iXBT Labs - Computer Hardware in Detail






Multi-Core Processors in 3D Games

<< Previous page

     Next page >>

Lost Planet

It's another multiplatform game, which engine can use multiple processors. The game demonstrates good graphics, so the main load falls on a graphics card. But in case of many dynamic objects in a scene, the number of Direct3D draw calls grows together with the CPU load. This game has a built-in benchmark with two tests. Let's analyze results of both tests:

The first part of the benchmark (Snow) is limited almost solely by a graphics card -- we don't see any performance differences between configurations with different numbers of cores. But the second test, notable for many objects in a scene, it severely limited by CPU speed. One core is clearly not sufficient to squeeze at least 30 fps. In this case, two cores have a significant advantage. Curiously, it exceed two times. Perhaps, it's the effect of extra expenses for executing several threads on a single core.

Perhaps, values of average and maximum CPU usage in 1280x720 tests with AA and AF will help us sort it out:

Snow Core 1 Core 2 Core 3 Core 4
Average 9.6 44.6 12.2 22.4
Maximum 25.0 82.8 31.3 48.4
Cave Core 1 Core 2 Core 3 Core 4
Average 18.5 66.1 15.7 20.7
Maximum 65.6 90.6 43.8 68.8

So, Snow shows very low CPU usage again. Interestingly, the work is done by all cores, not by a single one. So this engine is parallelized well, and it simply stays idle in the first part of the CPU benchmark. The total usage would have been about 90%, so a single core is theoretically enough for this Snow test.

The situation in the Cave test is different. Although the CPU load for three cores out of four is similar, load on the main core is higher by 1.5. On the whole, the total CPU usage in the second test amounts to over 120% for a single core. That is, one core of the 2.4GHz Core 2 processor fails to cope with Cave. This test needs at least a dual-core processor.

Nevertheless, we haven't cleared up the issue with more than a twofold performance lag of a single-core CPU from the other configurations. We are inclined to assume that it's the effect of executing several threads on a single physical core, which leads to an extra performance drop because of switching threads.

Devil May Cry 4

It's the second game on the same engine as in the previous test. Multiplatform multithreaded engine from Capcom. However, this game generates a different load on the system. DMC4 offers a built-in benchmark with four tests. However, we've taken results of only two tests, because the others do not differ much, and it makes no sense to cram useless numbers to the already overcrowded article.

We'll analyze the second and the fourth tests, which have different load types. The second test (just like the first and the third, by the way) puts up high requirements to GPU in the first place. And the fourth test has less work for a graphics card and a lot of objects that generate a heavy CPU load.

We can see right away that Lost Planet and Devil May Cry 4 have the same engine, and their tests show practically identical results. So all conclusions about the previous game hold true -- the main bottleneck in the first three tests is a graphics card, in the fourth test it's a CPU. The latter contains a lot of objects, and its performance is limited by the speed of a processor. So one core is not enough for acceptable gameplay again.

Once again we see a performance gain from the second core that exceeds twofold. It confirms that there were no mistakes in the previous case. Perhaps, a single-core processor executes all threads together, and the system spends extra resources on switching between these threads, which results in such behavior. Let's take a look at the CPU usage in 1280x720:

Scene 2 Core 1 Core 2 Core 3 Core 4
Average 24.5 33.3 4.4 5.1
Maximum 48.4 64.1 68.8 73.4
Scene 4 Core 1 Core 2 Core 3 Core 4
Average 29.4 44.4 3.9 5.4
Maximum 67.2 79.7 65.6 59.4

Our conclusions agree with what we said about Lost Planet, only the difference in CPU usage between two scenes of the built-in benchmark is not that big as in the previous game. So we get less than 70% in the second test and only 83% in the fourth test of the total CPU usage.

If we judge only by the second value, a single-core should theoretically suffice. But in practice it does not happen. And the problem is in the multi-threaded nature of the multiplatform engine from Capcom. In fact, the game runs well on a dual-core processor.

Write a comment below. No registration needed!

<< Previous page

Next page >>

blog comments powered by Disqus

  Most Popular Reviews More    RSS  

AMD Phenom II X4 955, Phenom II X4 960T, Phenom II X6 1075T, and Intel Pentium G2120, Core i3-3220, Core i5-3330 Processors

Comparing old, cheap solutions from AMD with new, budget offerings from Intel.
February 1, 2013 · Processor Roundups

Inno3D GeForce GTX 670 iChill, Inno3D GeForce GTX 660 Ti Graphics Cards

A couple of mid-range adapters with original cooling systems.
January 30, 2013 · Video cards: NVIDIA GPUs

Creative Sound Blaster X-Fi Surround 5.1

An external X-Fi solution in tests.
September 9, 2008 · Sound Cards

AMD FX-8350 Processor

The first worthwhile Piledriver CPU.
September 11, 2012 · Processors: AMD

Consumed Power, Energy Consumption: Ivy Bridge vs. Sandy Bridge

Trying out the new method.
September 18, 2012 · Processors: Intel
  Latest Reviews More    RSS  

i3DSpeed, September 2013

Retested all graphics cards with the new drivers.
Oct 18, 2013 · 3Digests

i3DSpeed, August 2013

Added new benchmarks: BioShock Infinite and Metro: Last Light.
Sep 06, 2013 · 3Digests

i3DSpeed, July 2013

Added the test results of NVIDIA GeForce GTX 760 and AMD Radeon HD 7730.
Aug 05, 2013 · 3Digests

Gainward GeForce GTX 650 Ti BOOST 2GB Golden Sample Graphics Card

An excellent hybrid of GeForce GTX 650 Ti and GeForce GTX 660.
Jun 24, 2013 · Video cards: NVIDIA GPUs

i3DSpeed, May 2013

Added the test results of NVIDIA GeForce GTX 770/780.
Jun 03, 2013 · 3Digests
  Latest News More    RSS  

Platform  ·  Video  ·  Multimedia  ·  Mobile  ·  Other  ||  About us & Privacy policy  ·  Twitter  ·  Facebook

Copyright © Byrds Research & Publishing, Ltd., 1997–2011. All rights reserved.