PlayStation 5 vs. Xbox Series X: Next-Gen Console Face Off

The next generation of gaming consoles is coming at the end of 2020, and we now have key details on the hardware. Here's our breakdown of what to expect in terms of performance and features, and which console may reign supreme.

PlayStation 5 vs. Xbox Series X: Next-Gen Console Face Off : Read more
Specs aren't everything, personally I prefer Xbox but the PS4 has an install base of over double the Xbone. The machines are fairly similar in terms of spec and I doubt a difference will be seen until devs have started on games specifically for this gen so in like 18 months-2 years by which point the war has already been won. Pricing, bundles and services offered will be key. I suspect SSD size will also feature as Xbox has a 1TB model with easily expandable storage with what are essentially memory cards and the PS5 has 825GB RAW which will equate to well under 800 after OS and formatting happens and currently no PCIE 4.0 SSDs will fit into the console due to heatsinks and are expensive (like likely 1/3 of the console expensive) and need to be validated by sony.
 

Marnad

Distinguished
Apr 26, 2009
97
8
18,635
For the first time in years, I am tempted to return to console gaming. They will finally support high frame rates and variable refresh rates. The addition of an SSD is great, as well!

Still, one burning question; where applicable, will these new consoles support keyboard and mouse control? If so, it's a no-brainer for me. On paper, the XSX GPU looks comparable to an RTX 2080, which is anywhere from $950 to $1100 CAD. (I'm still using a GTX 1070).

If the XSX can consistently push 4k at 60 FPS, minimum, and offer keyboard & mouse support, count me in for this generation. Even 1440p at 120 FPS would be great.

Thoughts from anyone on the keyboard and mouse support? Perhaps I already missed this info...?
 

onemoretimex

Distinguished
Jun 29, 2015
210
4
18,595
PS5 also looks like it might be helpful for those that cant swim...

Ive always been an xbox guy but havent had a console since the 360.. however im tempted now...just wish i could have the old fat xbox controller.
 
I really wish they'd make this gen dual boot. Given the specs they'd make more than capable Windows/Linux PCs. Heck, even run an OS in a VM. As rebooting would not be necessary.

I think it would help drive sales. As a buyer could have a dedicated console and PC in one. Then people with a limited budget would not have to choose one or the other.
 
  • Like
Reactions: JarredWaltonGPU
Mar 20, 2020
2
0
10
A lot of people are overlooking the fact the file IO takes MUCH longer the floating point ops. This seems to be a huge misconception between the the two systems. After writing a simple test on my computer with 1000000 floating point ops (using * as it's the slowest, and few optimized codes use /) and reading a file (newline char seperate file of int) of 1000000 ints, the floating point ops came to 34ms and the fileIO came to 555ms. If Sony's is IO speed is roughly twice what the XBSX is, or roughly half the time, the PS5 could perform Many more flops in the time that XBSX is still loading the file. I should also mention that the file I tested was not even in .obj format which would take much longer to parse; only a little bit longer if you thread. Meaning while rendering the same scene, the PS5 will have already have finished rendering by the time the XBSX has finished loading the file. With rough numbers, benefitting XBSX, the PS5 can load a file the file and preform 18M flops in the time that the XBSX can load the file and do 4M flops. However this does mean that at a certain point the XBSX can load a scene faster, but it'd be around 50 light sources with physics already accounted for (wouldn't be more than 1000000 ops).

Another thing people miss is that both consoles will have close to instantaneous seek time, meaning game sizes will stay about the same as (unlike HDDs) the developer won't have to put hundreds (literally) of copies of textures and models on the harddrive.
 
Last edited:
A lot of people are overlooking the fact the file IO takes MUCH longer the floating point ops. This seems to be a huge misconception between the the two systems. After writing a simple test on my computer with 1000000 floating point ops (using * as it's the slowest, and few optimized codes use /) and reading a file (newline char seperate file of int) of 1000000 ints, the floating point ops came to 34ms and the fileIO came to 555ms. If Sony's is IO speed is roughly twice what the XBSX is, or roughly half the time, the PS5 could perform Many more flops in the time that XBSX is still loading the file. I should also mention that the file I tested was not even in .obj format which would take much longer to parse; only a little bit longer if you thread. Meaning while rendering the same scene, the PS5 will have already have finished rendering by the time the XBSX has finished loading the file. With rough numbers, benefitting XBSX, the PS5 can load a file the file and preform 18M flops in the time that the XBSX can load the file and do 4M flops. However this does mean that at a certain point the XBSX can load a scene faster, but it'd be around 50 light sources with physics already accounted for (wouldn't be more than 1000000 ops).

Another thing people miss is that both consoles will have close to instantaneous seek time, meaning game sizes will stay about the same as (unlike HDDs) the developer won't have to put hundreds (literally) of copies of textures and models on the harddrive.
The problem is that you're assuming IO happens constantly and stalls everything out. Most game engines are getting pretty good at putting all the important stuff into RAM, and if data needs to be pulled from storage, it's done asynchronously with a view to prefetch stuff that's likely to be needed soon.

You also need to have efficient file handling code. Back in the day, I had some code that initially processed lines of data in a file and it took about 5 seconds to run through a large file (thousands of lines). The file was about 2MB I think? Not massive, but with something like 20K lines (it was a CSV). I tweaked the file read/load code to use a 4MB buffer and grab 'chunks' of a file, and then process those. Suddenly the time was cut down to 1 second. Smart code for reading from storage will take this approach.

So, most of the file IO happens at level / game load time -- though even on PCs with the fastest storage around, there's a lot of processing of data that takes time. If you're looking at pulling 16GB of data from an SSD in an optimized fashion, that would take 3 seconds on the PS5 and possibly 5 seconds on the Xbox Series X. I don't think anyone will really care about that difference.

Plus we don't actually have hard numbers on the Xbox SSD. I've seen 4.8 GBps, which is nearly the same as 5.5 GBps for storage purposes. Having used SSDs for many years now, I can honestly say that the only times I actually notice the difference between a decent SATA SSD and the fastest NVMe SSDs are in storage benchmarks that exaggerate the gap, or in doing Steam game validations. Otherwise, load times are almost always within 1-2 seconds, and often a fast SATA SSD (eg, 850 Pro 2TB) will beat a modest NVMe SSD (eg, 950 Evo).
 
Specs aren't everything, personally I prefer Xbox but the PS4 has an install base of over double the Xbone. The machines are fairly similar in terms of spec and I doubt a difference will be seen until devs have started on games specifically for this gen so in like 18 months-2 years by which point the war has already been won. Pricing, bundles and services offered will be key. I suspect SSD size will also feature as Xbox has a 1TB model with easily expandable storage with what are essentially memory cards and the PS5 has 825GB RAW which will equate to well under 800 after OS and formatting happens and currently no PCIE 4.0 SSDs will fit into the console due to heatsinks and are expensive (like likely 1/3 of the console expensive) and need to be validated by sony.
The PS4 and Xbox One really aren't that close in raw specs. They have the same GPU architecture, but the Xbox used a 12 CU / 768 core GPU paired with a 'special' 32MB ESRAM buffer and 8GB of DDR3, with 1310 GFLOPS of compute. The PS4 had an 18 CU / 1152 core GPU with GDDR5 instead of DDR3, and 1843 GFLOPS. So the PS4 GPU was 41% faster, and the memory bandwidth was 2.59 times higher if you couldn't fit what you needed into the 32MB buffer (which you couldn't).

PS4 Pro and Xbox One X actually swapped places. The Xbox One X has 12GB GDDR5 and 326GB of shared bandwidth, with 40 CUs / 2560 GPU cores and 6001 GFLOPS of computational power. (Again, same GPU architecture, so GFLOPS is actually a very reasonable comparison point.) The PS4 Pro has 36 CUs / 2304 cores but with the same GDDR5 setup (plus an extra 1GB DDR3 for ... I don't recall), and 4198 GFLOPS. So, PS4 is faster than Xbox One, but Xbox One X is faster than PS4 Pro.
 
  • Like
Reactions: Makaveli

salgado18

Distinguished
Feb 12, 2007
928
373
19,370
The thing about storage is that no game has used high-bandwidth storage yet. The only exception may be Rage, with its streaming textures. Fast SSDs can change the way a game is made, to the point that a game level doesn't need to have a fixed list of assets, since it's so fast to load them. In some cases, the game could unload an asset an load another while the player is turning the camera! Games today consider only RAM as valid memory, but all could change in this generation. Considering this, the SSD on both consoles could be a real game changer.
 
  • Like
Reactions: Chung Leong
The thing about storage is that no game has used high-bandwidth storage yet. The only exception may be Rage, with its streaming textures. Fast SSDs can change the way a game is made, to the point that a game level doesn't need to have a fixed list of assets, since it's so fast to load them. In some cases, the game could unload an asset an load another while the player is turning the camera! Games today consider only RAM as valid memory, but all could change in this generation. Considering this, the SSD on both consoles could be a real game changer.
This is what Cerny is suggesting, BUT ... I sort of think the reason Sony / Cerny are playing up the SSD stuff so much is because they know the Xbox has a faster CPU and GPU. The big problem is that even though 5.5 GBps is 50-100X faster than the 50-100 MBps that the older hard drives could do, storing stuff in RAM is going to be almost 100X more bandwidth than the SSD. All the real-time rendering of graphics absolutely must have data already in RAM or it will choke, and that can be loaded and cached, especially with 16GB of RAM.
 
the Xbox Series X has a large SSD for storage

you forgot a BIG asterisk...it is proprietary storage!

remember how many ppl disliked sony's handhelds and their proprietary storage?

they were costly as F.


got some gen 4 storage sittign around? too bad. you cant use it casue proprietary :|
 
Mar 20, 2020
2
0
10
The problem is that you're assuming IO happens constantly and stalls everything out. Most game engines are getting pretty good at putting all the important stuff into RAM, and if data needs to be pulled from storage, it's done asynchronously with a view to prefetch stuff that's likely to be needed soon.

You also need to have efficient file handling code. Back in the day, I had some code that initially processed lines of data in a file and it took about 5 seconds to run through a large file (thousands of lines). The file was about 2MB I think? Not massive, but with something like 20K lines (it was a CSV). I tweaked the file read/load code to use a 4MB buffer and grab 'chunks' of a file, and then process those. Suddenly the time was cut down to 1 second. Smart code for reading from storage will take this approach.

So, most of the file IO happens at level / game load time -- though even on PCs with the fastest storage around, there's a lot of processing of data that takes time. If you're looking at pulling 16GB of data from an SSD in an optimized fashion, that would take 3 seconds on the PS5 and possibly 5 seconds on the Xbox Series X. I don't think anyone will really care about that difference.

Plus we don't actually have hard numbers on the Xbox SSD. I've seen 4.8 GBps, which is nearly the same as 5.5 GBps for storage purposes. Having used SSDs for many years now, I can honestly say that the only times I actually notice the difference between a decent SATA SSD and the fastest NVMe SSDs are in storage benchmarks that exaggerate the gap, or in doing Steam game validations. Otherwise, load times are almost always within 1-2 seconds, and often a fast SATA SSD (eg, 850 Pro 2TB) will beat a modest NVMe SSD (eg, 950 Evo).


I am not assuming that they are happening constantly. I count them as two seperate operation. The way I calculated it at the end assumes file IO comes before the calculation of doubles - which it would. Assets are never all loaded at the beginning, especially in open world games. We do have the numbers for the xbox ssd, it's on the official Microsoft website. I know smart code and file IO, it's my job and what I teach. Using the posix fork and a semaphored mmap in one thread and processing it in another is quite faster than buffering a file.
 
Last edited:
Microsoft winning in games? Are you out of your mind! THEY HAVE NO EXCLUSIVES!

I take Uncharted, God of War, The last of us, Spider-Man, Horizon Zero Dawn, Death Stranding, Shadow of the colosus... and now there is rumor for Sony to buy Konami franchises...

MS is NOT having better library, period...

You guys are a joke.

tumblr-mjihuv58qb1r9j9xbo1-500.gif
 
Mar 20, 2020
1
0
10
Plus we don't actually have hard numbers on the Xbox SSD. I've seen 4.8 GBps, which is nearly the same as 5.5 GBps for storage purposes. Having used SSDs for many years now, I can honestly say that the only times I actually notice the difference between a decent SATA SSD and the fastest NVMe SSDs are in storage benchmarks that exaggerate the gap, or in doing Steam game validations. Otherwise, load times are almost always within 1-2 seconds, and often a fast SATA SSD (eg, 850 Pro 2TB) will beat a modest NVMe SSD (eg, 950 Evo).


I thought the Xbox Series X I/O throughput was announced already - am I missing something? I also noticed that the article made it seem like we this data wasn't readily available.

From the link below - SSD I/O Throughput:

2.4 GB/s (Raw), 4.8 GB/s (Compressed, with custom hardware decompression block)

 
Mar 20, 2020
1
0
10
I can believe it on Games. At least MS enhances all its prior titles. I have been play GamePass for the last two years and have not run out of titles to play. They look so much better than the PS4 counterpart. I also added GamePass Ultimate and have even more to play on my PC. I have just started playing on xCloud and it is sweet to go someplace and resume a game on my tablet or phone. Face the facts, Sony is not optimizing games prior games for the new PS5. In addition, no PS1, PS2, or PS3 games to play because it is not backwards compatible and on top of that only top 100 of PS4 games will play not optimized. If you used the ultra fast SSD on older games, they get very little optimizations on the PS5 and this has already been stated.
 
Mar 20, 2020
1
0
10
That xbox will look great sitting in a corner while the owner wishes they has playstation exclusive to play. Xbox the winner. Pssh gimme a break.
 
On paper, the XSX GPU looks comparable to an RTX 2080, which is anywhere from $950 to $1100 CAD.
The RTX 2080 is going to be more than two years old by the time these consoles come out, and there should be a new generation of graphics cards launching around that time, possibly offering significantly better value. We already have the 2070 SUPER, a card that typically performs within 5-10% of a 2080 for under $700 CAD, so the 2080 is a pretty poor value at this point (and arguably always has been compared to last-gen hardware). It's possible that there could be new cards offering a similar level of performance for around the price of an RTX 2060 or RX 5700 not too long after these consoles come out. Nvidia will be moving to a new process node, AMD will be moving to the same updated architecture as the consoles, and Intel should be joining the graphics card market as well, so I would expect a lot of competition in that space relatively soon.

I really wish they'd make this gen dual boot.
The entire point of a console is to lock people into a software ecosystem. The manufacturer isn't making any significant money on the base hardware, and is in many cases losing money on it. They make their money on game licensing and distribution, online services and proprietary hardware peripherals, so if people were using their consoles as Windows PCs and largely ignoring the console side of things, that would only be costing them money.

A lot of people are overlooking the fact the file IO takes MUCH longer the floating point ops.
It's yet to be seen how the new consoles might use the SSD for things like caching data, but in general, I doubt that would be much of a limitation. Today's open-world games manage to load data in the background relatively smoothly even on the ultra slow, 5400RPM laptop drives that the current consoles use. No doubt they need to optimize within the limitations of that hardware to make it work, but moving to an SSD will give them a massive amount of headroom to work with, no matter the exact amount of performance a drive manages to offer. And of course, both consoles have 8-core, 16-thread processors to work with, so they should be able to dedicate some threads to loading and processing this data in the background without substantially impacting a game's performance.

Anyway, as these hardware "face offs" usually go, this one is rather silly. Perhaps more so than usual, since we are comparing systems based on incomplete specifications, and they are still more than half a year away from launch. How much cache will each processor have? Will they offer significantly different levels of raytracing capability? What sort of flash memory will each SSD be using? Will one system be significantly more expensive than the other? At this point, it's anyone's guess, at least among those who are not actually working on them.
 
As far as I know, Sony hasn't revealed what the PS5 will look like yet. Any images floating around are either fan concept renders, or images of the devkits, which rarely end up looking anything remotely like the final consumer product. See here for some examples of what prior console devkits looked like...


The PS1 devkit shown there looks a lot like a regular PS1, though I believe the actual pre-release PS1 devkit looked significantly different, more like a chunky PC.
 
I am not assuming that they are happening constantly. I count them as two seperate operation. The way I calculated it at the end assumes file IO comes before the calculation of doubles - which it would. Assets are never all loaded at the beginning, especially in open world games. We do have the numbers for the xbox ssd, it's on the official Microsoft website. I know smart code and file IO, it's my job and what I teach. Using the posix fork and a semaphored mmap in one thread and processing it in another is quite faster than buffering a file.
That's fine. Then I still point to the fact that memory bandwidth is about 100X faster than even the 5.5 GBps of the PS5 SSD. It will only help with load times and perhaps making things pop in faster without noticeable delays. If you're doing asynchronous loading of assets, you're perhaps get a bit of hitching right at the start of a level (happens a lot on PC), after which things smooth out. And 16GB is enough to cache quite a lot of data, even on open world games. So you load in areas that aren't not yet visible before the player hits the boundary where they're needed, and unload areas as the player moves away from those areas, and in general the storage won't need to read more than a few hundred MBps at a time.
 
I thought the Xbox Series X I/O throughput was announced already - am I missing something? I also noticed that the article made it seem like we this data wasn't readily available.

From the link below - SSD I/O Throughput:

2.4 GB/s (Raw), 4.8 GB/s (Compressed, with custom hardware decompression block)

Crud. I missed that bit. Got busy with other stuff and focused more on the CPU and GPU. I've updated the text a bit. Thanks.
 

Chung Leong

Reputable
Dec 6, 2019
493
193
4,860
Fast SSDs can change the way a game is made, to the point that a game level doesn't need to have a fixed list of assets, since it's so fast to load them. In some cases, the game could unload an asset an load another while the player is turning the camera!

On the consoles, the NVMe controller can write directly to video memory without intervention from the CPU. Latency will be very low. And it sounds like both consoles will have hardware for on-the-fly texture decompression. In theory, a game can draw upon 100GB (or even 0.5TB) of assets at any time. That's not something that can be done on a high-end gaming rig--no sane developer is going to go with a 100GB RAM requirement.
 
D

Deleted member 14196

Guest
The Xbox will be just as good if not better and it will be better because way more games are backward compatible with it and the online is second to none
 

Chris Fetters

Distinguished
BANNED
Dec 6, 2013
31
12
18,535
How the hell is the memory a "tie"??? The Series X has a MUCH wider memory bus (320bit vs 256bit on PS5) that allows it to feed it's larger GPU with significantly more memory bandwidth, as well as not having to share said bandwidth with the CPU/audio/etc... which has its own prioritized addresses & traces to the other 6GB. So please explain to me how a wider bus pushing much more bandwidth to the GPU is not a Microsoft/Xbox win again???

And if you're going to round up the PS5's TFLOP number to 10.3 (from 10.28), then you should do the same for Series X & list it as 12.2 (from 12.16). Only fair.