Digital Foundry Article Technical Discussion [2018]
- Thread starter Shifty Geezer
- Start date
-
- Tags
- digital foundry
- Status
Ubisoft is really starting to lead the pack here and showing AAA Multiplatform muscle. The technology for their open world, lighting, shading and textures.
Really surprised by the showing across all platforms, really no one was left behind and all of them performed better than my expectations.
Most of the time I would suspect that (and earlier in the gen @Shifty Geezer also noted, if XBO was getting 720p back then how crappy would the resolution and graphics be years later when we start pushing boundaries. And here we are, probably amongst the best looking games on XBO, higher resolution than 900p. Really impressive that neither base machine hasn’t has the absolute limit yet.
This DF video sold me on buying Fc5. Much like the other ones sold me on AC:O.
This was also very close to 6x output from XBO to 1X here. 1440*1080 is only slightly higher resolution than 1600*900.
Really surprised by the showing across all platforms, really no one was left behind and all of them performed better than my expectations.
Most of the time I would suspect that (and earlier in the gen @Shifty Geezer also noted, if XBO was getting 720p back then how crappy would the resolution and graphics be years later when we start pushing boundaries. And here we are, probably amongst the best looking games on XBO, higher resolution than 900p. Really impressive that neither base machine hasn’t has the absolute limit yet.
This DF video sold me on buying Fc5. Much like the other ones sold me on AC:O.
This was also very close to 6x output from XBO to 1X here. 1440*1080 is only slightly higher resolution than 1600*900.
And once again the X1X delivers greater performance per FLOP, per Watt, per $$$, per mm^2, per unit of BW, per ... everything .... than PS4 Pro.
The narrative that the PS4 Pro focused on efficiency and being "smart" while the X1X is a "brute force" machine is demonstrably false.
X1X isn't just by the fastest machine on paper and in reality, it's also the most efficient. And that is a sign of MS really understanding the interaction between software and hardware, and then really nailing their design.
The narrative that the PS4 Pro focused on efficiency and being "smart" while the X1X is a "brute force" machine is demonstrably false.
X1X isn't just by the fastest machine on paper and in reality, it's also the most efficient. And that is a sign of MS really understanding the interaction between software and hardware, and then really nailing their design.
You're right on everything, but if we compare the PS4 to the Pro, we can basically say the same thing. All of that with a bandwidth that is only 24% greater than the base PS4... so, i see a sign of nothing... just a hardware launched later at a higher price and which copies the PS4 architecture.
For a real comparison, we need to see consoles launched at the same time and price... this is what we saw between the Xbox One and the PS4...
mate theres pro games that already go from (ps4) 1080p -> (ps4 pro) 1800p eg Final Fantasy XIV, No Mans Sky etc or prepare to have your mind blown even (ps4) 1080p -> (ps4 pro) 2160p eg MBL the show 17, Diablo 3 etc
Theres lots of examples of this, how is this 'impossible feat' happening?
Simple, in most cases the ps4 version could of been done at higher res eg 1200p but the developers decided just to stick to 1080p cause A/ its native for a lot of TV's + B/ res jump is not going to provide *that* much of a visual upgrade, C/ less likely to have lower framerate
Why do we have to have everything released at the same time and at the same price? Comparisons between the Pro and XOX are valid because they are the highest tear of consoles available on the market right now. People compared the SNES to the Genesis even though the later launched 18 months earlier, the N64 to the Playstation even though there was a year between those as well. Same with PS3 and 360. PS3 came out a year later at a higher price point and people still compared it to the 360 through that console generation.
I thought the whole point of these comparisons was to help inform consumers on what they're getting if they buy something? Also, copies the PS4 architecture? The amazingly original PS4 architecture that is an x86 cpu, a gpu and ram, practically unheard of until Sony invented it.
Indeed, you can compare what you want. However, this kind of sentence implies something special that you simply can't measure : "And that is a sign of MS really understanding the interaction between software and hardware, and then really nailing their design".
Yet we don't know what Sony could have done at the same date and at the same price. So the comparison used by function to make his conclusion is pointless : "the X1X delivers greater performance per FLOP, per Watt, per $$$, per mm^2, per unit of BW, per ... everything .... than PS4 Pro."
Everything is right but the reasoning is flawed...
The most obvious reason for the performances disparity is known by everyone instead of supposed secret optimizations : bandwidth.
If the X had a comparable increase of bandwidth to the Pro according to the GPU power, this would translate in 236 GB/s... and i'm not certain that you would obtain the same performances... inversely, the Pro would have needed 281 GB/s to match the X.
Everything is here, there's no magic...
Ok but you can't judge efficiency without considering the price and the launch date. It's obvious that the Pro efficiency has been limited by a compromise on bandwidth.
Actually, the Pro could be very efficient for its price and its launch date but we can't really compare.
Edit : we could also see it this way for FC5 : with +100$, you have a resolution that is x2,25 higher than the base PS4. With +200$, it is only x4 times higher instead of x4,5 higher. With a literal interpretation, this means that the X would have been less efficient than the Pro at the same price.
Last edited:
I still don't see why not. Look back at history, the PS3 was released after 360, was held back by bandwidth in many cases compared to 360, and was more expensive. In most cases, the 360 was the more efficient machine. But by your logic it's unfair to compare the PS3 and 360 because the PS3 came out a year later, was more expensive, and was generally the worse performer.
We are comparing things that exist in the real world with real world performance. When we start putting arbitrary rules on that comparison is when we invalidate the objectivity of the comparison. I agree that because the XOX came out later and is more expensive that it should be better. But that isn't always the case. Furthermore, I have enough faith in Sony's engineers to know what type of performance they wanted out of pro to have designed the system to achieve their goals. They knew the shortcomings of the system and designed things into it to mitigate those shortcomings. Things like CBR and RPM are examples of this. You are right that Pro is held back by bandwidth, but it isn't like the technology for higher bandwidth didn't exist when the Pro was released. Sony chose not to use it.
In that case, PS3 would be rated even worse efficiency. Recop's saying you need to factor time interval. Which is true, but only depending on what comparison one is trying to make. As ever, people are asserting points on an undefined metric, using their own assumed point of reference.
Two different comparisons:
Given two products present on the market, which is presently the most efficient. An absolute comparison.
Given two products present on the market, which was most efficient for the technology available at the time of release. A weighted comparison.
Given two products present on the market, which was most efficient for the technology available at the time of release. A weighted comparison.
We can go on.
Which is the most efficient for a given price tag?
Which improved most efficiently on its predecessor given the relative in RAM BW?
Which renders the most pixels per second per gigabyte of RAM?
Which improved most efficiently on its predecessor given the relative in RAM BW?
Which renders the most pixels per second per gigabyte of RAM?
"Most efficient" is the younger sibling discussion to "most powerful". Technical comparisons are redundant without defined metrics.
The Pro is a nice machine, and I think the extra time MS had did have a very large impact on what they were able to deliver. We know they did lots of work profiling software and measuring the impact of various potential tweaks on virtual hardware. We can also see that they didn't just increase raw BW, for example, they went wider and also increased the number of memory channels per unit of bus width too (doubled), probably to reduce the impact of contention and to allow more room for scheduling reads/writes.
So yeah, the time, the money, and the extent of their profiling (which I expect Sony did some of too!) will all weigh into this.
But my point was still, as Brit says, that the idea of Pro being "efficient" (i.e. smart, good, clever) while the X1X is only faster because of "lol M$ brute force" (i.e. dumb, clumsy, not as smart) is nonsense. Any way you can try and look at "efficiency" - and I've given examples of the kind of things I think you can at least try to assess - X1X wins. And yes, time and money are a big part of that. And they allowed things like the CPU tweaks to boost IPC, the "Hovis method" for increasing clocks within the same power envelope, the unique memory controller setup, the further enhanced command processer etc.
But just because it's by far the fastest console doesn't mean it's a "brute force" machine. It's also on balance - and in probably every area you can see based on Digital Foundry and VG Tech findings - turning out to be the most "efficient".
That's not really comparing any type of "efficiency" though, it's normalising performance and cost around PS4 and then comparing an absolute price on one axis and a performance multiplier on the other.
Judging how X1X performance would scale with price using Sony products that use cheaper optical drives, coolers, and probably cheaper PSUs and cases seems ... tricky. And far more abstract than looking at the actual performance and characteristics of the real hardware that's actually out there.
Yeah but in this case we have a clear anwser. If Sony wasn't even able to match the 360 one year later, then we can be almost certain that in the same conditions than the PS3, MS would have produce an even better machine than the 360 and thus an even better console than the PS3.
But when we have the expected result : later + more exepensive = better hardware, then it's harder to make a relative comparison.
Last edited:
It's nonsense in a strict comparison. We still don't know which constructor would have been able to make the best hardware with the same constraints.
Another interesting point is that the X usually fails to match its GPU difference with the PS4 in the most demanding games = those running at a resolution of 900p or lower on Xbox One, or a resolution lower than 1080p on PS4.
The X GPU is x3,3 more powerful than the PS4 GPU. Yet in a game like Homefront, the resolution of the X version is only x2,8 times higher...
It's certainly proven and expected X is much more efficient in terms of wattage. (I did predict the actual power consumption before it was known but... I'm sorry, I cheated.)
More importantly, we are seeing third parties with patches providing outstanding improvements, beyond napkin math.
But the idea of making a "cost per performance" or "cost per release date" argument is pointless. The business targets were different, release date was a year apart, and even though the retail price is $100 more, we don't know the real BOM difference, not even estimates by IHS. We can't quantify what a year delay means in terms of market impact, certainly not as a percentage. It was easier with ps3/360 or ps4/xb1 since it wasn't a case of more power for more money, nor more power a year later. In those cases one of them clearly messed up.
More importantly, we are seeing third parties with patches providing outstanding improvements, beyond napkin math.
But the idea of making a "cost per performance" or "cost per release date" argument is pointless. The business targets were different, release date was a year apart, and even though the retail price is $100 more, we don't know the real BOM difference, not even estimates by IHS. We can't quantify what a year delay means in terms of market impact, certainly not as a percentage. It was easier with ps3/360 or ps4/xb1 since it wasn't a case of more power for more money, nor more power a year later. In those cases one of them clearly messed up.
Last edited:
Sony's objective was to propose a console which could provide a better IQ for 4K TV, and a substantial improvement for VR games.
MS objective was to crush sony after all the criticism they got with their "underpowered" Xbox one.
So they brought a nice piece of engineering, to cake the crown again as the most powerful console.
MS objective was to crush sony after all the criticism they got with their "underpowered" Xbox one.
So they brought a nice piece of engineering, to cake the crown again as the most powerful console.
Long term cost reduction will be interesting. Right now the X SoC probably have lower yield because of the clocking, so it should improve a lot on 7nm, meaning the difference in SoC cost between them will be much smaller once they both go slim. OTOH the memory is a fixed ratio of 1.5x more expensive.
The other thing I wonder is how mid-gen will cause a problem with next gen adoption. If mid-gen ramps up to sell a lot, the next gen will not have a big difference and most gamers just bought a new console relatively recently. This can't be good for studios wanting to make next-gen exclusive games. I guess cross-gen will be the norm and adoption will be slow.
The other thing I wonder is how mid-gen will cause a problem with next gen adoption. If mid-gen ramps up to sell a lot, the next gen will not have a big difference and most gamers just bought a new console relatively recently. This can't be good for studios wanting to make next-gen exclusive games. I guess cross-gen will be the norm and adoption will be slow.
That's undisputed. What we're seeing are some alternative comparisons trying to gauge some measure of efficiency, and as MrFox points out, we've no idea what the BOMS are. So another comparison that can be made, which is the most efficient per dollar profits?
AFAICS people are picking their concept of 'most efficient' based on which brand flag they prefer to wave. *shrug*
- Status
Similar threads
