Current Generation Hardware Speculation with a Technical Spin [post GDC 2020] [XBSX, PS5]

Status
Not open for further replies.
But that would mean in some cases it could be faster than 2.4 GB/s, which would only be a good thing.

Yes, like I said earlier, I'd be surprised if they'd actually limit the speed to that, since the controller they're using goes up to 3.75GB/s (both read & write speeds)

Its decompressor has 6GB/s limitation. At 2.4GB/s raw speed, the speed with compression is up to 4.8GB/s which is 80% speed of the decompressor.

There may not be much headroom for increasing raw speed of SSD.
 
They may also be limited by sharing a bus with the SSD expansion cartridge and be limited in bandwidth that way. It would make sense as that would give them two VME3 level storages on one bus which is efficient in terms of cost, thermo, etc. And the 2.4GB/s fits snugly with that, which is still very fast vs the previous generation after all, so I bet it didn’t feel like a big compromise.
 
I can't understand how people do math on opinions.
We do not know how much the I/O is fast, all we know is the raw speed of SSD, we don't know how the hardware subsystem works, how much the decompressor and the BC_Pack HW accelerated impact on speed, we don't know if this is peak value or average, we don't know how fast is the I/O when Mixed-size assets are requested on read (big file+lot of little files) or when read is mixed with writes.
the same things applies to the ps5 ssd, we don't know nothing
we don't know if it's sequential read or not
we don't know how temperature impact on those ssd.
we really can't say who is faster and how much, how can people do MATH on opinions?
 
I can't understand how people do math on opinions.
We do not know how much the I/O is fast, all we know is the raw speed of SSD, we don't know how the hardware subsystem works, how much the decompressor and the BC_Pack HW accelerated impact on speed, we don't know if this is peak value or average, we don't know how fast is the I/O when Mixed-size assets are requested on read (big file+lot of little files) or when read is mixed with writes.
the same things applies to the ps5 ssd, we don't know nothing
we don't know if it's sequential read or not
we don't know how temperature impact on those ssd.
we really can't say who is faster and how much, how can people do MATH on opinions?

We do have a rather good idea of how a large part of the Series X I/O actually works. The only thing that is really a mystery at this point is Directstorage. Is it related to NVIDIA's GPUDirectsorage initiative? Or is it completely something else?
I will leave the maths to others.
 
Last edited by a moderator:
Its decompressor has 6GB/s limitation. At 2.4GB/s raw speed, the speed with compression is up to 4.8GB/s which is 80% speed of the decompressor.

There may not be much headroom for increasing raw speed of SSD.
Sorry I missed where the 6gb/s limitation comes from? Couldn’t this be just another “guaranteed” number as everything they have revealed so far are sustained numbers.

Bandwidth will be vastly different depending on type of thing it is handling, for instance audio could be compressed up to like 200X, so compressing/decompressing audio it will be much faster than lets say texture.
 
Sorry I missed where the 6gb/s limitation comes from? Couldn’t this be just another “guaranteed” number as everything they have revealed so far are sustained numbers.

Bandwidth will be vastly different depending on type of thing it is handling, for instance audio could be compressed up to like 200X, so compressing/decompressing audio it will be much faster than lets say texture.

This the maximum of the decompressor like the maximum for the decompressor is 22 GB/s on PS5.
 
This the maximum of the decompressor like the maximum for the decompressor is 22 GB/s on PS5.
Where did the maximum come from? Didn’t Cernys exact words say “in some instances where it reached 22GB/s” and I’ll bet that the instance was Audio file. Audio is easier to compress/decompress requiring less compute.

Just stating that 22GB/s is maximum (in context of amount of data possible) for hwcompressor is same as saying: how many people can you transport in an hour through 2 lane highway which has speed limit of 22mph? You are missing what you are working with are you shuttling people in 2 story buses(Audio) or in a large 2 door lambourghini (textures)
 
Where did the maximum come from? Didn’t Cernys exact words say “in some instances where it reached 22GB/s” and I’ll bet that the instance was Audio file. Audio is easier to compress/decompress requiring less compute.

Just stating that 22GB/s is maximum (in context of amount of data possible) for hwcompressor is same as saying: how many people can you transport in an hour through 2 lane highway which has speed limit of 22mph? You are missing what you are working with are you shuttling people in 2 story buses(Audio) or in a large 2 door lambourghini (textures)

Average should be around 5GB/s. Steady state for Xsx is 2.4, and average compression ratio for complex stuff in Zlib is more like 2 to 1 (see the graph from the above post). Same should go for PS5 in most cases, for while I'd expect faster burst loading most SSDs run at around half general max throughput under heavy load.
 
Last edited:
This talk about ssds has people trying to fudge numbers to make themselves feel better. What started as a technical discussion has now devolved into ways we can talk ourselves into believing that the I/O gap between both platforms is smaller than it seems. Its the same thing I've observed in the other thread about the PS5 when the GPU is discussed. Its silly and its why I left the other forums to come here.

What is clear is that both companies have differing design goals and have tailored their systems to meet those goals. I for one am happy that the hardware is not a copy and paste like last gen allowing us to discuss the different approaches. However, this fudging of numbers reeks of fanboyism and has caused this thread to take an embarrassing turn.
 
This talk about ssds has people trying to fudge numbers to make themselves feel better. What started as a technical discussion has now devolved into ways we can talk ourselves into believing that the I/O gap between both platforms is smaller than it seems. Its the same thing I've observed in the other thread about the PS5 when the GPU is discussed. Its silly and its why I left the other forums to come here.

What is clear is that both companies have differing design goals and have tailored their systems to meet those goals. I for one am happy that the hardware is not a copy and paste like last gen allowing us to discuss the different approaches. However, this fudging of numbers reeks of fanboyism and has caused this thread to take an embarrassing turn.

Agreed. Far more interesting to talk about what is ultimately going to pass through the pipe than the size of the pipe itself. Trees,forests and all that.
 
What is clear is that both companies have differing design goals and have tailored their systems to meet those goals. I for one am happy that the hardware is not a copy and paste like last gen allowing us to discuss the different approaches. However, this fudging of numbers reeks of fanboyism and has caused this thread to take an embarrassing turn.

I believe/suppose that both I/O solutions will have different strong points:
Sony SSD is all for fast sequential read giant assets (UE5 techdemo, based on primitive shading, raytracing simulation using compute)
Microsoft SSD is all for fast picking a lot of little files (Advanced Mesh Shaders, direct texture tiles adressing Sampler Feedback, etc)

different strong points for two different approachs, UE5 and others engines can use both
no needs to SSD-war on internet forums.
 
I believe/suppose that both I/O solutions will have different strong points:
Sony SSD is all for fast sequential read giant assets (UE5 techdemo, based on primitive shading, raytracing simulation using compute)
Microsoft SSD is all for fast picking a lot of little files (Advanced Mesh Shaders, direct texture tiles adressing Sampler Feedback, etc)

different strong points for two different approachs, UE5 and others engines can use both
no needs to SSD-war on internet forums.

This. They both are good and will enable graphics like shown, even a gaming laptop. Everyone happy.
 
This talk about ssds has people trying to fudge numbers to make themselves feel better.

Don't need to fudge numbers to be honest. Just do the logical Math at the numbers given and the results speak for themselves.

MS said that compression bandwidth with BC is over 6GB/s.

Quoting myself in another thread.

To contrast with XBSX, the XBSX uses BC for texture compression is it seems more efficient. I think I read somewhere that it's up to 30/40% more efficient than Kraken's 10% gain.

4.8GB/s + 40% is 6.72GB/s. Again this seems to align with what MS said about the SSD and the BC compression ratio i.e. over 6GB/s.

Likewise, 30% is still 6.24GB/s. Of course these are with the information given and what I read regarding BC over Kraken.

That said, yeah you have fanboys taking the information given and changing it for their own purposes and exaggerating things way out of proportion. Silly.
 
This the maximum of the decompressor like the maximum for the decompressor is 22 GB/s on PS5.

Its more like 25-27 GBs. The bandwidth of the SDD is the limiting factor as the decompressor can handle over twice the volume of compressed data than the SSD can provide. 22 GBs is a rate provided data that lossly compressed at a ratio of 4:1 (5.5 X 4).
 
Last edited:
Status
Not open for further replies.
Back
Top