It’s easy to forget just how powerful the SX will actually be. It is beyond all expectation from about 18 months ago. I personally think it will render any talk of power, performance or multiplat comparison to be minimised because these things will not be in favour of the PS5 at all.
Written article available here for those that prefer to read these sorts of things.
Is it wrong to assume that XSX has 52*4 = 208 RT cores?
And then if multiply 208 RT cores with there frequency of 1.825 GHz i.e,
208*1.825=379.6 ~ 380
This calculation is equal to what RT performance is mentioned for XSX.
Will it be wrong to apply the above formula for the 2080ti’s 68 RT cores running at 1.545GHz max frequency reported by nvidia as well? I.e, 68*1.545= 105
I got the impression that there are 52 Rt cores each able to perform 4 ops per clock. But I dunno about Nvidia setup.
It really is the dream wish list of specs: An 8 core CPU clocked up to 3.8Ghz, high speed RAM, SSD, ray tracing and 12tf of GPU power. The only disappointment is that we didn’t get a massive leap in memory but unfortunately that was not on the cards this year.
I made a thread on digital foundrys xsx hotchips deep dive, maybe it would be a good idea of that thread was merged into this one?
I’m kinda on the fence with Tommy Fisher. He nailed the XSX specs. No one was talking about 52CUs, and he nailed the speed as well. I think he had some kind of inside info, but he was quoting the original GCN PS5 dev kit which were apparently around 13tflops.
Reminder: No insider drama friends. This isn’t the forum for checking receipts on insiders.
The whole system seems so well designed and put together. The design flair is definitely on the inside.
I guess the process is they make a chip, each CU gets tested. Then they bin the ones that don’t have enough usable and keep the rest. Couple of silly questions.
If a chip has more than the required usable does the chip know this and can switch to one if one fails?
Does binning mean literally throw them in bin it could they be used elsewhere?
For Binning, if you look at pc GPU’s, this is how you get the 3080 and the 3080ti. Some are capable of higher speeds or have less production faults so you can use them to the fullest capacity. Those that do not work for a lesser product are then pulled together after testing.
If you have enough that fit inside a category envelope, you can use them in other products or potentially design around the restrictions. You have to pay the same amount for a wafer of chips, no matter how many fail so utilising as many as possible is how they can pull costs down per chip.
For the failed CU question, this is outside my knowledge wheelhouse unfortunately. It may be worth throwing the question to Richard Leadbetter of Digital Foundry
The XSX is really such a well designed console. MS redeemed themselves on the hardware front with the Xbox One X, and the Series X is just an evolution of the amazing work they did on the XOX. I owned the PS4 Pro, and while it was ok, it was never the quality of hardware the XOX was.
I now really hope MS and the third parties can exploit all of the XSXs abilities.
Microsoft was always about the most powerful console, and they nailed that on the OG Xbox, the 360 and now the XOX and XSX. Now lets get that software nailed bois.
Believe it or not but I was actually a believer in klee, he seemed quite credible, he was former writer for a quite popular gaming mag. I was all set the PS5 was homage to crazy ken and it was going to be some 14tflop beast. It was kinda funny and shocking that its a 9-10tflop machine with a variable cpu + gpu.
Watching the Digital Foundry breakdown of the Hot Chips presentation makes it even more apparently that the backwards compatibility team is the centerpiece of Xbox hardware design
Those people are freaks of nature and Sony should be doing all in their power to poach them
I was thinking there would be an actual number in NV metrics for RT performance in the video.
NV should list those raw numbers with the ray cast/s lol.
Series X dev kit with 40GB ob RAM
You are still a bit generous with 9-10tf, better get used to 8,4 wait till Sony reveal the base clock.
To write for a popular game magazine doesn’t mean you understand the context of information given. It was clear from the beginning that for a correct understanding he lacked the foundations. Just like with Reiner and Schreier.
That is what happen frequently unfortunately everywhere. People use their authority to prevail on the logic authenticity of an information.
I tend to agree. I remember AV Watch’s Interview with Mark Cerny about the PS4 Pro. Mark responded to the question about True 4k graphics and comparing the specs against Project Scorpio. Paraphrasing at best here but he said that to achieve 4k resolution at current rendering fidelity, you would need 8 tera flops at a minimum.
I don’t personally think that the PS5 was changed in clocks as a reactionary step against the Xbox. I believe that the technology group designing the PS5 had the current design as a option that they could target. It involved stepping the target release date back and they made the call so they could make the changes needed. @Colbert did his lovely poem sometime ago that talked about Sony dealing with heat issues. That, If he can expound upon what happened, Could be the biggest part to the story of why the PS5 is a 8-9 TFlop machine.
With everything coming out about the consoles, It is hard to compare apples to a Kiwifruit that is locked in a safe at the top of The empire state building. With nothing coming out from Sony on a full breakdown, It may indicate a unwillingness to allow the public to see that Xbox have the advantage in hardware because while the faithful will still stand behind SSD speeds, It hamstrings any potential to put out a marketing message targeting the casual market to overlook that the other box is prettier.
These two twitter threads go into some crazy detail (at least in my mind lol) over all of this. Some good reading.