Trying to replicate Forza Tech Demo on PC for Reasons *spawn*

[pjbliverpool]

What data point is this? If you're referring to the DF Forza Apex analysis it makes no such conclusion.

Further to this, thanks to the wonder that is Rivatuner 7.0 beta along with Afterburner I was able to measure GPU usage over time in Forza on my GTX 1070. At 4K, all Ultra settings, 4x MSAA, vsync 60fps on Brands Hatch (day) with 16 cars I was averaging between 50-60% GPU usage at a locked 60fps. No where near the 88% claimed for Scorpio. Unfortunately I couldn't test Nuremberg as it's a paid DLC but Brands Hatch seems to be the most taxing of the pure race circuits (i.e. excluding Rio).

Interestingly though my frame times were a little all over the place and thus there was the occasional dip below 60fps. This was down to CPU limitations though as I'm pegging both core 1 and core 3 of my 2500K at 100% throughout the race. At no time does GPU utilisation go above 70% so the dips are clearly not GPU related. This can't be compared to Scoprio performance as DF noted that the CPU performance related settings were not ramped up to PC ultra settings in line with the GPU related settings. It does suggest it's maybe time to upgrade my CPU though.

 

[Jay]

At 4K, all Ultra settings, 4x MSAA, vsync 60fps on Brands Hatch (day) with 16 cars I was averaging between 50-60% GPU usage at a locked 60fps

what was weather set to? Was it also the same?

 

[pjbliverpool]

what was weather set to? Was it also the same?

The screenshots in the Forza Scorpio demo showed day time weather and DF didn't mention anything about rain in the piece so I assume Scorpio was running in the day.

 

[Ike Turner]

what was weather set to? Was it also the same?

The comparisons are useless anyway as there's no way to replicate what was running on the Scorpio demo with Apex on PC. It was a custom bench with a full grid of cars at there max LOD all packed together during the all thing etc. Let's all wait until E3 before wasting time jumping to conclusion.

 

[Ike Turner]

The screenshots in the Forza Scorpio demo showed day time weather and DF didn't mention anything about rain in the piece so I assume Scorpio was running in the day.

Day-time wet track was the screenshot. Anyway, the comparison is meaningless right now imho.

 

[pjbliverpool]

Day-time wet track was the screenshot. Anyway, the comparison is meaningless right now imho.

Which is a weather setting that doesn't exist in the PC game. It's either day time dry or raining. I agree that direct comparisons are very difficult at this stage, which is why claims of Scorpio being as powerful or moreso than an GTX 1070 seem unfounded. Certainly if we are looking at a Polaris class architecture then based on pure specs it should be a maximum of 25% faster than the RX 480 (as that's how much more memory bandwidth it has, ignoring contention issues). And that put's it well below a 1070 in terms of real world performance. I'm ignoring the CPU efficiencies the improved command processor brings in that analysis as we're specifically talking GPU performance. It seems clear that Scorpio will punch comfortably above it's weight in CPU terms when DX12 is properly used.

 

[ramr]

What data point is this? If you're referring to the DF Forza Apex analysis it makes no such conclusion.

The data are the data.

 

[pjbliverpool]

The data are the data.

I don't understand what this means. There is no data which points to the conclusion that you made.

 

[mosen]

This can't be compared to Scoprio performance as DF noted that the CPU performance related settings were not ramped up to PC ultra settings in line with the GPU related settings.

That's not true.

 

[pjbliverpool]

That's not true.

Yes, it is.

"The crazy story here is that we've gone over our PC ultra settings and for everything that's GPU-related, we've been able to max it - and that's what we're running at, 88 per cent,"

although i agree, if gonna try to replicate what DF was saying may as well get setting as close as possible.
can't remember the settings, but it sounded like they turned everything up including weather.

Nope they don't mention anywhere that rainy weather is part of the Forzatech demo on Scorpio and rain isn;t classed as a graphics setting on the PC, i.e. you can have rain at either Low settings or Ultra settings. The only point it's mentioned is when Richard says: " the evidence seems to suggest we're looking at performance in the same ballpark as an Nvidia GTX 1070-class GPU here and even that can drop frames at ultra settings when wet weather hits. Turn 10's stringent budgeting would ensure that this would never happen in a console release."

The last sentence there suggests that the Scorpio demo wasn't running with weather and that he's making the assumption that with weather it would still hit 60fps because of the environment budgeting. In other words - if the GPU can't hit 60 fps with weather then other elements of the environment would be dialled back to ensure that it does.

It's also worth noting that despite 88% being the headline figure for Scorpio at PC ultra settings, the video version of the article also mentions that Scorpio peaked at 100% GPU usage a few times and thus must also have been dropping frames.

 

[mosen]

Yes, it is.





Nope they don't mention anywhere that rainy weather is part of the Forzatech demo on Scorpio and rain isn;t classed as a graphics setting on the PC, i.e. you can have rain at either Low settings or Ultra settings. The only point it's mentioned is when Richard says: " the evidence seems to suggest we're looking at performance in the same ballpark as an Nvidia GTX 1070-class GPU here and even that can drop frames at ultra settings when wet weather hits. Turn 10's stringent budgeting would ensure that this would never happen in a console release."

The last sentence there suggests that the Scorpio demo wasn't running with weather and that he's making the assumption that with weather it would still hit 60fps because of the environment budgeting. In other words - if the GPU can't hit 60 fps with weather then other elements of the environment would be dialled back to ensure that it does.

It's also worth noting that despite 88% being the headline figure for Scorpio at PC ultra settings, the video version of the article also mentions that Scorpio peaked at 100% GPU usage a few times and thus must also have been dropping frames.

NO, it's not.

"This is a stress case. Like I said, we do a ton of work to prove out the performance of any of our titles," explains Tector. "What we're doing is we're letting the cars drive. The full AI is running, the full physics is running for them and then we're snapping them back into this fixed grid... And then we let the AI player drive around and this is one of the cases we use to enforce the [rendering] budgets. We do a validation pass since all the cars or all the tracks or whichever components are running within budget. That's one of the biggest reasons we can hit those smooth frame-rates. We get super-stringent about how we follow those."

 

[pjbliverpool]

NO, it's not.

The article is extremely clear that only the GPU-related features were ramped up to PC ultra settings. That's a direct quote from Turn 10's Studio Software Architect. I don't understand your denial in this matter? The quote above means nothing since there are graphics settings on the PC (reflections for example) which specifically state they have a heavy impact on CPU performance. This could easily be one of these settings that was not ramped up to Ultra levels.

 

[Ike Turner]

Gals/Guys...only a few weeks to wait...lets not get stressed out about this. Scorpio / PS4 Pro have plenty enough horsepower for everybody to enjoy what devs can do with them. Comparison to PCs is always going to end up the same as usual; Flops vs Flops the consoles will once again be a bit more efficient.

 

[iroboto]

Yes, it is.
It's also worth noting that despite 88% being the headline figure for Scorpio at PC ultra settings, the video version of the article also mentions that Scorpio peaked at 100% GPU usage a few times and thus must also have been dropping frames.

Like most ultra settings, they aren't ideal for game playing. They are designed very much for benchmarking and/or just draining whatever is left on the GPU. I'd hardly call Ultra settings optimized for visual performance nor indicative of any feature levels we would normally see in a console game.

On PC, just the step down on a handful of features at 4K resolution will make or break the difference between being playable and not playable.

 

[Ike Turner]

Like most ultra settings, they aren't ideal for game playing. They are designed very much for benchmarking and/or just draining whatever is left on the GPU. I'd hardly call Ultra settings optimized for visual performance.

Not always, but in the case of Forza Tech (as it also applies to FH3) ultra reflections (windshield in cockpit view and outside on the cars) are horrendous and totally unrealistic (too clean and sharp). We are drifting off-topic tho..

 

[mosen]

The article is extremely clear that only the GPU-related features were ramped up to PC ultra settings. That's a direct quote from Turn 10's Studio Software Architect. I don't understand your denial in this matter? The quote above means nothing since there are graphics settings on the PC (reflections for example) which specifically state they have a heavy impact on CPU performance. This could easily be one of these settings that was not ramped up to Ultra levels.

My denial or your denial? They clearly said that it's an stress test and they ramped up everything on CPU and GPU even if that wasn't necessary and they did it without any specific optimizations.

 

[pjbliverpool]

My denial or your denial? They clearly said that it's an stress test and they ramped up everything on CPU and GPU even if that wasn't necessary and they did it without any specific optimizations.

WTF? The quote is: "The crazy story here is that we've gone over our PC ultra settings and for everything that's GPU-related, we've been able to max it - and that's what we're running at, 88 per cent,"

Exactly where do they specifically say "they ramped up everything on CPU"? Hint: they do not.

 

[mosen]

WTF? The quote is: "The crazy story here is that we've gone over our PC ultra settings and for everything that's GPU-related, we've been able to max it - and that's what we're running at, 88 per cent,"

Exactly where do they specifically say "they ramped up everything on CPU"? Hint: they do not.

You said this:

This can't be compared to Scoprio performance as DF noted that the CPU performance related settings were not ramped up to PC ultra settings in line with the GPU related settings.

And I quoted another paragraph from the same article which said something else:

"This is a stress case. Like I said, we do a ton of work to prove out the performance of any of our titles," explains Tector. "What we're doing is we're letting the cars drive. The full AI is running, the full physics is running for them and then we're snapping them back into this fixed grid... And then we let the AI player drive around and this is one of the cases we use to enforce the [rendering] budgets. We do a validation pass since all the cars or all the tracks or whichever components are running within budget. That's one of the biggest reasons we can hit those smooth frame-rates. We get super-stringent about how we follow those."

Which shows that T10 really showcased a stress test. Having all cars with full AI/physics on screen with AI player is the heaviest only-CPU-related setting that one could achieved and I don't think that any player can run something similar (stress test) on his/her PC in this manner. On the graphics related setting DF described it this way:

Turn 10 literally ramped up everything to ultra and it just worked, with the game retaining a 4K60 performance level. As well as validating the capabilities of the new console, we also get an insight into how some ultra-level settings may actually amount to a rampant misuse of PC GPUs' capabilities.

So they ramped up everything (every setting) to ultra. And then we have this quote from T10:

"The crazy story here is that we've gone over our PC ultra settings and for everything that's GPU-related, we've been able to max it - and that's what we're running at, 88 per cent,"

Which means every display setting, since all of those display settings are GPU related. And you read it like this:

The article is extremely clear that only the GPU-related features were ramped up to PC ultra settings. That's a direct quote from Turn 10's Studio Software Architect.

Which isn't correct. They said "we've gone over our PC ultra settings and for everything that's GPU-related, we've been able to max it". Please show me where they said that "only" GPU-related features were ramped up to PC ultra setting? This is your assumption:

The quote above means nothing since there are graphics settings on the PC (reflections for example) which specifically state they have a heavy impact on CPU performance.

Which is in contrast considering what DF said (all settings ramped up to ultra) and there is no direct quote from T10 in support of your assumption. Instead they said "everything".

Also reflection setting (as mentioned in the game setting) has heavy impact on both "GPU and CPU".

 

[Silent_Buddha]

Which means every display setting, since all of those display settings are GPU related. And you read it like this:



Which isn't correct. They said "we've gone over our PC ultra settings and for everything that's GPU-related, we've been able to max it". Please show me where they said that "only" GPU-related features were ramped up to PC ultra setting? This is your assumption:



Which is in contrast considering what DF said (all settings ramped up to ultra) and there is no direct quote from T10 in support of your assumption. Instead they said "everything".

Also reflection setting (as mentioned in the game setting) has heavy impact on both "GPU and CPU".

I'd agree with this. As stated it would imply that anything that could impact the GPU was turned up (regardless of whether the CPU would be turned up as well). The quote being "everything that's GPU related" and not "everything that is GPU only." That part could certainly use more clarification.

The only things that may not have been turned up would be CPU only graphics rendering features. I noted in my comment on the article that since I don't have either Apex or FH3, I couldn't go in and see what settings, if any, were CPU only settings.

Something else to consider is that CPU load generated by the game is going to be lower on Project Scorpio than on PC regardless as the bulk of the draw calls will be handled by the command processors on the GPU and not the CPU.

Regards,
SB

 

[pjbliverpool]

You said this:

Yes that's exactly what I said. i.e. that based on the statement by Turn 10's studio software architect, some CPU performance effecting settings were not ramped up to PC maximums and thus PC CPU performance isn't necessarily comparable to Scorpio with everything turned up.

And I quoted another paragraph from the same article which said something else:

Which was totally irrelevant since it referred to only some specific features which would impact CPU performance and is not an exhaustive list of every feature in the game that can be dialled up which has a direct impact on CPU performance.

Which shows that T10 really showcased a stress test. Having all cars with full AI/physics on screen with AI player is the heaviest only-CPU-related setting that one could achieved and I don't think that any player can run something similar (stress test) on his/her PC in this manner. On the graphics related setting DF described it this way:

They were running a stress test (CPU included) at Xbox One settings. So yes, that was the maximum CPU load that the Xbox One version of the game is capable of delivering. The PC test was never supposed to be representative of what the Scorpio version of the game will look like, it was a simple quick and dirty test to see if the GPU could cope with PC ultra settings.

It doesn't matter how DF described it, because Turn 10 themselves described it specifically as:

we've gone over our PC ultra settings and for everything that's GPU-related, we've been able to max it

So you tell me why they make the very specific qualifying statement of "for everything that's GPU-related" as opposed to just saying "everything"? There's clearly a reason why they make that qualifying remark.

Which means every display setting, since all of those display settings are GPU related. And you read it like this:

Except there is no "display settings menu" in the XBO version of the game. So this clearly isn't a simply matter of going to the graphics menu and ramping everything up to Ultra. There may be some settings in the PC menu which merge different graphical values together which they are able to separate when manipulating the XBO version.

Which isn't correct. They said "we've gone over our PC ultra settings and for everything that's GPU-related, we've been able to max it". Please show me where they said that "only" GPU-related features were ramped up to PC ultra setting? This is your assumption:

Ok, I'll grant that my wording wasn't perfect there. I didn't mean to imply that they didn't touch any PC exclusive setting which might have had an impact on CPU performance. However I am saying that there must be some settings which they didn't ramp up which have a detrimental impact on CPU performance. Whether they also have a detrimental impact on GPU performance or not (like reflections) is irrelevant. Which is why my original statement that CPU performance between the two versions can't be compared is entirely valid - unless you have proof - contrary to Turn 10's statement, that as well as everything that is GPU related, they also ramped up everything that is CPU related to the PC maximums as well. Do you have that proof?