Support NeoGAF

[Finalizer]

The 7790 is the closest approximation, because it's a lot more architecturally similar to the Xbox One GPU compared to a 7770. People are focusing on the gap between a 7850 and the 7770, and using that as a guide, but considering how the Xbox One GUP has a 7770 beat, it's obvious that the numbers would be even better for the Xbox One GPU by comparison. Plus it will benefit from console level optimizations. Even in the very edge article they refer to 20fps or so on 1600x900 before any optimizations whatsoever. The Xbox One being a system that has to be more carefully managed by developers, would obviously suffer more from lack of proper optimization, and the more powerful PS4 that is simpler to design for would be much better off in such an un-optimized scenario.

Keep in mind similar things could be said about the 7850 as a representation of the PS4 - you're talking about a GPU with two less CUs.

The point is, neither GPU in the example gives the full, true picture - and that's besides the point in the first place. No one (reasonable) would use these metrics as an actual judge of either of the new systems; they simply do not, and cannot, give the full story. They're just an example for a very general idea of the performance difference; or at least, the resolution difference necessary to achieve rough performance parity on these GPUs.

Again, not meant to be wholly accurate. You want that, there was another post with a nice write-up detailing the exact GPUs and respective over/underclocks on them needed for a more accurate test. Even then, it'd still be an inherently muddy comparison that I'm not even sure the effort would be worth it, though it'd still be interesting to see nonetheless.

 

[artist]

"Completely useless, stupid, irrelevant, why bother, no way"

I'd rather be informed and stupid rather than being intentionally oblivious and idiotic.

 

[SenjutsuSage]

Keep in mind similar things could be said about the 7850 as a representation of the PS4 - you're talking about a GPU with two less CUs.

The point is, neither GPU in the example gives the full, true picture - and that's besides the point in the first place. No one (reasonable) would use these metrics as an actual judge of either of the new systems; they simply do not, and cannot, give the full story. They're just an example for a very general idea of the performance difference; or at least, the resolution difference necessary to achieve rough performance parity on these GPUs.

Again, not meant to be wholly accurate. You want that, there was another post with a nice write-up detailing the exact GPUs and respective over/underclocks on them needed for a more accurate test. Even then, it'd still be an inherently muddy comparison that I'm not even sure the effort would be worth it, though it'd still be interesting to see nonetheless.

I understand all that, but it's just really misleading to try and pass off a 7770 as somehow accurately representing the perf dif between the PS4 and Xbox One. There's too many factors that clearly suggest the Xbox One GPU will definitely perform better in most every situation compared to a 7770. It will definitely be much more capable.at higher resolutions also. Man, off to work! I hate having to get up early

 

[Liabe Brave]

Toss in the move engines that can do various kinds of copy operations that help save bandwidth....

The move engines do not save bandwidth. They're part of main memory access, not over and above it. When the move engines are operating at 25.6 GB/s (max speed), you can only access the DDR3 at 42.4 GB/s, not 68 GB/s.

 

[EmptySpace]

PS4: 8 ACE/64 queues


Verses


Xbone: 2 ACE/ 16 queues

ouch.

 

[Finalizer]

I understand all that, but it's just really misleading to try and pass off a 7770 as somehow accurately representing the perf dif between the PS4 and Xbox One. There's too many factors that clearly suggest the Xbox One GPU will definitely perform better in most every situation compared to a 7770. It will definitely be much more capable.at higher resolutions also. Man, off to work! I hate having to get up early

I mean, fair enough if you're worried about people trying to run away with those graphs to deliberately mislead folks on the performance difference between the new consoles, but my point is that I personally find the information interesting on a very general level, with the caveat that I'm not one to trounce it as real proof of anything, heh.

 

[bonus_sco]

I understand all that, but it's just really misleading to try and pass off a 7770 as somehow accurately representing the perf dif between the PS4 and Xbox One. There's too many factors that clearly suggest the Xbox One GPU will definitely perform better in most every situation compared to a 7770. It will definitely be much more capable.at higher resolutions also. Man, off to work! I hate having to get up early

There's no PC equivalent to either console. The front end of the PS4 GPU is also significantly different to the PC cards, has GPU cache modifications and coprocessors. It should also perform significantly better than the PC GPU chosen here.

Any comparison to PC parts is fairly pointless.

 

[nib95]

A 7770 is a more architecturally flawed comparison to the Xbox One GPU than a 7790 is because the 7770 lacks dual geometry engines, and the 7770's bandwidth is even lower than the 7790's. The Xbox One's bandwidth is quite a bit more than the 7790's.

Less flawed than a comparison to the 7790. Architecturally the XO is somewhere between the 7770 and 7790 though raw performance wise it is exactly the same as the 7770. Performance wise the 7790 is actually closer to being on par with the PS4. Architecturally the PS4 is somewhere between the 7850 and 7870, but with raw performance numbers that are above the 7850.

You're accounting for the negative differences with the 7770, but not the positive. The fact that the 7770 has a much faster clock than the Xbox One (enough to make up the architecture differences in raw numbers) and the better bandwidth on the bulk of its main ram, 74 GB/s versus 68 GB/s. Obviously the Xbox One and PS4 have considerably more ram.

The move engines do not save bandwidth. They're part of main memory access, not over and above it. When the move engines are operating at 25.6 GB/s (max speed), you can only access the DDR3 at 42.4 GB/s, not 68 GB/s.

That puts a further hole through SenjutsuSages argument.

 

[TheD]

Less flawed than a comparison to the 7790. Architecturally the XO is somewhere between the 7770 and 7790 though raw performance wise it us exactly the same as the 7770. Architecturally the PS4 is somewhere between the 7850 and 7870, but with raw performance numbers that are above the 7850.

You're accounting for the negative differences with the 7770, but not the positive. The fact that the 7770 has a much faster clock than the Xbox One (enough to make up the architecture differences in raw numbers) and the better bandwidth on the bulk of its main ram 74 GB/s versus 68 GB/s. Obviously the Xbox One and PS4 have considerably more ram.

Don't forget that the HD7770 also does not have to share it's RAM bandwidth with the CPU, unlike the GPU in the XB1.

 

[randiangamer]

"Best looking" is subjective.
I think Beyond: Two Souls is the best looking console title but that's not out yet so for me it would probably be The Last of Us/Uncharted 3 (they look about the same) or Killzone 3.

Exactly, thats why i said "if I have to choose..."

 

[malboroking]

The move engines do not save bandwidth. They're part of main memory access, not over and above it. When the move engines are operating at 25.6 GB/s (max speed), you can only access the DDR3 at 42.4 GB/s, not 68 GB/s.

Its amazing what people will believe when you give something a fancy name. Xbox One had the move engines and PS4 has hUMA, and probably 90% of the people that talk about these two things have absolutely no clue how they actually work =P

 

[Amir0x]

We know the PS4 GPU is more powerful. We'll find out soon enough just how fast when the multiplats start shipping.

that's not how you find out how fast something is lol

 

[NingenJanai]

Its amazing what people will believe when you give something a fancy name. Xbox One had the move engines and PS4 has hUMA, and probably 90% of the people that talk about these two things have absolutely no clue how they actually work =P

Yeah, but hUMA is pretty much automatic, while the Xbox One move engines require a lot of effort from the programmers, as far as I know, I might be wrong.

 

[nib95]

that's not how you find out how fast something is lol

Comparing GPUs that are architecturally and performance similar to the console GPUs is useless, even for a rough estimation of differences, but comparing launch multiplatform titles isn't. Oh dear...

 

[Skeff]

Comparing GPUs that are architecturally and performance similar to the console GPUs is useless, even for a rough estimation of differences, but comparing launch multiplatform titles isn't. Oh dear...

up ported cross gen games right?

 

[randiangamer]

nope, not technically, not artistically, not anything, plenty of other games are way beyond Halo 4 in the "best looking" category.

Thats your opinion. Halo 4 artwork is , IMHO, the best i have seen up to date.

 

[bonus_sco]

Yeah, but hUMA is pretty much automatic, while the Xbox One move engines require a lot of effort from the programmers, as far as I know, I might be wrong.

Not that much, they're asynchronous. You start an operation and do something else until it's finished.

 

[bonus_sco]

Yeah, but hUMA is pretty much automatic, while the Xbox One move engines require a lot of effort from the programmers, as far as I know, I might be wrong.

What part of PS4s coherent memory access can't the Xbox One do?

 

[NingenJanai]

Not that much, they're asynchronous. You start an operation and do something else until it's finished.

Thanks, I think I got it mixed with the ESRAM.

 

[Mung]

If first gen multiplats have a big difference that will be surprising.

If they are identical though, that does not mean that the specs difference is meaningless. There could be many reasons why these games are similar e.g. forced parity.

I would expect differences to become greater over time, especially if the ps4 userbase becomes much larger.

 

[gofreak]

If you're discounting the move engines, then you know a lot less than I thought. The various copy operations and especially the dedicated texture decompression on one of the move engines is something that helps the GPU in ways that a 7770 doesn't benefit from. Texture decompression is either performed on the CPU, or can be done on the GPU through computation.

Textures are decompressed by a vanilla gcn TMU automagically, not on CPU or in gpu alu.

 

[Vizzeh]

Comparing GPUs that are architecturally and performance similar to the console GPUs is useless, even for a rough estimation of differences, but comparing launch multiplatform titles isn't. Oh dear...

Comparing multiplats at all won't be a good test even in year 3 when some difference should be evident. You will need to compare the 1st party x1 vs 1st party ps4, year 1, 2, 3 as they will tweak for their own hardware inc maximising their GPU not a port for both systems that likely wont push the cards they will prob only entail resolution, particle, fps differences.

 

[Majmun]

So the gap in power between hardware is much more wider this time around? The Ps3 and X360 were pretty close, if I understand correctly.

Ps4 exclusives are going to look like crazy. I remember Ps3 exclusives looking better than X360 exclusives. Maybe that has to do more with devteam talent instead of hw power, though

 

[longdi]

Any insider sources know why MS chose such an underpowered box if they wanted to release head to head with PS4 and still charged unsuspecting gamers 100 bucks more?

Albert's claims of smart people working on Xbox/DirectX...how could they make such a big mistake?

50% difference, i believe we are going to see something like GC and Xbox type of difference in graphics and performance.

 

[nib95]

Textures are decompressed by a vanilla gcn TMU automagically, not on CPU or in gpu alu.

Sometimes I don't know why I bother with SenjutsuSage... that's two big points in his argument he's had shot down in one page. It's like every other week he'll trawl some new B3D/TeamXbox forum posted titbits to form a new argument in defence of the Xbox One that also gets shot down almost immediately.

No joke, I've seen this happen dozens of times with him now over the months. He was a key defender and promoter of many of the now infamous secret sauce potentials and possibilities.

 

[artist]

Any insider sources know why MS chose such an underpowered box if they wanted to release head to head with PS4 and still charged unsuspecting gamers 100 bucks more?

Albert's claims of smart people working on Xbox/DirectX...how could they make such a big mistake?

50% difference, i believe we are going to see something like GC and Xbox type of difference in graphics and performance.

Different priorities.

Sometimes I don't know why I bother with SenjutsuSage... that's two big points in his argument he's had shot down in one page. It's like every other week he'll trawl some new B3D/TeamXbox forum posted titbits to form a new argument in defence of the Xbox One that also gets shot down almost immediately.

No joke, I've seen this happen dozens of times with him now over the months. He was a key defender and promoter of many of the now infamous secret sauce potentials and possibilities.

I dont even bother to read through his posts anymore. CTRL+F, CTRL+F!

 

[Cesar]

Notsureifserious.gif

because is pure BS

It's amazing how the hardest of hardcore MS fans were hanging their hopes on latency in the RAM comparison, yet somehow don't think latency applies to the cloud.

We have had Turn 10 say they use the could to calculate the AI in Forza 5, loading of more power to use for graphical effects. It IS true guys, except it.

 

[bonus_sco]

Textures are decompressed by a vanilla gcn TMU automagically, not on CPU or in gpu alu.

They sure are, the Move Engines can swizzle textures during copy. Useful for packing data for optimal GPU access and unpacking for optimal CPU access (if accessing linearly on the CPU).

 

[Chobel]

What part of PS4s coherent memory access can't the Xbox One do?

eSRAM, CPU can't access eSRAM.

 

[nib95]

We have had Turn 10 say they use the could to calculate the AI in Forza 5, loading of more power to use for graphical effects. It IS true guys, except it.

Did you hear guys, a Microsoft studio said it so 'except'!!

 

[KKRT00]

This topic really got to 50 pages overnight? Console wars, serious business.

PS4 is more powerful, end of story, but its not 50% and wont be probably more than 30% in most real world scenarios.

 

[Majmun]

We have had Turn 10 say they use the could to calculate the AI in Forza 5, loading of more power to use for graphical effects. It IS true guys, except it.

I really doubt it's something significant. Improving HW capabilties throught the internet sounds like hocus pocus

Why upgrade HW when we can upgrade our internet connection instead?

 

[Skeff]

We have had Turn 10 say they use the could to calculate the AI in Forza 5, loading of more power to use for graphical effects. It IS true guys, except it.

It's playable offline just FYI.

 

[bonus_sco]

eSRAM, CPU can't access eSRAM.

Right, but that's an extra on the Xbox. The DDR3 is coherent if you want it to be. The ESRAM should be used as a GPU scratch pad.

I can't imagine many scenarios where you'd want the CPU accessing it. That's the point of the move engines with their swizzling, to transfer data between optimum cache access layouts.

 

[Finalizer]

We have had Turn 10 say they use the could to calculate the AI in Forza 5, loading of more power to use for graphical effects. It IS true guys, except it.

Consider the game has to function offline. It has to be functional without those cloud-based resources. It's certainly not getting increased graphical fidelity through some remote server.

 

[nib95]

This topic really got to 50 pages overnight? Console wars, serious business.

PS4 is more powerful, end of story, but its not 50% and wont be probably more than 30% in most real world scenarios.

According to Edges sources it's 50% difference at the moment. Raw performance numbers put the PS4 at just under 40% more powerful. Out of curiosity, what is your evidence to support the notion that it won't even be a 30% difference in real world circumstances despite what the OP and several other insiders have said?

 

[TacticalFox88]

This topic really got to 50 pages overnight? Console wars, serious business.

PS4 is more powerful, end of story, but its not 50% and wont be probably more than 30% in most real world scenarios.

lol, once Guerilla Games, Santa Monica, Naughty Dog, and Polyphony Digital figure out all the tricks and kinks, it's going to leave EVERYONE in the dust.

 

[Hollow]

We have had Turn 10 say they use the could to calculate the AI in Forza 5, loading of more power to use for graphical effects. It IS true guys, except it.

I don't know why I found this as funny as I did.
*accept, by the way.

 

[bigboss370]

as far as polyh3dron is concerned, 60FPS is very important for racing games.

if its looking and running so smoothly like this:

then 60fps really isn't that important to me.

 

[Kaako]

Looking forward to the first party games which make the best use of the given system at hand. Fun times ahead gentlemen, fun times ahead.

 

[Mung]

We have had Turn 10 say they use the could to calculate the AI in Forza 5, loading of more power to use for graphical effects. It IS true guys, except it.

So presumably the AI cars start racing like morons if your internet cuts out?

 

[bigboss370]

lol, once Guerilla Games, Santa Monica, Naughty Dog, and Polyphony Digital figure out all the tricks and kinks, it's going to leave EVERYONE in the dust.

with Killzone SF looking as it does as a launch title, I can't even comprehend what we're getting later on. The moment they showed KZ in Feb, I was sold. Next gen is here.

I'm sure MS will have some nice looking games too. Ryse looks nice. But yeah when it comes to Sony first party, they seem to operate on another level. Oh, along with Quantic Dream.

 

[Amir0x]

We have had Turn 10 say they use the could to calculate the AI in Forza 5, loading of more power to use for graphical effects. It IS true guys, except it.

let me ask you a simple question, Cesar.

What do you think is going to happen to Forza 5's "graphical effects" when you play it offline, since the game is completely playable offline? Do you think it's going to be a downgrade? Do you believe when playing offline, all these snazzy visual effects will disappear due to the lack of the POWER OF THE CLOUD™?

 

[malboroking]

We have had Turn 10 say they use the could to calculate the AI in Forza 5, loading of more power to use for graphical effects. It IS true guys, except it.

Meridian 59 had server side AI in 1996, this isn't anything new. Anyone who believes the cloud is going to help the visuals of any Xbox one game is fooling themselves. Not to mention the fact that the PS4 could do it as well if they wanted. Hell, PS3 and 360 could do it too.

 

[EmptySpace]

This topic really got to 50 pages overnight? Console wars, serious business.

PS4 is more powerful, end of story, but its not 50% and wont be probably more than 30% in most real world scenarios.

ps4 games will have better systems, not just prettier graphics and faster performance.

the 8 aces will definitely work their magic. ps4 titles might not *look* like they are 50% better, but the execution of better ai, or more complex collision detection or other physics sim, etc. i.e. better looking liquid or fire effects, will be possible on ps4. the sum of those parts will enable gamers to have a better, more immersive experience.

 

[Mung]

let me ask you a simple question, Cesar.

What do you think is going to happen to Forza 5's "graphical effects" when you play it offline, since the game is completely playable offline? Do you think it's going to be a downgrade? Do you believe when playing offline, all these snazzy visual effects will disappear due to the lack of the POWER OF THE CLOUD™?

Don't bother Amirox.

Not one person who talks about 'the power of the cloud' has been able to give an answer to this simple and obvious question so far.

 

[Coulomb_Barrier]

This topic really got to 50 pages overnight? Console wars, serious business.

PS4 is more powerful, end of story, but its not 50% and wont be probably more than 30% in most real world scenarios.

It could be slightly bigger than 50% with PS4's superior Compute functionality being fully taken advantage of.

 

[PairOfFilthySocks]

if the "PC" version of the Xbox One controller coming out in 2014, is wired and works with the console as well, would be a huge plus.

wait, is the regular pad not compatible with PCs?

 

[TheD]

We have had Turn 10 say they use the could to calculate the AI in Forza 5, loading of more power to use for graphical effects. It IS true guys, except it.

You have no idea what you are talking about.

What Forza 5 does is upload replays and then tries to come up with a profile of people's drive styles based on those replays.
It has nothing to do with off loading the in game AI calculations (and even if it did, the AI is done by the CPU and thus it would not have an effect on the graphics!)

 

[Chobel]

Right, but that's an extra on the Xbox. The DDR3 is coherent if you want it to be. The ESRAM should be used as a GPU scratch pad.

I can't imagine many scenarios where you'd want the CPU accessing it. That's the point of the move engines with their swizzling, to transfer data between optimum cache access layouts.

Mind explain more about the bolded part? why do you need to swizzle?