[PCgameshardware.de] Assassin's Creed Origins

[Spjut]

It's still work in progress.
http://www.pcgameshardware.de/Assassins-Creed-Origins-Spiel-61043/Specials/Benchmark-Test-1242105/2/

CPU wise, performance benefits markedly up to eight cores/16 threads.
They mention that when testing with fewer than six cores, there's occassional stutter and fluctuating framerates appearing shortly after the game has started.

 

[PhonakV30]

I Installed Assassins Creed but finally I gave up due to game play.I got bored fast but I still enjoy Gear of war 4 !

 

[Udgnim]

wow a game that actually scales up to 8 cores

 

[Krteq]

wow a game that actually scales up to 8 cores

Yeah, it's nice to finally see a progress in multi-threaded performance in games. I hope this will be a trend in a near future.

 

[nurturedhate]

Yeah, it's nice to finally see a progress in multi-threaded performance in games. I hope this will be a trend in a near future.

Not only using the extra cores but using them well. Consistent performance increases from 4 to 6 to 8, nice. I would really like to see 10 and 12 cores tested to see where the scaling stops.

 

[amenx]

wow a game that actually scales up to 8 cores

At 720p. A res from last century. Dont know why they bother with these stupid irrelevant benches in this day and age.

 

[nurturedhate]

At 720p. A res from last century. Dont know why they bother with these stupid irrelevant benches in this day and age.

Because it helps remove potential gpu bottlenecks...

 

[Carfax83]

Not only using the extra cores but using them well. Consistent performance increases from 4 to 6 to 8, nice. I would really like to see 10 and 12 cores tested to see where the scaling stops.

AnvilNext can definitely scale to 10 core CPUs. This is from Ghost Recon Wildlands which uses the same engine. It's likely though that AC Origins has an improved version of the engine with better scaling. Hopefully PCgameshardware.de will revisit CPU scaling again with AC Origins and test a 6950x or better CPU.

 

[amenx]

Because it helps remove potential gpu bottlenecks...

Who cares about GPU bottlenecks on resolutions that nobody utilizes? How practical is this info for enthusiasts on their 1440p or higher displays?

 

[nurturedhate]

Who cares about GPU bottlenecks on resolutions that nobody utilizes? How practical is this info for enthusiasts on their 1440p or higher displays?

How practical is a study on economic policy when I'm trying to balance my checkbook? Just because it's not the data you want does not diminish its importance or relevance overall.

 

[Carfax83]

Who cares about GPU bottlenecks on resolutions that nobody utilizes? How practical is this info for enthusiasts on their 1440p or higher displays?

They're testing CPU scaling, not GPU performance. Why on Earth would they use higher resolutions for, when it would defeat the purpose?

 

[Spjut]

For instance, 720p benchmarks are very useful for users with 144 hz monitors.
720p benchmarks is also useful if you're deciding for a new CPU. If you see one CPU performing much better than the other when not limited by the GPU, it's a big hint that's the CPU that will age the best.

IMO, I simply find it interesting to see that AC:O utilizes eight core CPUs. And the way PCgameshardware worded it, it sounds like you'll have to get used to the stutter and framerate fluctuations if just having four cores.

 

[Samwell]

IMO, I simply find it interesting to see that AC:O utilizes eight core CPUs. And the way PCgameshardware worded it, it sounds like you'll have to get used to the stutter and framerate fluctuations if just having four cores.

That's not fully right. They write that after the start of the game you get stutter and fluctuations. So probably it ends after a short time.

 

[IllogicalGlory]

Who cares about GPU bottlenecks on resolutions that nobody utilizes? How practical is this info for enthusiasts on their 1440p or higher displays?

It shows the CPU's headroom (so you know what to expect when you play at a high resolution on your 1180 Ti in the future), and it could be useful to show which CPU runs the game best overall, inclusive of stress points not present in the 720p benchmark run, without having to do the extra work of seeking those out.

If they tested at 1440p, showing every CPU performing the same, what good does it do anyone? I want the one with the most headroom.

 

[krumme]

Impressive scaling.
Would be nice to know a bit of how they got it. Anyone knows?

 

[amenx]

It shows the CPU's headroom (so you know what to expect when you play at a high resolution on your 1180 Ti in the future), and it could be useful to show which CPU runs the game best overall, inclusive of stress points not present in the 720p benchmark run, without having to do the extra work of seeking those out.

If they tested at 1440p, showing every CPU performing the same, what good does it do anyone? I want the one with the most headroom.

In most games there is virtually no difference between mid-range and high end CPU scaling at 1440p or higher. There are no current games of note that are NOT GPU bound. 720p benches do show considerable scaling, but this is not applicable to modern gaming on high res scenarios. I'm annoyed by this because too many people are suckered by such useless benches and may upgrade their CPUs only to find little to no difference on their high res screens.

Most benching sites know this and do not include these misleading low res benches due to this. The ones that do imo are just blindly carrying over benching methods of 15-20 years ago when it was relevant at the time. Kyle of [H] still does 640x480 res CPU benches . Its why I never read his reviews. 1080p should be the minimum.

 

[Paul98]

If you have both CPU and GPU results you can better see where your rig will be bound, what performance you will get.

Thus if you are lowing your settings to get like 144 fps, you will know if you will even get there with the CPU you have

 

[frozentundra123456]

It's still work in progress.
http://www.pcgameshardware.de/Assassins-Creed-Origins-Spiel-61043/Specials/Benchmark-Test-1242105/2/

CPU wise, performance benefits markedly up to eight cores/16 threads.
They mention that when testing with fewer than six cores, there's occassional stutter and fluctuating framerates appearing shortly after the game has started.

That scaling test is at a fixed 3ghz Ryzen. Would be interesting to see if the scaling is the same with a 4.5 or higher cpu.

 

[Guru]

720p is still relevant for CPU testing, because say 2 years from now you get the 2080ti and the games libraries you have now are not GPU bound, then you'll want the fastest processor to be able to pull more FPS.

Also it shows the theoretical maximum, so in those CPU bound situations you'd know that a certain CPU will be pushing more FPS.

But overall if money is in question, CPU is not an important choice, you can easily get anyone between 6600k to R5 1600 and be 100% great in most games, especially since you are going to be GPU bound even at 1080p.

 

[MajinCry]

Impressive scaling.
Would be nice to know a bit of how they got it. Anyone knows?

I wonder if they are using that fiber threading model. Rather than having several "big" threads that work on a task, have several threads that work on micro-tasks (tasks, that are part of tasks, that are part of tasks, that are part of tasks). With this implementation, there is no context switching overhead, so there's not a bunch of wasted CPU time each frame.

Naughty Dog uses it in their engine: https://www.gdcvault.com/play/1022186/Parallelizing-the-Naughty-Dog-Engine

 

[frozentundra123456]

720p is still relevant for CPU testing, because say 2 years from now you get the 2080ti and the games libraries you have now are not GPU bound, then you'll want the fastest processor to be able to pull more FPS.

Also it shows the theoretical maximum, so in those CPU bound situations you'd know that a certain CPU will be pushing more FPS.

But overall if money is in question, CPU is not an important choice, you can easily get anyone between 6600k to R5 1600 and be 100% great in most games, especially since you are going to be GPU bound even at 1080p.

I agree that cpu testing at gpu bound resolutions does not give a valid test of cpus, except that to say the all of them are "good enough" to allow full use of the gpu. I also agree that low resolution testing gives valid measures of relative cpu performance *under those conditions*. However, I am not really sure that one can say that it will accurately predict cpu performance in a few years with a more powerful gpu. It *may*, but IMO, it is an oversimplification to assume relative cpu performance at 720p in game A in 2017 will accurately predict relative cpu performance at 1440p,with a more powerful gpu, in game B in 2020. Personally, I agree with another poster that 1080p should be the minimum for cpu or gpu testing these days. If one wants a less gpu bound scenario, they can turn down the settings instead of going to a resolution that is lower than most smartphones these days.

 

[frozentundra123456]

AnvilNext can definitely scale to 10 core CPUs. This is from Ghost Recon Wildlands which uses the same engine. It's likely though that AC Origins has an improved version of the engine with better scaling. Hopefully PCgameshardware.de will revisit CPU scaling again with AC Origins and test a 6950x or better CPU.

This is an even more unrealistic test than the one from Assasin's Creed. Look closely, and you will see that the cpu is underclocked to 2.0 ghz, as well as lowering the resolution to 720p. I would assume they underclock the cpu to avoid being gpu bound, a reasonable thing to do in theory, but we get back to same argument about whether it is relavent to real life at all.

 

[Carfax83]

This is an even more unrealistic test than the one from Assasin's Creed. Look closely, and you will see that the cpu is underclocked to 2.0 ghz, as well as lowering the resolution to 720p. I would assume they underclock the cpu to avoid being gpu bound, a reasonable thing to do in theory, but we get back to same argument about whether it is relavent to real life at all.

Its unrealistic as far as actual gaming performance is concerned yes, but as a test to discover the parallelism in the engine, it's excellent.

 

[Spjut]

If one wants a less gpu bound scenario, they can turn down the settings instead of going to a resolution that is lower than most smartphones these days.

The advantage with just changing the resolution however is that it's strictly GPU bound, whereas turning down the settings from high to medium or low can include settings that are CPU bound as well. It's more work for the testers to identify what settings affect the CPU load as well.

 

[Carfax83]

OK I installed it and I've been playing for about an hour and a half. My impression is that the game is very well optimized considering the large amount of detail, density and simulation that's in the game. It's EXTREMELY efficient when it comes to memory usage. You'd think a game like this would use tons of RAM and VRAM, but it doesn't, as my screenshots will show. Actually, it's hard to believe that Ubisoft tweaked LoD and draw distances to the degree that's shown in AC Origins, whilst simultaneously reducing memory usage

As for CPU usage, I haven't detected anything out of the ordinary so far, at least for a DX11 game. Yeah core 1 is getting slammed more than the others, but that's typically ordinary in a DX11 game. As long as the other cores/threads are getting some usage, it's not an issue I think.

The most important things though, is that there is no stuttering at all, and stability seems to be very good as I haven't had a single crash. Overall, I'm glad I bought it so far. It's been a while since I've played an AC game, and this is as good a time as any to jump back in. It's going to be months before I beat it, as I don't get much time to play games these days.

Here is a screenshot at 1440p max settings with a Titan Xp in a busy area in Siwa. The crowd density and overall detail is far greater than in the base console versions:

​