Next Generation Hardware Speculation with a Technical Spin [pre E3 2019]

Status
Not open for further replies.
I don't know if this is really true. Personally I think the only console hardware that's launched that was already outclassed by PC hardware has been Xbox One and PS4. Sure, PC hardware has outpaced consoles, but I think if you look back and think of the PC hardware available in, say, 1994, the PS1 would have been considered high end. 3Dfx Voodoo was at least a year away, PC hardware lacked the advanced video decompression of PS1, and it's lighting and geometry engine is very impressive for it's time. PS2 was mighty impressive. It's pixel fillrate is insane, as was it's geometry engine. Xbox launched with what amounts to a Geforce 3, along side the launch of the Geforce 3. Even then, Xbox had twice the vertex shaders. 360 had a tesellator, unified shaders, 4 sample per pixel per cycle MSAA, and competitive fillrate. And the PS3... Well Cell was paper impressive. If you go back further, console's ability to scroll backgrounds and draw sprites destroyed what was available on PC's when they launched.

Anyway, consoles have launched with high end hardware. They just don't anymore. I think it's partly because the PS3 / 360 generation was so long that even midrange parts were a huge upgrade, plus PC hardware's pace, especially in the ultra high end, has really gotten out of hand. I think traditionally (since the launch of 3D hardware) the highest end PC graphics cards have been close to price range of a console, at launch. An X800XT was roughly $450 when the 360 launched, IIRC. A 20GB 360 was $399.99. Now, we've got the ultra-high end PC graphics cards at $1300+. There's no way a console can launch at that price unless they want to get 3DO'd. Which also launched with impressive hardware at the time.
Power draw. High end PCs today draw five times that of what is suitable for a box tucked away somewhere discreet in a living room.
Cost of IC development. Coming up with fully custom IP and hardware design is more of an effort obviously at 10 billion+ transistors per chip. Even though it could be reasonably argued that the ARM eco system would allow it, simply walking up to a PC chip vendor and saying "make one for us", makes sense too.

(Pricing is volume and margins.)
 
hmqgg

xCloud is a "white-label" service.
Not only Sony, many JP devs/pubs, and ofc including western ones, consider using xCloud or rebrand it as own service.
Some of those business talks even begun before 2018.
To meet the requirements, we'll see the size of xCloud deployment in JP would be even close to the whole EU (maybe larger).
xCloud is "mainly" based on Xbox hardwares, but those GPU instances in Azure datacenters won't stand by and do nothing, and Azure is open to partners. So xCloud is not only "XboxCloud".

https://www.resetera.com/threads/so...based-tech-for-gaming-ai.117337/post-20841665
 
Depends how they define xcloud.
At the moment xcloud is probably the name of the project.
I'm pretty sure it's not just the blades, it covers lot more of the infrastructure.
So using xcloud could mean data centers that are equipped for low latency paths etc, not the blades themselves.

I can see the Scarlett blades being used for azure work, but running PS games I find harder to believe.
ID buffer v2 (even current ps4 for BC), other customizations would need to be emulated. It's easier for MS to do compatibility compared to Sony due to the software stack they implemented.
Pretty sure PS5 was in development way before this was being considered.
Maybe cloud versions of games get recompiled?

I could see other companies (dev/pubs) that don't have a custom console running on Scarlett though.
 
The only way Sony can keep on its promise of no loading times (including the initial loading of games) is if they ship PS5 with 1TB of non-upgradable ultra fast SSD. They can either have an empty slot for user upgradable HDD/SSD sata3 or allow external HDD/SSD.

Initially, about 8-10 AAA games can be installed on the non-upgradable ultra fast SSD drive, but once an HDD is installed all games will be transferred to that storage and the whole ultra fast SSD will be turned into a scratch pad.

250 GB = scratch pad for the actual game being played
750 GB = 20GB of each games are installed

Up to 35 games (20 GB each) can be installed on the SSD and there will be zero loading times, not even the initial start up.

If 256 TB SSD + 2 TB HDD is a lot cheaper, maybe Sony will still go that route.

Would it count if Sony loads the game during bootup? Around the time the PS/PS5 Logo starts right after a cold boot. Would you still consider that loading to give us that illusion that there's no initial horrid initial load like Days Gone, Uncharted, Horizon?
 
I don't expect a 60 CU GPU I expect anything from 42-48 CUs. I just don't see next gen having the same CU count as the Xbox X.

Not only that, but AMD parts are known to have been terrible with power/performance scalling. We are not talking about a 60 CU pushed through hell to achieve 13.44 TFlops. We are talking about a 48 CU, which by itself, all things equal, would already go over 10 TFlops, but with a lower clock and voltage (undervolting GCN almost always led to a decrease in power consumption with minor drops in performance) could still be in the 10TFlops ball park with a power consumption of 200W.
 
God forbid being conservative and realistic! You won't be popular.... Seriously, i think you're right there, imo. People are expecting a huge jump from the mid-gen refreshes, foremost the One X which is a very powerfull machine for a console. We should be looking at the 1.8TF PS4, from there the jump to even just 8TF isn't that bad i think.
People are dreaming away and being very power-hungry, that is not why we buy consoles. Consoles never really had high-end hardware on release for the most. Some buzzwords from Mark Cerny seems to have upped some expectations and hype, which probably was their intention anyway.
The issue is how much of the increase in power is needed to migrate from 1080P to 4K? If we were still using 1080P displays, people would be very happy with 8TF next generation consoles. But Cerny's comments about amount of TF needed for 4K has people nervous.
 
Historically, most consoles didn't make much money and the companies that operated them went bust...

Edit - This answer is more to your rebate to my assessment that most consoles were sold at a loss. I noticed after that that was not what you actually wrote, but I'm leaving the post anyway as it is a continuation of my statement about consoles being sold at a loss.

The PS3 and X360 were sold at a loss initially and both players are still here? Even the PS4 was initially sold at a small loss. Selling at a loss is not a requirement for a business failing. The consoles not selling at all and therefore no software being sold is.

https://www.investopedia.com/articles/investing/080515/economics-gaming-consoles.asp

https://www.trustedreviews.com/news...d-from-ps-plus-subs-and-launch-titles-2905846

More on topic, I believe the Wired article implied that the PS5 will be more expensive than the PS4. The PS4 was sold with a 60 dollar loss at the beginning. If the PS5 comes at 500, this could get us a budget of 560-600 per console.
 
Last edited:
Edit - This answer is more to your rebate to my assessment that most consoles were sold at a loss. I noticed after that that was not what you actually wrote, but I'm leaving the post anyway as it is a continuation of my statement about consoles being sold at a loss.

The PS3 and X360 were sold at a loss initially and both players are still here? Even the PS4 was initially sold at a small loss. Selling at a loss is not a requirement for a business failing. The consoles not selling at all and therefore no software being sold is.
The history of consoles goes back to the 1970s. Console makers include Atari, Intellivision, Sega, etc. PS3's loss-leading wiped out all the profits from PS1 and PS2 combined and another disaster like that would definitely have seen Sony crash out of the console business. PS4 being nigh profitable hardware meant far, far better economics, although networked services are now a big boon too. XBox lost MS $5 billion by being lossy hardware (and having crap contracts) and if they didn't have cash to burn and a mission to compete, could have easily bailed on the console space as a bad idea.

Historically, it's very hard to make money successfully with games consoles. That's probably no longer as true given network services, but no company wants to wipe out their income from those networked services on lossy-hardware if its not necessary.
 
The only way Sony can keep on its promise of no loading times (including the initial loading of games) is if they ship PS5 with 1TB of non-upgradable ultra fast SSD.

Explain to me why they have to go non-upgradeable rather than 4x PCIe 4.0 NVMe? The interface is faster than 1TB of commodity flash.

If 256 TB SSD + 2 TB HDD is a lot cheaper, maybe Sony will still go that route.

It will be more expensive. Not only will the price of flash fall, and flash that is fast enough to provide fast access when you pack only 256GB of it be more expensive, the smallest hard drives they can find will be more expensive in 2020 than they are today.

The company that makes most of HDD spindle rotors just released that they expect that consumer HDD sales will fall by half next year. This will make low-end drives more expensive.
 
Explain to me why they have to go non-upgradeable rather than 4x PCIe 4.0 NVMe? The interface is faster than 1TB of commodity flash.



It will be more expensive. Not only will the price of flash fall, and flash that is fast enough to provide fast access when you pack only 256GB of it be more expensive, the smallest hard drives they can find will be more expensive in 2020 than they are today.

The company that makes most of HDD spindle rotors just released that they expect that consumer HDD sales will fall by half next year. This will make low-end drives more expensive.

I copypast my post from resetera here. Like you said economy of scale will favor SSD not HDD, they will be more expensive later. The current situation shows than 1Tb SSD is competitive with 256 Gb SSD + 2To HDD.

I don't believe any rumors with a SSD + HDD. It is a bad idea from a cost perspective. NAND flah cost is falling from a cliff. QLC SSD are currently available at 107 euros without IVA(company don't pay IVA in Europe), if you took the price for some of the logistic, retailer profit and the packaging for a mass order the cost is probably 50 euros or dollars and SSD are cheaper in US.

https://www.pccomponentes.com/cruci...MI96qqvJGe4gIVSUHTCh0JhwEmEAQYASABEgLE7_D_BwE

For a 512Gb SSD the cost is half the price 25 dollars but you need to add the price of the HDD, HDD have incompressible cost because of mechanical element, an internal 2To HDD cost 70 dollars in retail probably 35 dollars of cost in a console with mass order. The cost of the two elements combined is more expensive than the SSD alone. If you take a 256 Gb SSD the cost is only 12,5 but it is only a few dollars(2,5) under the cost of a SSD 1tb Nvme alone.

The other problem is this incompressible 35 dollars cost for the HDD at the end of the gen because of mechanical element. Price of 128 Gb SSD are very low much more than a HDD. And we can imagine it will be the case with 1Tb SSD at the end of the gen.

At the end of the gen console can cost less than 199 dollars/euros using only a 1Tb SSD and maybe less than 149 euros/dollars at the end of the generation(sure at 100% for a console without bluray other element without big economy of scale biggest part is now consoles and some mechanical element).

https://www.pccomponentes.com/kings...MIyfC98MWf4gIVaijTCh05HQahEAQYAyABEgIskvD_BwE

The price is 17 euros without taxes, and without all the element I talke about before I doubt the cost is more than 75% of the final price or maybe 50% of the final price.

NAND flash are supposed to contnue to drop until end of the year, probably most of the drop are done and it wll drop a little bit more.

https://www.tomshardware.com/news/ssd-prices-drop-2019-prediction,37982.html
 
Have Sony actually promised no loading times?

This. I'm unsure where the fixation has come from. Cerny mentions the want to remove loading times and shows a tech demo where the loading speed is increased by ~18 times. Logic suggests we'll see loading times dramatically decreasing next gen and Sony may use 'logo splashes' to hide loading (IMHO). As it stands based on what's been shown a 2 minute load will be under 10 seconds, you could say that's pretty close to no loading for most games.

The history of consoles goes back to the 1970s. Console makers include Atari, Intellivision, Sega, etc. PS3's loss-leading wiped out all the profits from PS1 and PS2 combined and another disaster like that would definitely have seen Sony crash out of the console business. PS4 being nigh profitable hardware meant far, far better economics, although networked services are now a big boon too. XBox lost MS $5 billion by being lossy hardware (and having crap contracts) and if they didn't have cash to burn and a mission to compete, could have easily bailed on the console space as a bad idea.

PS3 was an outliner though, also it wiped out all the profits of the loss leading PS1 and PS2 :p
 
Last edited:
Not only that, but AMD parts are known to have been terrible with power/performance scalling. We are not talking about a 60 CU pushed through hell to achieve 13.44 TFlops. We are talking about a 48 CU, which by itself, all things equal, would already go over 10 TFlops, but with a lower clock and voltage (undervolting GCN almost always led to a decrease in power consumption with minor drops in performance) could still be in the 10TFlops ball park with a power consumption of 200W.

We also know AMD sent engineers from Ryzen to help improve power usage.
 
I copypast my post from resetera here. Like you said economy of scale will favor SSD not HDD, they will be more expensive later. The current situation shows than 1Tb SSD is competitive with 256 Gb SSD + 2To HDD.

I don't believe any rumors with a SSD + HDD. It is a bad idea from a cost perspective. NAND flah cost is falling from a cliff. QLC SSD are currently available at 107 euros without IVA(company don't pay IVA in Europe), if you took the price for some of the logistic, retailer profit and the packaging for a mass order the cost is probably 50 euros or dollars and SSD are cheaper in US.

https://www.pccomponentes.com/cruci...MI96qqvJGe4gIVSUHTCh0JhwEmEAQYASABEgLE7_D_BwE

For a 512Gb SSD the cost is half the price 25 dollars but you need to add the price of the HDD, HDD have incompressible cost because of mechanical element, an internal 2To HDD cost 70 dollars in retail probably 35 dollars of cost in a console with mass order. The cost of the two elements combined is more expensive than the SSD alone. If you take a 256 Gb SSD the cost is only 12,5 but it is only a few dollars(2,5) under the cost of a SSD 1tb Nvme alone.

The other problem is this incompressible 35 dollars cost for the HDD at the end of the gen because of mechanical element. Price of 128 Gb SSD are very low much more than a HDD. And we can imagine it will be the case with 1Tb SSD at the end of the gen.

At the end of the gen console can cost less than 199 dollars/euros using only a 1Tb SSD and maybe less than 149 euros/dollars at the end of the generation(sure at 100% for a console without bluray other element without big economy of scale biggest part is now consoles and some mechanical element).

https://www.pccomponentes.com/kings...MIyfC98MWf4gIVaijTCh05HQahEAQYAyABEgIskvD_BwE

The price is 17 euros without taxes, and without all the element I talke about before I doubt the cost is more than 75% of the final price or maybe 50% of

Why shouldn’t Lord Vader be able to alter the plans any further? :p

Just because they might launch with some SSD and mechanical drive, doesn’t mean they can’t change the configuration in a refreshed SKU as prices become more amenable. The most important thing is to maintain the SSD portion for developers.

Mind you, I believe it’s a bit dubious to have both to begin with, although it will be interesting to see what requirements they impose for any external drive performance, which MS has been grooming for years.
 
Last edited:
They could be using e.g. a 128-256GB chunk of SLC connected directly through PCIe, and whether they use QLC or a mechanical drive as main storage depends on how low the price-per-GB on QLC gets by the end of this year.

Though to be honest I think they should get rid of mechanical drives and just use replaceable NVMe M.2 sticks.
 
https://www.dramexchange.com/WeeklyResearch/Post/2/7284.html

the NAND flash industry this year is clearly exhibiting signs of oversupply, and SSD suppliers have gotten themselves into a price war, causing SSD prices for PC OEMs to take a dive. Average contract prices for 512GB and 1TB SSDs have a chance to plunge below US$0.1 per GB by the end of this year, hitting an all-time low.
The problem is that the oversupply is going to be temporary. Next couple years might either drop more or get a manufactured price gouging again.

The good thing is that pcie controllers are now tied in price against sata. So hopefully the console expansion will be nvme.

If they considered a 256GB + 2TB, at this point they might as well put a 1TB on board and an empty nvme slot. If they can negociate a long term contract to be impervious to the next price hike.

Some power users will want to add a 4TB costing as much as the whole console.
 
Explain to me why they have to go non-upgradeable rather than 4x PCIe 4.0 NVMe? The interface is faster than 1TB of commodity flash.

I was thinking Sony want the SSD bandwidth to be uniform in all consoles as games will be developed around that fixed ultra fast SSD for streaming of assets.

To add to the idea of 20 GB of assets of each game being cached to the ultra fast non-replaceable SSD. They could enable the latest save of the game with its game assets to be saved on the alloted 20 GB. You could pick your game and in 5 seconds you could be playing your game where you left it off. (I'm thinking 4-5 GB/s to a 20 GB of RAM, so 5 seconds)
 
I was thinking Sony want the SSD bandwidth to be uniform in all consoles as games will be developed around that fixed ultra fast SSD for streaming of assets.

To add to the idea of 20 GB of assets of each game being cached to the ultra fast non-replaceable SSD. They could enable the latest save of the game with its game assets to be saved on the alloted 20 GB. You could pick your game and in 5 seconds you could be playing your game where you left it off. (I'm thinking 4-5 GB/s to a 20 GB of RAM, so 5 seconds)

That is something I would really like as I got spoiled with the instant play on the Switch.
 
I was thinking Sony want the SSD bandwidth to be uniform in all consoles as games will be developed around that fixed ultra fast SSD for streaming of assets.

To add to the idea of 20 GB of assets of each game being cached to the ultra fast non-replaceable SSD. They could enable the latest save of the game with its game assets to be saved on the alloted 20 GB. You could pick your game and in 5 seconds you could be playing your game where you left it off. (I'm thinking 4-5 GB/s to a 20 GB of RAM, so 5 seconds)
I think the only problem with a set amount of space per game is that you'll either have games outgrow the space, which will lead to compression, or things being omitted from the cache. That means loading things from slower storage, or perhaps the decompression would cause loading times to increase. I think a fast storage disk cache is the best solution if you are including a mechanical drive for installs. Storage speed is important, obviously, but I game on a PC with a SATA SSD as a boot drive, a mechanical drive with games installed to it, and an M.2 drive that I move games to when I'm going to play them. It's faster, for sure, but there are plenty of games that still take time to load, startup, and even fast travel. Unreal engine games are some of the worst offenders. Even if the drive was 2 or 3 times as fast, and it was the limiting factor, there are a couple of games that take 2+ minutes to start up so you are still talking about the better part of a minute. I doubt any hardware solution is going to eliminate loading times for the length of the generation. Maybe Sony will take a nintendo style quality control approach and force developers to limit loading to a certain amount of time to pass certification, though.
 
Status
Not open for further replies.
Back
Top