Let's benchmark CPUs!

[Redwolf]

I am curious how CMAK performs on different processors. Obviously, that mostly affects the combat resolution, so lets benchmark that if you are interested.

if this is your thing, please post how long your computer takes to compute this example turn.

I made a test scenario to have mostly defined conditions:

http://schlepper.hanse.de/tmp/bench3a.cmg

a copy is on

http://redwolf.dyndns.org/tmp/bench3a.cmg

Instructions to use:

- get a stopwatch

- load file (it's a saved game)

- [EDIT] Press Shift-S until sound is "off"

- press "GO" without giving any orders

- start stopwatch AFTER the "computer player thinking" box disappears, that means when the blue bar at the bottom starts

- stop the stopwatch when the blue bar disappears. Not when it is full, when it disappears

Please run it a number of times, about 5, then post:

- the times you got for the blue bar without computer player

- what kind of CPU you have, clockspeed

- overclocked, if yes what is original and what is overclock speed?

- how much RAM?

- (EDITED to add: how fast is your RAM, e.g. "333 MHz" or "800 MHz dual-channel")

- (EDITED to add: if you know, the processor core you have (Prescott, Thunderbird etc.) or if you don't know that, post your cache sizes and socket)

My results:

- AMD Athlon XP 2500+ (runs at 1800 MHz), Barton core

- not overclocked

- 512 MB RAM at 333 MHz

My times: 55, 58, 58, 55 and 59 seconds.

[ August 31, 2004, 07:35 PM: Message edited by: Redwolf ]

[Pzman]

Results after five tests

-Power Mac G4 1.2Ghz (Sonnet Encore/ST G4 upgrade card)

-not overclocked

-512MB RAM

TImes: 1m:35s, 1m:29s, 1m:35s, 1m:30s, 1m:31s.

[tabpub]

Pentium 1.6

1 GB Ram

3:04

2:55

I got tired after that...bed time...

[_UXcva]

-Power Mac G4 800DP

-Not overclocked

-1.6GB RAM

Times w/sound = off_1:58, 2:01, 2:07, 2:09, 2:11

Times w/sound = on_2:55, 2:58, 3:00, 3:02, 3:11

[ August 31, 2004, 11:11 AM: Message edited by: _UXcva ]

[Juju]

P-IV 2.4 (not overclocked) / 1GB Ram / W2K

with sound: 2:03 / 1:59

without sound 1:54

Then I got bored...

Those times are quite a bit longer than Redwolf's and Panzerman's, even though their CPU's are slower than the one I have, and they both have half the RAM I have. Now why is that?

[Redwolf]

512 MB or 1 GB doesn't matter for this test. Only if you are very short it should become an issue. I didn't measure how much this turn takes but it is very likely much less than 128 MB.

I wonder why you have a different time with and without sound. The combat resolution computation shouldn't be affected and the game is playing no sound. Very odd. I am not in front of my computer to see whether it's the same for me.

Are you sure you only stopwatched the blue bar progress, without "Computer player thinking"?

[Juju]

I figured RAM wouldn't have too much to do with it, but that still leaves the CPU speed difference.

I tried it once without sound because _UXcva's times (the post above mine) with and without sound differ tremendously. Almost a minute!

And yes, I'm sure I timed it just as you intended.

[Redwolf]

It is not surprising that the older Pentium-4s perform bad.

I had longer posts on this in the GF, but the Pentium-4 was very radically designed to make some code run very fast, and leave other code behind. On top of that, P4s depend heavily on fast RAM but early P-4 systems have slow RAM, some every 133 MHz RAM (and not RDRAM either). My employer's main product performed as fast on a 2.4 GHz Pentium-4 as on a 1 GHz Pentium-3 before I went optimizing it. By now, things are fine on 2.8GHz Xeons for us, but if a codebase was optimized for earlier CPUs and never P4'ed you can expect bad performance on older P4s.

There is a 322 page manual about how to code so that it runs fast on a P4.

http://www.intel.com/design/pentium4/papers/24943801.pdf

%%

I would be very curious to see how CMAK performs on an AMD-64 processor (in 32 bit mode obviously) and on Celerons.

[ August 31, 2004, 10:12 AM: Message edited by: Redwolf ]

[Ales Dvorak]

-I book G3 600 MHz

-RAM 640 MB

-not overclocked

1) 4.35

2) 4.31

3) 4.45

[Pzman]

Originally posted by Juju:

P-IV 2.4 (not overclocked) / 1GB Ram / W2K

with sound: 2:03 / 1:59

without sound 1:54

Then I got bored...

Those times are quite a bit longer than Redwolf's and Panzerman's, even though their CPU's are slower than the one I have, and they both have half the RAM I have. Now why is that?

Juju, it is generally know that a Mac's speed is faster than a compairable PC, and they may not messure the CPU speeds the same way, although I don't know why. At the peak, I believe CM was useing only 85-100MB of RAM.

[PseudoSimonds]

It took me 3:04 to benchpress my CPU. Of course, most of this time was spent opening the case and taking out the CPU. I hope this helps!

[Pzman]

He said benchmark, not press.

[AOK. 11]

AMD Athlon XP 3200+

1034MB Ram

Geforce 6800 GT 256MB

______________________

52 secs

53 secs

55 secs

53 secs

54 secs

[AOK. 11]

By the way,

not overclocked

[_UXcva]

Originally posted by Juju:

I tried it once without sound because _UXcva's times (the post above mine) with and without sound differ tremendously. Almost a minute!

Perhaps a Mac guru can pipe in with a better explanation but having the sound on and not having a dedicated sound card eats CPU resources with bigger turns like this one. Also the game doesn't take advantage of the 2nd processor which might handle some of the burden as it does in some OSX apps. I generally turn on sound only for replays on larger turns.

[Pzman]

That is one reason my 1.2Ghz's upgrade card does well compaired to some other Mac's since it only has a single processor. Since CM doesn't take advantage of the sceond CPU, its of little use. If I turned off the sound I might find my times go faster as well.

[Platehead]

Dual Intel Pentium III 800

512 MB RAM

Not overclocked

Times:

1) 2:50

2) 2:45

3) 2:47 - No sound

4) 2:54 - No sound

5) 2:49 - No sound

[ParaBellum]

AMD 2600+, not overclocked, 512 MB Ram.

Average time after 5 runs: 57 seconds.

[Mies]

Desktop machine:

- AMD Athlon 1200 Mhz (overclocked to 1300 Mhz)

- Windows XP Pro

- 640 MB RAM

Times:

1 => 1:26

2 => 1:24

3 => 1:30

4 => 1:28

5 => 1:25

Notebook:

- PIII 1 Ghz

- W2K

- 256 MB RAM (32MB allocated for video memory)

Times:

1 => 3:12

2 => 4:37

3 => 4:57

4 => 2:18 (after restarting CMAK)

5 => 3:39

6 => 2:39 (after restarting CMAK again)

[flamingknives]

Mac G4, 1.25GHz DP

Not overclocked

512MB RAM

64MB Radeon 9000

1:28 or near enough for each run.

Of all my gadgets, I don't have a stop watch, so I just pressed go at the minute and timed it from there. Computer thinking was up for about 2 seconds, so it's hardly a considerable error.

[Pyewacket]

XP / PIV 2.8 MHz / 500MB Ram

1:30 average benchmark

1:00 to loose the game

[flamingknives]

Athlons are pretty hot at this, aren't they.

[Redwolf]

Originally posted by flamingknives:

Athlons are pretty hot at this, aren't they.

The Pentium-III posted above is in the same ballpark, relatively speaking, it uses one 800MHz CPU. It is just the Pentium-4s with less than 2.8 GHz (which usually have slow RAM and old cores) that don't like the CM code.

And the Macs, surprisingly :confused:

[flamingknives]

Above 1GHz, the Macs are coming in roughly on a par with the P4s (1:30) Considering G5s won't run CM, they're not doing too bad.

[tar]

Juju, it is generally know that a Mac's speed is faster than a compairable PC, and they may not messure the CPU speeds the same way, although I don't know why

You have to know a little bit about what the "speed" numbers are measuring. In the case of processors, it's the clock speed. This is roughly comparable to RPM measurement of an automobile engine. It is only directly useful for comparing processors of the same type, since then the other parts of system that affect performance don't vary.

That is why most cross-computer benchmarks published by Mac and PC magazines usually rely on measuring the time to do something useful like compute Excel Spreadsheets or do Photoshop filters or something like that.

To push the automotive analogy a bit further: Can you figure out if a Corvette with its engine running at 2500 rpm is faster or slower than a VW Golf with its engine running at 3500 rpm? Of course not, since rpm is only one of the measures of engine "speed" and not generally the one you most care about.