Predict: The Next Generation Console Tech
- Thread starter Acert93
- Start date
- Status
Am I being overly cynical, or am I right to expect "Official ___ Magazine" to write whatever gets their fanbase excited and so they can't be used as any source for real material?
Prophecy2k
Veteran
Yeah it looks like the magazine sourced all their rumours from Neogaf as well ;-)
PSM AFAIR is an unofficial Playstation Magazine. Regardless from a magazine designated for fans of a particular console it will do the same, unofficial or not
Anyways if we assume there is some truth in these there are a few things there that make me wonder such as
-There will be an online pass as a security measure
-Games have to be registered in the PSN account
I wonder what that would mean for people who wont have access to internet or dont have their consoles connected in general.
I mean...you wouldnt expect from kids and teens to know all the procedures and do them just to play their games. Accessibility is a necessity.
The "- no BC" also worries me, not in terms of my BR owned games but my PSN owned content.....hmm
This of course, but if I was a ps3 fanboy I would not be excited about the imposition of buying only fullpriced games
That is entirely realistic. I think whilst it would only yield about 1-1.5/2 Tflops on the outside the overall design would be pretty solid. I think the most important thing is the memory architecture rather as that will be key in terms of getting the most performance out of the system and RAM has become a very power intensive and expensive part of the average console. I think having 64MB of eDRAM embedded in an interposer which is connected to main memory might be a very good idea as they aren't likely going to want to shrink the eDRAM much going by past actions and it will allow them greater flexibility in manufacturing the APU. It'd be a very good use of eDRAM having a fast pool right on top of the main memory bus, it should give good performance and allow them to integrate newer memory architectures without requiring them to do too many changes to the Kabini Microarchitecture. I don't think they'd have time to validate Jaguar cores with DDR4 but if they use an interposer that won't be a problem and they can go right up to the wire on DDR4 availability whereas if they had to validate a CPU it'd likely need to be nearly finished now based on the timescales I have heard for validating CPU designs.
Any such security is unlikely to be 'buy new only'. Rather, it'll be a 'pay $x license fee to the publisher' akin to Online Pass. When you consider the retail price of a disc includes production, distribution, pressing fees etc., and these are covered with the initial purchase, the only fiscal consideration in preowned is lost revenue to the publisher (and maybe console company) from not selling a new game. Forcing gamers to buy new is going to turn many away from the platform to the rival, and maybe get your console dropped from gaming stores. Adding a $5 license fee wouldn't be a massive burden to gamers but would win a lot of friends from developers. It's success/failure would depend on how much the fee is.
My speculation on a completely unqualified rumour.
COPS N RAPPERS
Regular
the specs look to be interesting, some ideas work in places with current speculation others not so much.
........with each game the eyebrow sort of raises, especially if it's 1080p 60 fps. all 3 of these games were not running on AMD hardware. they were core i7s with 680 gtx and 10 gbs of memory.
i hope that's just a theory, because a lot of people's devices brake (do to whatever) and they end up transferring. not to mention when gamers have multiples of the console in their house.....what will they do if they don't have a connection to stream games?
no good.
to protect against theft i hope, that would help because iv'e had people brake into my house.
that'll be the death of them, in this economy....there'd be no point in APUs if the range was that high.
kudos if it's a 28 nm gpu, but the whole console should be at the right price.
New xbox next leaks for today.
http://www.computerandvideogames.co...ltimate-issue-to-tell-you-everything-we-know/
Interesting. This might have some additional info instead of just regurgitation.
From the link:
Seems like another suggestion that Durango will be revealed a lot earlier than E3. I'm hoping we see it 2-3 months.
Does that mean 16 hardware threads? It seems kind of excessive and I was thinking in terms of quality over quantity and Amdahl's law and all that. It doesn't seem to make sense in light of AMD's CPU architectures as Bulldozer etc is dual thread per single 'core' and Jaguar is one thread per core. Perhaps it might be wise to discount their theories given what we already know about AMD architectures?
Don't hold your breath...
Don't hold your breath...
No no, let him hold it. The amount of ridiculous speculation and hyperbole in this thread has become rampant.
"My ex-girlfriends' former roommate's nephew's brother in law's veterinarian's IT department head said that Durango will feature 800 quadrillion bits and a total linear computational speed of 60 trillion operations per second. So it must be true."
Don't hold your breath...
If it's shipping in 2013, I'd be surprised if they waited to E3 to announce, it's too risky if you think your competitor is shipping in the same year.
The absolute last thing you'd want is for a competitor to have a better showing within a day or so of your unveil.
If we factor in, that 4 Jaguar cores form a "module" with a common L2 cache, it could mean 4 "modules" or 16 Jaguar cores. That would be more computing power than I had expected (but one would need a lot of threads to use it efficiently). And it would somehow fit to the first devkits allegedly containing two quadcore Xeons (16 threads with HT). I don't know.
Can 16 Jaguar cores be "enough" for next gen gaming? (talking about CPU).
If their source is good, I think that would guarantee that the CPU is not from AMD, but is instead an IBM one made as a derivative/successor to the A2.
In many ways this would suck. A2 is a CPU designed for supercomputers and routers. It's in-order 2-issue, and IBM says it has ~0.7 IPC (in one benchmark chosen by them -- probably a best case) per core, divided by 4 threads. Even if you assume that they get a very good clock boost from moving to a new process, if you use all threads, each of them is potentially slower than a Xenon thread. You'll have 5 times as many of them, of course, and at least the lsu doesn't utterly suck this time.
On the plus side, anything A2-related would probably get transactional memory, which should make writing code for those absurd threadcounts much more pleasant.
I think this is stretching it pretty far. "Cores" are pretty unambiguous when talking about Jaguars. I think it is much more likely that either the source is good and it's an IBM CPU, or that the source is bad and it has no value.
Easily. If it is an A2, 4 Jaguar cores could probably give it a run for it's money on normal code.
I consider it pretty much confirmed, that Durango is a complete AMD package.
I would tend to the latter version. But do you remember the discussion if a BD module is really only an enhanced single core or indeed two? I only brought it up, because someone mentioned "modules" in a reference to Durango. If their source heard the same terminology, he might have transferred that discussion about BD to Durango without knowing details about the cores and the difference Jaguar cores actually make. He was probably not aware that it would be Jaguar derived cores (it probably does not get advertised with that codename to devs
) which could explain the confusion.But yes, I know it is somehow far fetched, which is why I emphasized a word in "it could mean" by setting it italic in my post
.
Last edited by a moderator:
I dont understand, what is the problem again with it being 16 Jag cores? Is there some reason that number is not feasible?
Otherwise this rumor sounds in line with past ones. Albeit with maybe a but more CPU, but 8-16 cores has been rumored.
Seems very reasonable. Especially since the mag is unlikely to be sophisticated enough to parse the whole cores vs modules thing.
If I'm not mistaken this mag has seemed to have a good Durango source in the past...I believe I've referenced one of their editors dropping Durango hints on a podcast before.
Edit: and yes here is my previous post: http://forum.beyond3d.com/showpost.php?p=1646446&postcount=11977 You have to copy and paste the podcast link in that post if you're interested, it's broken for clicking. Or you can just go directly to the youtube correctly timestamped here: http://www.youtube.com/watch?v=kCCrRDtC4a4&t=3m0s
I'm also in the process of grabbing a pdf of that issue to see if there's anything juicy besides whats been reported, in the "8 page spread".
Otherwise this rumor sounds in line with past ones. Albeit with maybe a but more CPU, but 8-16 cores has been rumored.
Seems very reasonable. Especially since the mag is unlikely to be sophisticated enough to parse the whole cores vs modules thing.
If I'm not mistaken this mag has seemed to have a good Durango source in the past...I believe I've referenced one of their editors dropping Durango hints on a podcast before.
Edit: and yes here is my previous post: http://forum.beyond3d.com/showpost.php?p=1646446&postcount=11977 You have to copy and paste the podcast link in that post if you're interested, it's broken for clicking. Or you can just go directly to the youtube correctly timestamped here: http://www.youtube.com/watch?v=kCCrRDtC4a4&t=3m0s
I'm also in the process of grabbing a pdf of that issue to see if there's anything juicy besides whats been reported, in the "8 page spread".
AMD is always repeating the Lego-principle of their [CPU] designs since years. It can also mean a 4 INT-core module if the design stands up to their "promise". Considering they where brute-force re-routing the initial BD-design lately, gaining 20% space, one could ask the hen & egg question: did they re-route to get 4 INT-cores or could they just go for 4 INT-cores because of the re-routing.
As the FPU/GPU fusion is planned anyway the 4:1 ratio isn't really such a problem if the software to run the chip quasi-heterogenous is good enough.
- Status
Similar threads
- Locked
