[PCgameshardware.de] Assassin's Creed Origins

[Carfax83]

A game like this could benefit heavily from DX12. It's a shame that Ubisoft couldn't get it ready in time for release. I really hope they implement a DX12 patch later down the road though.

 

[psolord]

Hello. Sorry for the bump. I did some Assassin's Creed Origins benchmarks and I think the findings are semi interesting.

My benchmark consists of an 8 minute run, doing various (but the same on all runs) stuff. Except on the 5850 which the performance was giving an unplayable experience, so I stuck with the built in benchmark. The built in benchmark is run on all the other systems as well, for reference.

Latest 1.5 patch for all systems with the same latest drivers amongst same gpus.

Everything is documented in the following videos.

Assassin's Creed Origins 1920X1080 Ultra GTX 1070 @2Ghz CORE i5-8600k @5Ghz - 82fps

Assassin's Creed Origins 1920X1080 ultra GTX 1070 @2Ghz CORE i7-860 @4GHz - 66fps

Assassin's Creed Origins 1920X1080 Ultra GTX 970 @1.5Ghz CORE i7-860 @4GHz - 49fps

Assassin's Creed Origins 1920x1080 High 7950 @1.1Ghz CORE i7-860 @4GHz - 40fps

Assassin's Creed Origins 1920x1080 Medium 5850 @900Mhz CORE i7-860 @4GHz - 10 fps

For the 5850 the performance was roughly the same for Very Low, Low and Medium. It is quite probably hitting a vram limit so hard, that the performance stays fixed at that framerate for all these settings. The HD630 of the i5-8600k, which is not documented here, gave me 8fps for medium. That does not mean that this is the usual performance delta of the old 5850 with the HD630, but that a story for another time.

I also did a couple of 720p benchmarks, in order to better evaluate , cpu differences.

Assassin's Creed Origins 1280X720 Ultra GTX 1070 @2Ghz CORE i5-8600k @5Ghz - 113fps

Assassin's Creed Origins 1280X720 ultra GTX 1070 @2Ghz CORE i7-860 @4GHz - 76fps

We may be having a 49% performance difference between the 8600k and the 860, but that's not the worst I have seen. In Grand Theft Auto V, the delta was at 60% and that at 1080p maxed graphics bar msaa.

Grand Theft Auto V 1920X1080 V.High fxaa GTX 1070 @2Ghz CORE i5-8600k @5Ghz - 115fps

Grand Theft Auto V 1920X1080 V.High fxaa GTX 1070 @2Ghz CORE i7-860@4ghz - 72fps

(this guy 4ghz user should change his username :lol

In GTA V, the run was more cpu limited, but the cpu load was lower. The reason is the much better cpu utilization of the physical and logical cores of the i7-860, in Assassin's Creed Origins.

Just look at the differences in the two cpu loads of the 860, during my runs of the two games at 1080p.

Assassin's Creed Origins



Grand Theft Auto V


I can't remember when was the last time I saw such huge gaming cpu load on the 860. Probably never. Anything above 50% is impressive to be frank. Most of the time, hyperthreading does very little. Even so it lost badly, due to the 6 threads of the 8600k being much stronger.

Of course I have seen worse, in AotS for example but also have see better, in CoD WWII for example, where the difference was minimal. But I digress.

Also here are some relevant 8600k vs 860 graphs, from the 720p run, to better highlight the differences.



I am very curious to see how my 2500k@4.8Ghz will fare here, which will be determined in a couple of weeks, since I have to finish some 860 testing on the 1070, before decommissioning it.

I am currently actually playing the game on the stock 8600k and my 970@1.5Ghz, at Ultra minus one tier lower for the shadows and game plays absolutely fine. By fine I mean borderline 60fps.

Quite interesting game I might add. Someone could easily call it, Assassin's Creed Origin Witcher 3 edition with Far Cry Primal DLC, but they would just be mean! I am very OK if games borrow good elements of other games.

 

[Carfax83]

@ psolord, great testing! Though I would advise that the beginning section of the game has a moderate load on the CPU. When you get to the big cities like Alexandria, the CPU load is even greater. In fact, I would expect that your Core i7 860 will probably choke and start to drop frames.

 

[slashy16]

Its too bad they didn't include a 7700K or 8700k in that comparison to show how a CPU with half the core count but much faster will obliterate slower 16 thread CPU's.
I know my 8700K averages 120+ fps at those settings

 

[Carfax83]

Its too bad they didn't include a 7700K or 8700k in that comparison to show how a CPU with half the core count but much faster will obliterate slower 16 thread CPU's.
I know my 8700K averages 120+ fps at those settings

What comparison are you talking about exactly? Care to do a comparison with me? I have a 6900K at 4.2ghz. It would be interesting to see how your higher clocked 8700K compares against my lower clocked 6900K with quad channel RAM. Lets run the internal benchmark at 1080p high quality global setting to make it more CPU bound, with MSI Afterburner enabled and take a screenshot of the final result.

 

[slashy16]

What comparison are you talking about exactly? Care to do a comparison with me? I have a 6900K at 4.2ghz. It would be interesting to see how your higher clocked 8700K compares against my lower clocked 6900K with quad channel RAM. Lets run the internal benchmark at 1080p high quality global setting to make it more CPU bound, with MSI Afterburner enabled and take a screenshot of the final result.

 

[StinkyPinky]

AC:O is a very interesting game both from a technical and gameplay perspective. Very optimised and stable, credit to unisoft where it is due. Would love to see a DX12 patch in the future but since the game is so smooth/good looking already I don't think they will bother.

 

[psolord]

Here's a friend's 4590 with a GTX 1070 at 720p Ultra.



And here are my screencaps from the above vids, of the i7-860@4Ghz and i5-8600k at 720p ultra.



Notice how my i7-860 gets a very high rating at 64fps although it's only a little faster from the 55fps of the 4590, which got a low rating in return.

My friend (internet friend really), said that his run was stuttering like there was no tomorrow, while my 860 was running like a champ.

That's why I never buy locked cpus man... Also I am avoiding hyperthreaded cpus after the i7-860, but I do have the feeling that the 860 will be faster than the 2500k OC. According to my testing this happens very rarely however, but it still remains to be seen.

If anyone has a 2500k OC and can test 720p Ultra with a 1070 until I do, please inform us.

edit: and of course any other 720P ultra test would be welcome for cpu comparisons. Especially Ryzen ones, to see how their extra threads are doing. Technical interest. Not e-pen. Also please make sure you use the fullscreen mode. Borderless had severe stuttering problems up to the version and driver I tested. One of the very few games I have tested, that exhibited this behavior. Very weird. Also please note that even at 720p, a 1070 with a 8600k does have a very high gpu load, which indicates that it could be a little gpu limited at times, but it would still make an interesting test.

 

[Carfax83]

@ psolord, have you tested your Core i7 860 in Alexandria yet? That will be the measure of your CPU. Also I don't know if you've noticed, but the game tends to aggressively stream a LOT of data from storage. I've seen it reach around 50MB/s or more in some cases, which is very high. This might be a factor in the high CPU usage, as it's the CPU that streaming and decompressing the data. This might be a memory saving feature of the engine, since the game is so large. Only the area around the protagonist is fully loaded in memory, and everything else is streamed from storage.

 

[psolord]

Hello my friend.

No unfortunately I did not have the time to progress in a satisfying way. I have many tests to finish on the 860, so I can dismantle it and give it to a friend.

I don't think the Alexandria level would impact the delta however. It may be heavier, but if it brings say a 20% performance impact on the 8600k, it will bring the same impact on the 860. I mean if the code becomes more complex in Alexandria, the i7-860 should cope with the same percentage performance loss, since it has the same L1 and L2 caches as the 8600k per core, but it has an advantage for its L3 cache per core, but a disadvantage per thread. I think HT is all about resources sharing and operates on data already available but I am not sure. I think the 8700k would have even more cache if the increased threads would cause a disadvantage compared to the 8600k (12MB/12threads vs 9MB/6threads).

I will try to do a final test at a later point however, before giving the system away. Alexandria should be what, 3-4 hours from where I am now.

You are right about the crazy streaming, but I have seen it subsides once you are on the same area for a while. Actually one of the reasons that I do so long benchmarks, is exactly because I want to push the systems to push as much data as possible, something that is not possible in 30-60sec runs.

In my runs of Origins, for 1080p ultra, the RAM usage was at around 7GBs and Vram usage at 3.5GBs. My system alone would have lots more RAM and Vram to spare if the game requested it, so if the devs took this approach, it's bad coding really. The game should allocate more resources if it found them.

I believe it has to do more with the protection though. It must be doing some on the fly encryption/decryption or whatever. I intend to test the cracked version once it comes out, mostly out of curiosity to see how the protection affects system resources. If it does that is.

 

[Carfax83]

I will try to do a final test at a later point however, before giving the system away. Alexandria should be what, 3-4 hours from where I am now.

Just to let you know, you can visit Alexandria or any other area on the map whenever you want. You don't have to wait for the storyline to progress.

In my runs of Origins, for 1080p ultra, the RAM usage was at around 7GBs and Vram usage at 3.5GBs. My system alone would have lots more RAM and Vram to spare if the game requested it, so if the devs took this approach, it's bad coding really. The game should allocate more resources if it found them.

Yeah I agree with you here. But I can see both sides of the issue. It really depends on how efficient the algorithm is. Take a game like the Witcher 3 for instance. It has aggressive streaming, just like AC Origins. And just like AC Origins, it has low comparatively low VRAM and RAM usage.

However, unlike AC Origins, the Witcher 3 has bad pop in and the LoD over distance is terrible; especially for inorganic objects and structures (uses the same LoD as the console versions). AC Origins is far superior to the Witcher 3 in that regard.

I know that the developer stated that the area surrounding the protagonist is fully loaded in memory, so I suppose it's only the outer areas that are being streamed from storage.

I believe it has to do more with the protection though. It must be doing some on the fly encryption/decryption or whatever. I intend to test the cracked version once it comes out, mostly out of curiosity to see how the protection affects system resources. If it does that is.

Honestly, I doubt it's being caused by any DRM. The game is comparatively moderate in open areas when it comes to CPU usage. It's in the congested areas like the cities where there is a very large spike in CPU activity.

This is to be expected though, due to more A.I and objects. Just like how Novigrad was basically the only CPU intensive area in the Witcher 3. And Alexandria etcetera has a lot more A.I, higher object count and more geometry than Novigrad by a long shot.

DX12 would make a huge dent in the CPU usage I have no doubt as it would dramatically lower the rendering overhead.

 

[psolord]

Hello. Sorry for the bump. I did some Assassin's Creed Origins benchmarks and I think the findings are semi interesting.

My benchmark consists of an 8 minute run, doing various (but the same on all runs) stuff. Except on the 5850 which the performance was giving an unplayable experience, so I stuck with the built in benchmark. The built in benchmark is run on all the other systems as well, for reference.

Latest 1.5 patch for all systems with the same latest drivers amongst same gpus.

Everything is documented in the following videos.

Assassin's Creed Origins 1920X1080 Ultra GTX 1070 @2Ghz CORE i5-8600k @5Ghz - 82fps

Assassin's Creed Origins 1920X1080 ultra GTX 1070 @2Ghz CORE i7-860 @4GHz - 66fps

Assassin's Creed Origins 1920X1080 Ultra GTX 970 @1.5Ghz CORE i7-860 @4GHz - 49fps

Assassin's Creed Origins 1920x1080 High 7950 @1.1Ghz CORE i7-860 @4GHz - 40fps

Assassin's Creed Origins 1920x1080 Medium 5850 @900Mhz CORE i7-860 @4GHz - 10 fps

For the 5850 the performance was roughly the same for Very Low, Low and Medium. It is quite probably hitting a vram limit so hard, that the performance stays fixed at that framerate for all these settings. The HD630 of the i5-8600k, which is not documented here, gave me 8fps for medium. That does not mean that this is the usual performance delta of the old 5850 with the HD630, but that a story for another time.

I also did a couple of 720p benchmarks, in order to better evaluate , cpu differences.

Assassin's Creed Origins 1280X720 Ultra GTX 1070 @2Ghz CORE i5-8600k @5Ghz - 113fps

Assassin's Creed Origins 1280X720 ultra GTX 1070 @2Ghz CORE i7-860 @4GHz - 76fps

We may be having a 49% performance difference between the 8600k and the 860, but that's not the worst I have seen. In Grand Theft Auto V, the delta was at 60% and that at 1080p maxed graphics bar msaa.

Grand Theft Auto V 1920X1080 V.High fxaa GTX 1070 @2Ghz CORE i5-8600k @5Ghz - 115fps

Grand Theft Auto V 1920X1080 V.High fxaa GTX 1070 @2Ghz CORE i7-860@4ghz - 72fps

(this guy 4ghz user should change his username :lol

In GTA V, the run was more cpu limited, but the cpu load was lower. The reason is the much better cpu utilization of the physical and logical cores of the i7-860, in Assassin's Creed Origins.

Just look at the differences in the two cpu loads of the 860, during my runs of the two games at 1080p.

Assassin's Creed Origins



Grand Theft Auto V


I can't remember when was the last time I saw such huge gaming cpu load on the 860. Probably never. Anything above 50% is impressive to be frank. Most of the time, hyperthreading does very little. Even so it lost badly, due to the 6 threads of the 8600k being much stronger.

Of course I have seen worse, in AotS for example but also have see better, in CoD WWII for example, where the difference was minimal. But I digress.

Also here are some relevant 8600k vs 860 graphs, from the 720p run, to better highlight the differences.



I am very curious to see how my 2500k@4.8Ghz will fare here, which will be determined in a couple of weeks, since I have to finish some 860 testing on the 1070, before decommissioning it.

I am currently actually playing the game on the stock 8600k and my 970@1.5Ghz, at Ultra minus one tier lower for the shadows and game plays absolutely fine. By fine I mean borderline 60fps.

Quite interesting game I might add. Someone could easily call it, Assassin's Creed Origin Witcher 3 edition with Far Cry Primal DLC, but they would just be mean! I am very OK if games borrow good elements of other games.

Hello. Sorry for the bump.

As promised, I did some quick testing on the cracked version of the game, compared to my original version of the game. I was particularly curious to see how things would change, especially since this game is presenting a crazy cpu usage.

Both runs were done with the 1.21 game version, same 390.77 driver and same clocks. 8600k stock and 1070 stock, at 1080p Ultra. RAM at 3200Mhz CL14 however. The system is patched for Spectre and Meltdown with both the windows patch and the asrock bios.

Here is the non cracked version


and here is the cracked version


Here is the cpu usage of the two runs.


and the two cpu usages overlayed with each other.


I don't see any severe difference in the two runs. For all intents and purposes they seem about the same. There is 1fps difference in favor of the cracked version, but that may be within margin of error.

However the 8600k is a very strong gaming cpu. Maybe the protections are not enough to even make a dent, especially at the settings I am using, so the gpu limit may came into play. However I did want to test real gaming settings. That's why I used stock clocks this time, to limit the power of the 8600k. Maybe I'll do some lower res test later. I do see some bigger spikes in the cpu usage of the original version, but they don't seem so awful. Certainly not 25% higher cpu usage as it has been rumored.

Unfortunately I still didn't have the time to test my 2500k, which should provide a cleaner picture, since it is only a quad core and with much lower IPC too.

ps Since my picture editing skills are very basic, if someone can change one of the two cpu usage graphs to a red color and then overlay the two runs, it would be helpful.

 

[TheELF]

Previous Denuvo cracks would create a key and just tell the game that the key is valid anytime it checked,if this one does the same then all the negative impact should still be there.
Completely disassembling the game code to remove all the denuvo calls seems pretty far fetched.