-
Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.
Digital Foundry - Several sources indicate Neo will launch this year
- Thread starter DieH@rd
- Start date
You can switch to zen and not need a break. You just have a few years with some restrictions on content due to neo/Scorpio before the zen machine becomes the baseline. tbh next gen probably will be chasing 4K whether we like it or not, and that will suck up a lot of available power. So a PS5/Xb2 with zen running 4K could still be compatible with games running on a neo at 1080p for example.
-ImaginaryInsider
Member
Supporting the Neo for a traditional lifespan is counter to why it exists. It's a supplemental hardware refresh to augment the original PS4 for the remainder of this generation.
If you are buying a Neo, you are deliberately signing up for a shorter lifespan, but the trade-off is not waiting until the PS5 for better graphics and certain features. That trade-off is not going to be for everybody, thankfully it's entirely optional.
No Sony have been clear that Neo is current gen, there will not be games on Neo that's not on the original PS4. Neo is literally a supped up PS4 architecturally. Your argument could apply for the very existence of Neo as well for owners of PS4, it doesn't hold.
Edit: instead of a Slim were getting Neo. That's a better way of putting it.
10nm chips will be out early next year in phones, GPUs won't be that far behind.
Apple is being supplied 7nm chips for its iPhone in 2018, again I don't think GPUs will be that far behind.
By holiday 2020 we will have at least 7nm SOC in PS5.
He was talking about frequency of updates in regards to NEO. He still talks about complete' cycles' as an alternate thing to what the NEO is.
Go back and read what he actually said.
A 2019/2020 Zen CPU with 16GB ram HB2 with 1+tb/s and 14+TFLOPs is absolutely not something that can be held back by what the PS4 and NEO will be offering, which are marginal jumps by comparison.
The investment into that kind of tech, which goes far beyond NEO's investment, would have been a waste if they limit the games to the previous hardware.
I tend to agree.
7nm SoCs in consoles by holiday 2020. Zen + HBM2/3 + 20 TFLOP GPUs.
Hopefully PS5 and the equivalent Xbox will put both Neo and Scorpio to shame as far as being able to handle highend AAA games at native 4K.
You'd have to think by 2020 we'd be able to get really good 10TF machines.
Proelite
Member
7nm means twice the density of 14nm, and may or may not allow a higher clock. Most optimistic case for AMD is a ~25% clock increase for "free" on 7nm finfet. Best case is something like 15 teraflops.
You do have a point. They may not follow the same path as Sony. Or maybe Sony will give in some day and allow them.
Liabe Brave
Member
The 1TB and 500GB variants are listed for December release on the Microsoft store. In October, it seems only the 2TB will be for sale, at $399.
And even if all the SKUs are out when Neo launches, Sony's machine is going to be positioned as a premium offering. That likely means a bigger hard drive than the base PS4. Alongside 4K video and greatly improved specs, I'd expect a slightly higher price than the highest Xbox One.
geordiemp
Member
God I hope they put a Sata 3 in and some space so we can put in our own HD.
My Ps4 2 TB is full, been deleting lots of games.
And false advertising
How so?
Commanche Raisin Toast
Member
I'm predicting the same HDD tray/dock as regular PS4 but it comes with 2TB instead of 500GB, and with the announcement of external HDD support.
Hopefully that means a few extra USB ports. I'm fine with 2, but an extra one in the back for PSVR and/or the external HDD would be welcome.
geordiemp
Member
The tight fit for the HD with no space and the slower Sata 2 is just poor thinking by Sony IMO.
Sata 3 and space for a 4TB drive would be nice.
SEGAvangelist
Member
It would be kind of like calling Xbox One the Xbox 1080. Not really false advertising, but not necessarily a good idea considering the AAA games won't be targeting 4K.
7nm may be further than 2020, lithography is already a problem. Process leader Intel is expected to only debut on 2020 earliest based on its new cadence, putting other foundries at 2021 for premium mobile devices. GPUs one year later. Even then, does AMD have a billion dollar capital to design chips on 7nm? Intel+Nvidia could kill AMD off by then.
Not really at all.
The system will(assuming of course)play 4k blurays and upscale games to 4k. Not false advertising at all.
SEGAvangelist
Member
But it won't play a ton of games at 4k. That's why I clearly said it's not really false advertising.
Since it was rampant speculation on my part I don't think we should expect anything in particular. The question was merely whether Sony has some end game in mind for a cost-reduced PS4 based on the work they've done for Neo. The previous assumptions about how a console generation plays out have been disrupted so it will be interesting to see what they have in mind beyond this holiday.
It isn't, actually. If a defect is isolated to a functional unit you're disabling then there's no problem. I'm also curious where your assertion about the number of CUs on Neo's SoC comes from. If we accept it as fact then you can have up to 4 non-functional CUs and still produce a viable Neo. My thought experiment was what you do when you have 5 that are defective? Or 6? 7? You also have marginal failures parts that are within spec at lower clock rates but fail when you push them harder.
Binning is a technique that has been used before in silicon production, so I was just wondering whether it would apply here to any effect. Perhaps, and perhaps not ... but I'm hardly alone in pondering possibilities without being in possession of solid facts, and at least I'm not pretending otherwise.
The Praiseworthy
Member
MilkyJoe
Member
If you have to ask...
if it plays 4K video and upscales games to 4K they could get away with calling it PS4K even if games do not play in native 4K
geordiemp
Member
Many years ago I worked at Motorola semi fab, testing chips and fault finding. But things move on, just giving my opinion.
It really depends what the faults are, as once you get so many faults then your also possibly affecting the reliability and confidence in that silicon wafer as a whole. Maybe one of the processes was slightly off.......If the fault is a foreign particle then fine, but what if its more ?
Given 18 separate fails in a tiny chip.....You can get so many modes of silicon failure which can make the wafer suspect - you would not advise to put it in the next NASA shuttle lol. Do you strip the silicon and analyse each defect on each wafer to make sure each occurrence affects no other part of the wafer ??? doubt it.
Would you be comfortable paying for a 50 % binned chip to be made up , fully assembled. and providing a 2 year warranty ? Maybe you could bake it in the ovens for a few weeks to get more early bath tub fails and then assemble it.....But its not automotive class and that costs more....
....Or lets face it you would not go forward with a process or batch with that many issues anyway..
MilkyJoe
Member
"Get away with" being the operative phrase
Clear
CliffyB's Cock Holster
All depends on pricing.
To which point I'm not entirely sure why exactly people are expecting Neo to be so damn expensive. Given the One S is pretty much the same price as the One despite being upgraded to be able to handle 4k video... why exactly is this modestly upgraded PS4 going to be so much more than the base unit?
Is changing the APU and Ram really going to add $100+ to the BOM? Really can't see it myself.
Neither of us has enough information to make that call, but I'd at least want to consider my economic options. You keep pushing the 50% failure case, but there's a lot of room in the middle. If you're happy to ship a chip with four disabled CUs out of 40 as a high end unit, would you really balk at shipping one with five as a base model that doesn't need the full CU count?
If I'm trying to cost reduce a product I'm looking for every possible win and I'm doing some pricey stress testing on a representative sample to see what my options are. Binning isn't novel in silicon production, and designing around the inevitability of imperfection turns out to be far more cost effective than insisting on flawless parts.
fair enough, just saying it's not false advertising
serversurfer
Member
I disagree. House said Neo will be sold alongside PS4 as a "premium option." Shu just called Neo, "the high-end PS4." It isn't meant to be the sales volume focus any more than the Corvette is meant to sell more than the Camaro. Or Civic Type-R versus standard Civic, if you'd prefer.
Oh? Sorry, but I don't really recognize your name apart from this thread. Have I talked to you about trolling in the past?
If you're happy with the performance of games on your PS4, then no, Neo isn't for you.
Ah, now I see where you're getting this stuff. The frame rate increases you're talking about came not from bumping the i3 to i5, but rather bumping the 750Ti to a GTX960.
Read this DF article instead. First note that in the official sysreqs for Witcher 3, CDRP say Core i5 2500K minimum. Now look at this table:
See how even a Pentium does better than 60 fps? If you're talking about minimum frame rates, then even the i5 doesn't hold 60; you need an i7 for that. It sounds like rather than do a good job of optimizing performance of their game in cities and such, CDPR's response was to say, "/shrug Just get a better CPU." Hardly an elegant solution.
Thanks for that. This is what I meant by CDPR doing a poor job of handling the cities. I understand that it may take more than 16ms for 1000 NPCs to sequentially decide what they're gonna do next, but there's no need for the rendering engine to sit around and wait for them all to figure it out. Really, NPCs don't need to be making 60 decisions a second in the first place. 10 per second should be more than sufficient. Probably even less for NPCs that are further off and not even interacting with the PC. So yeah, make that stuff asynchronous and crank the LOD way down. That should solve most of your crowd-induced performance issues.
I've talked about stuff like this before, but I mostly get ignored. =/
Didn't you say 61% earlier? Where are you getting this figure? All I can find is AMD saying its IPC is 40% better than Excavator.
lol This specifically says the CPU for Scorpio isn't confirmed yet. ;p
I think that rather than mandating it, they're merely enabling it. When Peggle: Ad Nauseam is a PS5 launch game in 2020, I see no reason the same executable shouldn't be allowed to run on Neo and PS4.
Outside of GAF, nobody is supposed to care about Neo at all. Sane people are supposed to continue buying PS4 instead. ;p
They already have an entire production line to produce nothing but PS4 APUs. Nobody else uses them.
I agree with this and it's worth adding that House actually talked about nine years of support for PS4. That means that he expects CoD:2022 to still support PS4 when it comes out. Maybe not CoD:2023, but by then PS5.5 will be launching, so hopefully by that time most PS4 buyers will have at least upgraded to a PS4.5 (Neo).
Yeah, this sounds reasonable to me too. Keep the margins on PS4 as slim and aggressive as possible, and let reduced production costs on Neo bolster your profits. "PS4 is PS4," so it doesn't matter to Sony which one you buy, because they're gonna sell you exactly the same software either way.
No, I think Neo will see 9+ years of support, just as all of their devices are intended to receive.
I only read a translation, but I actually got the impression he was simply saying they wouldn't have new hardware every year, because they wanted to give developers a stable target to study/learn for a few years. So, smartphone-like in the cross-comparability, but not in the annual hardware updates.
I think it will be held be back for the first few years, just as the need to reach a larger audience dictates cross-gen support at the start of any new generation.
I think the argument is that you'd need to have a horrific number of defects before fusing off as many as 17 CUs was more cost effective than simply making a smaller die for the PS4. Also, if your defect rate was really that high, I'd think it'd be affecting the CPU section of the die equally.