TSMC has fully validated their design tools for 5nm (PDK, transistor models, etc.) They are in risk production already. It remains an insane long shot, but if some SKU were to be 2021...
You mean PS5 Pro in ~2022 or so? :)
TSMC has fully validated their design tools for 5nm (PDK, transistor models, etc.) They are in risk production already. It remains an insane long shot, but if some SKU were to be 2021...
You're asking for the one thing that is going to leave you unhappy.Honestly all it needs for me to be happy is true 4K at 60fps. Graphics have finally gotten to the point where we are splitting hairs when it comes to comparisons.
Common mistake a lot here seem to be making. You look at next gen as if it's going to be current gen with more power. As if they are making better hardware to run the games we have now at a higher resolution and higher frame rate.If next gen will be around the 1080ti's performance, it runs Destiny 2 @75fps in native 4K so it could happen. Thing is, Bungie always went for graphics over resolution and performance so I'm not sure it will happen.
___________________________
Apparently there was another Reddit "leak" for Anaconda dev kit (Dante)
Specs are on the edge of believable, but clearly a $500 box.
So a 8C/16t CPU, 64CU GPU at ~1500Mhz, 24GB of GDDR6 at around $7/GB....- That would mean 2x more memory and 2x faster bandwith than XBX.
- Assuming Navi has 64 CUs with same tflops as GCN that would mean almost exactly 12 Tflops (~12.095), again 2x more than XBX GPU.
- That would mean 2x more memory and 2x faster bandwith than XBX.
- Assuming Navi has 64 CUs with same tflops as GCN that would mean almost exactly 12 Tflops (~12.095), again 2x more than XBX GPU.
The specs are believable considering the rumoured Anaconda 12TF...
1475Mhz with 64CU would reach exactly 12TF.
Concerning the memory, the 48GB would be for the devkit... You would need to slash that by 2 for a consumer device, giving us 24GB of GDDR6. Doable.
The bandwidth would indicate a 384bit memory bus clocked at 15gbps. That's quite a bit.
Finally, the storage amount needs to be slashed by 4 to give us 1TB. Most likely, the speed would be lower too.
I'm not making this mistake, it's obvious. I've said it with this in mind:Common mistake a lot here seem to be making. You look at next gen as if it's going to be current gen with more power. As if they are making better hardware to run the games we have now at a higher resolution and higher frame rate.
That's not going to be the case. We will have native 4k games running at 30fps with "next gen assets" and "tech"..... Improvements in lighting, AI, geometry complexity and/or interactivity and a slew of other features that would cripple current gen hardware but be possible on next gen hardware.
And as usual, majority of the devs will opt for 30fps to allow for more eye candy in their games and again as usual only high end PCs will be able to take those same games and run them at a higher frame rate.
If they are talking about Destiny with a very small visual upgrade, it sounds too easy to get it to 4k/60 considering a parallel desktop GPU runs Destiny 2 easily.Considering the fact that i'm not expecting a huge visual update to the renderer in Destiny 3, I would think Zen CPUs can make a 60fps target a very real possibility.
It is a dev kit. Likely with double the RAM, just like the X dev kit.You're asking for the one thing that is going to leave you unhappy.
Common mistake a lot here seem to be making. You look at next gen as if it's going to be current gen with more power. As if they are making better hardware to run the games we have now at a higher resolution and higher frame rate.
That's not going to be the case. We will have native 4k games running at 30fps with "next gen assets" and "tech"..... Improvements in lighting, AI, geometry complexity and/or interactivity and a slew of other features that would cripple current gen hardware but be possible on next gen hardware.
And as usual, majority of the devs will opt for 30fps to allow for more eye candy in their games and again as usual only high end PCs will be able to take those same games and run them at a higher frame rate.
How does that allign "clearly" with a $499 box though? That looks more to me like a Dev kit.
No it is not.
Navi is based on GCN still, and from what the sources I TRUST say is it's much better efficiency than raw power. RADEON 7 is based on instinct cards using vega chip as it's basis but is basically chips from instinct cards that more than likely didn't meet certain criteria. NAVI will be a line of GPU's, and is just the code name, for that line. Just like Polaris was for the RX 500 line.
It's going to have variations, and be much more efficient than vega, with new memory controller. I expect something better than the RX VEGA's going into 2080 territory possibly performance wise for certain task's, but not compete on 1:1 perfomance. There's just no way in terms of what they have R&D wise compared to NVIDIA.
I expect a highend version, but from what AMD has been open about, they consider the Radeon 7 for now to be their high end. With Navi line filling in the gap for replacing the RX 500- RX VEGA line at much better power/performance and in price.
Even with anexanhume being much more knowledgeable than I am at this stuff, it's still just speculation. But from what we know in terms of Radeon being short handed, and having these already in play roadmap wise it's hard to believe that they will be able to compete with a 2080ti so quickly with a more efficient GCN arch, that chip wise is a evolution to Polaris.
I mean if it is more powerful than a Radeon 7 with better performance/watt. Then gravy, it's just hard to swallow knowing what Radeon group has had available for resources, and knowing this was all part of a roadmap way back when Raj was with the company.
So they never alter these things? I know it's not the same company, but I was thinking of how Naughty Dog was initially promising 60fps Uncharted 4 when they realised how good it looked in TLOU Remastered, but they just couldn't get it to work reliably and eventually had to accept (a really good and hard-locked) 30fps.
Where are you getting your info from? Everything that is out there from leakes that show up to Navi 10 has shown where they lay performance wise around 1080/2070.
If there is a Navi 20 which some recent leaks which should be taken with a grain of salt say it will beat a 2080ti. but I believe that is in specific workloads like certain rendering techniques.
If there is a card with 72 cu's jesus the price won't be cheap fora consumer card. You said I was confused. I am not I know NAVI is a architecture, but also know that Consumer cards based on that which have somewhat leaked are lining up to be around 1080/2070. With Navi 20 which will come later being possibly their new render card like the instinct which they might have a desktop consumer version of possibly.
Someone said Navi would beat a radeon 7, I was strictly talking consumer video cards. Not the entire arch that has other chips based on it being made being used in lets say consoles or professional render cards.
More than likely they have something on the high end in the works.
But I highly doubt they will launch it this year, and mainstream gpu's will be what they focus on more like something to replace the rx 500 line and vega 56/64 line.
- That would mean 2x more memory and 2x faster bandwith than XBX.
- Assuming Navi has 64 CUs with same tflops as GCN that would mean almost exactly 12 Tflops (~12.095), again 2x more than XBX GPU.
We don't know for sure if this is still based on gcn architecture. Navi has been in the pipeline for so long that things could of changed.
I don't really see any way a console will have an SSD as the main storage device, it makes no sense. If they will have some form of SSD, it will be for caching only. Next gen will have 1TB as a baseline if not more, it's too expensive for a console. It also means that external HDDs are irrelevant because whatever you install on them will never be able to keep up. A cache drive is the only thing that makes sense, it will both be reasonable in terms of costs and you will still be able to use an external drive or replace your current drive while maintaining the advantages of the SDD cache drive.
Microsoft already has ML for dividing game blocks in a smart way for enabling games to launch before they've finished downloading, they can use the exact same tech to cache the most relevant parts of every game installed on your HDD to keep it always ready for the next loading screen or streaming data. You can also keep a snapshot of the system's memory when you are done playing, that way you can jump right back in the game quickly even if you've launched other games before, just like on a smartphone.
So in order to use an external drive, you will have to copy a 200GB game into the main drive just to launch it? Isn't that the exact opposite use of an external drive? Also, SSDs are just too expensive when you can have a smart caching system with a small SSD and a large magnetic drive.The idea would be to have the 1TB SSD for storage, and if you plugged in an external HDD, any games on it would be transferred to the SSD when you wanted to play them.
True, but Uncharted 4 was also rebooted after Amy Hennig and co left the studio. Different director could have meant different priorities.
I'm discussing hypothetical Navi GPUs because the so-called leaks you insist on quoting are tenuous and incomplete at best.
How many CUs does Navi 10 have from the leaks? What clocks?... No idea? Exactly my point.
Leaked performance figures of Navi-based parts could be from underclocked, undervolted engineering samples that clearly aren't final or are aimed at completely different market segments than the leaker thinks they are.
Trying to conclude anything concrete from them is useless and making sweeping statements about an entire unannounced GPU microarchitecture's possible performance, like "Navi won't be faster than Radeon 7", is more than a little misguided.
Also, you have no idea how much a 72CU part will cost because you have no idea of:
a) the 7nm production yields for a new Navi chip
b) the die size and TDP limits of said chip
c) the price AMD pays TSMC per silicon wafer
d) how many product lines the first Navi chip production runs will be binned to
e) how much margin AMD will put on it when it provides a RRP.
Without any of the above, your assessment on price or even cost is simply as good as numbers plucked from thin air.
Look, i'm not trying to talk down to you or be condescending, but I think you are making conclusions about things you don't properly understand and then arguing till you're blue in the face when corrected, instead of acknowledging your lack of understanding and looking to learn. This thread would be much better if more people asked questions to knowledgeable folks like anexanhume and Miniature Kaiju, rather than bullishly pushing incoherent arguments becuase they read a few PC benchmarks on Anandtech.
Sounds about right.
So in order to use an external drive, you will have to copy a 200GB game into the main drive just to launch it? Isn't that the exact opposite use of an external drive? Also, SSDs are just too expensive when you can have a smart caching system with a small SSD and a large magnetic drive.
Do me a favor and google the prices of magnetic drives in google. Using an SSD in a console as the main storage device is just stupid, a console budget allows for ~35$ for the storage solution.
Do me a favor and google the prices of magnetic drives in google. Using an SSD in a console as the main storage device is just stupid.
You could do 72CU and deactivate 2 CUs per Shader Engine (= 8 CUs) -> 64 active CUsWhy do people keep saying "64 CUs". 64 is the most random number. I mean I do get why, because right now it's AMD's glass ceiling. There are only two scenarios with Navi:
1) They break the 64 CU limit - So why 64 CUs? I mean it could be 64, but it could also be 56 or 72. 64 Is a totally random number.
2) They don't break the 64 CU limit - Consoles will never leave every CU active, they will always turn off some CUs for yield. So again, 64 makes no sense.
So why everyone here is using the number 64? If they break the limit it's a totally random number and if they don't it's an unachievable number.
a 256GB cache drive will perform just like a main SSD drive if the OS knows what it is doing and a 256GB SSD + 2TB HDD will cost less than a single 1TB SSD drive.I hope you get your spinning drive machine and It get my SSD machine.
You can do any combo you would like if they break the 64CU limit, it's still the most unlikely number statistically because if they don't break the 64CU limit, it's unachievable.You could do 72CU and activate 2 CUs per Shader Engine (= 8 CUs) -> 64 active CUs
Why do people keep saying "64 CUs". 64 is the most random number. I mean I do get why, because right now it's AMD's glass ceiling. There are only two scenarios with Navi:
1) They break the 64 CU limit - So why 64 CUs? I mean it could be 64, but it could also be 56 or 72. 64 Is a totally random number.
2) They don't break the 64 CU limit - Consoles will never leave every CU active, they will always turn off some CUs for yield. So again, 64 makes no sense.
So why everyone here is using the number 64? If they break the limit it's a totally random number and if they don't it's an unachievable number.
Do me a favor and google the prices of magnetic drives in google. Using an SSD in a console as the main storage device is just stupid, a console budget allows for ~35$ for the storage solution.
Can someone give me a simple layman explanation for the follow two queries?
1. Why has it been a trend for AMD GPUs to always have lower performance/watt compared to nVidia GPUs (since forever)?
2. Why do AMD GPUs always have higher raw "TF" value which do not translate into performance compared to their nVidia GPU counterparts?
A) I suspect Sony and MS will keep the die size the same as the X1X assuming they are going for a $499 price tag. Costs will initially be higher but isnt the entire point of shrinking die sizes is to reduce cost?So A) "should be able to double the CUs" in a purely physical sense of being capable of putting about twice as many transistors in a chip of the same size on 7nm as it was possible on 16nm -- sure. This doesn't say anything about the cost of said chip though which might be quite a bit higher than that of the chip on 16nm of the same size and costs are way more important than process capabilities - as you can probably guess from the fact that current gen consoles all have APUs with 200-400mm^2 die sizes while the same 16nm process (or family of) they are using allows building something like a 815mm^2 sized Nvidia GV100 GPU.
B) The capability of doubling the number of transistors wrt GCN GPUs is largely irrelevant at this point as GCN is unable to make use of these additional transistors due to hitting the power ceiling well before reaching the maximum die size on both 14nm (Vega 10 is consuming 300W at 482mm^2) and 7nm (Vega 20 is consuming 300W at 331mm^2). For them to make any use of what 7nm process provides in GPU complexity increases they have to make *huge* gains in power efficiency first.
C) 28nm to 16/14nm was a move from planar to FinFET transistors, and this was the main reason for most designs gaining clocks on 16/14 compared to same(ish) designs on 28nm. This won't occur again with the transition to 7nm and thus nobody should expect significant clock improvements on 7nm parts when compared to 16nm predecessors - unless said parts will be specifically re-engineered to reach significantly higher clocks of course.
So yeah, 64 is totally random just like 72 and 56.64 CUs are "random" in the same way that the 8 CPU cores or 8GB of RAM in PS4 are random - deliberately selected to balance performance, cost, and power consumption.
The "limit" of 64 CUs has to do with how the Vega architecture scales. Looking at benchmarks shows that you don't get a performance improvement commensurate with the CU increases going from Vega 56 to 64 adjusted for clocks. The architecture has inherent inefficiencies that make it not scale well. Exceeding 64 CUs is completely possible, but a fool's errand given diminishing returns.
AMD has acknowledged this limitation and you can read about that over at Anandtech. I've linked the article many times in the past.
Saying consoles will be limited to 64 CUs is pure speculation. It's entirely dependent on Navi's architecture. Given "scalable" is in AMD's marketing materials, I'm guessing we're going to see bigger parts. There's also no theoretical limit to how many extra CUs the die include that will be disabled for yield. That just becomes a cost factor. Using that as a basis to decry 64 as having no foundation in reality is severely misguided.
No, it doesn't. a small SSD for caching makes sense, a full main storage SSD doesn't. You would get the same performance (except maybe some edge cases) and it will be much cheaper. Even if you could place a 1TB SSD in a console, it would be much smarter to use a small SSD for cache and have x2 or x4 the storage space on a magnetic drive.Following this logic, consoles would have never got hard drives at all. They wouldn't have got WiFi chips added. Why include new technology when it's going to raise the price of your console? Because the benefit is worth it. It's very much platform strategy and having a vision for the future. Making conservative or "safe" choices gets you the Xbox One design. We saw how that turned out.
I just got a new PC with an SSD and the damn thing loads up windows before i even have the chance to sit down on my couch. It's crazy.Anyone who doesn't want an SSD as the storage medium is either nuts or never used one. You can always hook up an external HDD for more space and just copy games back and forth as you wish. What's wrong with that setup?
The first response is more nonsense. The second makes a lot of assumptions and largely ignores my points. I don't think it's constructive to continue.So yeah, 64 is totally random just like 72 and 56.
No, it doesn't. a small SSD for caching makes sense, a full main storage SSD doesn't. You would get the same performance (except maybe some edge cases) and it will be much cheaper. Even if you could place a 1TB SSD in a console, it would be much smarter to use a small SSD for cache and have x2 or x4 the storage space on a magnetic drive.
A) I suspect Sony and MS will keep the die size the same as the X1X assuming they are going for a $499 price tag. Costs will initially be higher but isnt the entire point of shrinking die sizes is to reduce cost?
B) What exactly do you mean by power ceiling? I dont see the PS5 APU consuming anywhere near 300W. The X1X has a TDP of 170 with 40 CUs and clockspeeds of 1.17 Ghz. do you think they will hit the power ceiling for an APU thats under 150W? Also, what do you make of the graph below? To me, this means they can get 30% more clock speeds for the same power consumption when going to 7nm.
C) Balls. Day ruined.
Can someone give me a simple layman explanation for the follow two queries?
1. Why has it been a trend for AMD GPUs to always have lower performance/watt compared to nVidia GPUs (since forever)?
2. Why do AMD GPUs always have higher raw "TF" value which do not translate into performance compared to their nVidia GPU counterparts?
You have no idea what improvements Navi will have over current AMD architecture, no idea what are the costs of 7nm in each point in time, no idea what are Microsoft/Sony's budget or targets for the APU, no idea what are the yields but somehow 64CUs makes more sense to you than 56 or 72 because... zero knowledge? Amazing! So yeah, 64CU is random just like 56CU or 80CU. I mean you can say 64, just like you can say 56 but for some reason, people started taking 64CU as a dogma for no reason at all.The first response is more nonsense. The second makes a lot of assumptions and largely ignores my points. I don't think it's constructive to continue.
Ok. How about we make this simple.....I don't really see any way a console will have an SSD as the main storage device, it makes no sense. If they will have some form of SSD, it will be for caching only. Next gen will have 1TB as a baseline if not more, it's too expensive for a console. It also means that external HDDs are irrelevant because whatever you install on them will never be able to keep up. A cache drive is the only thing that makes sense, it will both be reasonable in terms of costs and you will still be able to use an external drive or replace your current drive while maintaining the advantages of the SDD cache drive.
Microsoft already has ML for dividing game blocks in a smart way for enabling games to launch before they've finished downloading, they can use the exact same tech to cache the most relevant parts of every game installed on your HDD to keep it always ready for the next loading screen or streaming data. You can also keep a snapshot of the system's memory when you are done playing, that way you can jump right back in the game quickly even if you've launched other games before, just like on a smartphone.
Sony spent 37$ on the PS4's HDD during launch, not 25$; in 2016 Microsoft spent on a 1TB drive 32$ and on a 2TB drive 55$. your numbers are just wrong so some sources could be nice. Even if they will be able to have a 1TB SSD for ~37$, current X-enhanced games are 100GB+ so next-gen games are going to be ~200GB. You will be able to install ~5 games on a 1TB drive, that's crazy.Ok. How about we make this simple.....
How much do you think Sony/MS spent on a 500GB HDD back in 2013, how much do you think they spend on a 1TB HDD now, and how much do you think a 1TB M.2 SATA SSD will cost them (OEM pricing) in 2020?
For some context..
2013, 500GB HDD $60 retail (~$25 OEM)
2016, 1TB HDD ~$60 retail (~$25 OEM)
2019, 1TB SSD ~$100 retail (~$40 OEM? If looking at OEM pricing for HDDs)
So how much you think an SSD will cost at retail in 2020.... And how much do you think it will cost sony/ms at OEM pricing. And at what cost would you consider it to be unreasonable for them to Still use an SSD?
Lastly, have you thought about why an SSD is even needed? And what it would mean to not have one?
And the cache + HDD combo thing.... That only makes sense if they CANNOT get a 1TB SSD at under $40. But if they can then the cost and complexity of having a cache SSD and a 1/2TB is no longer worth it.
So in order to use an external drive, you will have to copy a 200GB game into the main drive just to launch it? Isn't that the exact opposite use of an external drive? Also, SSDs are just too expensive when you can have a smart caching system with a small SSD and a large magnetic drive.
Sony spent 37$ on the PS4's HDD during launch, not 25$; in 2016 Microsoft spent on a 1TB drive 32$ and on a 2TB drive 55$. your numbers are just wrong so some sources could be nice. Even if they will be able to have a 1TB SSD for ~37$, current X-enhanced games are 100GB+ so next-gen games are going to be ~200GB. You will be able to install ~5 games on a 1TB drive, that's crazy.
One major upside of an M2 SSD, that many ignore. You don't need a nice mounting bracket anymore to reduce vibrations from the HDD.Lastly current projections have flash memory prices falling fast to the point that a 1TB NVMe SSD would likely cost Sony and MS about $35 to put into a console releasing late 2020. Also, using just that SSD would mean no heavier HDD and no drive bay, shrinking the console, simplifying manufacturing and lowering shipping costs. So no, everything indicates it wouldn't be too expensive at all.
Even if they will be able to have a 1TB SSD for ~37$, current X-enhanced games are 100GB+ so next-gen games are going to be ~200GB. You will be able to install ~5 games on a 1TB drive, that's crazy.
Sony is looking at prices over the long term, not just at launch. Today SSDs are much more expensive than HDDs, but the prices are coming down and the technology is getting better. There is something inevitable, SSDs with no moving parts are going to become more affordable than HDDs with moving parts. If this happens in the next 2-3 years, then it would have been a mistake to forgo SSDs for HDDs if Sony had the ability to absorb the initial loss in 2020.Sony spent 37$ on the PS4's HDD during launch, not 25$; in 2016 Microsoft spent on a 1TB drive 32$ and on a 2TB drive 55$. your numbers are just wrong so some sources could be nice. Even if they will be able to have a 1TB SSD for ~37$, current X-enhanced games are 100GB+ so next-gen games are going to be ~200GB. You will be able to install ~5 games on a 1TB drive, that's crazy.
And for the 30th time, I do think that an SSD is a good idea for next-gen, but having an SSD as the main drive is a waste of money so they need to use a smaller SSD for caching. It's not a PC, in a console caching SSD will work great even with the simplest OS implementation.
My numbers are based on this. And I even made them more expensive. At least for the 2013 ones. As for the 1TB HDD ones I just made estimates based on how much a 1TB HDD cost back in 2016.Sony spent 37$ on the PS4's HDD during launch, not 25$; in 2016 Microsoft spent on a 1TB drive 32$ and on a 2TB drive 55$. your numbers are just wrong so some sources could be nice. Even if they will be able to have a 1TB SSD for ~37$, current X-enhanced games are 100GB+ so next-gen games are going to be ~200GB. You will be able to install ~5 games on a 1TB drive, that's crazy.
And for the 30th time, I do think that an SSD is a good idea for next-gen, but having an SSD as the main drive is a waste of money so they need to use a smaller SSD for caching. It's not a PC, in a console caching SSD will work great even with the simplest OS implementation.
The pro was never "beastly" to me even at $400. It was what the ps4 should have been.
Oh? You don't need to transfer all the data? You mean... using the SSD for caching? :)Firstly, 200GB game? Really? More like 100GB I'm guessing. Something like the complete FFXV is an outlier, and not all that data would need to be transferred.
People play from their HDD, they like having their libraries installed, it's not for backup. Most launch PS4 games are less than 40GB and it had 500GB so you could have like 12 games and still, everyone wanted 1TB and called the 500GB drive too small.Secondly, any game you buy will need to be installed from a disc or downloaded from the internet anyway, and they'd go straight into the 1TB SSD first. So you'd only be transferring the games from HDD to SSD after you had already transferred them to the HDD because you had been currently done with them. And given the size of the SSD you'd be able to hold like 8 games in it on average, so the supposed problem of constantly having to transfer back and forth wouldn't actually be that much of an issue.
You've just described caching again, another point to team "small SSD drive for caching".Thirdly, plenty of devs would be able to set the game up so that it only needs some of the data on the SSD, not all, and BC PS4 games (as well as, probably, some indie titles) could run from the HDD if you wanted anyway.
Removing a driver bay and saving 40g of weight will never offset the huge difference in pricing between an SSD and a magnetic drive. Anything you can do with an SSD you can do cheaper with an SSD+HDD combo.Lastly current projections have flash memory prices falling fast to the point that a 1TB NVMe SSD would likely cost Sony and MS about $35 to put into a console releasing late 2020. Also, using just that SSD would mean no heavier HDD and no drive bay, shrinking the console, simplifying manufacturing and lowering shipping costs. So no, everything indicates it wouldn't be too expensive at all.
Yes, 200GB games. X enhanced games are close to 100GB, next-gen games will be bigger. Gears 4 is 121.23GB, how much will Gears 6 on next-gen will take? 100GB will probably be a pretty small AAA install size.You can't just assume that, as though developers just won't bother to work on better data compression. Also, Sekiro literally just came out at a cool 15GB, so that's hardly universal.
If in some magical way SSDs will become so cheap that they will be cheaper than a mechanical drive, Sony can use a full SSD for the slim or pro. For now they better stick with an SSD+HDD combo.Sony is looking at prices over the long term, not just at launch. Today SSDs are much more expensive than HDDs, but the prices are coming down and the technology is getting better. There is something inevitable, SSDs with no moving parts are going to become more affordable than HDDs with moving parts. If this happens in the next 2-3 years, then it would have been a mistake to forgo SSDs for HDDs if Sony had the ability to absorb the initial loss in 2020.
They could have an onboard 1TB SSD to reduce costs with an expansion bay like the PS2, where the user could put in a drive with two scenarios: Performance is good enough to run games (an SSD) or performance is good enough to be used as storage only (HDD). It would allow them to keep the costs down, but future proof the system.
What ever a 1TB SSD will cost them in 2020, a cache SSD and a 2TB HDD will cost them less and preformance will be the same.My numbers are based on this. And I was even made them more expensive. At least for the 2013 ones. As for the 1TB HDD ones I just made estimates based on how much a 1TB HDD cost back in 2016.
Regardless of the accuracy of my numbers tho. My point stands. We can safely say these companies are willing to spend anywhere from $22 to $40 for storage. So the question is simple.
How much do you think a 1TB SSD will cost them in 2020?
I am willing to bet that there is no way next gen consoles don't come with an SSD of some sort.XBX dev-kits also have a SSD which is normal for a dev-kit. Regardless of the veracity of the reddit leak, XB2 dev-kits will most certainly have a SSD too. But retail consoles certainly won't. It's way too expensive for what you get.
It would be much better to use the budget on GPU clock, number of CU, or GDDR6 clock increase or a combination of those. Also 1TB is too low for next gen. I expect a minimum of 2TB (which is what the latest models of Pros have) which would be even more expensive if SSD.
Ok. How about we make this simple.....
How much do you think Sony/MS spent on a 500GB HDD back in 2013, how much do you think they spend on a 1TB HDD now, and how much do you think a 1TB M.2 SATA SSD will cost them (OEM pricing) in 2020?
For some context..
2013, 500GB HDD $60 retail (~$25 OEM)
2016, 1TB HDD ~$60 retail (~$25 OEM)
2019, 1TB SSD ~$100 retail (~$40 OEM? If looking at OEM pricing for HDDs)
So how much you think an SSD will cost at retail in 2020.... And how much do you think it will cost sony/ms at OEM pricing. And at what cost would you consider it to be unreasonable for them to Still use an SSD?
Lastly, have you thought about why an SSD is even needed? And what it would mean to not have one?
And the cache + HDD combo thing.... That only makes sense if they CANNOT get a 1TB SSD at under $40. But if they can then the cost and complexity of having a cache SSD and a 1/2TB is no longer worth it.
Oh? You don't need to transfer all the data? You mean... using the SSD for caching? :)
People play from their HDD, they like having their libraries installed, it's not for backup. Most launch PS4 games are less than 40GB and it had 500GB so you could have like 12 games and still, everyone wanted 1TB and called the 500GB drive too small.
You've just described caching again, another point to team "small SSD drive for caching".
Removing a driver bay and saving 40g of weight will never offset the huge difference in pricing between an SSD and a magnetic drive. Anything you can do with an SSD you can do cheaper with an SSD+HDD combo.
Yes, 200GB games. X enhanced games are close to 100GB, next-gen games will be bigger. Gears 4 is 121.23GB, how much will Gears 6 on next-gen will take? 100GB will probably be a pretty small AAA install size.
If in some magical way SSDs will become so cheap that they will be cheaper than a mechanical drive, Sony can use a full SSD for the slim or pro. For now they better stick with an SSD+HDD combo.
What ever a 1TB SSD will cost them in 2020, a cache SSD and a 2TB HDD will cost them less and preformance will be the same.
Of course there are advantages to both. But we are not talking about a charity organization here.But if you want bigger storage you'll need to grab a bigger ssd only?
Cab see advantages to both methods.