• Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.

Digital Foundry: Hitman performance analysis (PS4/XB1)

http://www.eurogamer.net/articles/digitalfoundry-2016-performance-analysis-hitman

Starting off with the PlayStation 4 version, we were immediately pleased to see a noticeable performance increase right out of the gate. Compared to the beta, we're looking at a boost of 10 to 15 frames per second, with performance that remains above 30fps the vast majority of the time. The second training area fares even better now, and manages to turn in something approaching a stable 60fps throughout: a taste of what could have been. So while performance remains rather variable, the more serious drops in frame-rate we encountered in the beta have been completely eliminated here.

That's all well and good for PlayStation owners, but how does Hitman fare on Xbox One? First impressions suggest a near identical presentation; we see the same 1080p rendering resolution with all of the same visual features intact. Unfortunately, when it comes to performance, the results aren't quite where they need to be. The first two areas on Xbox One exhibit frame-rate numbers more in line with the PS4 beta than the final game on Sony's platform. This means drops well below 30fps in the busiest areas, with jarring spikes up to 60fps. Thankfully, with the addition of the 30fps cap, it's possible to eliminate these spikes, but the drops below 30fps remain an issue - at least in this first mission.

Once we ventured into the meat of the game, however, things took an unexpected turn. The introduction sequence for this map is mildly disappointing, both consoles dropping below 30fps with PS4, predictably, commanding a slight lead. Things look up once we're handed control, with both versions crossing the 30fps barrier as PS4 maintains its advantage. This continues as we explore the surrounding garden, but just as we make our way indoors something changes - Xbox One takes the lead.

That said, there is one other bone to pick, and it applies to both consoles equally - loading times. Using stock hard drives, both consoles weigh in at over 90 seconds of loading when starting up the Paris map.

Ultimately, Hitman turns in respectable performance on both consoles. It's not quite as consistent as we'd like, but the majority of the game can be experienced at a stable 30fps. It's an interesting take on the Hitman series that shares more with the older releases than the more linear Hitman Absolution. Next time, we're going to delve deeper into Glacier 2, the engine powering the game, and examine the PC version. In the meantime, just know that, aside from some minor issues, Hitman is solid and worth checking out.

more at the link including the vid
 

dark10x

Digital Foundry pixel pusher
It's definitely a fascinating game, engine wise. It has a subtle beauty that I quite like.
 
Somewhat off topic, but i wonder what is causing load time issues in this generation of console games. Is it just hard drive type, is it the quality of the hard drive? Does RAM type or amount factor in? I know those who think they're well knowledgeable will say "Its the low end CPUs" as if it was a reflex.
 
Good stuff, I'll be using the 30fps cap. I hope they optimise it further so that it almost never drops below 30fps on either machine.
 

_machine

Member
[Agent]ZeroNine;198046577 said:
Somewhat off topic, but i wonder what is causing load time issues in this generation of console games. Is it just hard drive type, is it the quality of the hard drive? Does RAM type or amount factor in? I know those who think they're well knowledgeable will say "Its the low end CPUs" as if it was a reflex.
Much more level data and bundles/packs to load without necessary a major steps forward in technical process of loading data (same HDD technology), higher complexity (in case of needing to compile something) etc.
 
If it's just the initial load of the big area, I won't mind.
Those kind of load times could be 5 minutes for all I care, I usually don't jump between games (at least not within missions etc.) so thanks to Resume/Suspend mode games start up in an instant.
 

On Demand

Banned
Much more level data and bundles/packs to load without necessary a major steps forward in technical process of loading data (same HDD technology), higher complexity (in case of needing to compile something) etc.

Wouldn't a higher speed BD drive eliminate the need for installing to the HDD and improve load times?
 

ps3ud0

Member
Wouldn't a higher speed BD drive eliminate the need for installing to the HDD and improve load times?
No because BDs data rate would still be nowhere as close to HDDs - a 12x BD speed is only around 300Mbps while SATA 2 theoretically is 3Gbps (though mechanical drives wont hit that; SSDs would)...

ps3ud0 8)
 
For me the beta looked like PS3's Hitman with higher res, more NPCs, but really dated overall, felt closer to PS3 than PS4 imo.

I do think Absolution looks better but that game went out of its way with colour filters and whatnot so it had a much more stylish look.

This looks a bit flat in comparison but maybe it has more impressive tech behind it.
 

dark10x

Digital Foundry pixel pusher
For me the beta looked like PS3's Hitman with higher res, more NPCs, but really dated overall, felt closer to PS3 than PS4 imo.
The training missions are tough to judge as they are actually meant to be fake reconstructions of environments within a giant subterranean bunker.

The Paris map, which actually takes place in Paris, is quite beautiful in comparison.
 

derExperte

Member
[Agent]ZeroNine;198046577 said:
Somewhat off topic, but i wonder what is causing load time issues in this generation of console games. Is it just hard drive type, is it the quality of the hard drive? Does RAM type or amount factor in? I know those who think they're well knowledgeable will say "Its the low end CPUs" as if it was a reflex.

In other games using an SSD doesn't improve load times nearly as much as it should even considering X1PS4 only have SATA2 so it's reasonable to assume that the CPUs are indeed the main culprit. Still doesn't excuse those load times.

I want to see the performance on pc, the beta on NVIDIA cards was unacceptable.

Early benchmarks but with very few cards, only highest settings and the DX12 benches are completely unreliable: http://www.pcgameshardware.de/Hitman-Spiel-6333/Specials/DirectX-12-Benchmark-Test-1188758/
 

derExperte

Member

seph1roth

Member
First DX12 performance reviews are far from good.

Maybe future "DX12 only" games will be different.

Deus Ex Mankind Divided probably will be a good example.

EDIT: Ok, seems like they have problems with AMD. 15-20% improvement on AMD with DX12, same performance or even worse with Nvidia.

Nvidia at it again, don't believe their lies...Maxwell DX12 compatible, my ass.
 

etta

my hard graphic balls
Yesssss, Xbox version is 1080p as well.
So those sub-30 drops only occur in the training area, while Paris holds 30 much better?
 
Nvidia at it again, don't believe their lies...Maxwell DX12 compatible, my ass.

Come on. You know it is more nuanced than that for all three GPU makers.

edit: in fact (although this thread is not the best for this), I am not really sure what you are referencing.

I am seeing higher mins (75 vs 70, etc.) and higher average for DX12 for 980Ti@ 1080p
as well as marginal gains for a 980. The 960 must be seeing a bottleneck other cards are not seeing @ that res. Probably due to its set-up differences (GM206 vs GM204 and GM200) as well as its comparatively aenemic bandwidth.
 

seph1roth

Member
Come on. You know it is more nuanced than that for all three GPU makers.

edit: in fact (although this thread is not the best for this), I am not really sure what you are referencing.

I am seeing higher mins (75 vs 70, etc.) and higher average for DX12 for 980Ti@ 1080p
as well as marginal gains for a 980. The 960 must be seeing a bottleneck other cards are not seeing @ that res. Probably due to its set-up differences (GM206 vs GM204 and GM200) as well as its comparatively aenemic bandwidth.

980 Ti probably is the exception due to its newest revision of Maxwell architecture.

But those "marginal" improvements of the 980 against those 15% performance improvements of the R9 390 (a GPU that competes with the 970), are embarrasing.

Let's face it, Maxwell is a shitty architecture compared to AMD GCN, with expensive GPUs that are not prepared for DX12.
 
980 Ti probably is the exception due to its newest revision of Maxwell architecture.

But those "marginal" improvements of the 980 against those 15% performance improvements of the R9 390 (a GPU that competes with the 970), are embarrasing.

Let's face it, Maxwell is a shitty architecture compared to AMD GCN, with expensive GPUs that are not prepared for DX12.

I honestly still have no idea what you are talking about, as DX12 is more than async compute as far as GPU features go. Is it that surprising that an AMD sponsored game seems to use that as its headline optimisation?

I think the answer is obvious.

Imagine another game that has order indepedent transparencies or some voxel structures in it that run an order of magnitude slow on AMD hardware due to their lack of some GPU features. It would be a similar yet reversed situation where a hardware specific feature tips the scale in a cross-vendor comparison.

You are misrepresenting the situation in its entirety and generalising for some reason. I am not sure why.
 

Kssio_Aug

Member
The unblocked frame rate seems a total mess! It seems to me that locking it is the only decent option for both console versions!

Yet, I'm glad to know that locked at 30 it seems pretty steady overall.
 

dark10x

Digital Foundry pixel pusher
Woah, really?

Fuck yes. The review thread made it seem like reloads were a minute long.

That plus the improved performance means I'm going to go buy this right now.
Maybe I was too optimistic (going from memory). :\ It's not actually that fast. It's actually more like 35-40 seconds. It's still much faster than the initial load, but not particularly short.
 
Maybe I was too optimistic (going from memory). :\ It's not actually that fast. It's actually more like 35-40 seconds. It's still much faster than the initial load, but not particularly short.

That's still fine by me. I had no issues with the beta reload timing so if it's along the same lines, the timing should be fine.
 
Top Bottom