PlayStation 4 (codename Orbis) technical hardware investigation (news and rumours)
- Thread starter Love_In_Rio
- Start date
- Status
One LED in the proper housing can create the effect seen with a gradual turn on in brightness. You don't need to have multiple LEDs.
Sorry, I thought the LED "flowed". I just looked up a few videos and found that it didn't. Ya it could be done with one with some control to the brightness.
Blame Asia for not allowing me to get one on my hands yet!
It could be an optical delusion, but it looks to me as gradient is moving when LED is pulsing.
And that's what you get for using PC hardware. Could have told you that back in 2006 
Can someone answer me this;
So I was looking at Xbox One and PlayStation 4 die shots and I noticed that XO has SRAM 4MB while PS4 has none, is 4MB of SRAM L2 Cache or L3 Cache in Xbox One?
If its L2 Cache then it means that PS4 has no L2 Cache, if so then how it would work?
Thats all...
--
Anyway I read this interesting interview;
http://gamingbolt.com/project-cars-...splitting-renderer-across-threads-challenging
It seems Project CARS on PlayStation 4 uses only(for now?) 4 cores and it seems that have some minor difficulties, nothing serious to worry about...
(anyway I may not be able to respond since it seems I am going to be banned because "I am someone else who is banned in here". I bet I have a different provider, IP address, email and other things than that individual with which they confuse me thus I apologize in advance if I am unable to respond...)
L2 is a standard component of the Jaguar modules that are in both consoles so it will certainly be present in the PS4 also.
Being discussed here.
About the only thing I think it points to regards PS4 is that it seems 7 cores are available compared to XB1's six cores (14 MB/s versus 12 MB/s. Detailed discussion in the above thread).
hesido
Regular
Project Cars Dev also talks about having to copy mem to Garlic mem, I think this is to maximize bandwidth usage of the GPU, the garlic mem is not coherent with CPU caches so GPU can read mem at the max bandwidth available.
Here: http://gamingbolt.com/project-cars-...splitting-renderer-across-threads-challenging
Don't know about the final verdict on PS4 having HSA. but it worries me that they have to shuffle memory around. Well, "worry" is an over statement but I was expecting more..
I never had to program a language where I had to allocate mem, just a JS hobbyist here, so can someone explain why they have to shuffle things around?
Since they are working on a unified pool of RAM, can't they re-assign a certain fragment of memory as "garlic"? I mean, let the CPU do its thing, write back to RAM, declare that portion of memory as "garlic mem", and have the GPU read at full speed? Since there's ample amount of RAM available, can they "double buffer" this and have the GPU work on Garlic while the CPU is hammering another portion of that mem for the next frame? Or is this probably what they are doing and complaining about? Can't understand what the following means:
I don't know that that LCUE is Is the "memory copying" they mention just the CPU writing results back to RAM, or are they copying onion mem to garlic? Is re-assigning a mem portion as garlic/onion very costly, costly enough that copying data between two "pools" is cheaper?
Can somebody enlighten this poor soul?
Here: http://gamingbolt.com/project-cars-...splitting-renderer-across-threads-challenging
Don't know about the final verdict on PS4 having HSA. but it worries me that they have to shuffle memory around. Well, "worry" is an over statement but I was expecting more
..I never had to program a language where I had to allocate mem, just a JS hobbyist here, so can someone explain why they have to shuffle things around?
Since they are working on a unified pool of RAM, can't they re-assign a certain fragment of memory as "garlic"? I mean, let the CPU do its thing, write back to RAM, declare that portion of memory as "garlic mem", and have the GPU read at full speed? Since there's ample amount of RAM available, can they "double buffer" this and have the GPU work on Garlic while the CPU is hammering another portion of that mem for the next frame? Or is this probably what they are doing and complaining about? Can't understand what the following means:
I don't know that that LCUE is
Is the "memory copying" they mention just the CPU writing results back to RAM, or are they copying onion mem to garlic? Is re-assigning a mem portion as garlic/onion very costly, costly enough that copying data between two "pools" is cheaper?Can somebody enlighten this poor soul?
Last edited by a moderator:
Until someone more knowledgeable than me replies, I'll just say that the cache-coherent bus is there to reduce latency. Instead of the GPU having to go all the way out to main RAM to get access to some recently CPU-used data, it's in cache and immediately to hand. If the data you want isn't in the caches as its conventional data in RAM, you want to bypass this bus. Of course, whatever data the CPU holds in cache is data also being written to RAM, and the GPU can access the same data from RAM instead of from the cache which might be what they're talking about regards the copy, although I'm very uncertain about this and quite probably wrong.
Project Cars Dev also talks about having to copy mem to Garlic mem, I think this is to maximize bandwidth usage of the GPU, the garlic mem is not coherent with CPU caches so GPU can read mem at the max bandwidth available.
Don't know about the final verdict on PS4 having HSA. but it worries me that they have to shuffle memory around. Well, "worry" is an over statement but I was expecting more
I never had to program a language where I had to allocate mem, just a JS hobbyist here, so can someone explain why they have to shuffle things around?
Since they are working on a unified pool of RAM, can't they re-assign a certain fragment of memory as "garlic"? I mean, let the CPU do its thing, write back to RAM, declare that portion of memory as "garlic mem", and have the GPU read at full speed? Since there's ample amount of RAM available, can they "double buffer" this and have the GPU work on Garlic while the CPU is hammering another portion of that mem for the next frame? Or is this probably what they are doing and complaining about? Can't understand what the following means:
I don't know that that LCUE is
Can somebody enlighten this poor soul?
You might as well link to the original article so people have a better idea of the context.
hesido
Regular
It was quoted in the pjbliverpools quote, sorry:
http://gamingbolt.com/project-cars-...splitting-renderer-across-threads-challenging
Who? The benchmark coders? If the point of the benchmark is to measure performance of one single core then it's obviously not poor multithreading to not use more than one core.
Doh I'm sorry, some why I thought about Project Cars interview while typing this
- Status
