Developer of A Way Out Claims PS4 is Crippling Performance

One of the most unique games shown at E3 this year is called “A Way Out.” It’s the second title by director Josef Fares, and it’s explicitly designed for multiplayer, either via the PlayStation Network or sitting side by side on the couch. Games with extensive co-op modes are comparatively rare, but games that literally require co-op play are basically unheard of. The game doesn’t take place entirely in split screen, but much of the action and gameplay uses that presentation method (the video below showcases how things work).

The game itself looks interesting enough, but it’s the comments by director Josef Fares that are likely to cause some controversy. “You want the honest truth? This machine is not so strong as you think,” Fares told Engadget, pointing to the PS4 running his game. “This is like a five-year-old PC. If consoles were as powerful as PCs are today, you would see all different games. Most of the work developers put out there is to make them work on consoles.”

I’m not in a position to comment on whether Fares is accurate in saying most of the optimization work that goes into console games is to make them work properly. But I think there’s an important point he’s overlooked. I’ve said before that the Steam Hardware Survey has problems, particularly when it comes to ranking video cards, but since it’s the best data set we’ve got, let’s take a look at what it can tell us. After all, this is an interesting question — does targeting a GPU like the PS4’s result in lower-quality games than we would see, on average on the PC?

The on-average is important. No one is arguing that you couldn’t make games prettier if you only targeted the top-end cards from Nvidia (and hopefully AMD, before too much longer). The question is, how much additional freedom would this give developers?

To that end, we’ve tallied up the most popular 10 video cards on Steam from both AMD and Nvidia, and ranked them relative to the PS4. The PlayStation 4 has 1,152 cores, 72 texture units, and 32 render outputs (1152:72:32 is the way we normally write this). It runs at a clock speed of 800MHz. Since the author’s original quote mentioned the PS4, that’s the platform we’ll compare today. Because we can’t exactly match the specs of the PS4 in any PC hardware, we’ve compared it with the HD 7850. That GPU has a slightly higher clock speed (860MHz instead of 800) but fewer cores and texture units (1024:64:32 @ 860MHz vs. 1152:72:32 @ 800MHz). Note that all of our comparisons are against the PS4, not the PS4 Pro.

Here’s what the color codes below mean, and bear in mind that these are reasonable projections, not pronouncements from on high:

Red: The PlayStation 4’s integrated GPU is probably more powerful than the PC equivalent.
Blue: The PC GPU is faster than the PS4 equivalent.
Purple: The situation is unclear, either because comparative performance data is not available, or because there are multiple GPUs grouped under the same name. When Steam refers to the “Radeon R7,” for example, some of the cards sold under that brand are faster than the PS4 GPU based on specs alone, but others may not be.

Without further ado, here’s our results. Performance comparisons, when available, were made using Anandtech’s Bench tool, comparing the HD 7850 to the various GPUs listed in the chart below. First, we’ll look at the Top 10 most common GPUs from both vendors according to the Steam Hardware Survey.

DataSet-Beta

Data from the Steam Hardware Survey (6/21/2017). The “Est. GPU Perf” figure refers to the number of GPUs in our survey that we’d expect to outperform the GPU inside the original PlayStation 4.

This graph shows us several interesting things. First, the Top 10 AMD and Nvidia GPUs (DX12) hold 46 percent of the combined market. But even if we assume that the two purple SKUs are actually at the top end of their product class, there are a number of popular GPUs on Steam that aren’t likely to match the PS4’s GPU. The GTX 750 and AMD’s Radeon R5 chips are far less powerful than the chip inside the PlayStation 4.

As some of you have undoubtedly noticed, however, Nvidia commands most of the PC market and most of the most popular SKUs (at least, according to Steam). So I decided to run the numbers again, this time using the Top 20 Nvidia cards. Again, relative ratings against the AMD Radeon HD 7850 were confirmed using Anandtech’s Bench tool. While I don’t claim this represents some ironclad, guaranteed result, it should put us solidly in the ballpark. In this case, the GTX 1050 is listed in purple because I wasn’t able to find results that pitted it against the HD 7850 (not that it would change the numbers much either way).

Data from the Steam Hardware Survey (6/21/2017) — Nvidia only. The “Est. GPU Perf” figure refers to the number of GPUs in our survey that we’d expect to outperform the GPU inside the original PlayStation 4.

Nvidia is in a stronger position here than AMD, but not by as much as you might expect. In both cases, there are a significant number of GPUs that aren’t necessarily more powerful than the PS4’s — and that matters to developers when they’re working on next-generation titles. A game that only a fraction of your potential audience can play isn’t exactly going to sell very well. And finally, for those of you who want the whole enchilada, the Top 20 Nvidia cards and Top 10 AMD cards collectively account for 61.33 percent of the GPUs listed in the Steam Hardware Survey.

The console-PC relationship is more complicated than people think

I’m not claiming to have disproven Josef Fares’s remarks or arguing that optimizing for consoles is anything less than difficult. In fact, everything I’ve ever read about optimizing console titles suggests that a great deal of work goes into keeping frame rates up and gameplay smooth. Companies like DICE and Naughty Dog have given detailed presentations on how they used various tricks, ideas, and innovative programming techniques to squeeze maximum performance out of comparatively modest hardware. It’s entirely possible that Fares is referring to CPU horsepower, where both the PS4 and Xbox One haven’t quite packed as much of a punch as some might have wished.

The PS4 Pro and Xbox One X both improve this situation modestly. While no one will ever mistake Jaguar for a true high-performance CPU core, cranking up the clock speed by ~1.35x (both Sony and MS) will improve overall system performance. It should even help reduce L2 latencies, since Jaguar’s L2 is typically clocked at 1/2 clock speed.

NDPS4_8

Cross-cluster cache latencies on the PS4 are bad enough that companies had to come up with innovative ways to keep data local while simultaneously keeping their cores fed.

I’m not denying Fares’ point that he and his team could build a much more ambitious game if everyone had access to absolutely cutting-edge hardware. But this has been a fact of life since Pong debuted in your (once) local arcade. Even if we look at the PC ecosystem, a significant number of Steam gamers are still playing on hardware that doesn’t match the minimum capabilities of the PS4. Since game developers always want to sell their hardware on the largest set of systems they can, there’s never going to come a day when we see games only offering compatibility with the Radeon Fury XXV or the GeForce Alpha Titanium 9000.

Whether you play on a console or prefer a PC, game developers will always have to target a variety of price points and graphics abilities to keep their titles accessible to the largest audience possible. Given that consoles have practically become PCs in this latest generation, it’s not surprising to see some of the same issues and concerns about upgrading, compatibility, and appropriate development targets.

Now read: The Best Free PC Games

Let’s block ads! (Why?)

ExtremeTechGaming – ExtremeTech