Baseless Next Generation Rumors with no Technical Merits [post E3 2019, pre GDC 2020] [XBSX, PS5]

Status
Not open for further replies.
Nintendo's own software dominates their platform and gives them huge margins. So their is a volume game on HW which then drives the SW. They have no reason to make high end HW that has chance of limiting HW sales due to a price barrier.
 
Phil Spencer has said in interviews that he has no idea what the name of Scarlett is going to be and they they have not even started making a list of names yet, but in this Mixer interview he says that even though the physical ID is not done that he knows what it will look like and that it is a "cool looking, "new", thing... Which will be interesting-": https://mixer.com/Mixer?vod=110357039 (around 50 minutes 20 seconds in).

If they really don't even have a list of names but are already set on the design then that design must be something special, right?

This got my brain stuck on overthinking whether there something with the spec of Scarlett or some new tech that would enable, or maybe even require, the console to have a unique "new" look?

But maybe it's just has big windows and is fully RGB kitted.
 
Except high(er) end hardware could well mean reaching a bigger audience of core gamers who presently buy the high(er) end Sony/MS option because Nintendo's is inadequate for their high(er) end gaming interests. Looking at Nintendo console sales, from NES to Wii U, ignoring the anomalous Wii, sales decreased generation to generation as the relative performance dropped. Their rivals willing to use more hardware at a higher price gained a larger share of the market; massively so. The low end price is only lucrative if coupled with a USP audience, so it worked for Wii and is possibly working with NSW.

Even when you look at handheld performance, there's a reasonable positive correlation between cost and sales performance. GBA Color was cheaper than GB (inflation adjusted) and sold worse. DS was more expensive than GB and sold more. 3DS took a dive, costing a lot more and selling a lot less in the long run, though it's still going. Interestingly 3DS sales were really strong at first, ahead of NSW's present sales it seems.
 
This got my brain stuck on overthinking whether there something with the spec of Scarlett or some new tech that would enable, or maybe even require, the console to have a unique "new" look?
What about a round box with a central vortex column of air sucked in from the bottom and blown out the top, like the Mac Pros? The problem with going too far away from the pizza-box design is clashing with the CE arrangement of the typical home.
 
Well, we have, inflation adjusted. They really didn't sell well though! (Barring the Atari 2600 which was anomalous on account of being the first video console ever.)


I had the Intellivision, good times, but wow what a cost at that time. Glad I no more have the telephone handset wires to the controllers. What's up with the cost of PS3? it was 599$ wasn't it?
 
I had the Intellivision, good times, but wow what a cost at that time. Glad I no more have the telephone handset wires to the controllers. What's up with the cost of PS3? it was 599$ wasn't it?
$499 for the cheapest available model. Same lower price option, the Core, is used for the 360.
 
I think Sony may have the need to jump from MS Servers to Google Servers. Just in case the agreement with MS fails. As Stadia's remote console is 10,7 TF and is Vega (or Navi10) Sony needs to be max 10,7 TF (and Vega or Navi10)... This to let the same software run (eventually) on Stadia's remote console. Also for Sony PS4 support for long long time is important. So many games for long time will be developed with PS4 in mind, then ps4-pro enhanced, then PS5 enhanced at double FPS than ps4-pro. In really easy, cheap way. That just needs 8,4 TF of calc power... No more. And the doubling the ps4-pro bandwidth. That is also needed. Look at Stadia remote Console. It uses 16 giga HMB2... I think also Sony will in PS5 use same memory structure: HMB2... or maybe HMB3. Just for marketing purposes (as as stated 8,4 would be enough) PS5 will be IMHO slightly above 10 TF (fp32). Really limited RT support
.. if any. Again not really needed. Different story is Anaconda. I believe MS is really preparing an expensive monster to exite fans and make America great again. Ready to sell by also loosing money... Maybe even 200 $ for each console.... I think 14 TF (Navi20)... 20 giga machine.... Real RT HW support.... 7nm+ much more expensive than 7nm process. So maybe Sony & Google will be 7nm and MS 7nm+.....
 
Last edited:
One interesting thing I remember was one leak talking about PS5 having custom Navi with 56CUs and 12.9TF. Funny thing about this leak IMO would be that it could have actually been Vega56 that was included in there, but at 1.8GHz (Gonzalo boost clock). This would bring us exactly at 12.9TF that where leaked, because there is no way in hell any consoles is having 3.5TF more then 5700XT.

Since we got several confirmations back in January that Sony sent dev kits to developers, I am thinking maybe they wanted to emulate Navi with Vega card, since it was surely early for 7nm Navi retail SOC to be included inside the dev kit.

This would nicely tie up into expected final specs which would be 3-4TFs lower, but at considarably higher IPC + RT hardware. Will see, but when I think back in 04' and 12' we had plenty of similar leaks where jurnos and leakers didnt exactly know the difference between early dev kits and final chips found in retail consoles.

Dont have edit button somehow but here is the link :


I think dev kit numbers with Vega56 where mixed up with what will be in final hardware. Both consoles will be under 10TF with Navi GPU and hw RT IMO.
 
Last edited by a moderator:
His croissant picture was taken on April 17th, 9:20:00 AM. EXACTLY.
Fake leaker.

So far the only info that I work with:
AMD Gonzalo
Ariel PCI-ID
RuthenicCookie's original posts
Official info from AMD, Sony, and Microsoft.
 
When I look at the current Navi block diagrams from the AMD slide deck, it looks like they put 2 shader engines each with 20 CU's.

If this is true, I think a potential scaling would be to 3 shader engines for 60 total CU's. The question is how many CU's do they need to disable for reasonable yields? Let's assume they have to disable 1 pair of CU's in each shader engine (since CU's are paired up now a dual compute unites), then we're probably looking at 54 total CU's. 54 CUs @ 1.5 GHz ~ 10TF.
 
In Desember 2019 it's the 25th anniversary of the PlayStaion family. I think Sony will celebrate this with more PS5 info, maybe even show the console design, introduce coming games and release date.
 
The question is how many CU's do they need to disable for reasonable yields? Let's assume they have to disable 1 pair of CU's in each shader engine (since CU's are paired up now a dual compute unites), then we're probably looking at 54 total CU's. 54 CUs @ 1.5 GHz ~ 10TF.

I agree... That is gonna be rasonably cheap, not too power hungry.... Structurally BC with possibily no OS interventions or little...

well 10.7* 56/54 = 10.32 TF.... Looking at Stadia's specs.... (that in my view share with ps5 the same NAVI10) ... NAVI10 sweet spot is probably as you said 1.5 ghz

then it can maybe run Ps4-pro turbo mode by using 36 of them at 1822 mhz.....8.4 TF

BC with original ps4... Just one shader engine at 800 mhz... So easy
 
Last edited:
One interesting thing I remember was one leak talking about PS5 having custom Navi with 56CUs and 12.9TF. Funny thing about this leak IMO would be that it could have actually been Vega56 that was included in there, but at 1.8GHz (Gonzalo boost clock). This would bring us exactly at 12.9TF that where leaked, because there is no way in hell any consoles is having 3.5TF more then 5700XT.

Since we got several confirmations back in January that Sony sent dev kits to developers, I am thinking maybe they wanted to emulate Navi with Vega card, since it was surely early for 7nm Navi retail SOC to be included inside the dev kit.

This would nicely tie up into expected final specs which would be 3-4TFs lower, but at considarably higher IPC + RT hardware. Will see, but when I think back in 04' and 12' we had plenty of similar leaks where jurnos and leakers didnt exactly know the difference between early dev kits and final chips found in retail consoles.

Dont have edit button somehow but here is the link :


I think dev kit numbers with Vega56 where mixed up with what will be in final hardware. Both consoles will be under 10TF with Navi GPU and hw RT IMO.

If 1,8 Ghz clock was real the only way to do it is in 7nm+. We already see that clock speed is far from the sweet spot at 7nm.
 
When I look at the current Navi block diagrams from the AMD slide deck, it looks like they put 2 shader engines each with 20 CU's.

If this is true, I think a potential scaling would be to 3 shader engines for 60 total CU's. The question is how many CU's do they need to disable for reasonable yields? Let's assume they have to disable 1 pair of CU's in each shader engine (since CU's are paired up now a dual compute unites), then we're probably looking at 54 total CU's. 54 CUs @ 1.5 GHz ~ 10TF.
60CU NAVI would be round about 370mm2 without ANYTHING else included (70mm2 for Zen2 + RT hardware). Really, only thing that fits in die size budget would be 40-48CUs. I think they will be gunning for higher clocks and less CUs, duo to Gonzalo leak which is pretty much 100% PS5 SOC (1.8GHz and 40CUs would be 9.5TF).
 
If 1,8 Ghz clock was real the only way to do it is in 7nm+. We already see that clock speed is far from the sweet spot at 7nm.

you use only 2/3 of the GPU in that ps-pro turbo mode... I know this is gonna a bit noisy

maybe this time GPU and CPU are in different dies... As actually Stadia is....
 
Last edited:
Status
Not open for further replies.
Back
Top