It means "does not equal." Read my post again with that knowledge and let me know if this question still makes any sense at all.

Well I'm not sure what Acert meant. "Some important metrics" is incredibly vague after all. It depends what you consider an important metric. It's a vague statement.

What's not vague is the GPU gap, etc. I would mostly agree that it's a minor issue for now. We have yet to see real world scenarios where we can really compare performance.

I think people would be less keen too focus this sharply on the performance gap if the Xbox wasn't $100 more expensive than the Playstation.

Microsoft participating in that discussion makes it a less minor issue for them hah..

Maybe when Microsoft stops the terrible damage control. Maybe.

I think the software will speak volumes, but I don't think we've seen enough for it to speak very loudly at all. Unless you're comparing low resolution, highly dithered GIF's of exclusive games from different developers.

That's always fun. And highly accurate.

I am hesitant to post because nuances are ignored (like "some important metrics" are just flippantly ignored) and basic math... but here it goes.

Grim, an example where the PS4 is 100% faster is fill rate. 32 ROPs (8 RBEs if you like) in the PS4; 16 in Xbox One.

If Xbox One = 100%

Then PS4 = 200%

In the context of qualitative comparisons "faster" with a percentate assumes the 100% baseline. Hence

If PS4 is 33% faster = PS4 is 133% speed of Xbox One (baseline 100%); which is equivalent to saying the Xbox One = 100% and PS4 = 133%

If PS4 is 100% faster = PS4 is 200% speed of Xbox One (baseline 100%); which is equivalent to saying the Xbox One = 100% and PS4 = 200%

This is why 18 CUs is 50% more than 12 CUs (or 150% of 12) but 12 CUs is 33% less than 18 (or 66% of).

Anyways, yes, in a strictly fill rate bound game the PS4 would be 100% faster (200% speed of, 2x faster, etc). But games are not strictly bound by one metric--and reality is they are bound variously across a frame by many components, some of which are not easily testable and highly variable by the design of the workload (e.g. one program could feed ALUs at full speed but on the same system and the same size of workload, but due to layout, could become bound).

I think there are a lot of ways for Xbox One devs to address the performance pit falls: e.g. for graphically bound situations

* Run the PS4 version at 1920*1080 (~2MP) and Xbox One at 1386*780 (`1MP). That is extreme. 1600x900 is 30% reduction and, frankly, few people would miss the 320 and 180 pixels. In fact you could even cheat on vert/horiz based on what people are more sensative too (e.g. 1344x1080)

* Use dynamic resolution at a locked framerate. PS4 may run at 40Hz-60Hz but lock into 30Hz (super stable, quite a few games do this), Xbox One may cut closer to 30Hz, dynamic resolution could then assure it stays at 30Hz at the rare scenarios it drops too much (e.g. fill rate bound scenarios).

* Run the Xbox One at 720p with high quality MSAA and top end AF; run PS4 at 1080p with post process AA and moderate AF.

* Cheat in ways people won't notice. Cut your particle buffer. Slightly reduce shadow map resolution and lower the QA on edge dithering. Cut your top end texture size. More aggressive LOD. Cut off 20 pixels on the top and bottom. Basically all the ways PC games can be tweaked and few would ever notice, especially because a) no one is playing both versions side by side and b) users won't be able to touch the "quality sliders" to see what they are missing.

The fact both designs are both 8x Jags and both GCN means a decent developer can plan ALL of these approaches into the engine at the beginning and use the last 3 months to performance tweak.

The bigger problems for MS in the above regard is not the raw specs but

1) Their split pool design. 1 large fast pool is always better than 2 (especially 1 very slow and large and 1 fast and very small). It isn't that the PS4 is 33%-100% faster in some metrics--the issue is even getting to that baseline may pose a challenge due to the memory design.

2) MS's internal studios are ... well, I called it **years** ago. They are a problem. Deft even remembered my prediction ;) And as I said back then 3rd party exclusives would be harder because Sony would probably co-launch (they did) taking away incentive. Devs cashed in on the Xbox 360 because they were assured even in 2006 the PS3 base would be very small and the 360 base much larger. That is not the case this time around so fewer publishers are interested. Taking it a step further developers have a hard time ignoring a competitive system that is cheaper--Mr. Bean counter keeps telling me the $100 difference could = an extra game sale or two. You are NUTS to go Xbox One exclusive unless MS pays for everything.

3) 3rd party performance. PS4 is easier to get predictable performance. PS4 is overall faster. Almost all my games were 3rd party on the Xbox 360. This is an area where, having less than PS4 stability/eye candy (even if minor), hurts. Especially as there is a big ? mark on 1st party.

4) Price. People are paying more for something that is slower. Ok, you get Kinect...

I don't hate Kinect. Kinect 2.0 is a big bowl of Awesome Sauce (TM). The gerbils at MS just cannot realize that, "Hey, why not make a killer app ... like a boxing game? Maybe toss in some fitness into it. And on fitness games... Kinect has workout junk and waggle minigames. Why not make a game that provides fitness so, you know, it is fun???? What a concept!!!"

Which all boils down to poor management. Don is gone. Balmer will be gone. Bungie ran away. MS killed ES, most of Rare, FASA, and a number of other studios. They just were not ran well. And this lack of concrete vision can be seen in the flip-flopping of the Xbox One user policies.

If someone told me in 2010 MS would launch a 1.3 TFLOPs console in 2013 at $500 I would have laughed. Another design faux paux: MS planners totally missed GDDR5 densities and bet a HUGE chunk of their APU on embedded cache and slow DDR3.

Just a basic design acumen tells you: Xbox One was not designed to be cheap out of the gate. It was designed for significant cost reduction down the road, at the expense of performance.

I think it is insulting that they want people to believe their design choices were for better performance. They were trying to be cute and get as much performance out of a price reduction conscious design. They did that, but it did come at a cost. Sony will pay through the nose later this gen for GDDR5. But right now MS is aiming to lose NO money on Xbox One at launch and make a killing on Media. They believe Kinect 2.0 with media is their killer app.

The dev documents clearly outlined this path.

Let me distill that into terms this thread can understand:

Whoa, whoa there. Are you saying it's 100% + 33% + 50% faster?!

You heard it here first folks. Acert said the PS4 is 183% faster than the Xbox 1.

But don't worry. It's not all Grim ;)

A game running at 20fps and 30fps looks "very similar." Which would you want?

There are performance gaps that big this generation with a smaller performance gap between consoles.

But who knows? The Xbox should at the very least be very easy to develop on. And developers have good reason to target some kind of parity between platforms.

That article is fundamentally flawed. It is NOT a very good article. It is just as bad as those above appealing to MS PR, specifically appeals to authority (We have smart people, best in the industry, trust him!). What DF presented has some facts involved with some false assumptions that negate the purpose of the article because it asks the wrong question to arrive at the desired conclusion.

To Richard's benefit the sub-title is about teraflops (TFLOPs). If the point was finely restricted to the fact, "FLOPs alone don't determine performance" then there would not be an issue. When it is broadened by (a) looking at hanging with multiplat titles and (b) begins using benchmarks without firm controls that account for everything but the desired variable it is a fundamentally flawed experiment. This is a pure "drive traffic" fluff piece catering to the lowest common denominator as the core assumptions brought to the table are inadequate to address the issue.

Just as my post above outlined it isn't an issue of having just less flops (33% less). It has half the ROPs and substantially less texturing as well. Throw in the variance in memory architecture and how GCN performs under various clock speeds and the article serves no real value above what is generally known. On the flip side now that Xbox One has a higher clock the Xbox One has a higher triangle setup rate (2 verts / clock) -- but a substantially lower triangle transformation (due to fewer shaders).

These aren't things easily accounted for. We know in general most software is not significant setup limited outside mass tesselation scenarios, hence few people have mentioned this issue this gen (last gen with RSX having 1/2 Xenos and Xenos having a magnitude higher peak triangle transform rate on the GPU due to Unified Shaders this was an issue--as I said in 2005 on these very forums: each generation poses different bottlenecks.)

Beyond the obvious (to anyone who understands these issues at a remedial level) there is no such thing as "real" software as a valid performance metric unless you are a developer (in which case the only metric you care about is the metric of how it works in YOUR game). We have on the PC examples where on GCN Game.A. is 20% faster on GPU.A. over GPU.B. but 50% faster in Game.B. Did the GPU magically become a LOT faster? No. Workloads matter. HW and SW design differences matter.

Hence anyone really wanting to discuss the HW in any intelligent fashion would (a) try to restrict that performance issue as finely as possible within the bounds of the synthetic performance and how the general software/hardware model makes that metric relevant and (b) looks to how important that metric is to the performance of a game and, if it is vital (a LOT of stuff is NOT vital, hence a lot of code is very inefficient because it simple CAN be) the question is how, if at all, can you work around it?

That is exactly how I laid out my previous response. I noted some key metrics in GPU performance and then discussed a handful of ways they could impact performance and more importantly some not so obvious ways to address the short comings.

From that angle I find Richard's article offensive. It is dumbing down the topic and leaving out key information that informs the shape of the "Question." Furthermore is trivializes the issue down to a lowest common denominator as if any real truth has been found.

Far too many variables were left uncontrolled for (poor science) and no substantial information is communicated (e.g. synthetics vs. real world by, say, testing PC settings by lowering all settings and then cranking up LOD, texturing, resolution, etc to see where specific weaknesses are).

Ironically PC testers often do this; e.g. HardOCP used to give you "best settings" at a set framerate/resolution/feature target. Many do this by testing Resolution, AA settings, and Quality settings. By proding the HW in these ways certain underlying issues are exposed: CPU limited? Bandwidth or Framebuffer limited? Slow AF? etc.

And, again, I never said Xbox One is slower in every way than the PS4. But it is obvious that in general the PS4's GPU is in most cases for typical workloads a bit, or substantially, faster. But, as I also said, there are corner cases where Xbox One will be faster. And MS internal studios will probably target those areas, if they have good visual return, for differentiation by designing to a platform's strengths. But it will take 2-4 years to really get a handle on what to avoid and what to target. This is why later software looks better. Most people didn't realize that later PS3/360 games cut down on some IQ and features and buffed others because the return on investment was better. So PGR3 had some great HDR but PGR4 buffed almost all the other elements at the cost of the better HDR. Same reason a game like Kameo had a lot of snazzy effects and few later titles mimicked--they were too expensive and designers wanted the cycles elselwhere as the return on investment they felt was higher.

And with all things the most effort will be put into the lowest common denominator that matters for sales to hack away to bring it up to par (just as a lot of effort was put into PS3 titles to get them to match the 360 for most of the last gen).

It is unfortunate that these things cannot be discussed intelligently. Too many, "OMGBBQ!!!! dGPU secrete sauce" posts.Thanks Grift.

What I am more "afraid" of is how my posts are being trolled over to other forums. Grim mentions posting my stuff elsewhere to test me (lolz) but I have a hard time believing what I post is represented fairly/accurately if it cannot even be quoted correctly on the forum it was posted on.

Major lolz though at running over to B3D or other sites. There are devs there, and non-devs (it helps if you were a mod and know who people work for wink wink!!) Funny when Gamersyders run over there and take the side of fans and ignore the posts of developers. But us old timers haven't bothered much this gen as with the same GPU and CPU general architecture the synthetic differences and how it could play out were pretty obvious outside to various unknowns (mainly negative gotchas) and an eye toward future best practices. After a while you just get tired of the dribble. e.g. After John Stokes left Ars their tech archtiecture analysis went down the crapper. I have talked with their writers a handful of times about their poor console reporting (e.g. an article about mobile GPUs being better than RSX a couple years back was a blatant hack piece aimed at BUZZ) but most of them don't understand the things they write... or worse, know it, but are driving a certain angle to capture readers/clicks/advert money.

Hence why MS's recent PR is a win in the short term: it makes the Xbox fan base feel better. Nevermind logical people who spoke of the many merits of MS past products calling out MS PR as misleading, even false, the fact is fans want to be assured. So MS assured them. Which makes anyone who disagrees an enemy.

It is no different than the SDF last gen. My favorite was seeing technically inferior software tech getting called "better" due to artistic reasons (I prefer the color was my favorite line) on multiplatform titles. This gen the software is so complex and the IQ differences easy to obfuscate that preferences will more often than not be on artistic grounds, not technical. That will be assisted by the fact the consoles are in many ways nearly identical. Except in graphics (and sound, but people don't care about that). Art > Technicals.

But I have been saying that for over a decade. TF2 is a classic example of art allowing a game to live well beyond its technical prime.

I would like to add to this:Crytek has published a number of whitepapers over the year. Some of their BEST actually broke down each render segment on the PC/360/PS3 and showed where each one performed. You could clearly see places where all were the same (regardless of HW differences), how each had various strengths, and also areas where Crytek had to cut quality (or use a different approach) to meet the performance window.

They broke it down by m.sec. and detailed each segment. Very informative. And a good gateway to understanding how complex the issue is--and why looking at each area is relevant and how it can be addressed.

Those trying to "capture" a specific metric "A is X times faster than B in real software" really don't understand the complexities of the question at hand. And should be ignored. This is why if my posts are reviewed I have tried to always talk specific comparisons and what it means for a certain work load.

How it impacts a specific game is really up to the designers.

For those who really care about these issues look up the Crytek stuff and pay special attention to how they approach the hardware and try to resolve issues and challenges. It is a good lesson in how 0.1ms extra for a feature can spell doom -- or just get the feature disabled (lolz) or a different approach to satisfy the challenge.

As a lead SW engineer at a Fortune 500 told me last year: the new consoles are less about what can/cannot be done, but at what trade-offs. There will be a lot of trade offs behind the scenes that users will never notice.