• 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.

44alltheway

Member
zKvIdt0.png
what is Dante?
 
The XSX is 432 cubic inches³
The 1X is 263 cubic inches.

Theres a reason why the XSX has more space.
My guess is 48 - 52cus @ 1600-1700mhz.

We have to remember the SoC also has a 8core Zen2 which will be putting out some heat too.
Even an SOC with a 8 core zen2 + a 5700xt would need a rather serious cooling system

If designed like this with a huge heatsink and fan, 52,56cus @ 1600-1800mhs is possible.
Makes you wonder why this design has not be done before.

f282f91327921762.png

It was. In 1994. It's called the PC-FX ;)

latest
 


DANTE ( XDK )

CPU: Custom AMD Zen 2 8C/16T @ 3.2GHz
GPU: Custom AMD Navi @ 1475MHz
MEMORY: 48GB GDDR6 @ 672GB/s
STORAGE: SSD 4TB NVMe @ 3GB/s

- 2x more memory and 2x faster bandwidth than XBX.
- Navi has 64 CUs with same tflops as GCN that would mean almost exactly 12 Tflops (~12.095), again 2x more than XBX GPU.

This is definitely going to be $499.
 
If these specs are real and the retail console matches it (GPU only):

DANTE ( XDK )

CPU: Custom AMD Zen 2 8C/16T @ 3.2GHz
GPU: Custom AMD Navi @ 1475MHz
MEMORY: 48GB GDDR6 @ 672GB/s
STORAGE: SSD 4TB NVMe @ 3GB/s

Only at 64CUs it'd be 12.08TFs. That is very unlikely imo.
 
Last edited:
RTX 2080 Ti uses GDDR6. Does it have a latency penalty for RT/BVH acceleration?

I'll just quote here what he said to me in pm.
"ReRam being only 10 times slower than main memory sounds awesome, but remember even main ram is usually considered much too slow.
Accessing main memory from GPU (speaking about the dedicated memory on a discrete PC GPU here) often means a latency of 800(!) cycles.
GPUs (try to) solve this with caches and using massive hyper threading so other work can be done while waiting.
Both does not work well with RT, which is the major technical reason it is so expensive (aside the problematic time complexity of the algorithm in general)."



ReRAM is closer to NAND though, I think. GDDR6 is much better.

I was thinking ReRAM as a supplement to main ram, not a replacement of it. So exactly like Amigo Tsuishui said, a storage between RAM and NAND. Basically like Intel Optane solution but stepping it up a little further by implementing AMD Solid State Graphics solution by connecting ReRAM directly to the APU/GPU like it's a ram.

I'll quote here what he also told me about how ReRAM could be used. But he said the problem is actually content creation and games ballooning in size.
"Assuming magic compression exists, CPU could uncompress insane details to ReRam so it is available quickly, e.g. when player turns view.
Then my initial Megatexture argument would make more sense again, if we make some assumptions:
* decompression very expensive, so need to cache full environment around player, not just what's currently on screen. 10 x more data.
* Sub millimeter texture resolutions everywhere. 100x100 more data?" {/QUOTE]
 
Last edited:

quest

Not Banned from OT
If these specs are real and the retail console matches it (GPU only):



Only at 64CUs it'd be 12.08TFs. That is very unlikely imo.
That I'm assuming would be the old one not the one that shipped recent the 12tf was probably leaked to kleegamingfan from. Probably had a few silicon revisions since then. I don't think it is running at the insane Sony speeds but faster that 1.4. That thing is not 64CUs active in retail. Both are going to be in the 50s active imo.
 

vpance

Member
That I'm assuming would be the old one not the one that shipped recent the 12tf was probably leaked to kleegamingfan from. Probably had a few silicon revisions since then. I don't think it is running at the insane Sony speeds but faster that 1.4. That thing is not 64CUs active in retail. Both are going to be in the 50s active imo.

I think MS went wide with slow clocks which would make it easier to follow their design philosophy of late of keeping their cases simple to cool.

Would not be surprised to see 64CU in there. It's the obvious solution to achieving high performance on Navi as even we found out from tests done by DemonCleaner DemonCleaner on here. ~1.5Ghz is the optimal clock for perf/watts.

powerscalingfactor5kkho.png
 
Exactly. Especially since navi performs better at higher clocks. Having 64 cu and clocking it so low will be expensive and ineffecient compared to a lower CU count with higher clock

Honest question. Would lower wattages mean more reliability in the future?

Also, I'm wondering if Ms could push clocks up to +100Mhz just to play catch up if/when Ps5 confirms they are stronger. Would also help a bit to improve Navi's performance a little bit,
 
Last edited:
Honest question. Would lower wattages mean more reliability in the future?

Also, I'm wondering if Ms could push clocks up to +100Mhz just to play catch up if/when Ps5 confirms they are stronger. Would also help a bit to improve Navi's performance a little bit,
It just means the console will run cooler so cheaper solution is needed for cooling . But looking at design of xboxsx
It has been designed around the idea of cooling a hot gpu. So the clocks should be around 1.8 imo.

ofcourse gpu and cpu can be up clocked if the TDP allows. But there was a rumor saying MS increased the clocks for the gpu beginning of 2019 in the target spec sheet as the anaconda was behind ps5 target specs . So if that’s true the target specs given in 2019 from beginning is up clocked and probably will not have much room left to be over

ofcourse all of these are fairy tales for the time being till more stuff leaks after their reveals.
 
I'm treating what you say here as your speculation just to be safe. (My hype are so damn high right now.)

Anyway, what are your thoughts about STT-MRAM? Sony is already mass producing that kind of memory for a company called Avalanche. I also read Intel is planning to put L4 cache using MRAM in their CPUs. Isn't STT-MRAM a good candidate for L4 cache?
Definitely a good candidate due the higher densities at a lower cost compared to Toggle MRAM.
 
It just means the console will run cooler so cheaper solution is needed for cooling . But looking at design of xboxsx
It has been designed around the idea of cooling a hot gpu. So the clocks should be around 1.8 imo.

ofcourse gpu and cpu can be up clocked if the TDP allows. But there was a rumor saying MS increased the clocks for the gpu beginning of 2019 in the target spec sheet as the anaconda was behind ps5 target specs . So if that’s true the target specs given in 2019 from beginning is up clocked and probably will not have much room left to be over

ofcourse all of these are fairy tales for the time being till more stuff leaks after their reveals.

Thanks a lot for the clear explanation, definately the design of the box is aimed at being very cooling efficient
 
So far, everything indicates that they will just use the same kind of cooling as the PS4 Pro had.
Do you know for a fact that they developed a new solution?
They definitely have made a new (evolutionary) cooling solution for PS5 that allows them to rise the Clock speed for GPU in a console size box like never seen before. FACT.

CameFromNearFuture CameFromNearFuture

I thought he said 13.02.2.0 ?
That's not correct.
 
They definitely have made a new (evolutionary) cooling solution for PS5 that allows them to rise the Clock speed for GPU in a console size box like never seen before. FACT.


That's not correct.
So let’s get a bit more specific about the date of meeting . Do you know which date in February ?
 

DJ12

Member
So let’s get a bit more specific about the date of meeting . Do you know which date in February ?
There were leaks as far back as august saying the meeting was on the 12th, which is a Wednesday, the same day of the week the PS4 reveal happened.

So I'm going to guess it's the 12th
 
There were leaks as far back as august saying the meeting was on the 12th, which is a Wednesday, the same day of the week the PS4 reveal happened.

So I'm going to guess it's the 12th
So a week or so before ps4 reveal in 2013 ? Which means mid Jan we should have the announcement. Less than a month till announcement .but I want to hear Tommy Fisher Tommy Fisher feedback as well so we know more about him :p
 
Pretty sure it was the 20th of Feb 2013.
Yea last time it was feb 20 and announcement for meeting came out 28 Jan . So if the meetings is a week ahead this time so will the announcement of the event. So we will hear the ps meeting announcement from Sony in mid Jan
 

Panajev2001a

GAF's Pleasant Genius
And it will cost $600 and decimated by the lockhart and Xbox will outsell Playstation easily.

There is a strong rumour that Playstation will release two consoles... so that would sure even out the race!

Haha, we shall see ;). I think, as usual, you are underestimating Sony.
It feels like two quite custom “semi-custom” HW designs AMD has cooked up for MS and Sony
 
Last edited:
i just read this threads tbh, i have 0 teraflops of all this tech speak, i do i have one question tho!

the xbox series x backward compatibility is it disk based or only digital?
 

TBiddy

Member
Haha, we shall see ;). I think, as usual, you are underestimating Sony.
It feels like two quite custom “semi-custom” HW designs AMD has cooked up for MS and Sony

I don't think the difference will be stark. Both of the consoles will hit about the same mark (TF wise), have about the same CPU-power and about the same amount of memory.
 
Status
Not open for further replies.
Top Bottom