-
For our US members, be sure to check on your voter registration status and if you are not registered it will only take a second to do so!
The PS5 GPU in PC Form? Radeon RX 6700 In-Depth - Console Equivalent PC Performance? [Digital Foundry]
- Thread starter Tsunami561
- Start date
OP
OP
I mean, that's what it's going to be + some custom HW like they always have, on the ps4 it was the checkerboarding stuff, on ps5 it was the ssd controller, and on ps5 pro it's rumored to be some sort of AI thing
PC ports really need to have console equivalent settings as a part of the game's settings menu. I think many (most?) PC players would be more than content with an option that made the game look like it does on PS5, and any performance gains vs PS5 would just be a bonus.
Genuinely baffled that it isn't a thing yet and hasn't been for years.
Thats pretty much how im going to look at buying GPU's from now on, just buying a console equivelent or just slightly above it. Im hoping it becomes a thing to have settings that target those platforms so its a no brainer for people new to PC gaming dont have to fuss with graphics settings too much. Happy with my 6700XT.
This 100%.
People have to look up optimized settings and do tons of research. Generally speaking, console settings already are the optimized settings (performance to visual tradeoff). Just give us the .ini file!
For a period of time during the tail end of the GPU shortage this was the best value GPU on the market.
They don't base their hardware configs on how interesting people find it lol.
while as someone who grew up during that era of exotic hardware weirdness I have a lot of nostalgia for it, the consoles switching to standardised tech has been the best thing for developers big and small.
When you don't need to spend hours working out how to draw a 2D triangle on screen (an actual story I read from a dev when they got their first PS2 devkit) and can dive straight into an architecture you fully understand, all your focus can be on the game development itself.
It would shut down so many "pOoRlY oPtImIsEd" accusations as well if folks could see whacking everything to ultra gives you many multiples of performance demand compared to console settings, too
We have yet to see totally flawless Saturn emulation, I'm ok with less exoticism in console architecture.
THIS 100% - Sometimes it's too much to remember all of these PC settings. I would love the option to just match consoles
Does everyone do this? I just set to ultra, see if I like the performance, if not I drop to high, check again, drop to medium etc. Once it runs well I just rock those settings. I can't remember the last time I spent more than 2 minutes setting up game visuals. Maybe CP2077 to check out the frame-gen FPS difference in the benchmark but that's about it.
I must be doing it wrong. Almost all my play is on PC and I never do this.
I was under the impression that PC was usually the lead platform these days. So wouldnt any cutting down and optimizing for consoles be done after the PC version was mostly done with. That would make extra work to go back to PC after if true.
I'm not opposed to the idea even if I'd never use it.
Yup. I usually do about 5 minutes whenever I get a new AAA game.
If you cant be bothered to do that then you can almost always just use a lower preset until you find one that suits.
This video missed a lot. What were the PC specs if you want a real comparison you have to use like parts nowhere is the cpu listed or any other specs other than a 6700. Never mind the pc port of the last of us is very mediocre.
I don't mind and encourage these comparisons but the details are important.
I don't mind and encourage these comparisons but the details are important.
It's the classic process if you want the best performance with the best graphics.
The generic presets are often arbitrary. The medium preset for instance might lower textures to medium and anisotropic filtering to 8x even though you gain nothing performance-wise unless you're VRAM limited. Setting steps aren't always linear either, with medium looking closer to low than to high instead of being a middle ground.
There's also the problem where settings aren't described well without looking up what they do. Witcher 2 Uber Sampling was an infamous one that brought ultra selectors to their knees because it was actually just an SSAA option with a funny name. A good recent example is Persona 3 Reload hiding the performance heavy hardware raytracing option under a basic "Reflections" on-off toggle.
The RE4make VRAM fiasco was another good recent example. The settings menu was unclear about how the texture quality system worked. People assumed it was a quality setting when in reality it was basically a dedicated cache setting that didn't necessarily affect how textures looked. If you allotted too much the game would randomly crash at parts with zero warning.
Obviously it doesn't matter if you have a 4080 or something and can max anything out, but generally speaking it's leaving a lot of performance on the table if you don't actually go through the settings and check what the major performance drains are. The level of control is part of what I've always loved about PC gaming.
It's comparing GPU's, not price/performance for an entire system. Regardless, the CPU is largely irrelevant in these tests, especially when you're testing RT against ~30fps console RT modes. Only the most woefully outdated CPU would then become the bottleneck in that scenario. None of the games he tested are a workout for any 6 core/12thread CPU, only Cyberpunk running unlocked in RT would be but the GPU is limiting it long before that point. The results would be identical with these games and settings used if the 6700 was paired with 12400f DDR4 systems vs. a 14900x DDR5.
Yes, TLOU is a very shaky port - but is Hitman 3 on PS5? Cyberpunk in raster mode - both games where the 6700 has a significant lead? What made this video particularly interesting was that it was showing even with roughly equivalent hardware, there can be nearly identical performance as you might assume - but also significant variances between end game performance on different platforms.
As such, if anything this video illustrates why you shouldn't make proclamations about assumed architectural advantages of either platform based on the performance of individual games. You can always use more data.
Very true, hopefully next one at least one of them move to a different GPU vender to at least make the console to console comparison more interesting.
Console settings being the most optimized ones are not always true for PC h/w which a) could be different from console h/w and b) is constantly evolving.
The idea that console settings is some "golden standard" when it comes to optimizing a game for PC is a baffling one.
Hence why providing such preset is only interesting if you're doing a PC to console comparison (like DF/Alex do) but for the actual PC player it can be just as misleading as any other preset option (Medium/High/Ultra etc.)
Thus generally speaking I don't get why "console settings" preset would be any better than what you have now in the form of "medium/high".
Watch people set to it then whine because their 8 year old PC chugs.
Bad idea in practice unfortunately.
Yeah, but does stuff like Cyberpunk 2077 on PC high equals PS5? Not even the XSX version.
You said a lot that had nothing to do with my critique. It's about testing methodology and transparency. Sorry I'm used to watching comparisons by the likes of gamers nexus.
It felt incomplete
Because this preset would also come with a specific hardware baseline which low/med/high does not.
Your critique had little to do with this video though, 'transparency' is only of value if what you want exposed is actually relevant to what's being benchmarked. It was a comparison of the closest match of GPU architecture between PC and Console. It was not a systematic PC vs. Console review, it had a specific focus to see what kind of performance the PC implementation of a GPU architecture you could get with the same game load and too see where the different bottlenecks may lie between the two. Yes, TLOU is a bad port - but is Hitman 3 a bad PS5 port because it performs so much better on the 6700? Who knows, but there's a wide variety of results favouring each and it's only fair to include the outliers for both.
You could maybe argue that the PS5 was potentially CPU bound in some tests which would have skewed those results, but Rich specifically focused on scenarios where the PS5 was GPU bound. The only two games where this might possibly have been the case were in Hitman3 when it drops below 60fps, and potentially Monster Hunter Rise with its sub-120fps 4K results. However Hitman3 is extremely light on the CPU (modern PC's can literally get upwards of 400fps in this title if you're trying to go for CPU bottlenecks) so that's very unlikely, and Monster Hunter Rise is also very light on the CPU - not to mention Rich tested the supersampling mode in the game as well, and the 6700 maintained the same % lead it did over native 4k.
The methodology for what was being measured was sound.
I can't vouch for how it feels on console, but on Steam Deck I'd probably consider those frame time dips to be too distracting from the gameplay experience.
If you're going by the frame time graph, I believe the discrepancy there from the more flat PC timings is because the consoles are using vsync vs no vsync on the PC. It won't really affect the framerate results, but it could possibly account for the most constant see-saw graphs.
It had plenty to with the video and I'll be blunt the only thing it really proved is 1 the last of us PC is a terrible port and Hitman could run on my refrigerator. Cyberpunk is the outlier as obviously Sony has good tools in their porting kit but also on PC side Nvidia magic sauce makes theirs work yet somehow AMD PC parts can't do what the PS5 does? Yeah that's some BS playing there. I mean get real a PS5 isn't more powerful in raytracing in cyberpunk than cards higher than a 6700 it's fishy for sure BUT it's IMO down to the port and also on PC some Nvidia dev shenanigans.
I'm just saying even in the description they should list what they used it's low effort and comes off like a puff piece.
Edit : I'll add they've done this video before too or talked about before what was the point of the rehash?
For $400, you're really getting a good deal with more ram to use. Then there's the Series X which is closer to the 6800
I don't see the reason to have a Pro version unless there's a huge demand for RT
I don't see the reason to have a Pro version unless there's a huge demand for RT
Or not to play games at 720p-1200p for 60fps what is currently happening on the consoles.
Gosh I cannot fucking wait for the PS5 Pro. I'm a complete sucker for more powerful hardware (especially in the console space) inject that power in my veins.
That's just standard Vsync judder and what all vsynced titles look like without VRR, but yes, it looks and feels horrible. It's one of the reasons why Sony need to hurry up and support a universal 120hz output mode so that LFC can kick in and get rid of Vsync judder regardless of what framerate a game runs at.
Other thing is that some graphical effects settings on console are lower than low and can't even be selected on the PC.
Surely you'd have to expose these as options if this is what someone would like to do.
I have a 6700xt which is 12% faster than the 6700 and get triple digit FPS with FSR2 with 1440p which is 960p
Obviously XeSS is better and while it's more expensive, I still get triple digit performance metrics
The problem is developers are targeting resolutions that gives very little in return
Agreed. As your PC gets old and can't handle the new games as it did when you bought it, this would be a good solution. Just give me the settings of the equivalent console and I'm good to go.
I'm hoping the Steam Deck will be the start of this, in a way.
yup. Give me performance and fidelity modes, and also dynamic resolution to mitigate for more heavy moments in games.
Yes its lovely that some peopel have the option to play with 100 options menus, but why don't I get the option to not?
They do. These baselines are explicitly specified in game's system requirements.
Yeah but why? Console get their exclusive settings precisely because these aren't needed on PC.
I've watched the video and I'd argue that generally it boils down to "PS5 wins when the rendering is mostly video memory bandwidth limited while 6700 wins when the rendering is mostly CPU limited".
There is a weird case of CP2077 but it's hard to tell what's going on there without a more detailed debugging.