Minecraft CPU benchmarking (Weirdest Bench Test Ever)

[Vindola]

What is the most CPU intensive game? 

 

Most of you would probably say Battlefield 3/4 or maybe an open world game like a heavy modded Skyrim or GTA IV... but are they really as tough as it gets?

 

The answer however, is no. While these games do indeed have high CPU usage, they in fact cannot use as much as Minecraft can. Now this can be surprising to you, "Minecraft? Are you kidding me?! The blocky looking game?!" And to that my friends I tell you one thing, you have obviously never blown up thousands of TNT blocks. If you have, only then you would witness the true usage demands of the Java based game.

 

 

In the tests that I conducted, I used a PC with a 3770K clocked at 4.21Ghz and a 7870 core clocked at 1.1Ghz.

 

 

Now, first things first. Battlefield 4:

This screen shot was taken 10 seconds after a bunch of stuff were blowing up and I minimized the screen. Circled in blue is the usage period of interest.

The total CPU usage was about 45% as you can see; Battlefield 4 by itself was using 35.64% (as reported by Resource Monitor) I have done this with BF4 several times, and these usage results were one of the highest I got. As you can see the entire PC usage among all 8 threads was between 35% to 50%, except for the 7th thread which is practically not being used.

 

 

 

 

Now onto the results I got from Minecraft:

This screen cap was taken 4 seconds after I started the chain reaction of blowing up 19.8 thousand blocks of TNT!

As you can there is high usage among ALL threads, ranging between 80% up to 100%

 

 

 

 

And according to Resource Monitor, of the 88% usage, 83.15% was Minecraft by itself; making Minecraft (in this case senario) over 180% more CPU intensive than BF4!

 

 

 

 

 

Now onto the benching methodology:

 

In usual PC benchmarking tests that a lot of you are familiar with, most of the time the question to be answered is this: "How many frames per second can you get on *game being used to test* using a particular hardware setup?"

In this benchmark however, the question that must be answered is  "How many blocks of TNT can you blow up on Minecraft, using a particular hardware setup, and having the CPU able to recover the game? (i.e: without the game fully crashing?)

 

So how do you setup this benchmark?

 

1. Make sure you have Minecraft, duh! And the RAM dedicated to Minecraft must be set to it's default of 1GB, no more or less. A lot of extra RAM will make it easier for the CPU to handle the workload, you don't want to do that to your CPU! You want to show that goddamn CPU who's the boss!!
 

2. Install the Single Player Commands Minecraft mod. (what it is/download link/instructions: http://www.minecraftforum.net/topic/94310-162-single-player-commands-v49-official-download-spc-noclip/)

This mod is the most valuable tool I have ever had in Minecraft. It adds so many useful commands that you will wonder how did you live without it. The reason this mod is necessary for this test is to help us quickly lay down the thousands upon thousands of TNT blocks. Doing it one by one is a dreadful thing.

Note for users running Minecraft 1.6.4 or higher: The downside of the mod is that the author has not released a version for Minecraft 1.6.4 or higher... yet. So you will need to get the 1.6.2 release version from the profile editor within Minecraft, and then install the mod; this guy explains it well: http://www.youtube.com/watch?v=MxqYEgfbK_A

 

3. We now set up the world that we're gonna test in:

• Name it whatever you want
• Set it to "Creative"
• Click on "More World Options"
• Set "Generate Structures" to "OFF"
• Select "Superflat" as a world type, then click "Customize" and remove all the layers
• Click "Done" and then "Create New World"
• Your world should look like the picture on the lower right.

 

 

4. Quit your world, and then go back to the Single Player section.

5. Select the world you created and then press "Recreate" (This is done because every time the test is conducted, you need a "clean" world. If you do a second or third test on the same world, you will run into some problems which will give inaccurate results. So we simply "Recreate" to make a copy of that world and run the tests on the copies, because remaking the world each and every time gets repetitive)

 

 

 

6.After you press "Recreate" press "Create New World" and you should be in.

 

 

7. Make your video settings identical to these:

 

 

8. Now comes the part where we make use of our mod.

• Dig one block into the ground
• Point the cross to the bottom of the hole
• Type //hpos1 into your keyboard and press Enter. On the screen, there should appear coordinates of the point you were looking at.

 

9. Now double tap the space bar to fly, and then fly off diagonally from that point for about 13 seconds or so.

 

 

10. Afterwhich,

• look down on the ground
• type //hpos2 into your keyboard and press Enter. On the screen, there should appear coordinates of the point you were looking at.
• These are the coordinates of the second point. Next to them, in brackets, you should get another number. This is the number of blocks that exist between point 1 and point 2
• For the first test, we want about 10000 - 12000  blocks in between. So if the number in brackets is a lot less than this range, go forward and repeat step 10. If it's a lot more than this range, go backward and repeat step 10.

 

 

11. As you can see from the picture above, there are 12000 blocks between point 1 and point 2 (first time I measured it was 8500 so I moved forward and increased it)

 

12. Now we're gonna turn all of these blocks... INTO TNT!!!

13. On your keyboard, type //set tnt then Enter, and the mod will take care of it for you:

 

 

Look at dat TNT!

 

The right picture is an aerial view of the thousands of TNT blocks.

 

14. Go stand in approximately the centre of all your TNT.

 

15. Press E to open your inventory, go to "Tools", and get the "Flint and Steel"

 

16. LIGHT IT UP!

 

 

 

Now as the chain of explosions goes on, your FPS WILL DROP BELOW 1 FPS. The question is, after a minute or so when the explosions should be over, will your CPU be able to handle all that's going on and recover without crashing the game?

 

If you crash, Minecraft will:

a] Just becomes unresponsive

b] Shows the message "shutting down internal server"

or

c] Shows a "Minecraft has crashed" screen.

 

 

 

If this happens:

• Restart Minecraft
• Go to single player
• Delete the "Copy of world" (This is the one you were testing on)
• Begin from step number 5, but this time decrease the number of TNT blocks by 1000.
• Repeat once more if necessary.
• When you reach a run that your CPU is able to recover without crashing the game, the number of TNT blocks you blew up is your score.

 

On the other hand, if you PC survives blowing up 12000 blocks of TNT, then the scene should look like this:

 

 

In this case, it means your CPU can handle more than 12000 TNT blocks. In which case:

• Quit your world
• Go to single player
• Delete the "Copy of world" (This is the one you were testing on)
• Begin from step number 5, but this time increase the number of TNT blocks by 1500 to 2500 blocks
• Repeat once more if necessary.
• When you reach a run that your CPU is NOT able to recover without crashing the game, then the number of TNT blocks you blew up in your last successful run, is your score.

 

Note: while the screenshots you see are set in windowed mode, do the tests in fullscreen. The only reason these shots are windowed, is because it's easier this way to take screenshot of them.

 

 

TA DAAAAAA!

 

 

Questions that might be asked about this benchmark:

 

 

Q: Why the hell did you do this whole unconventional benchmarking test?

A: Why the hell not?! I had about 2 hours to kill so I did this. And while I'm not expecting a lot of people to do this, if any at all, it was fun doing it either way. I will be doing more benchmarks using this test on other PC's and I will be reporting the numbers.

 

 

Q: Why Minecraft to test CPU? Why not other games?

A: Minecraft gives a unique opportunity that no other game can offer. Besides the fact that, as demonstrated, it can easily push CPU's to their limits, you can easily configure the world as wish. Which makes it easier to eliminate variables that may tamper with testing, and so I found it ideal for this test.

 

 

Q: But are there any other games to do something similar to this?

 

A: I suppose you can emulate this ridiculous test by another equally ridiculous test. There is a mod for Skyrim that I know of that makes chicken explosive. So you can spawn a few hundred chickens, and kill one of them and they will all blow up. Then go figure out how many chickens you can blow up before crashing your game. However, such test would be both intensive on the GPU and CPU, as Skyrim has amazing graphics. In our Minecraft test, I wanted something that focuses all of the workload on the CPU to test it properly.

 

 

Q: Speaking of which, why is this only intensive on the CPU and not the GPU?

 

A: Well to understand that, I think the best thing to do is compare this test to 3D rendering. In 3D rendering, there are 2 aspects of it that stress your CPU and GPU: Transforming the model data, and displaying the model data.

Transforming the model data is all about the action of the parts moving in space, how they move and interact with each other. This workload is done by the CPU, and since there are literally tens of thousands of independent points that are moving all over the place and bumping into each other, the CPU usage easily hits the roof. Now there are of course a lot of games and modeling software that utilize not just the CPU to do this type of work, but also makes the GPU (which also displays the model data) do CPU work as well. Minecraft however, by itself, throws practically all of it's work on the CPU. Also note that the number of cores is perhaps the most important thing for this type of CPU work.

 

And as I mentioned, displaying the model data is done by the GPU, using OpenGL for Minecraft. And since the data points are goddamn easy to render simple blocks, the GPU doesn't have to do a lot of work on this... in fact, during my tests, since I didn't modify the GPU settings to do some of the CPU workload, the core clocks remained around idle. The only way to make the GPU make extra work, without assigning it CPU load, is to turn off VSync, then typically (when there are no explosions going on) hundreds of useless frames will be rendered, which would be a waste since it would be extremely higher than the average refresh rates of most monitors.

 

 

Q: Lastly, what's your own score on this test?

 

A: The 3770K gave me an explosion of 21.5 thousand TNT blocks without crashing Minecraft, so 21500

 

 

 

 

Feel free to ask any other questions about this weird test. Do share your results if you do it.

 

The End.

[Jeroen1322]

Sorry to long did not read if i have time i will!  

[HeathCliff]

http://www.youtube.com/watch?v=5GgflscOmW8

 

Although I would like to see how many a w/cooled 290X 4960X will be able to handle...

[Hans Christian | Teri]

http://www.youtube.com/watch?v=5GgflscOmW8

 

Although I would like to see how many a w/cooled 290X will be able to handle...

 

Minecraft isn't very GPU intensive so that doesn't exactly matter...

[HeathCliff]

Minecraft isn't very GPU intensive so that doesn't exactly matter...

 

My bad, I'm getting things mixed up pretty badly today.

[Vindola]

http://www.youtube.com/watch?v=5GgflscOmW8

 

Although I would like to see how many a w/cooled 290X will be able to handle...

 

Hardly relevant... considering the test focuses mostly on CPU's

 

Sorry to long did not read if i have time i will!

 

That's fine The TNT and explosive chickens can wait.

[NewAndreas]

You should also allocate more ram to minecraft for extra fun

[megadarkwood]

Nice thread!

[juretrn]

@Vindola Why Minecraft? Try Supreme Commander!

[Vindola]

You should also allocate more ram to minecraft for extra fun

I did in some of my runs, and while it affected some aspects of the game, it did not have a significant change on the test I conducted specifically. So I just ran ran the RAM as is instead.

[Vindola]

@Vindola Why Minecraft? Try Supreme Commander!

I haven't played SupCom before (though I have a general idea of it and seen it's gameplay) but I think Minecraft gives more accurate results when comparing two systems (I go over why in the end of the main post) And I think Minecraft is a bit more popular than Supreme Commander :\  (no offense to any SupCom fans)

[FloRolf]

Lol. Im gonna try this some time this week ^^

[thedigitalmilk]

You Sir are a pioneer in modern benchmarking techniques:).

[juretrn]

I haven't played SupCom before (though I have a general idea of it and seen it's gameplay) but I think Minecraft gives more accurate results when comparing two systems (I go over why in the end of the main post) And I think Minecraft is a bit more popular than Supreme Commander :\ (no offense to any SupCom fans)

Well, supcom is the most cpu intensive game, at least in the single thread department.

[TheMail]

Very interesting results, thanks for doing this man!

[Vindola]

You Sir are a pioneer in modern benchmarking techniques:).

 

Thank you, I strive to innovate.

 

I might also do a guide for the Skyrim explosive chickens that I mentioned, for a CPU/GPU bench test. Stay tuned.

[juretrn]

I might give more info regarding Supreme Commander. 

In supcom, unit counts can reach 1000//player in large multiplayer games. In such cases, players often have many fighter aircraft lying around, at least several 100's. Because air supremacy is important, huge air battles occur. Because of those, the simulation speed of the game reduces drastically. (simspeed ranges from +10 to -10, where 0 means 1 game second equals 1 realtime second.) The supcom community has created a benchmark with such air battles that really stresses the CPU. It represents a good benchmark for all CPUs, as simulation WILL start to slow down no matter what kind of CPU you have. Results are given in time elapsed in a standard 2 minute replay. 

I give a cookie to the one who can find my result .

 

link to forum thread:

http://www.faforever.com/forums/viewtopic.php?f=2&t=3046

[Lays]

I did this with 5 people on my MC server, we put tnt down for legitly like 20-30 minutes all 5 of us, Im the only one who didn't dc/lag I got 60fps the whole time watching it lol, also helped that I dedicated 8gb of ram to the mc client and 20gb to the server's client lol

[Vindola]

I did this with 5 people on my MC server, we put tnt down for legitly like 20-30 minutes all 5 of us, Im the only one who didn't dc/lag I got 60fps the whole time watching it lol, also helped that I dedicated 8gb of ram to the mc client and 20gb to the server's client lol

Yeah I have actually done some tinkering around since I made the post.

 

When I first increased the RAM, I only dedicated 2 more GB extra to MC (so 3GB in total) so I didn't get any significant improvements. Now I have dedicated 10GB to it, and as a result the CPU can handle even more workload, so now I have actually manged to survive blowing up 34000 TNT blocks.

 

However, since this bench test is more about the CPU than RAM, I decided to set the dedicated RAM at it's default of 1GB and give the CPU the full blow and see what it can handle. And for that I got 21K to 22K TNT blocks blown up without crashing.

[Electronics Wizardy]

minicraft is no designed to take advantage of fast systems very well

[martinrox1568]

I did this with 5 people on my MC server, we put tnt down for legitly like 20-30 minutes all 5 of us, Im the only one who didn't dc/lag I got 60fps the whole time watching it lol, also helped that I dedicated 8gb of ram to the mc client and 20gb to the server's client lol

Sorry but I have to call bull. Don't care of you used the town super computer but 20-30 minutes of TNT with 60fps is impossible. I have a 4770k and I couldn't go above 15fps

[Vindola]

minicraft is no designed to take advantage of fast systems very well

 

Not really. It doesn't matter what game you're playing, if you have thousands of things that are blowing up and interacting with each other at the same time, your system will slow down.

[Vindola]

Sorry but I have to call bull. Don't care of you used the town super computer but 20-30 minutes of TNT with 60fps is impossible. I have a 4770k and I couldn't go above 15fps

 

 

It is somewhat possible. It depends on how quick one person lays down the TNT blocks. If on average they lay down 30 blocks a minute each, then that's under 4000 blocks. And if the rendering distance is a bit decreased, then after you light it, your FPS should drop to 10 - 20 FPS for less than a second, but for the most part of the chain reaction, the FPS can very well be at the VSync limit of 60.

 

However, if on average they laid down 150 blocks a minute or more each, then yeah bull.

[Lays]

Sorry but I have to call bull. Don't care of you used the town super computer but 20-30 minutes of TNT with 60fps is impossible. I have a 4770k and I couldn't go above 15fps

 

okie dokie

[Soarin']

Man, you haven't tried ArmA 2 with a 1500+ AI war in singleplayer.