• 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.

Next-Gen PS5 & XSX |OT| Console tEch threaD

Status
Not open for further replies.

DJ12

Member
88 ROPS looks a bit unusual, hope these specs end up in the retail version.
Nvidia 2080 to, 1080 ti and probably more had this exact number so not unusual at all.

It is however 24 more than the 5700 xt which again suggests that the GPU is not rDNA 1.
 

SmokSmog

Member
Shading Units: 3,328
TMUS: 208
ROPS: 88
Compute Units: 52
L1 Cache: 32 KB (Per CU)
L2 Cache: 4MB
Clock ~ 2.0Ghz
Memory Type: ---
Memory Size: ---
Bus Width: ---
tenor.gif
 
Last edited:

Ultradsa

Member
I'm curious what Sony is ok loosing on each console, It's almost certain that it will be sold at a loss but how much of a loss are we talking? I'm pretty sure PS3 was sold at a $300 loss per unit at launch and that was a time when Playstation wasn't Sony's main money maker. I'm guessing they could afford to take an even bigger hit now but god damn is losing $300 on every console sound insane.....yet it was done before!
 
I'm curious what Sony is ok loosing on each console, It's almost certain that it will be sold at a loss but how much of a loss are we talking? I'm pretty sure PS3 was sold at a $300 loss per unit at launch and that was a time when Playstation wasn't Sony's main money maker. I'm guessing they could afford to take an even bigger hit now but god damn is losing $300 on every console sound insane.....yet it was done before!

Sony won't do that insane ps3-level loss anymore. They don't have to. But a $100 loss in the first year I think is still possible.
 

vpance

Member
I'm curious what Sony is ok loosing on each console, It's almost certain that it will be sold at a loss but how much of a loss are we talking? I'm pretty sure PS3 was sold at a $300 loss per unit at launch and that was a time when Playstation wasn't Sony's main money maker. I'm guessing they could afford to take an even bigger hit now but god damn is losing $300 on every console sound insane.....yet it was done before!

Anywhere from $20-$100 seems realistic this time. They're in a better position now to do so, more than MS may be willing to even.
 

Handy Fake

Member
I'm curious what Sony is ok loosing on each console, It's almost certain that it will be sold at a loss but how much of a loss are we talking? I'm pretty sure PS3 was sold at a $300 loss per unit at launch and that was a time when Playstation wasn't Sony's main money maker. I'm guessing they could afford to take an even bigger hit now but god damn is losing $300 on every console sound insane.....yet it was done before!
I wonder if the investment in first-party studios is a means to lessen that impact. Like photocopier manufacturers actually making money from toner, or printers being dirt cheap because most money is made on ink refills.
 
Last edited:

quest

Not Banned from OT
Sony has proven they love to surprise gen after gen. The smoke is building 💨
I don't see the cost worth it right now. The estimates I seen are 80usd per 4gig plus 20-25 for imposer. If you only do 8 gigs your at 180 usd just for half the ram at launch and forcing back to split pool again. There is a good reason AMD ran from it except some of their enterprise cards they can charge through the nose for on the compute side.
 

PocoJoe

Banned
Why, all the numbers are correct for being 52 cus.

At least say what you think is wrong instead of indicating your current level of butthurt.

Right numbers proves nothing. Anybody whom knows how these specs are made aka how many of each one have to be compared to other can make it up.
 

ANIMAL1975

Member
$1000 console then.
I'm not serious of course, but it's not totally impossible. There was this post some time ago about Sony starting producing ReRam chips
https://tech.nikkeibp.co.jp/dm/atclen/news_en/15mk/092703022/

And HBM3 i remember people in here saying (even prior to the HBM2/DRAM combo leak) that it would be a better solution than HBM2 as secret sauce, because it is better in everything and because being the next step, its price will go down faster as it gets adopted and surpass soon the older HBM2 getting discontinued. So, going HBM, the smart solution for the future cut in components cost, is the more expensive in the present.
All of this is just my two cents on the subject. Like i said i was not being serious on my original post, but thought that could elaborate a bit on the subject for the sake of conversation.
 
52 CU would amount to about 325mm2 on current N7 (30% larger than Navi 10). If it uses N6 (which offers about 18% greater logic density without needing new photomasks as N7+ would) then that could shrink back down to about 270mm2. That's not Outlandish.

If it is indeed 52CU, then I would expect the full chip would be 60 with 6 disabled. Why 60? Cause that gives three shader arrays, each with 20CU with 2 CUs disabled per array. The various BC modes could use one or two of those shader arrays depending on whether it's in PS4 base or PS4 Pro mode.

A chip with 80 CUs (4 shader arrays) with 72 active could also work as it would take the same 'butterfly' strategy as the PS4 Pro (i.e use half the chip for PS4 Pro and a quarter for PS4). But that would be far too large and costly in terms of silicon.
 

ANIMAL1975

Member
Update:

Tommy Fisher Tommy Fisher didn’t provide sufficient enough information. Now this does not mean he is discredited (so no pile ons), we just couldn’t come to acceptable terms for verification.

@CameFromNearFuture and @XcloudTimdog have yet to provide any information.
^
Everyone can think for themselves and believe or not, but he wasn't banned nor "discredited (so no pile ons)"... that's mod of war words.

Everyone just believes this guy now?
 
I wish it was 64 CU instead of 52. Or better yet, 80-128 CU range. So here is my question:

Lets assume the following:
1) PS5 52 Compute Units at 2.0Ghz with Ryzen 2 3.2 GHz
2) Xsex 64 Compute Units at 1.6Ghz with Ryzen 2 3.6GHz

Which Console is better? People keep saying that the difference in TFLOPS would only be 5-10% range, but then whey were people shitting all over with Xbox One 1.3TFLOPS vs PS4 1.84 TFLOPS ?
 

SlimySnake

Flashless at the Golden Globes
I wish it was 64 CU instead of 52. Or better yet, 80-128 CU range. So here is my question:

Lets assume the following:
1) PS5 52 Compute Units at 2.0Ghz with Ryzen 2 3.2 GHz
2) Xsex 64 Compute Units at 1.6Ghz with Ryzen 2 3.6GHz

Which Console is better? People keep saying that the difference in TFLOPS would only be 5-10% range, but then whey were people shitting all over with Xbox One 1.3TFLOPS vs PS4 1.84 TFLOPS ?
it was actually 1.2 tflops at the reveal. they only went up to 1.3 tflops in august of 2013.

and 10% of the 1.2 tflops was reserved for kinect. so it was actually only 1.08 tflops. or 70% less powerful. 5-10% is nothing compared to the 70% at reveal or the 40% with esram they launched with.
 

THE:MILKMAN

Member
I wish it was 64 CU instead of 52. Or better yet, 80-128 CU range. So here is my question:

Lets assume the following:
1) PS5 52 Compute Units at 2.0Ghz with Ryzen 2 3.2 GHz
2) Xsex 64 Compute Units at 1.6Ghz with Ryzen 2 3.6GHz

Which Console is better? People keep saying that the difference in TFLOPS would only be 5-10% range, but then whey were people shitting all over with Xbox One 1.3TFLOPS vs PS4 1.84 TFLOPS ?

Because PS4 was >40% stronger on the GPU. If there is only ~10% between XSX and PS5 it won't be noticeable.
 
^
Everyone can think for themselves and believe or not, but he wasn't banned nor "discredited (so no pile ons)"... that's mod of war words.

He has no credentials and has been pulling their "info" from months-old Pastebins practically anyone can search up and research if they wanted.

As far as I'm concerned, they're no insider. This isn't a pile-on btw; just me stating my honest opinion on the matter.

He didn't get banned yet.

Then what's holding up the approval process? I could speculate my thoughts on the matter but that'd be derailing the thread. All of the insiders I normally take some faith in are either verified or have some kind of proven track record to their name; someone who has neither is a lot sketchier, just saying.

I'm not serious of course, but it's not totally impossible. There was this post some time ago about Sony starting producing ReRam chips
https://tech.nikkeibp.co.jp/dm/atclen/news_en/15mk/092703022/

This is just R&D research and testing talks. They have "produced" ReRAM chips i.e prototype chips aiming at certain targets, but nothing commercial has materialized, and we're heading into 2020. If they are doing this basically from scratch, even if they manage to commercialize in 2020 they won't have it in the capacities Optane is currently provided in.

In fact the only thing that makes me hopeful now that ReRAM (as a storage implementation) could happen next year is if Sony and Micron have worked out some kind of deal and Micron are taking their 3D Xpoint development towards ReRAM. The two are technically PCM technologies and share a lot in common, but there are still some key differences. Theoretically it's possible to "mutate" 3D Xpoint into ReRAM of some type given the similarites, but the differences would have to be addressed obviously in order for that to work.

So the question now is if they can make those changes successfully for a commercial ReRAM product at decent capacity sizes between now and, say, June of 2020. Which I assume would be the cutoff before they really need to start finalizing hardware specs and getting everything lined up for mass manufacturing in time for the launch.

Not even entertaining HBM3; it doesn't mesh with Sony's newer system design ideologies and likely also wouldn't be congruent with the larger gaming market (believe it or not platform holders do tend to design their systems with some relative compatibility of performance to competitor offerings these days since that makes things easier for developers to deploy on multiple devices and helps them with the production and workflow pipeline regarding things like costs). It would sooner show up in lower-end enterprise data servers than the PS5.

This whole "secret sauce" thing is just getting out-of-control and kind of unrealistic now. At least there is some basis for thinking ReRAM (or better to say, 3D Xpoint, even the persistent RAM variant) can happen; it's a semi-mature tech with some years in actual real-world operation in big data business markets and offers large capacities. They are also relatively cheap to produce per GB (very cheap to produce per GB, in fact), and in 3D Xpoint's case, cheaper than ReRAM will be whenever it is finally commercialized at near-similar capacities.

Basically, gotta tame those "secret sauce" expectations with some taste of realistic expectations ;)
 
Last edited:

Ar¢tos

Member
I wish it was 64 CU instead of 52. Or better yet, 80-128 CU range. So here is my question:

Lets assume the following:
1) PS5 52 Compute Units at 2.0Ghz with Ryzen 2 3.2 GHz
2) Xsex 64 Compute Units at 1.6Ghz with Ryzen 2 3.6GHz

Which Console is better? People keep saying that the difference in TFLOPS would only be 5-10% range, but then whey were people shitting all over with Xbox One 1.3TFLOPS vs PS4 1.84 TFLOPS ?
You can't even compare...
1.3 to 1.84 is a lake in distance
12tf to 13tf is a puddle
The higher the base number, the bigger the diminishing returns are.
 
You can't even compare...
1.3 to 1.84 is a lake in distance
12tf to 13tf is a puddle
The higher the base number, the bigger the diminishing returns are.

Thats what I am having a hard time understanding. Both Xbox One (only in TFLOP terms) and PS4 have base of 1.0. Then numbers after the base is only a difference of 0.5. But now you guys are saying that if Xsex was 12 TFLOPS and PS5 was 13-14 TFLOPs with the difference in base range of 1-2 is only 10% difference in power. Yes I know there is a lot more that goes into graphics-performance than TFLOPS, but I am a noob.

:(
 
Last edited:
Status
Not open for further replies.
Top Bottom