Digital Foundry Microsoft Xbox Scorpio Reveal [2017: 04-06, 04-11, 04-15, 04-16]

What do you mean "not really", that's the purpose of the HW flags in the DirectX APIs! :rolleyes: As for AMD, probably, because objectively that's the best comparison to make - so that as many things as possible are equal.



Objectively it's 8.4Tf of FP16 RPM operations vs. 6Tf of 32-bit operations. How many FP32 operations can be optimised as FP16 RPM operations? The same number as the length of a piece of string :yep2:

I see your point from a theoretical standpoint. But can you discount bandwidth, memory and other advantages when talking 16 vs 32fp?

And how readily can 16FP be deployed across gaming development? It's seems 16 FP inclusion was mostly motivated by developments outside of gaming. You would think if RPM can drastically increase performance across broad spectrum of instructions, it implementation would have become a standard a long time ago, AMD would be expounding its use across a broader set of features and nvidia would of done more than include 1 16fpx2 unit per SM on consumer level gpus.
 
First two layers standard BR up to 50GB, third layer 33GB 4k texture pack?
Oh, that's interesting and would be great Xbox. :yes: I wonder if dual-format discs like this would work.

I see your point from a theoretical standpoint. But can you discount bandwidth, memory and other advantages when talking 16 vs 32fp?
Theoretical maximum performance is predicated on the rest of the system being able to support it.

nd how readily can 16FP be deployed across gaming development? It's seems 16 FP inclusion was mostly motivated by developments outside of gaming.

That's the big question, and the answer is that until recently there has been no incentive to explore this. FP32 is overkill for a lot of things, the $64m question is will FP16 will sufficient? Hair, fur, minor vegetation... Time will tell.
 
Last edited by a moderator:
How quaint, people talking about physical media.
Do you believe digital game purchases outnumber physical game purchases? If so, I've not seen compelling evidence.
 
How quaint, people talking about physical media.

Yes, because I collect games. I like sometimes return to some game, and I don't want download again 100 GB each time I want play some game. And also 1 TB HDD isn't enough, that's enough only for 10 games if they are 100 GB, or some 7 Games with DLC. Also HDD isn't good for long time storage. I had some games on my Xbox 360 HDD, now that HDD doesn't work anymore, so I need to buy new and download again all my stuff.
 
The memory interface queues may have been sped up due the change to a wider GDDR5 interface, or if they're queues in the uncore they could have sped up due to everything else being clocked significantly higher as well. The prior consoles had a lot matching the GPU speed, which Scorpio did increase significantly. I think that might serve more to make CPU and GPU upclocks suffer less from diminishing returns, rather than standing out on its own.

Optimizing how memory translation operations happen is a fuzzy thing to interpret.
Some possibilities might be more capacity to cache translations between host and guest, or tweaks to the TLB or page table walker that reflect what Microsoft knows as the owner of the hypervisor and system memory arrangement.

In theory, the VM's overheads would have been reduced already, since Microsoft mentioned that it did a lot of work like this for the Xbox One and over time it's seemed like there have been enough examples where the Microsoft's CPU-bound performance managed to edge the PS4. Maybe there are specific functions that Microsoft wanted improved, but from the standpoint of how the Xbox One didn't clearly suffer in CPU terms versus its non-virtualized competition, it might not be as obvious because a lot of other work has been done to avoid the overhead.

Thanks. I hadn't considered parts of the uncore matching GPU speed, which has increased considerably. CPU is now - going by released specs - less than double GPU speed, where as for X1 (none S) it was more than double.

I suppose simply being less CPU limited than PS4 and PS4Pro despite using virtualised OS(es) would be considered "good enough" for the market in which they're operating.
 
The flag will be for for the game to know that it can serve RPM-tuned shaders instead of 32-bit shaders if the GPU/driver have set the flag in DirectX. Hair and fur look to be the easy wins. PS4 Pro will be 30% hairier and furrier than Scorpio :nope:
I knew you are into Furry Paradise.

as I said in another thread, I really fear for the price of the console.
 
For the PC, yes. For tablets, yes. For phones, yes. For consoles stuck in the past, no.
That's what I thought. Where there is choice, consumers prefer physical. Where there is no choice, digital. Like Dictatorship vs. Democratic processes :mrgreen:
 
And also 1 TB HDD isn't enough, that's enough only for 10 games if they are 100 GB, or some 7 Games with DLC.

Which is why it's great one can plug in external hdds, two external drives for xbox and at least 1 for PS4, and play games from there.

But I get other's preference for physical media even if it might but be around in the future for consoles.
 
Sadly almost nothing new, just describing the expected production and integration with hyperboles and car references.

Not sure if they have anything left to talk about once we know all the specs, and see the board, and the exploded view of all components.


Probably the only "new" info I get out of it is he's strongly hinting it will be small. Given the GAF measurements, it's likely it simply fits in an Xbox One S case and will be sold in one.

BTW, I'm dubbing double rate FP 16 the new mythical "secret sauce"...
 
Probably the only "new" info I get out of it is he's strongly hinting it will be small. Given the GAF measurements, it's likely it simply fits in an Xbox One S case and will be sold in one.
From my own estimates of the size of the board (based on comparing it to the One S board in an image editor and aligning usb/hdmi ports), the board itself is slightly larger than the S board in both directions, so it's definitely not going to just re-use the same case design. Not that they could anyway, given the completely different thermal design (rear-vented blower vs top-vented axial fan).

Also, the ports are reverse order compared to the One S, so based on that (and a few other things, such as the UHD drive position and apparent orientation) I suspect that it's going to be a design similar to the original PS4, with the cooler mounted below the board, rather than above it.
 
That's what I thought. Where there is choice, consumers prefer physical. Where there is no choice, digital. Like Dictatorship vs. Democratic processes :mrgreen:

Well that certainly applies to phones and tablets. But for PC, physical fought a losing battle to digital as when offered a choice, consumers for the most part were unwilling to continue buying physical copies once they got used to the idea and convenience of digital ownership.

In the end, I think it was the convenience of digital ownership that consigned physical distribution to being just a alternate distribution method for digital copies of a game. It was so convenient it even got many people to stop pirating games. :)

Regards,
SB
 
What I don't get about this simulation approach to improve their design is that shouldn't this be AMD's job during the design stages of their APUs and GPUs? It's as if MS would be the "only" one who has an idea how games use their hardware in the first place to localise critical areas and improve on them.

AMD and Microsoft have two completely different goals with some overlap for what they require from a GPU.

AMD is in competition with NVidia not only for the consumer market but for the professional market. As such they have to attempt to innovate WRT to features as much as possible or risk losing significant ground to their competition.

Microsoft on the other hand, with Project Scorpio, was focused entirely on increasing the performance of current methods of rendering games on consoles.

That means that while profiling of current generation games is helpful in increasing design parameters for a GPU to accelerate how developers are currently using GPUs, it doesn't address potential future changes to how games might be rendered. OTOH, designing with the future in mind is risky. It could mean that your GPU ends up being less competitive than a GPU focused on current graphics rendering in games. For example, in the PC space AMD GPUs (GCN in particular) have generally aged quite gracefully as they get older, while NVidia GPUs sometimes don't age as well. But at the same time, those NVidia GPUs had a performance advantage at the time when they were most relevant.

So, when it comes down to it. Yes it's helpful to profile current generation games extensively to make a GPU more efficient at rendering those games and presumably other games in the near future of that generation. No, it's not necessarily something you want to focus on if you are designing GPU technology.

So, looking at something like double rate FP16. It definitely can and will have advantages in the future once engines are designed for it.. It also has virtually no advantages "now" and won't for any engines that don't take advantage of it in the next few years.

So if we look at Project Scorpio, it'll have massive benefit with regards to all games currently releasing and releasing in the next few years. The addition of dual rate FP16 would have virtually no benefit to current games, and little to no benefit to most games coming in the near future with the possibility of more widespread benefit after maybe 2-3 (?) years once it's incorporated into general use engines (UE, Unity, etc.) and once development of games using those modified engines enter the market.

Had AMD focused too much on profiling current generation game rendering techniques they may have decided not to include dual rate FP16 and/or decided that it was appropriate for the professional market but of little to no use in the consumer gaming market much like NVidia.

Regards,
SB
 
How many FP32 operations can be optimised as FP16 RPM operations? The same number as the length of a piece of string :yep2:
In addition to that theoretical peak, how many will be optimised for PS4 Pro when the vast majority of the market won't benefit from that work? By not including 16 revolutions per minute, MS has made it a very niche feature, I think. Unless you can optimise with a couple of lines saying 'use FP16 please!'. (?)
 

All of it is pretty meh. But one thing that stuck out is how much attention they are paying to mounting the HDD and securing the HDD cable. Enterprise storage applications generally don't even go this far in securing their HDDs. It all seems a bit excessive, especially if they are using a standard 5200-5900 RPM HDD. Considering that all of this is focused on increasing the reliability and thus speed of data access from the HDD, I wonder if this mean's we might see a 7200 RPM drive as standard? All that work would make a bit more sense at that point, still excessive, but at least it makes a bit more sense. It is meant to be a premium machine, after all.

Just seems odd to me.

Regards,
SB
 
All of it is pretty meh. But one thing that stuck out is how much attention they are paying to mounting the HDD and securing the HDD cable. Enterprise storage applications generally don't even go this far in securing their HDDs. It all seems a bit excessive, especially if they are using a standard 5200-5900 RPM HDD. Considering that all of this is focused on increasing the reliability and thus speed of data access from the HDD, I wonder if this mean's we might see a 7200 RPM drive as standard? All that work would make a bit more sense at that point, still excessive, but at least it makes a bit more sense. It is meant to be a premium machine, after all.

Just seems odd to me.

Regards,
SB
have to admit i assumed it was 7200 drive, but maybe its not.
i know a lot of people wasn't happy with the article. But i found it interesting, all the attention to detail.
excessive? maybe, and i wouldn't have appreciated it without the article.
maybe it needed to be excessive to actually fit everything into the form factor that it is.
things that aren't a big deal in a bigger foot print gets magnified the more you compress it all together.

there was lots of laughs and complaints about the x1 looks and size, have a feeling won't be the case with this one.
 
have to admit i assumed it was 7200 drive, but maybe its not.
i know a lot of people wasn't happy with the article. But i found it interesting, all the attention to detail.
excessive? maybe, and i wouldn't have appreciated it without the article.
maybe it needed to be excessive to actually fit everything into the form factor that it is.
things that aren't a big deal in a bigger foot print gets magnified the more you compress it all together.

there was lots of laughs and complaints about the x1 looks and size, have a feeling won't be the case with this one.

What would be wild, if impractical, would be if all that excessive detail was to incorporate a 10-15k RPM HDD. :D However, that would increase the cost quite significantly. Also, it'd need beefier cooling for the HDD if that were the case. So that's obviously not going to happen.

Regards,
SB
 
Back
Top