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

Status
Not open for further replies.
Sony shouldn't know about XB1's innards beyond what MS have publicly revealed. You generally don't talk about your rivals hardware (MS haven't said anything about Sony's audio, for example). Also this is the PS4 thread, and not any comparison thread. XB1 is irrelevant in this thread except where ti gives clues as to what is in PS4.
 
Sony shouldn't know about XB1's innards beyond what MS have publicly revealed. You generally don't talk about your rivals hardware (MS haven't said anything about Sony's audio, for example). Also this is the PS4 thread, and not any comparison thread. XB1 is irrelevant in this thread except where ti gives clues as to what is in PS4.
It's something I've noticed. MS have not mentioned Sony at all, Sony have gone on the attack though.
Especially the last conference, I think it's also the first time I heard them say that they have "the most powerful console" also.
It did make me wonder if they felt more confident to say that now, due to having more details about the XB1 from under the table.
As I don't think you could make such a bold statement and be sure of, with what is publicly known.
 
I'm uncertain at this point because I don't know what set of features counts in the eyes of AMD for a hUMA setup. If Kaveri were released and its features well-publicized, that would be different.
AMD didn't really point to any specific set of low-level gotchas that makes an APU not hUMA.

At this point, the APUs being built this year should have much of the featureset already in place, and it's down to subsets of functionality like fault handling and what kind of coherence implementation AMD considers good enough between the GPU and CPU.

Another reaon for Sony not disclosing hUMA, if it's actually present, is that hUMA is an AMD evangelization term, and neither Sony or Microsoft follow that very strongly.

For example, Microsoft has Tiled Resources instead of PRT, and Sony hasn't really discussed it as its own thing.
 
Sony shouldn't know about XB1's innards beyond what MS have publicly revealed. You generally don't talk about your rivals hardware (MS haven't said anything about Sony's audio, for example). Also this is the PS4 thread, and not any comparison thread. XB1 is irrelevant in this thread except where ti gives clues as to what is in PS4.

well im bringing up the hUMA article and the validity of it. Since Sony has been more than eager to bash Microsoft and make claims about how PS4 is the most powerful console, I don't know why they wouldn't of pointed this out. That's all and that's relevant.
 
As I noted earlier, hUMA is more AMD's deal than either console manufacturer.
The other is that, at least officially, that is a detail Sony shouldn't know, and AMD probably shouldn't be discussing it.

This wouldn't be the first time there were alleged leaks from AMD that compared the two console projects, and it is not a good thing to repeatedly happen at a company that is supposed to juggle multiple confidential projects for competing clients without leaking information between them.

Leaks are bad enough, but leaks that reveal a less than adequate isolation of design groups hurts everyone involved.
AMD staff being named as blabbing a comparison is also not good.
 
This was touched on before, but what kind of real performance impact would a GPU OS reservation have? Would it be a flat figure, like 10% of its cycles? If so, isn't that overkill? That's practically an entire RSX's level of performance just for the OS.. Could it be less?

Or am I misunderstanding this- could there be no 'reservation' on the GPU for the OS and its just generated on the spot (this I find unlikely)?
 
I do not see any very strange matter in the article...

MS surely need something 'special' on northbridge level, to deal with the embedded ram, so the 'default' hUMA design would likely not be suitable.
But that hardly prevents a custom one, I think.
 
Chinese walls

If they were adequate, we wouldn't get leaks or statements that show a single individual in the company synthesizing data from both sides. Maybe, per South Park, there are Forum Mongol Warriors that pop up whenever someone builds a Console Chinese Wall.
 
We'll have to wait until we can compare Kaveri and Kabini to see where AMD has drawn the line. The fundamental building blocks should be present to do much of what hUMA promises, so now the question is what additional functionality is missing.
 

I'd say that article is just about as shady as the original German one this is all based on.

There's just one problem with these various reports (VR-Zone, ExtremeTech): they're incorrect. After sending some emails to our representatives at AMD I was told that "Kabini doesn't support hUMA" which is the APU that both the PS4 and Xbox One processors are based on. AMD further clarified with us: Our spokesperson made inaccurate statements about our semi-custom APU architectures and does not speak for Microsoft, Sony or the AMD semi-custom business unit responsible for co-developing the next generation console APUs.

Kabini has nothing to do with anything here.

As usual many times these articles strain to get something the actual statements dont support.
 
Both Xbox and PS4 have a custom APU, the fact that Kaveri doesn't support hUMA is irrelevant.
 
Kabini has nothing to do with anything here.

As usual many times these articles strain to get something the actual statements dont support.
I may be projecting, but if Kabini isn't relevant, I probably wouldn't bring it up when asked about the consoles.
 
I may be projecting, but if Kabini isn't relevant, I probably wouldn't bring it up when asked about the consoles.

Or these "representatives" aren't fully informed/aware.

Unless the representative is say, Dave Baumann. Then yeah I accept it :LOL: Because we know he's sophisticated enough to nuance this.

I was told that "Kabini doesn't support hUMA" which is the APU that both the PS4 and Xbox One processors are based on.

We know that's pretty false for sure in Xbox case, with the ESRAM.

PS4 and X1 are different SOC's to KAbini. AMD should refer to PS4 and X1 when speaking about them. Such as "no PS4 SOC does not have hUMA" or "Yes PS4 SOC does have hUMA"
 
We know that's pretty false for sure in Xbox case, with the ESRAM.
That would be a topic for the Durango thread, but there would need to be more than that. If the northbridge and CPU portions share the same base architecture, what happens in the GPU section shouldn't eliminate the family resemblance.

PS4 and X1 are different SOC's to KAbini. AMD should refer to PS4 and X1 when speaking about them. Such as "no PS4 SOC does not have hUMA" or "Yes PS4 SOC does have hUMA"

Why would the PS4 be that different? The Onion and Garlic bus organization looks well in line with Kabini, as do the CPUs. If the northbridge's system request queue and MMU is close for both SOCs, a reason for Kabini to not have hUMA can readily carry over.
 

AMD actual statement was (below), and not that gibberish BS in that link.

"During a recent Gamescom 2013 interview, an AMD spokesperson made inaccurate statements regarding the details of our semi-custom APU architectures.

"AMD will not comment on the Microsoft Xbox One and Sony PS4 memory architectures and will not speak for Microsoft, Sony or other AMD customers."

http://www.nowgamer.com/news/2052820/ps4_vs_xbox_one_amd_retracts_performance_comparison.html

Sounds like AMD is doing damage control ...for the 3rd time. :LOL:

FYI: I have this strong feeling Marc Diana is telling the truth, but got busted on it - like a few others.
 
Last edited by a moderator:
...just to understand better:
all APUs has garlic&onion bus, and fusion APU design (llano&co) is detailed here:

http://amddevcentral.com/afds/assets/presentations/1004_final.pdf

What the ... is hUMA over this?
Zero-copy is already supported in llano, as far as such slides can say. So Zero-copy is not hUMA.

GPU can already read shared CPU memory via onion.
CPU access to GPU/UC memory is already detailed in slides, and it is slow due to the need of WC flush down to memory, I think.

So, what is hUMA more than what presented in such slides... The ability to do GPU memory access without flushing WC? The ability to i.e. remap GPU&CPU memory dynamically? Page fault support on GPU?

I really cannot understand it well... any idea?
 
Maybe the ability to probe the GPU caches (for pages marked for fully coherent access, maybe it's enough if the GPU automatically writes through if it writes to such a page [currently, it has to be determined manually, so it is not impossible to maintain coherency, it's just not as simple]). At least that's how I would interprete the hUMA slides.
 
Status
Not open for further replies.
Back
Top