Yep, MS knew the specs of PS5 and wanted to set the narrative
-
Ever wanted an RSS feed of all your favorite gaming news sites? Go check out our new Gaming Headlines feed! Read more about it here.
Xbox Series X: A Closer Look at the Technology Powering the Next Generation
- Thread starter Deleted member 5028
- Start date
No, they saw an opportunity to claim part of their machine was better and they took it.
Obviously it remains to be seen if the variable clock rates on PS5 end up being a good or a bad thing in the end.
looks like we'll be getting a demonstration of ray tracing on RDNA2 tomorrow
DirectX Developer Day Schedule - DirectX Developer Blog
Tomorrow (3/19) is DirectX Developer Day! Join the Microsoft DirectX team, along with partners AMD and NVIDIA, for a series of talks and demos covering the future of gaming graphics: The New Features and Unprecedented Opportunities of DirectX 12 – Jianye Lu The New Standard for Next Gen Games –devblogs.microsoft.com
A bit disappointed that there doesn't appear to be a DirectStorage-specific session, but I'm looking forward to this. Especially if it's possible to get a rough idea of how the 20 series GPUs can be further optimized and how RDNA2 GPUs compare.
How would variable clock rates be a good thing performance wise? They are there to keep the cooling under control, not to increase performance. This is also the reason MS stressed this so much during their presentation. That basically means that XSX cooling can handle any load (and hopefully stay quiet).
I asked this in a different thread and didn't get an answer, but would the PS4 Pro's higher fillrate explain why a lot of games had better framerates on Pro?
They did, the XSX has some serious cooling for sure. We will find out soon with real games :)
I'm not convinced it will be a good thing and will just add needless complication for devs, but I also don't think Cerny and Sony are stupid so I'm going to wait and see how it works in practice.
Now that you mention, yeah it makes sense.
Anyway, the difference back then was higher than it is now (PS4 Pro 55% over OneX, Ps5 22% over XSX)
Yes. It was very quiet running games and demos.
The dev kit was quite a bit louder, curiously enough, as it has more of the Xbox One X form factor without the more advanced cooling system I suppose.
DirectX 12 presentation. We'll be seeing new demos and a first look at RT on radeon gpus
DirectX Developer Day Schedule - DirectX Developer Blog
Tomorrow (3/19) is DirectX Developer Day! Join the Microsoft DirectX team, along with partners AMD and NVIDIA, for a series of talks and demos covering the future of gaming graphics: The New Features and Unprecedented Opportunities of DirectX 12 – Jianye Lu The New Standard for Next Gen Games –
devblogs.microsoft.com
Yes, high pixel fillrate allows the machine to push more pixels, but bandwidth is also very important in that regard. So even though the Pro had a better pixel fillrate than the X, because of its' very low memory bandwidth, you saw games like RDR2 running at 100% higher resolution on the X even though the power difference between them was just 40% and the Pro had a better pixel fillrate. It doesn't matter how powerful your water pump is if your water pipe isn't wide enough.
I'm not too concerned myself, mostly because I have fairly low expectations of the first round of games on any new hardware. They're usually held back by a lack of experience building on the new hardware, so I'm not feeling like I'll miss out on much.
Same. I think it's also why I find it pretty exciting that the games we are used to playing now, for the most part, will just carry over. If you don't like the first round of new offerings, play the stuff you know. maybe even with benefits here and there.
Does the Pro's pixel fillrate explain why a lot of games have better framerates on Pro vs One X?
Facebook Gaming
Watch live video game streams from popular creators on Facebook. Join a community of players and streamers.
mixer.com
The higher res on One X is the reason. If you would keep the res as it is for the Pro the FPS would be stable and even go beyond then if not capped.
It's usually because the X version aims higher in terms of resolution/power ratio. It's like if a game will target 1080p on Lockheart, it might perform better than the XSX native 4K version because the resolution difference is X4 while the power difference is X3.
The next generation leap of even having SSDs at all in both of these consoles, let alone custom designed I/O approaches that are more efficient for gaming workloads than anything in PCs currently due to the closed console system architectures, is the thing that makes it possible for game design to change. This quote below from one of the Xbox Series X system architects shows just how far they went looking at actual game profiling data (they had custom monitoring hardware in the Xbox One X silicon) to be sure that their end to end storage solution would meet the needs of how games actually run. And remember that the memory bandwidth difference (112 GB/s) between the two is far greater in terms of GB/s throughput as part of this pipeline than a several GB/s difference in drive decompression speed.
looks like we'll be getting a demonstration of ray tracing on RDNA2 tomorrow
DirectX Developer Day Schedule - DirectX Developer Blog
Tomorrow (3/19) is DirectX Developer Day! Join the Microsoft DirectX team, along with partners AMD and NVIDIA, for a series of talks and demos covering the future of gaming graphics: The New Features and Unprecedented Opportunities of DirectX 12 – Jianye Lu The New Standard for Next Gen Games –
nice!
I'm trying to figure out something maybe someone can help. The 76mb of SOC cache listed for series X, how would that be allocated? Going by what we know about RDNA and the Xbox audio block I don't see how those could be using more than ~12Mb of cache combined. What is using the rest of the cache? Assuming they are using the full-fat Zen 2 cache that leaves 32mb of unaccounted for. What am I missing?
Peak fillrate only matters if you are actually bound by it, as with everything. If that's not your bottleneck it doesn't really matter.
In most passes you're more often memory or (if you've done your job well) ALU bound.
EDIT: Fillrate might even become less relevant next gen. More and more passes are moving to compute which doesn't use the ROPs for its output. RDNA allows compute writes without decompression in most cases so there's no major drawback in doing so and it can even be faster than a standard write going through ROPs.
Last edited:
Are you counting all levels of RDNA cache? It looks like you aren't including L2 and L1 cache for Zen.
It's possible there's on-die SRAM for the memory interface, like Sony's SSD interface.
Yeah. There is still a lot not yet known about the specs of these systems. Maybe a SIGGRAPH presentation or a dev doc leak.
I hope all their wireless interfaces will change
Im playing a lot with headset now that my gf is always at home
controller disconnect a lot but while using headset connected on the controller I haveso much stutter it's insane. Can't enjoy it at all
I have an arctis 9x. I don't even speak how many disconnect I have if I use it wirelessly
Im playing a lot with headset now that my gf is always at home
controller disconnect a lot but while using headset connected on the controller I haveso much stutter it's insane. Can't enjoy it at all
I have an arctis 9x. I don't even speak how many disconnect I have if I use it wirelessly
I forgot about the L2 cache, but there is still stuff missing.
My estimates:
3.25mb L1cache (64kb per CU 4x vega 56)
8mb L2 cache for GPU (2x vega 56)
672kb for the audio block (4x xb1)
256kb for CPU L1 (32 kb per core)
4mb for CPU L2 ( 256 kb per core)
32mb for CPU L3 (shared)
That leaves 27.85 unaccounted for. Mem cache?
Why are you referencing Vega? Navi white paper has cache values in it.
There's cache for the memory interfaces, and maybe the audio processor too. Do we know if that's on the SoC die?
Guys what's the current understanding on Ray Tracing capability for these devices? They way I understand XSX setup is they can achieve upto 13 TF worth of RT in parallel to the GPU workload. But the PS5 setup confused me during the presentation?
They're the same. PS5 has proportionally less because it has less TMUs.
The 13TF is just how much CU worth of compute they'd need to do the intersection testing the intersection engines do. They will fight for memory bandwidth though with the normal rasterization workloads.
XSX gives you 560 for 10 gigs..335 for 6..PS5 is 448 for 16. I don't the edge is that signficant in terms of bandwidth.
10.28 is argued as the typical clocks which fluctuates under load. Given how the architecture is supposed to work, that would be a more accurate metric to go by than 9.2 tf going by the specs on paper. Need to see how it performs obviously.
Ah yes, that's right. I think they did say just that - the devkits have 56 active CUs. The retail systems have four disabled to improve yields, I believe.