchmarks. We will see where it shakes out typically.
 

Marsy

Member
Dec 4, 2018
180
It's bad enough I wasted my time reading that but I must ask; why do they have "3.5 Ghz (SMT?)" in their fact sheet? I thought it was made clear it had SMT enabled?

The only compelling info in that article was the discussion of the silicon...

Essentially unless its N7+, under current manufacturing process it can’t reliably hit the boost speeds. I hadn’t seen this discussion before
It's about as compelling as forum post arguing about their favorite company countering with an even lower process node. It's not.
 

space_nut

Member
Oct 28, 2017
2,315
NJ
Just a 3% or greater goes into 9tf territory. Then if a game is only utilizing 8cores then XSX has 3.8ghz while ps5 can go below 3.5ghz to say 3.3ghz which would be a 15% cpu improvement

Can’t wait to see DFs new capture tools to show results
 

Hellshy

Member
Nov 5, 2017
610
Also don't forget that the exclusive DX12 on Xbox will give it another 200% boost and it's effectively like a 3 generation power difference. The PS5 is actually weaker than the N64 if you read between the lines of Cerny's lies.

I should become a writer for one of these sites.
Is dx12 actually exclusive. I remember people claiming things like that backn when this gen launched but it turned out to not be the case.
 

Unaha-Closp

Member
Oct 25, 2017
3,112
Scotland
I give no fucks about numbers other than the number on the end of a game that I have been looking forward to playing and the number of years I've been waiting for it. The games will speak for themselves - until then I will continue not giving any fucks about numbers. The number of biscuits left to snack on with a cup of tea is also a number I like.
 

klik

Member
Apr 4, 2018
709
People act like 10tf rdna 2 is too weak..

We have Ryzen 3.5ghz 8c/16t, 5.5gb/s SSD, GPU that is even better than Radeon VII/RX 5700XT.. Even for 499$ is an amazing machine.


Cerny should have ended presentation with some amazing next gen gameplay demo though, maybe that would make people stop comparing one another's teraflop
 

Hu3

Member
Oct 25, 2017
1,278
“Don’t attack the source”, even though the source is a clownfest, with a clear mix of rumors and non researched motions.
 

gofreak

Member
Oct 26, 2017
3,836
It's bad enough I wasted my time reading that but I must ask; why do they have "3.5 Ghz (SMT?)" in their fact sheet? I thought it was made clear it had SMT enabled?
They're suggesting that because MS alters their clocks with SMT enabled, maybe the PS5 does too.

We don't know - but I'm guessing the workload management stuff encompasses that on the PS5 side. I think naturally you're more likely to run into the thermal/watt limit and incur clock variation with high load SMT work (coupled to high load GPU work) than high load single threaded work. So I think they're probably letting that manage it, rather than having discrete clock steps when you have SMT enabled or disabled.
 

Jroc

Member
Jun 9, 2018
1,463
If the PS5 can run maxed out 99% of the time, and the 1% base clock drop is negligible, then why even bother labeling the specs as variable?

Surely 10TF Locked (10.28TF Boost) looks better than putting an ambiguous 10.28TF (Variable).
 

pswii60

Member
Oct 27, 2017
12,905
UK
The proof, as always, is in the pudding.

But I'm interested in seeing the cooling solution for a GPU that many believe will be operating at its peak 2.23GHz clock "almost all of the time". Even Nvidia's highest end GPUs run at way lower clocks when overclocked to their peak and with significant cooling solutions. Many doubted the GitHub leak because it suggested that the GPU would even be running at 2GHz never mind 2.23.
 

ploonkey

Member
Oct 25, 2017
138
PC GPUs have variable/boost clocks and factor those into their TeraFLOP calculations, too. What exactly is different here?
 

AllChan7

Member
Apr 30, 2019
1,271
How is it misleading when Cerny explained it? They aren't hiding anything lmao

I really want this to end and everyone just accepts next gen is gonna be incredible regardless of what console you get. The fact devs are extremely happy with both consoles should be enough
 

gofreak

Member
Oct 26, 2017
3,836
Some people really can't stand the thought that the PS5 might be inferior to the Xbox SeX this time around huh?
PS5 is inferior to XSX in GPU compute at either end of the range. I don't think anyone disputes that. And so it doesn't really matter wrt XSX superiority - it is superior here in any case.

The dispute is over the arbitrariness of the clock picked in the article, or what the rationale for that was, if there was one. Especially when the article seems to ignore the key suggestion of the approach about why close-to-max clocks may be 'typical' rather than rare. Cerny may be wrong about how typical that will be, but there's little in the article to argue why. (And, in fairness, I'm not sure anyone could argue why, without access to the kind of workload data even only Cerny has for the prior generation, and nobody yet has for the next).
 

Aquavelvaman

Member
Oct 25, 2017
1,991
How is it misleading when Cerny explained it? They aren't hiding anything lmao

I really want this to end and everyone just accepts next gen is gonna be incredible regardless of what console you get. The fact devs are extremely happy with both consoles should be enough
Not going to happen, gamers always have to be at war. The next console gen is going to be the best yet, though.
 

Munstre

Member
Mar 7, 2020
51
Even after the PS5 is out and we are playing amazing looking games on it, people will still be arguing that it's all a lie and using github as proof.
 

Bluelote

Member
Oct 27, 2017
1,221
with Sony only publicizing max clocks in a variable clock machine, they are not wrong at all in questioning it...
 
Status
Not open for further replies.