Current Generation Hardware Speculation with a Technical Spin [post GDC 2020] [XBSX, PS5]
- Thread starter Proelite
- Start date
- Status
Do you think it could be life of black tiger?
So now SeriesX can play Best of PS4? Amazing.
Series X doesn't limit clocks..... It's a fixed clock. Meaning there's no variance. Having a limit means that clocks fluctuate and reach that limit and then throttle... Quite different.
Now xsx has 6GB/s decompressor, and it is 1.25x of the SSD speed 4.8 GB/s .
In other words xsx SSD can have 80% of the decompressor speed while PS5 SSD can only reach 9GB/s which is only 40% of the 22GB/s decompressor speed.
Is there any limitation about SONY’ SSD technique? Or they expect there will be better compression algorithm to fully use 22GB/s decompressor?
In other words xsx SSD can have 80% of the decompressor speed while PS5 SSD can only reach 9GB/s which is only 40% of the 22GB/s decompressor speed.
Is there any limitation about SONY’ SSD technique? Or they expect there will be better compression algorithm to fully use 22GB/s decompressor?
Going by your logic every single cpu and gpu is limited/throttled unless running under liquid nitrogen.
Comparing ps5 frequency changing and xsx fixed frequency and saying its limited just doesn't make sense in the context of this discussion.
Yea, probably more accurate in terms of their process, but also probably unnecessary.
Last edited:
99% pff
They should say 150%, the power of series X, powerful enough to translate from ppc to x86 realtime, no more repack and relicensing.
Xbox arcade is now reopen!!
(where my personal BC dreams lay)
Was some dev (or engineer?) tweeting that the xsx using its BCpack can mitigate the difference with the ps5 ssd. Windows central backed that up, but didn't name bcpack. It's something unannounced from MS (and other things).
There is no limitation. Just as always Microsoft like to talk about theoretical max numbers as it those were typical, average numbers. PS5 SSD also reaches >20GB/s in ideal conditions. 6GB/s (or 4.8GB/s ?) would be also ideal conditions for MS. Some data cannot be compressed as much when you use a lossless algo.
Last edited:
I went back and tried to find my predictions from the earliest post I had for this gen. This is what I had:
https://forum.beyond3d.com/threads/...eration-edition-2014-2017.56353/#post-1814831
Missed completely on Nintendo.
I think I came close enough for Sony, missed MS completely, but successfully got SSD for both.
https://forum.beyond3d.com/threads/...eration-edition-2014-2017.56353/#post-1814831
Missed completely on Nintendo.
I think I came close enough for Sony, missed MS completely, but successfully got SSD for both.
Damn, that's close to when the current gen released when you made your guess
Quite close to the PS4 gpu yes, you see the future
anexanhume
Veteran
For me, the boldest part was assuming Nvidia would be back in a performance-oriented console.
Would be nice if they do a fallback for any non repackaged game. That it can run it off disc natively.
Same fallback for Digital only games so work natively also without repackaging.
As far as I can tell they are just reporting the same source, so no windows central hasn't really backed it up.
It's less likely that AMD is playing as fast and loose with its numbers at it did with the later 28nm GPUs, in part because its most recent strategies have cut back on how many GPUs are available in a given time interval to cherry-pick. The Nano example does show how much specific SKU choices can produce leading PR blurbs.
The video's discussion about what they had to do in order to handle the change from always online shows how much engineering work was done behind the scenes that wasn't visible to outside observers. A lot of good work went into sidegrades or unsuccessful directions with the first console, but we have seen what happens when there's a more cohesive plan with the consoles since.
There's an interchange about AMD's security apparatus starting at 54:50 that is interesting in terms of how he insisted that his team wrote the relevant firmware, and that it was clear that they went with AMD because they had a good GPU. Then he apparently ran out of nice things to say about AMD's default security offerings.
That could be related to a difference in how wide and fast the connection is between the decompression block and the data fabric, or an implementation choice in the decompression hardware.
A simpler and more standard algorithm could have a smaller decompressor, and it might be more straightforward to let it ramp to some silly high number in rare instances without incurring heavy hardware costs. A more complicated compression standard may have prioritized more general scenarios at the expense of rare cases with high peak ratios, and it might have led to hardware that would have been more expensive to scale for scenarios that are very rare. For the sake of argument with made up numbers, a block can decompress at ~5 GB/s at least 75% of the time but cannot run at 20 GB/s for 0.1% of the time is more generally useful than one that runs at 4.5 GB/s or lower almost all the time but can hit 20 GB/s in very rare instances.
Panino Manino
Newcomer
I remember Microsoft saying that the SMT off was there because enabled can mess with the emulation/backwards compatibility.
Not quite. From what I recall... They indicated because all older games are setup with only 7 Threads that they would have a larger benefit from 7 Threads running even higher static speed at 3.8 GHz instead of static 3.6 GHz. They didn't say having more threads (14) would mess with emulation / backwards compatibility.
EDIT: Exact words from MS in https://www.eurogamer.net/articles/digitalfoundry-2020-inside-xbox-series-x-full-specs
"From a game developer's perspective, we expect a lot of them to actually stick with the eight cores because their current games are running with the distribution often set to seven cores and seven worker threads," explains Microsoft technical fellow and Xbox system architect Andrew Goossen. "And so for them to go wider, for them to go to 14 hardware threads, it means that they have the system to do it, but then, you have to have workloads that split even more effectively across them. And so we're actually finding that the vast majority of developers - talking with them about the their choices for launch - the vast majority are going to go with the SMT disabled and the higher clock."
It ain’t magic. It’s Kraken. It’s most marketed advantage is its decode speed which is multiple times faster than zlib. It’s also been lauded because it offers better compression (~20% better) than zlib.
But the XSX’s disadvantage doesn’t come from not being able to decode as fast as chip in the PS5 or offer less compressed data. MS is claiming 1:2 compression (2.4->4.8) for typical workloads while Sony isn’t claiming 5.5->11.
Again it’s Kraken and is not a compression scheme thats marketed as being regularly capable of losslessly compressing data at a 1:4 ratio. Kraken could claim way better compression than zlib if it was doing that level of compression for any significant portion of the data.
The PS5 biggest advantage is its raw bandwidth and the biggest benefit of kraken is probably that Sony didn’t have source a more costly solution than MS since Sony compressor is handling more twice the amount of compressed data.
Last edited:
Oh, I missed it by a mile with my 1.5GB/s six stacks HBM for 48GB of ram.... Launching in 2019!
Cross gen engines will be coded for 7 cores and 7 threads, so running at a higher frequency without smt will see a bigger benefit.
So not actually about BC titles, more about the engines themselves that are for in development games.
BC games probably works fine with smt on.
- Status
Similar threads
- Locked
