Matrix Games Forums

NOTE: This is an unofficial exercise just because, I was curious.

* Introduction *

Want to take the BFTB performance test? Here is how to play.

I have created three save games (2 BFTB & 1 COTA).

In general, the series simulation rate performance is dominated by available CPU.

* Other Factors *

On screen messages are known to slow things down.

The number of units selected can slow things down.

Of course, AI replans and heavy contacts can slow things down.

* Methodology *

Three test saves. One BFTB save reflecting the max number of units. One BFTB save reflecting the average number of units. One COTA save reflecting the BFTB average number of units.

The saves were created after all reinforcements would have arrived.

The side which was on defense was choosen for the player. Thus, guaranteeing without any orders issued that the forces would collide.

* Other Considerations *

Actual play with a multitude of orders given would probably contribute to CPU overhead.

It is not simply the number of units that solely impact performance, but more probably also influenced by the number of units in contact.

Although I was shooting for a certain number of units, beside reinforcements arriving units were also expiring from unnatural causes.

* Methodology to Generate Your Score *

Run BFTB/COTA at 1024x768 (the LCD res).

Load the test save.

Turn off on screen messages using the bottom toolbar.

Do NOT have any map units selected.

Make sure your machine is quiescent. If you have a multi-core, make sure, at least, two are allocated to BFTB/COTA.

Get a watch. Run the game at >>> for exactly one minute and see how many simulated minutes you get.

* Download *

http://home.comcast.net/~markshot/tempi ... rfTest.zip

(my results to follow)

NEW STUFF

I have added a new test save to the performance benchmark zip. An HTTR save reflecting the BFTB average number of units. See next post for those results as well.

MORE NEW STUFF

I've add a new file to the benchmark, BFTB120.

This represents the coming demo. 120 units is relatively on the low end for BFTB.

So, if you run the demo to 3D 10HR, you will have all the units on map. Then, if you conduct the benchmark as described here, you can apply the following ratios to your demo score to infer your probable BFTB performance for the retail game.

BFTB250 (average scenario size) = [your BFTB120] * 60%

BFTB600 (largest scenario size) = [your BFTB120] * 17%

It would be good if some other folks could also generate their score for BFTB120 and update their posts here. Thanks.

Here are my results. This represents number of minutes simulated in one actual wall clock minute.

BFTB600: 37 - the max scenario size you are going to see in BFTB.

BFTB250: 129 - the average scenario size for BFTB.

BFTB120: 214 - the demo scenario size for BFTB.

COTA250: 127 - intentionally chosen to match the above BFTB scenario size. For COTA, this is very close to the max scenario size.

HTTR250: 162 - intentionally chosen to match the above BFTB scenario size. For HTTR, this is also close to the average scenario size.

What you can do with this?

If you have a marginal machine (CPU), it is likely that even though the mega scenarios are beyond you that the average ones or below are still quite playable. As you can see that the performance drop off due to unit count increase is non-linear.

It's possible, but not enough real testing was done to be sure that COTA and BFTB performance are relatively comparable. For those, who already have been running COTA, this might give you some idea how BFTB might run. Of course, COTA has a greater number of smaller (less unit count) scenarios than BFTB.

One other thing to be aware when playing is that unit counts tend to climb during play due to reinforcements. So, simulation rate usually starts faster and trails off later in the scenarios.

Remember there will be a BFTB demo which should prove an excellent vehicle to test with. You can reference the test scenario's unit count and performance with what I have already posted already. (meaning the XLS of scenario stats)

My test system:

2007 Top End
Intel QX6700 Quad Core 2.66GHz OC'ed 3.20GHz
4Gb RAM
BFG NVidia 8800 GTX OC'ed
WinXP Pro SP3 32 bit

PS: Dave & Paul I would be interested in your comments. Thanks.

I just corrected my file names in the Zip.

If you want the real story, it was either this or test a new vacuum as my wife asked.

Now, I've run out of excuses and have to test vacuum!

BFTB600: 40 minutes

BFTB250: 133 minutes

Intel i7-860 (quad core)
4GB RAM
Radeon 4850/500MB
Win7 Home Premium

James,

That's very interesting. I would have expected more from your later CPU.

I am pretty sure that the simulation rates in the engine are free floating relative to machine cycles and not capped or scaled.

Thanks.

Well, keep in mind that I'm not a ninja at making sure everything is set up correctly!

When running multiplayer vs a Win XP box with an Athlon x64 3700+, the older Athlon box will move steadily along, i7 box will scream through 3 minutes in about 2 seconds, then stops and waits for the Athlon to catch up, then screams forward again -- it's weird to watch. [:)]

BTW, you didn't post your clock speed which be relevant within processor family comparisons.

Picky, picky... [;)]

2.8 GHz

BFTB250: 158
BFTB600: 42

I don't have COTA on this new machine yet or I would try that.
Your two year old machine is hanging on well Mark. I built mine back in February and it is top of the line.

Edit
I forgot my specs
Intel Core i7-920 Bloomfield 2.66GHz overclocked to 3.4 GHZ
12 GB of 1333 Mhz Ram
Nvidia 295 1.8 GB Gpu x2 in SLi (probably useless or this test [:)] )

Machine specs please.

I noticed Zoom level makes a difference in sim speed too. It is small but noticable. If I zoom in all the way the game simulates a little bit faster. Like 6%

Interesting. Well, I was using the standard zoom level at loading.

In past testing of this engine, video cards have turned out to be largely irrelevant along with display resolution. As such, this is a very good war game for laptops, since usually the video subsystem tends to be the most underpowered component.

Also, the memory foot print of this engine is pretty small compared to most games. Probably less than 200Mb in most cases.

BFTB250: 109
BFTB600: 30

AMD Athlon II X3 440 (3.0 GHz)
4 GB RAM
ATI Radeon HD 4670 1 GB

BFTB250: 120
BFTB600: 33
COTA250: 115

AMD Phenom II X4 3,20 GHz
4 GB RAM
NVIDIA 8800 GT
WIN 7 64 Bit

And for reference:

Ubuntu 10.04 (via Wine)
BFTB250: 107
BFTB600: 30
(Interesting on Linux: Processor speed set to "Performance": BFTB250: 107, "On Demand": BFTB250: 60 only)

Ok this test reveals the ugly truth of my old laptop.

BftB600: 8
BrtB250: 25
COTA250: 26

2Ghz Dualcore 2Gigs Ram

Guess that means that the big scenarios are almost unplayable as they would take days to finish, even at the highest speed. Talk about value for money. But the smaller ones are just fine. I'd like to add that the game might be running slow, but its absolutely running fine. No problems at all. I guess with this machine it would be impossible to play any graphics heavy game in 2010 at a good rate at all.

Folks,

Please remember to list your processor model and clock speed as that is the most important yard stick for comparison of these results.

Boogada,

Well, as you can see the CPU overhead increase is not a linear curve. So, a good number of lesser BFTB scenarios look well within your grasp. In fact, through much of HTTR's and COTA's development, I was also in that situation.

You probably find the UI smooth despite the slow simulation speed due to the way the engine uses two separate threads for each. That's why an AI replan can halt the simulation clock for a minute or two and yet the map interface is not jerky at all.

I also have an old 2004 laptop which is getting slower with each year.

Bump

I updated the first two posts with comparative information about HTTR as well.

I can only offer my machine's performance at HTTR, since I haven't bought BFTB yet (still waiting for the stipend, and frankly, with the issue of COTA possibly not being rebuilt in the engine, contemplating that game before BFTB, as I am genuinely interested in the theater).

So here goes:

HTTR250: 148

Machine:
Core 2 Duo Mobile 2.53 GHz
4GB DDR-II
Ati Mobility Radeon 3650 256 MB
Windows 7 Professional 64 bit

BFTB250: 147
BFTB600: 40
COTA250: 116

Intel Core2Duo 3.33Ghz + 3GB RAM + Windows 7 Home Premium

Edit: Fat-fingered BFTB600; corrected.

Nevermind ... PoE corrected his posting error.