Actually, I am a software developer and work for a major Cloud provider, so I have some clue when it comes to software development and computing hardware. Admittedly not a game developer, but this discussion is hardly a deep dive on low level API development or hardware design.

As for being a 'Pony' fanboy, I actually own a PS4 PRO, Xbox Series X, Gaming PC, and a Switch. However, I will freely admit the Xbox Series X was a shit purchase as everything on it I can play on my PC. However it is by far the best console in terms of hardware design. Microsoft really needs to push console exclusives for next-gen for me to bother with the Xbox Series X.

You are correct that the 3.5 GB would likely be used for functions that do not require higher memory bandwidth, but this fails to account for the fact that PS5 has no such limitation, and that the 3.5 GB will still be slower. What happens if a game requires 12GB of really high speed memory for it's engine? 2GB of that will then be faster on the PS5. That all being said, Xbox Series X is obviously the faster solution, as I even mentioned in my original post, however it would be dishonest to claim that the Xbox Series X has always the faster memory.

The reason I claim that the difference between the two in regular graphics rasterization is that we have the numbers right before us. We don't have to guess anything. The Xbox Series X has no "secret sauce" that will allow it to perform better then the raw numbers indicate. We have the specs right in front of us, performing performance differential calculations between the two is hardly rocket science.

As for developers claiming 30% difference, how about you back that statement up? This is the second time I am asking you.

As for SSD performance, the difference between Kraken and Velocity is meaningless as we have the performance numbers for post compression for both. It's a 70/80% advantage to the PS5. This is not up for debate. Velocity might very well be the better software compression, but that hardly matters when the PS5 SSD hardware itself is more then twice as fast as the Xbox SSD. Software cannot make up that difference itself.

I have always claimed the Xbox Series X is the faster console (other then SSD), but the difference is smaller then previous consoles and is hardly going to make a difference.

Please post any evidence of developers claiming otherwise. Oh look a developer just stated they see little difference between the two:

https://wccftech.com/keoken-on-next-gen-we-dont-see-many-differences-between-ps5-tempest-will-free-cpu-resources/

@greygoose12: I'm not believing any PR. Both companies have put out technical information that anybody with a basic comprehension of maths and computing architecture can compare. The Xbox Series X is obviously the more powerful console, but the difference is nowhere as big as some make it out to be. Its far less then the Xbox One and Ps4 or the Xbox One X and Ps4 Pro. A game that is 30 fps on the Ps5 is not going to push 60 for on the Xbox Series X. A 30 FPS game on the Ps5 would be 35 FPS on the Xbox Series X at best, assuming the same settings.

This is not random guesswork, the technical specifications are out there for anyone to read.

The Xbox series X is obviously going to use more then 10GB of RAM for games. Microsoft themselves has already confirmed this themselves: https://www.eurogamer.net/articles/digitalfoundry-2020-inside-xbox-series-x-full-specs

games get a total of 13.5GB in total, which encompasses all 10GB of GPU optimal memory and 3.5GB of standard memory

Thus, 3.5 GB of that RAM is going to be 33% slower then the PS5. Sure developers can choose not to use it, but then you only have 10GB vs 13.5-14GB usable in the PS5, which would mean the Xbox series X would have lower quality textures. Not gonna happen.

Additionally, 2TFLOP by itself is a meaningless number, the important thing is relative performance difference, which is only 18% and likely a bit lower due to Amdahl's law.

The Velocity statement is Microsoft just trying to spin that they too have a SSD and enhanced SSD speeds. It's impressive no doubt, but it's utterly blown away by what is in the PS5. Sony has dedicated I/O processing (on top of the SSD controller), enhanced flash translation layer, and a SSD that is more then twice as fast:

https://www.anandtech.com/show/15848/storage-matters-xbox-ps5-new-era-of-gaming

Give that article a good read, it goes into technical depth on both hardware solutions.

Pretending that the Velocity architecture is going to make up the difference is laughable when we have solid performance numbers for both. Feel free to post conclusive evidence proving me wrong.

Lastly, no developer has stated the Xbox Series X is 30% faster. That is obviously not true to anybody that has a brain and a calculator. Point me to a credible developer claiming this so I can laugh at them.

Sorry mate, 15% faster is the best case scenario for the Xbox Series X, which is frankly a tiny difference.

@tormentos: You can't compare the two, a 5700XT is RDNA on regular 7nm, while the PS5 GPU is RDNA2 on 7nm Enhanced. In addition, the 5700XT is slightly larger at 40 CU.

Even so, most 5700XT's can reach almost 2.1GHz stable when OC'd on the regular coolers (so not talking about Nitro):

https://www.techpowerup.com/review/asus-radeon-rx-5700-xt-tuf-evo/34.html

2079 MHz for the one.

So taking into consideration the slightly smaller GPU, improved process node, and newer RDNA architecture, 2230Mhz does not sound unreasonable. But the power draw would likely be quite high - hence the use of AMD SmartShift. If you need to max your CPU or GPU then one would limit the other to stay within the power envelope, it doesn't mean that both cannot run at max sustained speeds when required.

To give a example, suppose you are running a GPU intensive scene, tons of enemies on screen, shits going down in a epic way, so you need maximum GPU power! AMD SmartShift ensure the GPU is running at that 2230Mhz at all times while downclocking the CPU a bit (3.2 seems a likely lower limit). If you need intensive CPU calculations and not 100% GPU then the reverse happens, CPU is maxed to 3.5Ghz, GPU drops down to 2.1Ghz (or whatever).

In the scenario that you need to max the GPU and CPU at the same time, then AMD SmartShift (and likely other software mechanisms Sony has implemented) will adjust the CPU/GPU speeds for maximum performance at that paticular time. Increasing the CPU to 3.5GHz if needed or the GPU to 2230MHz otherwise, or even 98% of both. You have a total system budget and that gets allocated to the required resource that needs it the most.

However, and take note, as the vast, vast majority of games can be GPU limited (especially in todays 4k world) the GPU would be the component that would likely run at max clocks the majority (if not all) the time. Just as Mark Cerny said.

For the Xbox side of things, the reason they likely did not implement AMD Smartshift, is that due to the clocks and size of the GPU, they would not get much benefit out of it. That 52 CU monster is going to dramatically increase the power budget for a very minor increase in Mhz. Rather save the die space and lower costs.

@greygoose12: The problem with that is that a 52 CU part at 2230Mhz would be dramatically hotter and use way more power then the 36 CU part in the PS5. I'm not even sure a GPU of that size and clocks is achievable in a console form factor. The only reason the PS5 can get to that frequency (in a rather large console form factor) is due to the rather small size of the GPU.

Says who?

https://en.wikipedia.org/wiki/List_of_best-selling_PlayStation_4_video_games

All of the above numbers are also years out of date. Numerous titles don't get reported either. Uncharted 4 is likely closing on 20 million sold by now.

That being said, Nintendo 1st party games are usually excellent and have a devoted following. So they all usually sell great. I've got a number of titles in the list as well.

Likely just a joke. Just like John Oliver really doesn't want to f*ck that horse.

Breaking this down is rather simple. Lets compare the PS4 Pro vs the Xbox One X:

CPU: 2.1Ghz vs 2.3Ghz - 10% advantage to the Xbox One X.

GPU: 4.2 Tflop vs 6 Tflop - 42% advantage to the Xbox One X.

RAM Speed: 217.6 GB/s vs 326 GB/s - 50% advantage to Xbox One X.

RAM Amount: 8GB vs 12 GB - 50% advantage to Xbox One X.

Storage: The same, no advantage to either.

What was the result of all of the above? Checkerboard rendering on the PS4 Pro vs native 4k on the Xbox One X with minor graphical upgrades (Red Dead Redemption 2 and numerous others as an example. Games look very similar, and this was with a console that was 50% faster in almost all respects (other then CPU which has little impact on resolution).

Now lets compare the PS5 vs the Xbox Series X:

CPU: 3.2Ghz - 3.5Ghz vs 3.6Ghz - 3% - 13% advantage to the Xbox Series X. (I've lowered the PS5 CPU speed to account for AMD SmartShift)

GPU: 10.3 Tflop vs 12.1 Tflop - 18% advantage to the Xbox Series X. (GPU speeds should be at full bore almost all the time, considering the comments by Mark Cerny and other developers, and how AMD SmartShift actually works)

RAM Speed: 448 GB/s vs 10GB at 560GB/s, 6GB at 336GB/s (10GB 26% faster, 3.5 GB 33% slower, average looking at 11% advantage to Xbox Series X) (3.5 GB reserved by OS)

RAM Amount: 16GB vs 16 GB - The same, no advantage to either.

Storage: 8-9 GB/s vs 4.8 GB/s - 70-90% advantage to PS5.

So overall we are looking Xbox Series X being 15% more powerful the the PS5 in raw compute numbers, having no extra RAM advantage, and while having a massive disadvantage in SSD performance. But even this does not paint a clear picture. As your number of cores increase, so does your expected performance drop off. A GPU with 4000 compute units is not twice as powerful as a GPU with 2000 compute units. This is something called Amdahl's law [1], thus the amount of performance you can extract from the PS5 GPU is theoretically higher then what you can extract from the Xbox Series X GPU, as the PS5 GPU has 2304 compute units vs the 3328 compute units of the Xbox Series X. This will not have much impact, but the difference of a few % will be there. So you can conclude that the Xbox Series X GPU advantage is less then the 18% raw numbers would have you believe. Likely on the order of only 10%-15% at most.

So, to summarize, the Xbox Series X is on average 10-12% faster then the PS5 in available performance, while having a major disadvantage on the SSD front.

This is dramatically lower then the advantage that the Xbox One X had over the PS4 Pro. So forget all this nonsense about 4k120FPS or one console is going to "destroy" the other. Disregarding the SSD advantage of the PS5 (and that can be a major factor down the line, even for multi-plats) the raw numbers for a game looking the same on both platforms will roughly be as follows:

Xbox 3840x2160 (4k) vs PS5 3648x2052.

Big deal.

References:

[1] https://en.wikipedia.org/wiki/Amdahl%27s_law

@greygoose12: Well the difference between the two consoles is roughly 18% in raw GPU TFLOP numbers (likely less in real world, as performance with the number of SM units do not scale with perfectly linearity) and looks to be 10% at most in the CPU space. RAM Appears to be a wash between systems, and SSD is firmly in the PS5's favor.

All told, the difference between the two consoles is way smaller then the Xbox One / PS4 or the Xbox One X / PS4 Pro. My guess is for multi-platform games the advantage would be to the Xbox Series X, but the difference would be slight.

Xbox 3840x2160 (4k) vs PS5 3648x2052, hardly anything to get worked up over and something that is certainly going to have little impact on the sales numbers this coming generation.

Price, Exclusives, vendor lock-in, loyalty, and extra features (Gamepass, PSPlus) are all going to be way more important things.

Considering Pascal has way better power efficiency then Polaris, that would have just left us with a weaker performing NX.