CPU Performance Scaling
It would actually be quite interesting to take a closer look at how processors affect performance in AC. Specifically, I would have liked to test AMD Phenom as well as Core 2 Duo; this article is already long for a "game review", however, so I decided to stick with overclocking to see how it affects performance. The ATI and NVIDIA results are in separate graphs to help keep things legible. Since we couldn't enable anti-aliasing above 1680x1050, all testing was done at maximum detail with anti-aliasing disabled. We are also only reporting results for AC version 1.02, since the only area where 1.0 is faster is in anti-aliasing on ATI hardware.
If you're looking to get performance above 60 FPS, it's obvious that the first step is going to be purchasing the fastest CPU you can find. AC definitely supports dual-core processors, and even quad-core may be beneficial in certain situations. However, quad-core CPU usage often stays below 50% of the total CPU potential. Thus, an overclocked dual-core processor appears to be the best choice for maximizing AC performance.
During initial testing, we were a bit surprised to find that SLI didn't seem to improve performance. As this is a "The Way It's Meant to Be Played" game, that would have been another serious blow to NVIDIA's credibility. At the time, we were testing with a stock-clocked Q6600 at 1680x1050 and various graphics settings in order to utilize anti-aliasing. It was only when we began overclocking that we discovered the real culprit was CPU performance. There's CPU overhead associated with CrossFire and SLI, so with slower CPUs at moderate resolutions SLI and CrossFire will actually reduce performance in AC.
With an appropriately fast CPU -- at least 3.0 GHz would be our recommendation -- and running at 2560x1600, SLI and CrossFire are able to show substantial performance benefits. SLI and CrossFire both improve performance by 56%, but we still appear to be at least somewhat CPU limited. Increasing the CPU clock speed to 3.42 GHz on the X38 system (the maximum stable result for this particular system) further improves CrossFire scaling to 62%.
We should also take a moment to point out that in testing CrossFire did not work with both graphics chips until the latest catalyst 8.5 drivers. (That necessitated retesting all of our results, unfortunately, which is one more reason you're only reading this review now.) We heard a rumor that renaming the executable to COJ_DX10.exe (i.e. using the Call of Juarez CrossFire profile) would enable multi-GPU support; what we actually discovered was that renaming the executable did improve performance, but it improved it with single graphics cards as well. Whatever optimizations were taking place did not depend on CrossFire, and those optimizations now exist in an AC specific profile.
Finally, we need to caution against using the above results to compare ATI and NVIDIA graphics performance. The Dell system is running with DDR2-667 memory at 5-5-5-15 timings while the X38 system has DDR2-800 at 4-4-4-12 timings. The combination of BIOS tuning and memory performance could account for much of the performance differential. ATI looks to have a slight edge, but it's definitely not enough to make them a clear winner -- unless you're running 1680x1050 or lower with 4xAA.
Facebook
Twitter
RSS
32 Comments
View All Comments
geogaddi - Tuesday, June 10, 2008 - link
...now, what did i do with that babelfish...
ssgoten00 - Monday, June 2, 2008 - link
AC was only an average game overall. Graphics presentation was it's strong suit but gameplay was lacking and the game seemed to drag on as the player progresses through the game. Undoubtedly the repetitiveness was the worst part of AC. Not simply the fact that is seemed like some tasks were a redue of previously accomplished tasks but the shear fact that tasks were repeated verbatim with same characters and voices, only changing dialogue to create variation. Some characters players will have to kill multiple times under the guise of actually killing different characters in the game. AC was also disappointing in the fact that it mislead gamers in presenting its self a a somewhat stealth game. Nothing could be further from the truth. In AC players will often be forced into full on combat with multiple opponents to progress in the storyline. In vary select situations players have the choice of using stealth as a viable option. Ironically the last 5 or 10 minutes at the very end of the game are the most compelling. After the credits roll players are left in the main room to explore and decrypt code and hidden messages. It's unfortunate AC's developers couldn't have spent more time on puzzles that actually pertained to gameplay. Out of a possible 10 I give Assassin's Creed a 6.0 barely coming in at par, bordering on subpar.Donkey2008 - Monday, June 2, 2008 - link
It does have a Thief feel but after playing it on Xbox, I found it to be more Thief meets BloodRayne meets Splinter Cell with some of the best graphics I have seen in a while. It is sorta repetitive,but the violence cut away any boredom I had. I enjoyed it a lot.But I guess they could have done what other Rockin high-profile companies do and make an even more repetitive game exactly like its previous versions, but with a much worse soundtrack. Throw in some terribly low-res character models and reuse the same, bottom-of-the-barrel, cartoon-looking cutscenes and they would have a perfect 10 as well.
poohbear - Monday, June 2, 2008 - link
this is the kind of game reviews i'd like to see, wherein hardware is tested w/ a game to show performance. I mean, you guys ARE a hardware site and there are'nt many sites that do game reviews w/ hardware testing shortly after. I dont think u should do game reviews without hardware testing cause there are a ton of game review sites, but your niche shines when you do these hardware and DX analysis. cheers and thanks for a very informative article.DesertCat - Monday, June 2, 2008 - link
The article talks about wanting to check the performance of Assassin's Creed on a Phenom processor (and its 4 cores). I can speak to that to some degree.I have a Phenom 9600 (2.3 MHz) on an AM2 board (Asus M2N-SLI Deluxe) with an EVGA 8800GT OC (650 MHz). I play at 1280x1024 so I play in a letterbox mode. This processor is enough to run the game at acceptable frame rates but I would tend to think that a fast dual core would do just as well (like was found with the Intel processors in the article).
With the performance hurting TLB patch enabled, I noticed one area where frame rates truly took a nosedive: when doing the "look" pan from the top of one of those towers. I didn't have an fps counter on at the time, but I'm guessing it was in the 12-16 fps range based on the chunkiness I experienced. I got similar slow framerates when diving from those spots into the hay (especially the really high towers).
With the TLB patch disabled on my Phenom, those two low fps spots were much better. I'm guessing that those areas were in the low 20's of fps. The rest of the game was smooth as silk and probably above 40 fps. I do not, however, see high utilization on any particular core when I've checked.
If I was to point to areas that really stress a system in AC, I would say that the tower pan shots are the most common. (*minor spoiler ahead*) King Richard's speech from horseback about 3/4 the way through the game is also very intensive.
aguilpa1 - Monday, June 2, 2008 - link
I know I've had it for a long time?emboss - Monday, June 2, 2008 - link
The word you *are* looking for is possibly letterboxing?PrinceGaz - Monday, June 2, 2008 - link
I was going to mention that as well. Anamorphic means the pixels making up the image are stretched either horizontally (as with wide-screen DVDs) or vertically when displayed. If the game were anamorphic, it would be like it running on a monitor at 1920x1080 but being rendered internally at some other resolution such as 1440x1080 and stretched to the displayed 1920x1080.The correct description is what you said originally, that it allows only a 16:9 aspect-ratio view, so if I ran it on my monitor (1600x1200 native) the game itself would only use the central 1600x900 of that.
JarredWalton - Tuesday, June 3, 2008 - link
Sorry - I saw the original comment and thought I corrected it. Missed the other two occurrences. I wasn't thinking and just used the word after reading the thread on widescreengamingforum.com about AC. (I was hoping someone had found a way around the locked letterbox view.)AnnihilatorX - Monday, June 2, 2008 - link
This review is great! I have never read a game review that includes all the analysis, benchmarks, gameplay video conveniently presented.Excellent work!