Except if Sony decide to sell the system at loss. I don't see that happening.
Maybe the system could be sold as a mobile device. You take PS+ for 3 years and the console cost and you pay 20$ a month ( PS+ and part of the console ) and 300$ price.
-
Ever wanted an RSS feed of all your favorite gaming news sites? Go check out our new Gaming Headlines feed! Read more about it here.
-
We have made minor adjustments to how the search bar works on ResetEra. You can read about the changes here.
Digital Foundry: PlayStation 5 - When can Sony truly deliver a generational leap in power?
- Thread starter Theorry
- Start date
Except if Sony decide to sell the system at loss. I don't see that happening.
Maybe the system could be sold as a mobile device. You take PS+ for 3 years and the console cost and you pay 20$ a month ( PS+ and part of the console ) and 300$ price.
15TF will mostly depend on how high they can clock it (and what their cooling solution is). They should be able to get 64-72 CUs in a roughly 200mm^2 area based on known 16-> 7nm scaling.[/QUOTE]
Features which largely didn't impact raw tflops throughput aren't really relevant to a discussion of the realism of expecting cutting-edge throughout from a console launch when that has not, historically, been anywhere close to the case.
The PS4 launched with the throughput of a Q1 2012 card, call it 18 months prior. Based on the same extrapolation we'd expect a late 2020 PS5 to ship with no more than whatever's in the performance segment of January to March of 2019. I'd guess 10tflops, 12 at a push.
Interesting video from Digital Foundry.
Jaguar to Zen - Massive leap forward, AMD have come a very long way since 2013's slow mobile Jaguar tech. If 7nm Zen can deliver desktop level performance in a console it would be a true next gen leap for consoles and finally allow consoles to deliver 60fps while still pushing top end graphics, like a PC can.
10-15 Tflop GPU - Again a massive leap over current gen. Base PS4 is only 1.84 Tflop and all games are built for that, so going from a base level 1.84 Tflop to a base level of 10+ Tflop, that is a huge step forward.
The cost will be the tricky thing, depending on time of launch. $399 seems to be the sweet spot for mainstream success but if they deliver with the CPU and PS5 really offers a truly next gen leap (60fps gaming standard thanks to Zen) then I'm sure people would pay a little more.
2020 seems like the more realistic time though and Sony has no big reason to rush. If they somehow did this in 2019, at a good price, it will be an incredible achievement.
Jaguar to Zen - Massive leap forward, AMD have come a very long way since 2013's slow mobile Jaguar tech. If 7nm Zen can deliver desktop level performance in a console it would be a true next gen leap for consoles and finally allow consoles to deliver 60fps while still pushing top end graphics, like a PC can.
10-15 Tflop GPU - Again a massive leap over current gen. Base PS4 is only 1.84 Tflop and all games are built for that, so going from a base level 1.84 Tflop to a base level of 10+ Tflop, that is a huge step forward.
The cost will be the tricky thing, depending on time of launch. $399 seems to be the sweet spot for mainstream success but if they deliver with the CPU and PS5 really offers a truly next gen leap (60fps gaming standard thanks to Zen) then I'm sure people would pay a little more.
2020 seems like the more realistic time though and Sony has no big reason to rush. If they somehow did this in 2019, at a good price, it will be an incredible achievement.
40% power reduction compared to 10nm for TSMC.
http://www.tsmc.com/english/dedicatedFoundry/technology/7nm.htm
I agree, and was confused why you tacked on the 15TF comment to what I was talking about.
There will likely be no 7nm consumer GPU on the market at your advertised date, so I would say this comparison is not likely to hold. We are likely to see single chip 20TF+ by the time PS5 launches if it's 2020, I would venture a guess.
PS4 was stuck on the same 28nm as previous gen GPUs. No info exists on 7nm GPUs, so we are genuinely flying blind. When you start to talk about >50% area scaling, 40% efficiency improvements, etc. of one node to the next, you have to recalibrate expectations.
Last edited:
I'm fine with 2020, it far enough that it'll be a nice jump but still close enough. Plus by then I expect 4K TVs to be better priced and the expensive sets to be accessible.
Man I can't wait for PS Meeting 2020 =P
I hope Guerrilla will blow us away.
Just give me Horizon 2 / 3 with these kind of cities in an open world... I would my tell my grandchildren about this experience.
Don't ask a question you don't want the answer to. Prior to that? PS4 Pro.
Right. The PS4s specs were locked down when a 7870 wasn't top of the line, but close to it and cutting edge, 18-24 months prior to launch, so by the time it finally shipped its GPU was no longer so start of the art. There's nothing wrong with this, I'm not criticizing the PS4 for this fact, it's just the reality of product development and manufacturing.
The point being, the PS5 is locking down final specs somewhere between right now and 12 months from now. 15 TFLOPS is a real stretch for that timeframe.
The PS4 launched with the throughput of a Q1 2012 card, call it 18 months prior. Based on the same extrapolation we'd expect a late 2020 PS5 to ship with no more than whatever's in the performance segment of January to March of 2019. I'd guess 10tflops, 12 at a push.[/QUOTE]
PS4 launched with the latest GPU architecture GCN. It was AMD's fresh take on a new GPU architecture after several generations of VLIW. Consoles usually launch with the latest GPU architecture or one that's not been featured in the desktop GPU market yet, however they may end up performing.
Hawaii was the latest shipping GCN architecture in November 2013. PS4 was Pitcairn.
I know they have to get people hooked into a new gen with an approachable price point, but I'd rather not wait an extra 3 years for the 'real' console to show up. I'm not 13 years old with no job anymore - if the games are there, I'd be willing to drop the infamous $599 (or more). Give me that option at the start.
That said... if we're still going to dick around with 30fps and sub native res, I'm not going to be receptive to even $399. 4k120 oleds are a thing (next year they'll get that support over HDMI). I've been on the 120hz train since 2011, and have now experienced 2x, 3x, even 4x higher refresh rates than that.
My minimum standards have shifted and doing the eleventy billionth retreat of the standard 30fps console performance profiles (or worse if you are goddamn From Software) bores me to tears and is absolutely frustrating. Make the necessary graphical adjustments (except AF, keep it at 16x), I'm fine with that. Just get the refresh rate high and the res native.
That said... if we're still going to dick around with 30fps and sub native res, I'm not going to be receptive to even $399. 4k120 oleds are a thing (next year they'll get that support over HDMI). I've been on the 120hz train since 2011, and have now experienced 2x, 3x, even 4x higher refresh rates than that.
My minimum standards have shifted and doing the eleventy billionth retreat of the standard 30fps console performance profiles (or worse if you are goddamn From Software) bores me to tears and is absolutely frustrating. Make the necessary graphical adjustments (except AF, keep it at 16x), I'm fine with that. Just get the refresh rate high and the res native.
Last edited:
I'd say that is fair. I do doubt 15TF based on the 1.5GHz that would be required. I still say 12TF is the upper limit in a console size box under 200W even with the 7nm and power/watt advances.
I always favored a 2020 release for "next-gen". I also think that the main gain in spec will be on the CPU side this time and not so much on the GPU side. This is why I always was thinking about 10-12TF of GPU spec with at least 16GB of GDDR Ram. And even at that base line I think it will be priced above $399 in 2020.
What architecture had been available in the PC market that the Pro did not end up using?
With the leap in CPU power 4K+60fps could become the norm. I wonder how this will affect games. Especially a new Grand Theft Auto would benefit greatly from a powerful CPU. Liberty City with tenthousands of NPCs.
They could save 10-20W system power with an HBM solution.
PS4 is Liverpool, actually. We're not talking about codenames, anyway, unless you were expecting the PC GPU to somehow ship with the PS4. Both used GCN.
Thanks. I am unclear on a few things, though. First, do we have a figure for power reduction gains from 16 to 7 nm, or from 16 to 10 nm? Since 10 nm is unlikely to be utilized for either PS or Xbox successor, its usefulness as a reference point isn't particularly apparent.
Also, I don't fully understand their numbers for speed and thermal gains:
Aren't the two variables inextricably linked? If you reduce heat generation, you can raise clock, and ostensibly a chip designer would determine the thermal characteristics of a silicon process first, which would then determine clock frequency. In other words, quoting speed seems to put the cart before the horse. How should one use these two percentages in estimating 7 nm clock speeds? Do they add to 60%, or is the calculation more complicated?
GCN is a generic term AMD uses for an entire series of microarchitectures from 2011's Southern Islands to 2019's Navi. "Both used GCN" is about as meaningful as saying "both used silicon".
We're talking about which specific GPU micro architecture was current when, so we are of course talking about their code names.
GloFo has a similar power improvement guidance for their own 16/14 to 7nm, for what it is worth.
TSMC's 16-> 7nm number is 65%. Source: https://www.semiwiki.com/forum/content/6713-14nm-16nm-10nm-7nm-what-we-know-now.html
The numbers quoted are usually with everything else held constant. Speed would have a lot of indepdent variables compared to power, so you would become limited by some factor not critical to dynamic dissipation. Moreover, to increase speed, you increase voltage, which has a cubed relationship to power these days.
my latest theory:
Sony planned to release PS5 in 2019 and that is where parts of the rumours came from. But because chip manufacturers have a delay (as mentioned in the video) late 2020 is a real possibility. I can see 499$ being a lock now that MS showed to the world that they still care about power. Not to mention Sony has proven that quality exclusive games are a big focus this gen and the market reacted in a positive-money printing way to most of these products. This alone will give them enough faith besides having all other important current-gen titles on the console. A 100$ difference isn't as dramatic as 200$ when you look at it psychologically.
Sony planned to release PS5 in 2019 and that is where parts of the rumours came from. But because chip manufacturers have a delay (as mentioned in the video) late 2020 is a real possibility. I can see 499$ being a lock now that MS showed to the world that they still care about power. Not to mention Sony has proven that quality exclusive games are a big focus this gen and the market reacted in a positive-money printing way to most of these products. This alone will give them enough faith besides having all other important current-gen titles on the console. A 100$ difference isn't as dramatic as 200$ when you look at it psychologically.
I'm on your team!
And I think there will be another PS Meeting in February/March 2020 to show off the new console, before the November release. It worked splendidly with the PS4, no reason to change plan for the PS5.
I'm not convinced HBM is suitable for a mass produced product like a console? Also GDDR6 is likely pound for pound not far off HBM power consumption wise (are there any power consumption figures yet?) but more importantly I think it is much more of a known quantity
and an evolution of the RAM they already use.
Oh I know some are cheap but if both next gen systems launch at higher costs, then people won't have a choice if they want to continue to play new games.
Sony made PS4 cheaper than Xbox One while being more powerful, so who knows what they will be able to pull off for PS5. I still think they might aim for $399 though, even if it means the specs are a little more reserved than some people might want.
HBM will still have a distinct power advantage. The question is cost and interposer throughput. Here's a good source: https://www.gamersnexus.net/guides/3032-vega-56-cost-of-hbm2-and-necessity-to-use-it
Question is how different things look in 2020.
Codenames of PC GPUs? I'm confused now. What you're probably doing is using supposed console GPU equivalent PC hardware and then making an argument for their architectures. These consoles don't have separate codenames for CPU and GPU.
maybe people doesn't want to throw in the toilet their game collection?
Please let it be a desktop level CPU. Just having that with the existing GPU hardware alone would be a huge boost. It would do wonders next gen alongside even better GPUs than that exist in the current Pro and 1X
This current gen really wasn't a big jump from the last one. I want that "holy shit no way!" moment I got from seeing Fight Night Round 3, Gears of War and Ghost Recon again.
I thought 7nm was still (surprisingly) on track to be in mobile products for consumers to buy late this year? Richard even used the same phrase I did about this (first dibs) for the mobile manufacturers to get the new process up to speed and mature and then a console chip could if Sony/MS so choose have their chip made a year later for a late 2019 launch. Or they could wait an additional year to reduce costs and wait for even better yields.
This. It's not like they launch the system in 2020 and just slap in whatever's available a couple of months before launch. They will have a very good idea of the specs now, which means that, if the PS5 releases in 2020, it will be using late 2018 / 2019 tech. If it's got bleeding edge 2020 tech in it... It's not releasing in 2020.
AFAIK it is. Partially because GloFo/TSMC's "7nm" is marketing spin. It's closer to 10nm. They're really scrapping the barrel of process nodes at this point.
This is rule #1. $400 or go home.
We posting on this forum are enthusiasts and a ton of us are willing to pay up to $600 if the HW inside is truly worth it. However, we ain't shit in the bigger picture.
7nm will definitely be in phones this year. We'll get 7nm modems by next year (X24 from Qualcomm).
It's really meaningless. You need to look at tables of the minimum dimensions of channels, interconnects, etc. to really know how they compare. At this point, their 7nm is Intel's 10nm with a slight advantage in density.
Last edited:
That BS is whole other argument but yeah.....
I think the big difference with 7nm relatively is it is a full node jump (ignoring the nm games) versus say the 28nm planar > 16nm FinFet jump. I think the latter was 3D 20nm or similar? All confusing is all I know for sure!
The PS Meeting will maybe be in March 2020, but surely not the launch of the PS5. It's either October or November. Sony doesn't launch their consoles in spring.
How about no.
I know, I'm just talking about that type of stuff getting into motion.
Thanks for that.
28 -> 14/16 is a full jump (20/22 in between). 14/16 -> 7 is a full jump (10/12 in between).
