• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

New consoles = better CPU optimization?

just cause nvidia allow the instruction set to run on the amd stuff on the ps4s theres no reason to assume that they will release it for use on pc systems we all know that just cause something can be done doesnt mean that they will allow it in the drivers
 
I don't think the current SDK of PhysX is X87, they'd just be using that SDK on Jaguar.

Maybe they'd make it more threaded, but they were doing that anyway apparently.


just cause nvidia allow the instruction set to run on the amd stuff on the ps4s theres no reason to assume that they will release it for use on pc systems we all know that just cause something can be done doesnt mean that they will allow it in the drivers


You've got the wrong end of the stick.
 
I don't think the current SDK of PhysX is X87, they'd just be using that SDK on Jaguar.

Maybe they'd make it more threaded, but they were doing that anyway apparently.

That's the whole point, Jaguar is far to slow for Nvidia PhysX as it currently is, an FX-8350 can barely keep up 30 FPS on Nvidia PhysX, what do you think is going to happen when its introduced to a CPU with 70% less performance?

Nvidia will have to think again.
 
Software mode or not it still needs the support of hardware, just like emulators need hardware to run.

PhysX on consoles uses the IBM CPU instruction sets, which AMD CPU's do not have, I believe its x87, you know as well as I do how slow x87 emulation is on AMD CPU's
Nvidia will have to use the instruction sets on AMD CPU's, if it ends up good enough on an AMD Jaguar CPU core with 30% the performance of a Piledriver core, then its way more than enough for a Piledriver core.

PhysX on the PS4 for instance is a completely new fork and isn't a direct port of the build used on the xbox or the one built for the PS3. I don't know off hand how many cores it will utilise but multi-threading support with PhysX on the PC when processed on the CPU is to some degree down to the developer rather than completely down to the physics API itself so it doesn't automatically follow for definite that the version on the PS4 and/or xbox one will only run on one core.

I think AMD went down this rout quite deliberately (using 8 VERY slow cores)

http://www.eurogamer.net/articles/2...ross-platform-play-teases-character-transfers

With some games as they currently are, running them on the PS4 would be like playing Crysis 3 on a 7950 supported by a Sempron CPU.
even with 4 of those PS4 cores loaded up its still like running C3 with a SB Pentium.

Developers will have to get multithreaded, that will benefit Intel as well as AMD, but it will also make AMD much more competitive than what they currently are.

That would backfire spectacularly as the lack of multi-threading in games is as much or more so due to the impossibility or impractibility of threading a lot of the way game engines have to work as it is legacy or lazy coding.

That's the whole point, Jaguar is far to slow for Nvidia PhysX as it currently is, an FX-8350 can barely keep up 30 FPS on Nvidia PhysX, what do you think is going to happen when its introduced to a CPU with 70% less performance?

Nvidia will have to think again.

Software PhysX isn't that slow relative to other physics APIs for processing physics infact its faster than all but the most optimised libraries, they may be able to opptimise it more than it is on the CPU but you aren't going to see an order of magnitude more performance and for some reason they went with a GPU on those consoles that doesn't have the compute capabilities to properly handle rendering + next gen shaders + any decent level of hardware physics at the same time.
 
Last edited:
That's the whole point, Jaguar is far to slow for Nvidia PhysX as it currently is, an FX-8350 can barely keep up 30 FPS on Nvidia PhysX, what do you think is going to happen when its introduced to a CPU with 70% less performance?

Nvidia will have to think again.

Are you talking when the CPU is doing hardware acceleration? Because that's moot, as it wouldn't be on console.
And PD's performance doing hardware acceleration is moot again, as it shouldn't be.
 
Last edited:
PhysX on the PS4 for instance is a completely new fork and isn't a direct port of the build used on the xbox or the one built for the PS3. I don't know off hand how many cores it will utilise but multi-threading support with PhysX on the PC when processed on the CPU is to some degree is down to the developer rather than completely down to the physics API itself so it doesn't automatically follow for definite that the version on the PS4 and/or xbox one will only run on one core.



That would backfire spectacularly as the lack of multi-threading in games is as much or more so due to the impossibility or impractibility of threading a lot of the way game engines have to work as it is legacy or lazy coding.



Software PhysX isn't that slow relative to other physics APIs for processing physics infact its faster than all but the most optimised libraries, they may be able to opptimise it more than it is on the CPU but you aren't going to see an order of magnitude more performance and for some reason they went with a GPU on those consoles that doesn't have the compute capabilities to properly handle rendering + next gen shaders + any decent level of hardware physics at the same time.

I don't see how it will back fire, or on who. some devs are lazy or under skilled.

its kind of "this is what it is, deal with it" sometimes a good kick up the back side, where people are forced to change and modernise is the only way forward, we have had multicore CPU's for a decade, for whatever reason there are still some game devs out there stuck in the Pentium 4 era. they need that proverbial kick.

Are you talking when the CPU is doing hardware acceleration? Because that's moot, as it wouldn't be on console.
And PD's performance doing hardware acceleration is moot again, as it shouldn't be.

Its not moot as emulation does not happen without hardware. software emulators do not magically number crunch and render in an empty vacuum. they need calculation hardware like a CPU / GPU to do their intended job, the better that hardware the better it will do it.

A car does not function without its engine.
 
Last edited:
The PhysX that runs on CPU's now when you're running software mode doesn't run X87 as far as I know, if it did AMD CPU's would get crippled in games where software PhysX (Because you can't turn it off) is running and they don't.

They could pretty much just use the SDK we use now on Desktop.
 
I don't see how it will back fire, or on who. some devs are lazy or under skilled.

its kind of "this is what it is, deal with it" sometimes a good kick up the back side, where people are forced to change and modernise is the only way forward, we have had multicore CPU's for a decade, for whatever reason there are still some game devs out there stuck in the Pentium 4 era. they need that proverbial kick.

I mean literally unlike many workloads where you can leave threads in flight and catch up with them later there are much stricter orders to how you have to process a large amount of game relevant data and tight time constraints to get things processed within which makes it much harder and in many cases impossible to get significant gains from multi-threading*. There are some games where the nature of the game means you can make use of multi-threading to gain better performance but thats far from a majority case.

(I have actually coded a game engine from scratch and know the code of the idtech2-4 engines like the back of my hand so I do know a bit about what I'm talking even tho I wouldn't consider myself an expert on the subject).



* Thats not to say significant gains can't be made - but anyone betting their business on game engines in general being opptomised to run on lots of low powered cores is barking very much up the wrong tree. (If you look at the technical details of Steamroller its apparent AMD is very aware of this anyway - some of the architecture changes very much have this in mind).
 
Last edited:
I mean literally unlike many workloads where you can leave threads in flight and catch up with them later there are much stricter orders to how you have to process a large amount of game relevant data and tight time constraints to get things processed within which makes it much harder and in many cases impossible to get significant gains from multi-threading*. There are some games where the nature of the game means you can make use of multi-threading to gain better performance but thats far from a majority case.

(I have actually coded a game engine from scratch and know the code of the idtech2-4 engines like the back of my hand so I do know a bit about what I'm talking even tho I wouldn't consider myself an expert on the subject).



* Thats not to say significant gains can't be made - but anyone betting their business on game engines in general being opptomised to run on lots of low powered cores is barking very much up the wrong tree. (If you look at the technical details of Steamroller its apparent AMD is very aware of this anyway - some of the architecture changes very much have this in mind).

Your more of an expert than I am in that, so i'm happy to yield to you...

I'm certainly not under the illusion that suddenly AMD will start to show Intel up, but I do think that over all AMD will be more competitive than they currently are.
Which is what I said to start with, and I feel you would agree with me on that.

AMD absolutely MUST get their performance per core / per clock up, significantly, to remain relevant.
Yes I also think AMD do now realise that, here's hopping SR ends up a significant step in that the right direction.
 
Your more of an expert than I am in that, so i'm happy to yield to you...

I'm certainly not under the illusion that suddenly AMD will start to show Intel up, but I do think that over all AMD will be more competitive than they currently are.
Which is what I said to start with, and I feel you would agree with me on that.

AMD absolutely MUST get their performance per core / per clock up, significantly to remain relevant.
Yes I also think AMD do now realise that, here hopping SR ends up a significant step in that the right direction.

I have high hopes for steamroller, they've paid a lot of attention to the details, so I hope it pans out. I don't see it showing up intel but on paper it certainly should be competitive with the latest comparable intel CPUs when CPUs appear using it.
 
I have high hopes for steamroller, they've paid a lot of attention to the details, so I hope it pans out. I don't see it showing up intel but on paper it certainly should be competitive with the latest comparable intel CPUs when CPUs appear using it.

Yeah, AMD have a lot to catch up,

On the 8 core I'm looking for + 20% (per clock / core) and 20% better power efficiency, and overclocking to 5Ghz.
while even that is still some way from Intel's cores I'm happy to buy one of those for upto £200, considering the 8 thread Haswell is £270.
 
PhysX on Consoles will be as PhysX is on consoles now, software, not hardware accelerated.

I thought this 3.0 was meant to be a decent change :p?

or they will just use the alternative that is hardware accelerated on amd gpus and nvidia it's called havok and theres already ps4 demos of it
 
Back
Top Bottom