PlayStation 4 (codename Orbis) technical hardware investigation (news and rumours)

Status
Not open for further replies.
Never seen any claims or data that this arrangement whould impact bandwidth; TBH I don't think it does.
Without the technical knowledge to confirm/refute this, I turn to basic deduction. If PS4 now is 170 ish GB/s, and we're to believe that's half data rate, then the GDDR5 in PS4 was capable of 340 GB/s... That'll be a 'no' then. ;)
 
The 7th core has been unlocked in the lastest SDK
I knew it!

I need more info to believe it.

The clues were already there since around one year ago. Remember last year around october when PS4 games really suffered performance wise, notably with AC Unity and GTA5 (advanced warfare too) with ~4-5fps deficit at some points in both of those games, allegedly during CPU heavy moments?

Suddenly after 2 miraculous patches the performance improved dramatically in those 2 games but only for the PS4 version. It's less know but performance dramatically improved also in the subsequent Advanced warfare patches, again only for PS4 version (SP mode).
 
Well better understood of dev tool and/or better dev tool can give the same with this type of bottlenecks … so no "evidence" by this for a 7th Cpu unlock. ;)
 
Clamshell mode. :) Essentially, half the data bus goes to each GDDR chip with the address bus shared between the two. Each RAM chip sits on top of the other on either side of the PCB. Never seen any claims or data that this arrangement whould impact bandwidth; TBH I don't think it does.
Yes, clamshell mode's operation is well documented. Bandwidth remains a constant in either x16 (clamshell) or x32 mode and it's entirely about increasing the addressable memory, i.e. memory capacity. You can aslo vary the configuration of memory controllers to allow asymetric configurations.

We use a custom RAM configuration but it shares some of the design traits of GDRR5.
 
I recall 3dilettante saying that the original launch PS4 16x 4gbit gddr5 modules are in stripe mode or duplex mode or something of that sort,

This was known from the moment Cerny announced 8gig. 16 chips, eight on each side of console. Teardown pics confirmed that on launch.

CUH1200 model eliminated chips on one side, with 8 1GB chips being placed on the other side of APU socket.
 
Wonder if it's a full core or just some of it like X1, and that's also very recent like 10 days ago. Just in time for Uncharted 4 :smile:
 
Great find. I think this marks the time when Sony has finalised their plans for this-gen's development of PS4 OS services. I presume the last big thing they had to think about was PSVR.

What is the latest info we have about game/OS memory reservations?
 
I am pretty sure the ICE team @ND got a heads up for what was going to happen with the latest SDK though. Let's say this SDK was released early to mid November, i am pretty sure most first party studios would know about that in October if not earlier. You may still be right but i don't see why they'd let potential perf go to waste when we are still 4 full months away from release.
 
Now that i think about it, we got the same people working on it that went from this to this in 3 months. No, if they don't make use of it I'll be flat out disappointed, even if we still don't know what sort of an impact it'll have (major difference for some games, very minor for others i suppose).
 
@DieH@rd noticed you made a post about this here on Gaf http://www.neogaf.com/forum/showthread.php?t=1146956. Just wondering if you wanted to credit back to the gent who dumped it here and left; we often credit that forum as well when our information comes from there.

Also, some minor corrections on the verbiage around the SDK announcements for Xbox that you put in your OP, most of the important information were gathered from leaks. We shouldn't know today that XBO unlocked a 7th core if it weren't from the H4LTED leak. The two companies don't operate all that differently when it comes to revealing the capabilities of their systems.

Edit: thanks
 
Last edited:
Now that i think about it, we got the same people working on it that went from this to this in 3 months. No, if they don't make use of it I'll be flat out disappointed, even if we still don't know what sort of an impact it'll have (major difference for some games, very minor for others i suppose).

Probably minor everywhere. It's not even like unlocking 1/14 of the GPU or anything (assuming it's only half a core or something they're giving back).

Keeping up with the Joneses though. Now Xbox will only have a slight clock advantage (the part that will never go away) and it will be harder to detect in practice then when they had clock+7th core advantage, which was pretty minor to start. You might see a FPS or two difference somewhere when CPU limited...
 
Also, some minor corrections on the verbiage around the SDK announcements for Xbox that you put in your OP, most of the important information were gathered from leaks. We shouldn't know today that XBO unlocked a 7th core if it weren't from the H4LTED leak. The two companies don't operate all that differently when it comes to revealing the capabilities of their systems.

FMOD also have this in the same page
22/01/15 1.05.11 - Studio API patch release
Features:

  • Lowlevel API - Xbox One - Added access to 7th CPU core.
 
Probably minor everywhere. It's not even like unlocking 1/14 of the GPU or anything (assuming it's only half a core or something they're giving back).

Keeping up with the Joneses though. Now Xbox will only have a slight clock advantage (the part that will never go away) and it will be harder to detect in practice then when they had clock+7th core advantage, which was pretty minor to start. You might see a FPS or two difference somewhere when CPU limited...

debbie-downer.png
 
Is there any reason to think aren't using a job system similar to that on Cell that'll scale to available resources?

They are using fiber now but it is probably scalable. They will use it. The FMOD library is using it since 17th November. It is a good news. I hope they will do the same with Os reserved memory if they can give back 512mo to 1Gb to developer it would be a good thing.
 
Without the technical knowledge to confirm/refute this, I turn to basic deduction. If PS4 now is 170 ish GB/s, and we're to believe that's half data rate, then the GDDR5 in PS4 was capable of 340 GB/s... That'll be a 'no' then. ;)
Well obviously not 340GB/s. For me to believe that would be utterly utterly utterly utterly utterly ridiculous. Slight impact like I said in my original posts above. It was a single post over 1.5 years ago and I'm amazed I even remember anything about it, and the exact contents of his posts contents is now muddy and I dont exactly recall what he said.

This was known from the moment Cerny announced 8gig. 16 chips, eight on each side of console. Teardown pics confirmed that on launch.

CUH1200 model eliminated chips on one side, with 8 1GB chips being placed on the other side of APU socket.

Obviously it is known I was just giving as much description of the physical memory chips on the pcb so that I'd have the greatest possible success in getting the term I was looking for. The term I was looking for was clamshell mode.
 
Status
Not open for further replies.
Back
Top