With DirectX 10, the chips of graphics cards (GPUs) started to be able to do other tasks than just displaying graphics. Their big amount of “cores” enabled GPUs to compute different kinds of tasks much faster than a CPU would be able to. Nvidia offered CUDA to do video-encoding, matrix-multiplication or Folding@Home. ATI offered basically the same calling it Stream.
Now more and more developers are using these possibilities to enhance the visual quality of games, e.g. for physics calculation with Nvidia’s PhysX framework or possibly in the future with the open-source alternative OpenPhysics. But GPU acceleration can also greatly enhance the quality of water in games. The game Just Cause 2 offers support for GPU Water Simulation and shows quite impressively the differences between “normal” and “GPU” water.
Video Comparison:
Screenshot Comparison:
Let’s start with some screenshots for comparison (click on the image for full resolution):
As you can see, the visual quality improves quite a lot. But how does this GPU acceleration affect the performance?
Ingame Settings: 1920×1080, advanced settings as follows:
Performance Assessment
In order to measure the performance loss for enabling GPU Water Simulation in Just Cause 2, I took three common situations and measured the framerates for an interval of time:
Falling: Airlift to Bandar Arang Batu Besar (24676/27047), straight fall into death Flying: Flight from Kem Udara Wau Pantas (22217/23304) to Pelantar Gas Panau Seladan (29886/24713) for 30 seconds, high above the water Base Jump: Jump from highest skyscraper in Panau City – Docks District (1116/13726) in northern direction until death
The following diagram shows the actual testing results:
The framerates dropped quite significantly on my system (details above), on average the GPU Water Simulation lowered the FPS by 40%. When flying high above the water the game was still running fast enough, but especially in the city with an average of 19 FPS and a minimum FPS of 16, the game became unplayable.
Conclusion
GPU Water Simulation really improves the visual quality of games, but also needs a lot of computing power and thus lowers the performance significantly. On my system the game Just Cause 2 was not playable with GPU Water Simulation enabled in scenes with much action or geometry. Thus it takes a really powerful computer in order to be able to enjoy the great quality of GPU-accelerated water.
There are so many browsers out there, with Internet Explorer, Firefox, Chrome, Safari and Opera just being the most popular ones. Internet Explorer has been the king of all browsers regarding the market share, but mainly because it was pre-installed on all Windows versions after Windows 95. But over the last months, Internet Explorer’s marketshare has decreased to only slightly above 60% while Firefox almost managed to break the magical 25%-barrier, and even Chrome can claim more than 6% although it was released just one and a half years ago.
Since more and more users seem to consider switching from Internet Explorer to other browsers, they all have the tough choice of finding the right browser. I created this browser roundup to compare the latest versions of all major browsers and show their strengths and weaknesses.
The Contestants
I used the latest available final versions of all these browsers. Beta or Alpha versions are not considered as they may be unstable and thus do not qualify for everyday use. Internet Explorer was running in 64-bit, all other browsers in 32-bit.
Performance – JavaScript and Rendering tested with Sunspider, Peacekeeper, Dromaeo, V8
Standards Conformity – Compliance to HTML, CSS and JavaScript standards tested with Acid3, CSS3 Selectors Test
Loading Speed – Time to load websites tested with Numion Stopwatch on three websites
Startup Speed – Time it takes the browser to launch
Memory Usage – The amount of memory used with 1, 11 and 25 tabs
Customization – The amount of addons and themes available
Security – Vulnerability to different hacks, block rate of malicious websites
Testing the security was a really hard job. The only reliable data was the block rate. All browsers passed the hacking tests and the amount of vulnerabilities over the last years does not allow to judge the security. Data for the mean time of patch release was not available.
The Results – Overview
I analyzed and created charts for every single test, but these detailed statistics won’t be interesting for most people, so here are the overall results. A German version of this summary can be found on PCMasters.de.
The results of all tests have been summed up and weighted in order to create these final results. Further details on the weighting can be found below. All results based on time (Loading Speed and Startup Speed) and the memory usage are inverted so that higher scores are always faster. Security is left out due to the reasons mentioned above. Please remember: All tests were executed on the system named above. Results on other systems may vary. Numbers (e.g. addons, themes, vulnerabilities) were current at the time of creation of this benchmark (14th of April 2010) and may increase in the future).
The Results – Analysis
Performance: This measurement of performance indicates how fast websites can be drawn, scrolled or zoomed after all the content have been loaded, and how fast complex applications such as Google Docs are running. Opera and Chrome are clearly leading regarding the JavaScript and rendering performance, closely followed by Safari. The popular Firefox browser is quite slow compared to the ones named before, and Internet Explorer is really much slower than the others. This aspect is especially important for slow computers like netbooks.
Standards Conformity: There are some standards like HTML and CSS that define how the website is supposed to look like. Web developers should write their HTML-Code according to the standards so the website can be displayed correctly. But when the browser used by a visitor does not comply to these standards, the website may not be displayed correctly or some buttons may not work at all. This really slows down web development because developers have to create different versions of their scripts and have to test all browsers which is quite costly. Low standards conformity in Internet Explorer 6 caused Google to dis-continue support for this browser in order to force users to migrate to better ones. All browsers except Internet Explorer 8 comply quite well to the web standards, just Firefox takes a little hit by scoring just 95% on Acid3. Users should really choose a browser that scores high in this section in order to be able to view and use every website perfectly.
Loading Speed: This comparison shows how long it takes browsers to load websites using the same internet connections. As you can see there isn’t much difference between all browsers, Internet Explorer is slightly faster than the rest, but only by a few percent. Opera and Chrome could perform better using their features Opera Turbo or Chrome DNS Prefetching, but these features were disabled during the test.
Startup Speed: Determines how long it takes the browser to start up and load the front page (google.com in this case). Internet Explorer 8 is the fastest one here, but regarding the absolute values this doesn’t count much: Internet Explorer needs 0.8 seconds while Firefox, the slowest starting browser takes 2.4 seconds. Assuming the average user would start his browser 10 times a day, using Internet Explorer over Firefox would only save 16 seconds.
Memory Usage: Loading a website and rendering its contents always needs some of your system memory (RAM). But different browsers use their memory more or less efficiently, as the chart above is showing. Firefox and Safari were most efficient, Internet Explorer least efficient. Considering the fact that Internet Explorer used 521MB with 25 tabs open and 291MB with 11 tabs open, most users with modern computers (having 2-4GB of memory) won’t notice this much. Older and slower computers or netbooks mostly have limited memory (512-1024MB) and thus are more dependent on memory-efficient browsers like Firefox.
Customization: This includes the amount of addons and themes available for each browser. Many users don’t like the default interface of their browser and want to change it, or just want to add more features such as ad-blocking or notificiations (Facebook, Mail, etc) to their browser. This can be provided by using addons. The more addons and themes are available, the higher is the probability of finding just the ones you need. Firefox is most customizable with more than 12000 addons and 400 themes. Chrome and Opera with 4000/1400 addons and 400/120 themes can be adjusted as well, but Internet Explorer and Safari don’t offer many options here.
There is no way of summing up all these points to come to the conclusion “this is the best browser, this is the worst one”, because every user considers the aspects differently. Thus I will try to give some recommendations who should use which browser. I won’t recommend Internet Explorer anywhere due to its bad implementation of web standards and the shortcoming regarding performance.
The Results – Recommendations
Netbook-User: Netbooks have quite slow processors and thus need a browser with high performance, given by Opera and Chrome. But most Netbooks also have very limited memory, thus the browser should be quite memory-efficient like Firefox and Safari. Firefox is performing quite bad in all the performance tests, so Opera, Chrome or Safari would be the best choice for Netbook users.
Poweruser: If you surf the web every day for a long time using many different tabs you really need performance and memory efficiency. Again, Opera, Chrome and Safari offer these features in combination with perfect implementation of web standards. Safari’s loading speed is slightly better than Opera’s and Chrome’s, but Opera and Chrome have their Turbo and DNS Prefetching technologies respectively to boost page loading. It’s a difficult choice.
Customizer: If you want your browser to look exactly the way you need it, every button to be at the right place, and every feature to be directly available inside the browser, Firefox would be the choice. This browser offers more than 12000 addons and 400 themes for users for whom performance doesn’t matter that much.
Average User: The average user surfs the web only occasionally to check mails, research something or watch some video on YouTube. He doesn’t need any fancy features or designs and just wants to enjoy his web experience. Here almost every browser would fit, but most of them have a lot of features nobody actually needs. Chrome offers good performance, a simple interface with just the buttons required to do normal surfing and comfortable features such as an integrated history search inside the URL box.
The Results – Detailed Statistics and Analysis
The following part will show detailed statistics of the different tests and explain some results. All raw-data is stored in an Excel file you can download here, containing the exact scores and timings, URLs of tests and websites used and comments regarding errors or settings.
Performance:
Opera is leading regarding the performance, directly followed by Chrome and Safari. Firefox isn’t performing that well, but Internet Explorer clearly looses this battle. Its performance is far below the other browser’s, and the Internet Explorer 8 wouldn’t even start the Dromaeo test due to its bad implementation of web standards. Sunspider results are inverted to show runs per minute. On the testing system Opera was fastest with 6174 points in Peacekeeper and 434.52 in Dromaeo. Safari leads with an average of 595.4ms in Sunspider, and Chrome with 4397 points in V8.
Standards Conformity:
In order to check whether a browser complies to web standards, the Acid3 and CSS3 Selectors Test were used. Opera, Chrome and Safari pass Acid3 with 100%, Firefox scores 92% and Internet Explorer only 20%. All browsers except Internet Explorer passed the CSS3 test with 100%, Internet Explorer just passed 349 of 578 tests.
Loading Speed:
In order to measure how fast a browser can load different websites, the numion’s StopWatch was used. Loading speed was tested on three different websites and recorded as the average of two tests. Loading times for ESPN range from 2.2 to 3.5 seconds, from 0.5 to 0.9 seconds for Google and from 2.1 to 4.5 for this website. As you can see, the results are not really homogeneous, so one can’t conclude that one browser is loading faster than the other one.
Startup Speed:
When looking at the startup speed (starting the browser and loading the homepage google.com), Internet Explorer is clearly the winner and Firefox clearly the looser. But when considering the time it actually takes the browser to launch, from 0.8 seconds for Internet Explorer and 2.4 seconds for Firefox, this doesn’t really affect the browsing experience and can be neglected.
Memory Usage:
In order to measure the real memory usage (which can be quite difficult for browsers like Chrome who run multiple processes) the total memory usage of the whole system was measured before browser launch, after 1 tab, 11 tabs and 25 tabs. The memory usage results are inverted to show lower usage as higher score. Running one tab with just google.com in it, all browsers except Safari required between 55MB and 61MB, while Safari took only 32MB. Running more tabs, Firefox clearly leaves the concurrents behind with just 165MB for 11 tabs (others between 219MB and and 291MB) and 271MB for 25 tabs (others between 372MB and 521MB). For calculating the total score the memory usage for 25 pages was weighted only 50% because most users never open that many tabs. You can find the list of pages opened in all browsers here.
Customization:
To compare the ways of customizing the different browsers, the amount of addons and themes was assessed. Firefox has more than 12000 addons and, followed by Chrome with only above 4000. Similarly, Firefox has most themes with more than 400, closely followed by Opera. Internet Explorer and Safari didn’t have many addons and no official way of customizing the interface or theme. There were some third-party applications to chance their looks but they can’t be considered as official themes.
Security:
Comparing the security of different web browsers was a really hard job. The company scanit released a test that tries to crash browsers using various security leaks, but every browser managed to resist all of them. Then the block rate of malicious websites was considered: Internet Explorer blocks 85% of all malicious links tested, followed by Firefox and Safari with 29%, Chrome with 17% and Opera with only 1%. Just blocking links doesn’t make the browser safe from hacker attacks, so this does not really show the security of a browser. Then the public security vulnerabilities (leaks that are known to the browser developers and hackers that were not closed by a patch before being known publicly) were counted (inverted in the diagram). There was no data available for Chrome, and since Safari only had 4 of these vulnerabilities it is clearly leading here. The last diagram shows the amount of vulnerabilities found for each browser in the NVD. But since some browser exist for a very long time (e.g. Firefox and Internet Explorer) and some just for a few years (Chrome), I can’t really compare the amount of vulnerabilities ever found. This is why there is no total security score in the summarized diagram.
I hope this browser roundup gives an overview over the features and possibilities offered by each browser. If there are any suggestions, please comment! Screenshots to prove all results are available, but too big to be uploaded here. If you want me to prove any of the results, just comment or write an email and I will send the screenshots.
If you want to try a new operating system, but don’t want to put your computer and data to the risk of being corrupted, virtual machines can come in handy. They are small programs simulating to be a completely independent system with CPU, GPU, memory, HDD, DVD drive and so on and can run another operating system parallel to the one you’re actually using.
There are three major virtual machines out there: VMware Player, Sun VirtualBox and Microsoft VirtualPC. I took a look at all of them and compared their strengths and weaknesses. If you want to read the full review, have a look at the posts below:
In this series, I’m going to focus on comparing different Virtual Machines running Windows and their performance regarding different tasks such as single-threaded CPU-only applications, multi-threaded programs or even games requiring hardware 3D acceleration.
This article will show you when and for what to use VMware Player.
VMware Player is quite similar to Sun Virtual Box. You can also configure almost everything, from CPU count over memory to graphics acceleration. VMware Player can’t assign more cores than the computer actually has to the VM, but features better hardware graphics acceleration and up to 256MB of video memory. This also enables Windows 7 to run Aero and the glass effects smoothly.
I installed Windows 7 64-bit in the VM and compared the performance to my native Windows 7 64-bit. Like in VirtualBox, hardware acceleration is supported, but not required. The installation is even easier than in VirtualBox: Tell VMware Player which OS you plan to install, and it automatically suggests hardware configuration and manages the installation process by clicking the right buttons. After installation, it will install the VMware tools that include 3D driver support and other programs that make using the VM more comfortable.
A nice function that both VirtualBox and VMware Player support is the option to assign devices either to the host OS or the virtual one. This enables the user to connect two mice and keyboards and assign them to host and virtual OS. Now the virtual OS can be moved to a second screen and both users can use the computer simultaneously.
Performance:
Settings:
CPU count: 2
RAM: 1500MB
GPU acceleration: yes, 256MB, DirectX 9
First of all, the Windows 7 System Rating shows that both CPU and GPU performance in the VM are quite good, only the memory is a bit too slow:
To test how strong the VM really is, I ran SuperPI and got these results: It took the VM 0.773 seconds longer to complete the calculation, which equals a performance loss of 3.3%. This is a quite good result, similar to Microsoft Virtual PC, but not as good as VirtualBox.
Since the graphics acceleration really seems to work here, I tried some games to prove it. Older games such asCall of Duty, Couter-Strike 1.6 and Civilization were running nicely, Warcraft 3 had some texturing problems, and quite new games like Call of Duty 5 – World at War didn’t really run (I got around 4 FPS).
All in all, VMware Player is quite good in all fields: CPU performance, GPU performance and ease of use (installation and VMware tools). Moreover, working with it felt smoother than in VirtualBox or Virtual PC, but that may be due to Aero. This means one can work with VMware Player and even play older games. Regarding the possibility to connect keyboard and mouse to the VM and moving it to a second screen basically enables you to create a second, virtual computer that allows another user to work, play or watch movies on the same hardware at the same time.
In this series, I’m going to focus on comparing different Virtual Machines running Windows and their performance regarding different tasks such as single-threaded CPU-only applications, multi-threaded programs or even games requiring hardware 3D acceleration.
This article will show you when and for what to use Sun VirtualBox.
For these tests, I used a quite uncommon configuration: I ran a virtual Windows 7 on my native Windows 7. Most users probably won’t do this, but there is a reason why I chose this OS configuration: Sun VirtualBox allows you to use more CPU cores than you actually have.
The OS installation with Virtual Box is quite easy, just insert your install DVD into the real drive, create a new VM with the hardware configuration you like and tell VirtualBox which OS you’re planning to install. Then allocate your DVD to the VM and start the installation. Worked greatly with Linux, Windows XP and Windows 7 and VirtualBox automatically installed its tools to enable automatic mouse grabbing and this kind of stuff.
In contrast to Microsoft Virtual PC, VirtualBox comes with GPU acceleration, but only allows you to use up to 128MB of video memory. Moreover, the drivers are still beta, so the performance is not as good as you would like it to be.
Performance:
For this test, I didn’t choose settings that any serious user would try, but decided to test the option of using 8 cores.
Settings:
CPU count: 8
RAM: 2048MB
GPU acceleration: yes, 128MB video memory
I searched for a benchmarking program that can use any number of threads and found wPrime. Since it can be scaled to 4 or 8 cores, it was capable of running on both my real quad-core and on the virtual octo-core and still using all the cores.
The virtual octo-core was 0.303 seconds slower, which equals 1.9% performance loss. This is quite surprising when you consider that four of the cores are just simulated and the real OS and processor have to split up the work somehow. These results are even better than the ones I got from Microsoft Virtual PC where the performance loss was about 3.9%.
Hardware graphics acceleration should be there, but don’t expect too much. The drivers are still beta, the video memory is just 128MB and the graphics performance of VMs has never been close to the CPU performance. Unfortunately, I couldn’t get the DirectX acceleration to work, neither on Windows 7 64-bit nor on Windows XP 32-bit. That’s a pity because the only VM supporting graphics acceleration is VMware Player then.
All in all, VirtualBox is a really nice tool due to its possiblities of configuration such as using 8 cores, its easy installation and the CPU performance. There is no such feature as direct integration into the host OS as with Virtual PC, but nobody would expect such a feature, and the graphics power isn’t that awesome either, but still better than other VMs supplying just a standard VGA adapter without hardware acceleration or DirectX support.
In this series, I’m going to focus on comparing different Virtual Machines running Windows and their performance regarding different tasks such as single-threaded CPU-only applications, multi-threaded programs or even games requiring hardware 3D acceleration.
This article will show you when and for what to use Microsoft Virtual PC and the XP mode that comes with Windows 7.
The first thing to mention is the fact that Virtual PC requires hardware-assisted virtualization which must be supported by both CPU and mainboard. Many older and mainstream mainboards do not include this function and thus render the user unable to use Virtual PC. Most other Virtual Machines can profit from virtualization, but do not require it.
A nice feature Virtual PC has is that it can be directly embedded into the host OS (Windows 7). Like in other VMs, you can launch XP Mode normally in its own window with its own desktop and stuff, but all the programs you installed in XP Mode will be directly available from the Windows 7 start menu. Just click the shortcut, wait a few seconds for the VM to boot, and you will see your application running in an Windows XP environment, but the window is on your normal Windows 7 desktop.
Performance:
Settings:
CPU count: 1 (not adjustable)
RAM: 512MB (not adjustable)
GPU acceleration: no
Basically, Virtual PC doesn’t leave you the option to change any settings for your Virtual Machine, but if you really want to do so, you can edit the configuration file manually. Just open it with notepad and change the settings: C:\Users\Lennart\AppData\Local\Microsoft\Windows Virtual PC\Virtual Machines\Windows XP Mode.vmc
Since there is just one CPU and no graphics acceleration, I decided to test the only thing I can, CPU performance. I ran SuperPI first on the VM and then on my native Windows 7 and here is what you get:
The virtual Windows XP took just 0,848 longer than the native Windows 7, meaning a performance loss of 3.9%. This are really good values that basically enable you to work with applications running in the virtual machine as if they were running natively and you shouldn’t feel any slowdown (except for multi-threaded work such as video encoding).
All in all, Microsoft Virtual PC enables you to run old programs that can’t run on Windows 7 for some reason without any major loss in performance. Since there is no GPU support, you can’t use it for gaming, not even some very old games I tried were running. But if you just want to work, XP Mode will be a good choice.
As Google Chrome has been the fastest browser for a while and the alpha version of Opera 10.50 caught up quite well with Chrome’s performance but didn’t beat it yet, I decided to check the benchmark results given in the blog post by Opera. I ran all the benchmarks with the same settings as Opera did and also tested all four browsers. The only difference is that I used Chrome 5 (alpha) instead of Chrome 4 which is justified since Opera is also using non-stable builds of their browser. I summed up all the benchmark results in a diagram quite similar to the one used by Opera in order to make comparison easier:
The diagram published by Opera looks like this:
As you can see, the latest version of Opera really beats Chrome in Peacekeeper and Dromaeo, but Chrome is far ahead in both Sunspider and V8. I ran all the benchmarks several times to verify my test results and they are perfectly valid, but still differ from the ones published by Opera somehow. Now one has to ask why there is such a big difference regarding Sunspider and whether Opera faked its scores or the latest version of Chrome really improved in this sector.
All in all, it’s quite clear that Opera is really trying to beat all other browsers when concerning JavaScript performance and manages to catch up with Chrome. It will be interesting to see whether the final version of Opera 10.50 will beat Chrome in all tests or not.
In contrast to Opera, I want my benchmark results to be rock-solid and not to be challenged, so I took screenshots of each result and also created an Excel file that contains all results and the calculations made to create the diagram. You can download the screenshots and the Excel file to check my results.
EDIT: I spoke to the Opera development team and after some tests, we came to the conclusion that the missing support for SSE4.1 and 4.2 is what makes my Phenom slower than the Core 2 Duo in the Sunspider benchmark.
The latest version of Google Chrome, build 5.0.317.2, brings a new feature to improve the browsing experience when visiting foreign websites. It includes an automatic page translation tool which detects the page language and, if it differs from your browser language, offers to translate it to your language using Google Translate.
When Chrome offers you to translate a site, a blue bar will show up in your Chrome window:
When you click the Translate button, it takes a few seconds, and your page should appear in another language. I tried this on several websites with several languages, but it never really worked. Always some parts, like links and headlines were translated, but most of the text stayed in its original language.
Other languages are working, though:
As you can see, this feature is still very experimental, but since I’m using the alpha (dev) version of Google Chrome, I can’t expect every new feature to work perfectly. Google will surely fix this issue until the release of Google Chrome 5 final, and it will make browsing easier, as you will be able to view a lot more websites in languages you don’t understand, without having to mess with pasting the text to your favorite translation tool or guessing what it may say.
A newly released pre-alpha version of Opera promises to radically improve the browser’s performance and speed of displaying and loading websites. The official developer website even claims that “[Opera 10.50] is fast, more than 7x faster in SunSpider than Opera 10.10“. Since Google Chrome has been the fastest browser on the market so far, followed by Apple’s Safari browser, we may have a new player among the best-performing browser.
I downloaded and tested Opera 10.50 and found out that it actually is much faster than before, but also very unstable. I ran Peacekeeper three times, the first time it went well, the second time it was much slower and the third time it didn’t start the benchmark and showed just a blank page. I had to re-install Opera in order to fix this problem. But please keep in mind that this is a pre-alpha release and not meant for everyday use.
In comparison to Chrome 5.0.307.1, Opera 10.50 lost by 12.8%, which is actually a big improvement compared to Chrome 2.0.170 and Opera 9.64, where Opera lost by 60.2%. Safari, once the fastest browser, scored 28.6% less than Chrome, thus giving up its second place to Opera.
Considering the fact that Opera is still pre-alpha and that the development team will improve it before releasing the final version, one can expect Opera’s Peacekeeper score to increase again, bringing it dangerously close to Chrome.
Yesterday evening, the Closed Beta of Battlefield Bad Company 2 started. I managed to secure a beta key and get the game installer and immediately started testing. First of all, a few screenshots and performance information, but more information and a video will follow! There is a torrent download for Battlefield Bad Company 2 that’s working quite well in contrast to the official EA servers or other mirrors that are really overloaded and keep crashing all the time.
The beta supports multiplayer and includes only the gamemode “Rush” with the map “Port Valdez”. Official sources say:
The Battlefield: Bad Company 2 PC beta will feature Port Valdez, a new vehicle focused map where up to 32 players will compete against each other as either the US or Russian armies in the game’s Rush multiplayer mode. Players will have access to five land and three air vehicles where the Russian side can use everything in their arsenal including Main Battle Tanks as well as the fast-moving Quad Bikes and Mobile Armored AA. The fight will also be packed with plenty of infantry combat fought alongside the waterline towards a great oil industrial landscape in the Alaskan mountains.
Being based on the new Frostbite Engine, the game has quite high system requirements:
Recommended Frostbite PC Specifications for BFBC2 & BF1943: Processor: Quadcore Main memory: 2GB Graphics card: GeForce GTX 260 Graphics memory: 512MB
Especially explosions are really nice in this game, thus I created a “flip-book” of pictures (click to enlarge):
