Chobel
Member
Xbone CPU is clocked 9.37% higher than PS4 CPU, so it should be 107/98 dancers. Xbone has 6 more dancers than it should = more efficient.
-
Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.
Bink 2 runs faster on PS4 GPU than XB1 GPU, but same on either CPU
- Thread starter Insane Metal
- Start date
Xbone CPU is clocked 9.37% higher than PS4 CPU, so it should be 107/98 dancers. Xbone has 6 more dancers than it should = more efficient.
No. How would that even be possible? The benchmark is most likely running on debug consoles where the OS of both systems is interfering normally. The OS (i.e. the virtualization layer) was probably the reason why the Xbone's CPU was initially slower despite the clock speed advantage.
That's why I qualified it with "roughly".
The discrepancy is due to the Xbone's hidden dCPU (Dancer CPU) in the powerbrick.
IAmRandom31
Banned
And there will be a handful of devs who actually utilize GPGPU within their engine this generation.
Sony first party teams already are pushing GPGPU stuff. But yeh. Forget about it when it comes to 3rd parties. May be a few that do good stuff with it. But won't be many. Same as with the SPU's of Cell.
Cloth simulation is the type of thing SPE's would of blazed through and laughed at. Which is why it is right there with the current gen CPU's.
All will. It's becoming a standard technique that is available on all gaming platforms.
John Bender
Banned
Final results slide screencap:
fake edit: PS4 am 2 times more powerful confirmed!
Don't forget the hidden "Emotion Engine" inside the Xbone's HDMI cable.
IAmRandom31
Banned
Yeah, but it is extra programming. Extra work. From what I've seen over the years anything that isn't absolutely 100% necessary will be overlooked by the majority of developers.
Unless Sony builds a very easy API to take advantage of the Compute units of the architecture I just don't see it being used as much as it should. Hope I'm wrong though!
ShockingAlberto
Member
Wait
There's a Bink 2?
There's a Bink 2?
Canis lupus
Member
Glorious GPGPU future. Mark Cerny won all the gambles for this gen.
Yup.
But I only expect First Party/exclusive games to make full use of the feature.
Multiplats will always chase that parity going forward.
I wouldn't call them shit. From these numbers, it seems an i5 core is about 3 times faster, but that's why the PS4 and XB1 have 8 cores. As developers get used to using more cores, things will get better for the new consoles.
Yes, they are not nearly as impressive bang for the buck as the Cell or even the 360 CPU were, but that's the price we pay when both Sony and MS decided to break even from the start this gen.
Remember, folks, this is the GPU *COMPUTE* performance, and the advantage we are seeing for the PS4 makes perfect sense as we are talking 18 vs. 12 CUs. This DOES NOT mean the PS4 GPU is that much faster for graphics in games, as only 14 of the PS4's CUs can be fully utilized for those because of bandwidth limitations.
John Bender
Banned
Yup, like Dragon Age, Metal Gear Solid, Call of Duty, and so on.
shadowsdarknes
Member
Do the PS4 and Xbone allot the same number of cores for games?
Yes. There are multiple sources of good evidence for that.
AgentP
Thinks mods influence posters politics. Promoted to QAnon Editor.
This bullshit again? Ranger is that you?
Mr Moose
Member
Says 18 in the slides.
Yep. Don't think they used ESRAM
John Bender
Banned
Don't forget Guerrilla Games and Quantic Dream.
This has already been debunked, sir.
Canis lupus
Member
Are you talking about this
The slide says 2012, pretty sure this is outdated when it leaked. PS4 only had 4 gigs of gangsda ram 5.
No wonder, it is likely already crowded with more crucial data. From all we know, data synchronization between CPU and GPU is more costly, too.
lyrick
Member
How horribly unoptimized is Bink video if the differences between a GPU with 5.6 TF of compute and one with 1.8 TF of compute only offers a 12.5% decode difference?
And why is this not shown on the Developers site?
http://www.radgametools.com/bnkmain.htm
Insane Metal
Gold Member
That has been debunked since the console launch. All 18CUs are fully independent to be used in compute or graphics. No limitations. Anyway, it doesn't mean 100% more performance, yeah.
Developers can use all 18 CU's for rendering if they want. IIRC, Mark Cerny talked about that before.
Whether ESRAM is crowded most likely depends on what point in the process they are doing these calculations. Either way, it shows just how hard a shot to the knees the main memory bandwidth is on the Bone. ESRAM utilization is the key to performance on there and there isn't quite enough of it for most engines, lol.
Zalusithix
Member
A) TB is a measure of space - be it memory, or storage.
B) Just because hardware A has X times the theoretical total computational power of hardware B, not every algorithm can be expected to execute X times faster on hardware A. Things aren't that simple... nowhere near that simple.
They can, but the +4 donesn't help that much, if at all. It's not like they magically discovered more bandwidth since that 2012 slide was made.
(If they did, I missed it)
AgentP
Thinks mods influence posters politics. Promoted to QAnon Editor.
Let me know if you have evidence that every game engine has a bandwidth bottleneck at exactly 14CUs. You are pushing Mixtermdeia BS.
Show me one game with a memory bottleneck that leaves the CUs underutilized. I'm sure you have access to all major studio PS4 engine profiles.
lyrick
Member
Exchange the Bs with Fs, as brainfarts happen, then explain the shitty scaling.
I can almost see the 2.3ms to 1.4ms, but 1.6ms to 1.4ms with over 3x the computational power makes me think the scaling is shit. Or the revised numbers from the devs were just misinterpreted.
If they're supposedly seeing gains of over 30% [2.3ms to 1.6ms] when going from a 23% compute increase [1.3 TF GPU to a 1.8 TF GPU], but then only seeing a 12.5% gain with a 320% more computational power [1.8TF GPU to a 5.6TF GPU] something is a-fucking-miss.
You are wrong.
Mr Moose
Member
What's 1.6TF?
Zalusithix
Member
There could be a limit to how "wide" they can distribute the problem. Perhaps it maxes out at somewhere around the CU count of the PS4. From there the scaling to the PC might be a byproduct of the raw MHz jump and not from additional CUs.
lyrick
Member
Sorry 1.83TF... either way it's shit compared to the 5.6TF of the 290x
The CPU results already show massive gains on the 4 core i5 vs the jaguar, so it's not limited by CPU. In fact it somehow gets worse on the PC end when it comes to GPU compute.
Mr Moose
Member
But is this benchmark anything to do with the TFs of the GPUs? (I honestly have no clue lol).
GameSeeker
Member
Those numbers clearly show that for both next-gen consoles GPGPU is the way to go for physics simulations. The fact that the PS4 is almost 2X the Xbone shouldn't be surprising given that GPGPU was a key design point for the PS4. 2nd generation PS4 titles should really push GPGPU and do some amazing things.
Matt has spoken and that should end any and all discussion of 14+4. Thank you Matt.
Zombie James
Banned
Oh cool, it's almost like the PS4 has a better GPU or something.
John Bender
Banned
lol And we're done here!
lyrick
Member
It's certainly possible, but if that is the case that would basically mean that this particular Bench is worthless as it's almost a perfect situation for PS4s Architecture.
Yes number crunching for the code/decode process is directly attributable to the Theoretical FLOP/s maximum, probably even more so than the normal graphical bullshit bickering.
NeOak
Member
You seriously don't know how this shit works.
What about this?
http://www.cinemablend.com/games/PS...x-One-CPU-According-Benchmark-Test-61203.html
Claims the PS4 CPU is faster than the Xbox One.
Which would make sense given the many many rumors of PS4 having an uptick speed also.
http://en.wikipedia.org/wiki/PlayStation_4_technical_specifications#Central_processing_units
"The CPU's base clock speed is said to be 1.6 GHz, with an unknown uptick in speed for CPU-intensive usage"
http://www.cinemablend.com/games/PS...x-One-CPU-According-Benchmark-Test-61203.html
Claims the PS4 CPU is faster than the Xbox One.
Which would make sense given the many many rumors of PS4 having an uptick speed also.
http://en.wikipedia.org/wiki/PlayStation_4_technical_specifications#Central_processing_units
"The CPU's base clock speed is said to be 1.6 GHz, with an unknown uptick in speed for CPU-intensive usage"
I wish this was discussed earlier. I had no clue.
That Ubisoft dev in the other thread claims that Microsoft has freed some CPU allocation recently. http://www.neogaf.com/forum/showthread.php?t=913010
Funny, the source quoted in that Wikipedia articles doesn't say anything at all about an "uptick".
John Bender
Banned
By the way... is your statement regarding the PS4 CPU (Yes, you can get more out of the PS4's CPU than you can the Xbox's.) still up to date?
I think people in this thread have already figured that out.
LordOfChaos
Member
Really want to know the frame times for the Wii U CPU and GPU, are they available?
I loled at
I loled at
You know, apart from just about every cross platform game.
There you go.
Last year you said PS4 offered better CPU performance at the time. Is that still true after all the sdk updates and such?
Thanks.
Edit - Right thanks...