Current Generation Hardware Speculation with a Technical Spin [post GDC 2020] [XBSX, PS5]

[Betanumerical]

I also find it funny that Sony, while admitting that the PS5 is unable to run at max clocks at 100% ALU utilization (which is what it would take to hit the max TFLOP figure) 2.223GHz x 36 CU's x 128FP ops/clk, takes the liberty to round that ~10.25 result up to 10.3. While MS on the other hand simply discards .15 TFLOPS of actual compute and rounds down to 12 in their marketing.

I am not entirely sure that peak ALU utilisation is peak power draw. But it would be good to see some other people weigh in on this that have more knowledge.

 

[fehu]

It's because when the yields are known and they bump the clocks up to 12.85TF, they can just put 13TF instead.
Still April 1st.

Wrong! MS would never release a product with 13 in it.
Maybe if they increase the clock up to 13.2TF they would market it as 12.99+TF

 

[Deleted member 7537]

I am not entirely sure that peak ALU utilisation is peak power draw. But it would be good to see some other people weigh in on this that have more knowledge.

Well, his statement is not true to begin with.

 

[Shifty Geezer]

Well, his statement is not true to begin with.

Who's statement?

 

[iroboto]

Actually Nvidia states 1 to 2 GB. Consoles have 16. I would say its a fair ammount. RT on Xbox has to be done on the first 10 GB, so its up to 20% memory usage just for the RT structure. Wouldn’t it be nice if you could dump this, even if only part, to the ssd?
I would say so!

As for the SSD... not really. Just a faster SSD will not release you of all constrains. A 10 times faster ssd on the PS4 will only bring you 2x gains, and although proportions may change, this reality is common to all systems. To really take advantage of an SSD you need to get rid of a lot of other restrains. That was what PS5 did. Xbox has changes too, but as far as it’s public, not to the same extent.

We were talking about generic SSD gains. But we all very well know that tweet was a joke about the possible gains an ssd could bring to the ps5 over the X.
In fact this is the console forum!
And in that case, you cannot dissociate the fact that the SSD works in conjunction with those changes. That’s not something available on the PC space for a comparison to be made.

Also I would not say all SSDs would be enough to break any sort of I/O limit for some time. At least not in comparable ways. For instance, both consoles use dedicated compression and several optimizations on I/O.
Yet Microsoft games will support Xbox One. Can they really use these changes for anything meaningfull in game concept and design?
And when the current gen consoles are left behind? PCs cannot reach those levels of data compression without sacrificing CPU performance. They do not have dedicated decompressors and other custom changes.
And PCs are now part of the Xbox platform.
Heck, most PCs have no SSD, and most of the ones that have it, have 120 to 256 GB, half used with windows 10, and other installs.
I see complaints on The last Call of Dutty community over their last games, and the fact that with all the patches the game is now over 100 GB in size, and people simply do not have enough SSD.
Besides, most of them cannot even reach a 1 GB/s transfer speed.
So how will that work? Can we really compare those SSDs and the gains they can bring to performance and gaming design?

Any graphics work needs to be done on 10 chips, any 10 chips. Not 10 GB. There isn’t 2 separate pools unless they specifically state that specific chips are unable to give out their full bandwidth. As long as Xbox can freely leverage its 10 chips it won’t matter where the data resides. And no. You can’t dump this onto the SSD. It needs to leverage an enormous amount of bandwidth every single frame and likely do modifications and updates to the structure via asynchronous compute.

secondly. SSDs have the biggest leap this generation so far being 50 and 100X+ in difference from last generation. Whereas all other factors such as compute and bandwidth and cpu have barely doubled and up to quadrupled. So I would disagree with your statement there. The SSD could be 10x or 20X fast and unlock a significant difference already. The fact that we moved to 50 and 100+x respectively including compression to further it is dramatic.

PC space; there is a significant amount of CPU power available to do that work that consoles don’t have. So I would disagree about not being able to keep up. They have separate memory pools of which one can be significantly greater than the other. It can just grab data and decompress and send it over when it’s needed by the GPU.

or the alternative is to skip decompression and just have a significantly higher base system memory like 32GB.

It’s just that everyone needs to upgrade; and everyone will need to anyway the console specs are pushing on the high end of the spectrum.

 

[Betanumerical]

Any graphics work needs to be done on 10 chips, any 10 chips. Not 10 GB. There isn’t 2 separate pools unless they specifically state that specific chips are unable to give out their full bandwidth. As long as Xbox can freely leverage its 10 chips it won’t matter where the data resides. And no. You can’t dump this onto the SSD. It needs to leverage an enormous amino of bandwidth every single frame.

From my understanding the XSX's setup only contains 10 memory modules in total, to achieve 16GB with 10 modules MS used mixed density modules. To reach peak bandwidth you have to read from all of them at once. These 10 chips are all striped together on their lowest 1GB of address space making a total fo 10GB of memory.

The slower space comes from the upper memory space of the 6 2GB modules. These can only be striped with each other and not the smaller 1GB modules making up the 6GB of 336GB/s.

Thats where the 10GB fast space and 6GB slower space come from.

 

[Deleted member 7537]

Who's statement?

I also find it funny that Sony, while admitting that the PS5 is unable to run at max clocks at 100% ALU utilization (which is what it would take to hit the max TFLOP figure) 2.223GHz x 36 CU's x 128FP ops/clk, takes the liberty to round that ~10.25 result up to 10.3. While MS on the other hand simply discards .15 TFLOPS of actual compute and rounds down to 12 in their marketing.

They never said that.

 

[iroboto]

From my understanding the XSX's setup only contains 10 memory modules in total, to achieve 16GB with 10 modules MS used mixed density modules. To reach peak bandwidth you have to read from all of them at once. These 10 chips are all striped together on their lowest 1GB of address space making a total fo 10GB of memory.

The slower space comes from the upper memory space of the 6 2GB modules. These can only be striped with each other and not the smaller 1GB modules making up the 6GB of 336GB/s.

Thats where the 10GB fast space and 6GB slower space come from.

Some considerations:
There has always been an OS reservation and the game code will naturally take up memory as well.
In this case XSX has dedicated area of 6GB of memory for audio, file I/O and OS. But in this sense it’s not a separate pool because it’s not on it’s own. Where it physically resides I do not know. likely on the 2GB chips.

I think it’s very unlikely GPU data will suddenly Encompass the usage of the full 16GB such that it runs into a constraint issue. The SSD in the Xbox is still very fast and provides flexible options for them if they run into constraints. Not fast enough to crutch as additional memory but enough that they can play around with what they need. In addition to all sorts of other ways to bring in the right sized data closer to when they need it; they have more options to work with.

On a perspective: X1X manages 4K textures, 4K resolution with 326GB/s of bandwidth on 12GB and a very slow hard drive. 6 chips alone on XSX would exceed in bandwidth and equal the size.

if the argument is that I/O is the limiter for performance for games. I would say you can definitely design games in a way that it could be a limiter. But most games have also managed to work around that limit successfully. In the same way we have worked around the limits of limited VRAM for instance.

we will see how well the first year of Sony games will look. It’s on Sony to prove the drive performance for the first year of games. If they decide that it’s the right idea to leave behind PS4. MS has only committed to 1P titles being released on older hardware year 1. It doesn’t stop other studios from their ambitions.

 

[Shifty Geezer]

I also find it funny that Sony, while admitting that the PS5 is unable to run at max clocks at 100% ALU utilization...

But it presumably can when using less CPU. Dunno why some are making a meal of this, especially without the real-world device to consider in practice. All those specs we get like Bandwidth are peak theoretical, but the hardware is never used that much. XBox Series X is never going to perform 12 trillion multiply accumulates in one second in its life. The GPU is never going to process all 560 GB of its 560 GB/s bandwidth in one second, ever. Is MS's spec a lie, because in real life contention will stop the GPU accessing 560 GB/s? Course not.

Look at a graph of processor utilisation and you'll see plenty of holes. These specs, Flops and GB/s, are rates. If when the hardware needs to process the graphics maths for a fraction of a second, the GPU is able to process it at full pelt, then PS5 will be hitting 10 TF for that fraction of a second. Then, when it's got a different workload, the CPU can work full pelt. Somewhere between the workloads will be a balancing point of efficiency. Until we see what PS5's real-world balancing point and performance is, no-one should be making silly comments about engineering specs being lies.

 

[mrcorbo]

I think it’s very unlikely GPU data will suddenly Encompass the usage of the full 16GB such that it runs into a constraint issue.

To be most accurate there are two limits, though. There's a 10 GB limit of full bandwidth memory and an additional 3.5GB of lower bandwidth memory. FWIW, I'm not really worried about either the 10GB or 13.5GB being an issue.

 

[iroboto]

To be most accurate there are two limits, though. There's a 10 GB limit of full bandwidth memory and an additional 3.5GB of lower bandwidth memory. FWIW, I'm not really worried about either the 10GB or 13.5GB being an issue.

Thanks. Yea I wasn’t sure if it was 4.5 or 3.5 remaining after reservation so I didn’t write it.

 

[mrcorbo]

But it presumably can when using less CPU. Dunno why some are making a meal of this, especially without the real-world device to consider in practice. All those specs we get like Bandwidth are peak theoretical, but the hardware is never used that much. XBox Series X is never going to perform 12 trillion multiply accumulates in one second in its life. The GPU is never going to process all 560 GB of its 560 GB/s bandwidth in one second, ever. Is MS's spec a lie, because in real life contention will stop the GPU accessing 560 GB/s? Course not.

Look at a graph of processor utilisation and you'll see plenty of holes. These specs, Flops and GB/s, are rates. If when the hardware needs to process the graphics maths for a fraction of a second, the GPU is able to process it at full pelt, then PS5 will be hitting 10 TF for that fraction of a second. Then, when it's got a different workload, the CPU can work full pelt. Somewhere between the workloads will be a balancing point of efficiency. Until we see what PS5's real-world balancing point and performance is, no-one should be making silly comments about engineering specs being lies.

The way these all of these specs are being used is akin to looking at the "tale of the tape" before a boxing match and then using the reach and weight numbers alone to try to predict the outcome of the fight.

 

[Metal_Spirit]

It’s just that everyone needs to upgrade; and everyone will need to anyway the console specs are pushing on the high end of the spectrum.

Hehehehe... Good solution

Only 6.03% os steam users have more than 16 GB RAM, and 49.7% os users have quad cores, so I would have to agree...
But how many will do it... it's another matter!

 

[mrcorbo]

Hehehehe... Good solution

On the PC, that's always the solution!

 

[Shifty Geezer]

The way these all of these specs are being used is akin to looking at the "tale of the tape" before a boxing match and then using the reach and weight numbers alone to try to predict the outcome of the fight.

The daftest thing is we've been through this before, like, every generation. It's nigh impossible to predict real world performance form the paper specs because the actual operation is so infinitely more complex than the check-list of pieces.

For those desperate to claim a winner, accept your XBSX as the champ. For a technical consideration of how game look on these machines, we have to wait until they are out and in use, to see if there's a "PS3 moment*" or not. It's fun to speculate what the limits are, but the pitchforks and angry placards should be put away until something deserving of them.

* A machine that achieved below what would be predicted from its theoretical specs because the realities of the pieces coming together were very different than the raw numbers. And I am not saying XBSX might be turkey!! We know its architecture is same as PS5's so there can't be a Xenos versus RSX difference. It's more a case that perhaps Sony's solution can be proportionally more efficient similar to how 360 was. Or not. It's impossible to tell at this point.

 

[Metal_Spirit]

On the PC, that's always the solution!

Unfortunately, we are not talking about simple upgrade here. We are talking about a global change on the PC market, with total system replacement.
Just check steam stats!
I highly doubt it's worth supporting a PC market targeting about a pair of percentual points of the total global users.

 

[mrcorbo]

Only 6.03% os steam users have more than 16 GB RAM, and 49.7% os users have quad cores, so I would have to agree...
But how many will do it... it's another matter!

"If you build it, they will come."

PC hardware upgrades are driven by software. If MS along with other publishers/developers start releasing games with higher spec requirements on PC or games that derive large QoL benefits from upgraded hardware specs, PC gamers will upgrade to be able to enjoy them. We've seen it before. It's just that it's been a while since we saw such a sea change in performance potential. The SSD revolution is akin to the 3D revolution in the 90's. PC gamers will embrace this once they see what it can do to improve the gaming experience.

 

[iroboto]

Hehehehe... Good solution

Only 6.03% os steam users have more than 16 GB RAM, and 49.7% os users have quad cores, so I would have to agree...
But how many will do it... it's another matter!

unfortunately SSD or not, it's going to have to be the solution for the PC space unless 3P multiplatform games want to keep things restricted on console. PC is such a large market, I don't know how many would be willing to dive in to get the latest gear right now. Perhaps it's the right time to get a console and wait until the prices to come down to just re-build a whole new system. It's something that I'm considering because the wait for a new SSD, newer RAM, a new bus, new graphics cards that are DX12U compatible, a new processor as well. I'm not in a position to buy the pieces individually and incrementally and have a system that will function. My i5-6600 will likely be completely insufficient.

 

[BRiT]

DF's article on the PS5 discussion should be in the PS5 Thread: https://forum.beyond3d.com/threads/playstation-5-ps5-release-holiday-2020.61644

 

[pjbliverpool]

Unfortunately, we are not talking about simple upgrade here. We are talking about a global change on the PC market, with total system replacement.
Just check steam stats!
I highly doubt it's worth supporting a PC market targeting about a pair of percentual points of the total global users.

I don't understand why a full system replacement is needed?

It should be simple to scale a game designed for a console SSD down to a slower PC SSD by simply scaling down those easily scalable elements which take a lot of bandwidth. e.g. texture resolution, mip levels, LOD levels, animation complexity etc...

Scaling down to slower GPU's is even easier given the next gen consoles will run everything at 4K. A lowly 1660 GTX should be more than enough to provide an equivalent experience at 1080p.

Probably the most difficult part will be scaling the CPU requirements down from 8 to 4 cores, but 8 core CPU's have been around in the PC market for years now and can be picked up pretty cheaply, so if developers do need to make that the cost of entry into the next gen gaming for PC gamers then I don't think it's a huge ask.