Sony May Be Overselling Aspects of the PS5’s Hardware Performance
Sony is in a tricky position with the PlayStation 5. While it heads into the next console generation as the unquestioned winner of the current cycle, it looks as though the PS5 will be markedly less powerful than the Xbox Series X.
When Sony unveiled the PS5 last week, Mark Cerny told viewers that the PS5 wouldn’t be at a disadvantage against the XSX because a higher-clocked smaller GPU like the PS5’s could still outperform the wider, slower GPU on the Xbox Series X:
About the only downside is that system memory is 33 percent further away in terms of cycles, but the large number of benefits more than counterbalance that. As a friend of mine says, a rising tide lifts all boats. Also, it’s easier to fully use 36 CUs in parallel than it is to fully use 48 CUs – when triangles are small, it’s much harder to fill all those CUs with useful work.
We spoke to Dan Baker, Graphics Architect of Oxide Games, about the efficiency question and whether smaller GPUs would be a better fit for modern graphics workloads than larger ones.
“Small triangles are indeed inefficient,” said Baker, “Because you have to partially shade fragments that are ultimately discarded. However, this inefficiency is largely in CU execution because the CUs are being asked to compute more work, so you’d want more CUs to offset the inefficiency.
“However,” he continued, “This is specific to the type of renderer. In deferred renderers, which make up most of the market today, most of the shading computation is done in screen space, where the small triangle problem is minimized. Only the material setup really pays the cost for small triangles. For Oxide’s decoupled shading rendering technology, neither the setup nor the shading efficiency is affected by the size of the triangle, so we are impacted even less.”
According to Baker, the increase in memory latency that Cerny mentions is indeed a negative that can make smaller, high-clocked parts a bit less efficient than their wider, slower-clocked brethren.
What About Storage Performance?
Both Sony and Microsoft are delivering dramatic storage performance in their next-gen solutions, with Sony claiming ~2x the performance Microsoft does in terms of sustained streaming bandwidth. Sony has detailed a number of changes these improvements will deliver, including more efficient data loads, since objects don’t need to be duplicated dozens or hundreds of times in files across the game install.
There’s absolutely no question that upgrading from the HDD solutions inside the Xbox One X / PS4 Pro to PCIe-based SSDs will be an enormous improvement for both consoles. Swapping an HDD for an SSD is still one of the all-time best ways to improve performance, even in an old rig. When Sony talks about a 100x performance improvement compared with the PS4, that’s almost certainly true when measured against HDD latency, while storage bandwidth has improved nearly as much.
The question is, what aspect of gaming is this additional performance going to improve? Baker believes the high-speed SSD will be used as a giant page file. Texture data can be selectively streamed in and out of system RAM to eliminate things like load times and texture pop-in. What it probably isn’t going to be used for — not as such — is simply making the game world bigger.
To be clear, I’m not saying open-game maps won’t get bigger next generation, just that the use of ultra-fast SSDs probably won’t be the reason why they do. No open-world title loads the entire game world into RAM at once. Rather than attempting to cache an entire title in RAM, a PCIe SSD serves as a giant, texture-y RAMDisk. There are a lot of improvements developers can make behind the scenes to how assets to boost performance, but they’re also likely to be tied to some complex new methods of handling storage and data loads.
To some extent, this is par for the course. During the PS3 era, Sony even declared that the PS3 was deliberately difficult to program for because this ensured it took developers longer to unlock the full potential of the system.
“We don’t provide the ‘easy to program for’ console that (developers) want, because ‘easy to program for’ means that anybody will be able to take advantage of pretty much what the hardware can do, so then the question is, what do you do for the rest of the nine-and-a-half years?” explained Kaz Hirai back in 2009.
But this type of thinking has been less common of late and console manufacturers now provide more help than they once did. Microsoft’s Xbox Series X, for example, offers 8 threads at 3.8GHz or 16 threads at 3.6GHz and the manufacturer has predicted at least some developers will opt for higher clocks and lower threads due to the difficulty of parallelizing effectively.
The net effect of this is that it isn’t clear the PS5 will get a different benefit than the Xbox Series X from its faster storage, while the XSX is likely to be faster on the whole thanks to a wider GPU. The storage improvements on both platforms are more likely to improve data load times and things like texture loading, rather than by making game worlds larger in absolute terms. We don’t know how these factors will play into customer purchases, however, because we don’t know the price on either platform or how the impact of the worldwide pandemic will affect launch schedules. On the whole, it looks like the PS5 is set to be a bit less powerful — though not necessarily less popular — than the Xbox Series X.