How best to benchmark your PC

[Spuzzell]

Hello wise and handsome Hexus people!

I'm about to build a gaming PC from components for the first time and I'm looking for some advice on the best way to benchmark.

Is there a set way to go about it?

I would be interested in getting performance figures from my existing machine, then from the new machine at stock, then again from the new machine after OC.

The only component that will be shared between the two machines is the 128GB SSD I use as a OS drive. I'm keeping the same Windows install so don't want to format the drive.

Any advice on what best to do would be greatly appreciated, I don't know what software to use or how best to go about running benchmarks so everything is consistent between runs, or... anything, really.

Any help gratefully received!

Spuzzell

[shaithis]

Futuremark I find to be decent. They have PCMark for overall system benchmarks and 3DMark for gaming. When you upload the results you can compare them on the web and look at individual tests to see how each system compared.

[Spuzzell]

OK, thanks.

Is there anything I can do to ensure that whatever processes are running in the background is identical for each test?

[watercooled]

TBH it depends exactly what aspects of performance you wish to compare, as benchmarking isn't as straightforward as getting a number for two systems; e.g. it's possible (and common) for one system to be faster in one area, but slower in another.

Something like Futuremark is a more well-rounded benchmark, taking into account more than the speed of a single piece of code, and IIRC seems to track reasonably well with game performance. However if you want to get a more 'real world' comparison of performance for applications and games you run, then perhaps compare those applications and games too.

For background processes etc, you should generally be OK just closing anything which can be safely closed e.g. backup, instant messaging programs and perhaps antivirus as long as you're careful obviously.

[KeyboardDemon]

Sometimes, (and I'm not saying that this is the case here) scores generated by benchmarks have more to do with claiming bragging rights and getting that overall feeling of smugness when a system you've built does well against systems built by other people than they do in terms of giving you any really useful information.

Of course that's not the only reason to run benchmarks, if you know what the expected results are for a component then benchmarking that part will help you identify bottlenecks if it is not performing correctly or to see performance gains if it is performing above the curve. I use benchmarking a lot because in my case, when I have spent a lot of money replacing an expensive component, I want to be able to have figures that justify that expense, the bigger the performance gain the better I feel (see smugness above).

Synthetics like 3dMark are useful but they are synthetic tests and don't really have a big bearing in real situations, so if you have games with built in benchmarks that you frequently play it is worth testing with those games too, or simply run the fraps benchmark tool or a similar program to gauge the actual games performance and try and play the same level on each PC when you do, do a few tests on each rig and work out the average that way it will smooth out differences in recorded figures if you are getting scores from actual play rather than repeatable scripted play.

For startup times you can look into your system logs to get an accurate figure on how long it takes for your PC to boot and can be directly compared to other PCs using the same logs.

Here is a list of games with benchmark tools (taken from this link):

Alan Wake
Aliens vs. Predator
Arma 2
Batman: Arkham Asylum
Batman: Arkham City
BioShock Infinite
Company of Heroes
Crysis
DiRT 3
DiRT Showdown
Doom 3
Far Cry 2
Grand Theft Auto IV
GRID 2
Hard Reset
Hitman: Absolution
Just Cause 2
Lost Planet 2
Mafia II
Metro 2033
Metro: Last Light
Mirror's Edge
Natural Selection 2
Quake III Arena
Quake IV
Remember Me
Resident Evil 5
Resident Evil 6
Return to Castle Wolfenstein
Sleeping Dogs
Sniper Elite V2
Street Fighter IV
Thief
Tomb Raider (2013 version)
Total War: Shogun 2
Unreal Tournament 2004
Warhammer 40,000: Dawn of War 2
World in Conflict

[Spuzzell]

TBH it depends exactly what aspects of performance you wish to compare, as benchmarking isn't as straightforward as getting a number for two systems; e.g. it's possible (and common) for one system to be faster in one area, but slower in another.

Something like Futuremark is a more well-rounded benchmark, taking into account more than the speed of a single piece of code, and IIRC seems to track reasonably well with game performance. However if you want to get a more 'real world' comparison of performance for applications and games you run, then perhaps compare those applications and games too.

For background processes etc, you should generally be OK just closing anything which can be safely closed e.g. backup, instant messaging programs and perhaps antivirus as long as you're careful obviously.

I think really what I'm looking for is a unified test that measures everything through a series of hardware changes.

I'd like it to track startup time, time to open a browser, time in and out of hibernation, track and graph memory and processor usage throughout AND then test gaming performance.

I want data, dammit :-)

[Spuzzell]

Sometimes, (and I'm not saying that this is the case here) scores generated by benchmarks have more to do with claiming bragging rights and getting that overall feeling of smugness when a system you've built does well against systems built by other people than they do in terms of giving you any really useful information.

Of course that's not the only reason to run benchmarks, if you know what the expected results are for a component then benchmarking that part will help you identify bottlenecks if it is not performing correctly or to see performance gains if it is performing above the curve. I use benchmarking a lot because in my case, when I have spent a lot of money replacing an expensive component, I want to be able to have figures that justify that expense, the bigger the performance gain the better I feel (see smugness above).

Synthetics like 3dMark are useful but they are synthetic tests and don't really have a big bearing in real situations, so if you have games with built in benchmarks that you frequently play it is worth testing with those games too, or simply run the fraps benchmark tool or a similar program to gauge the actual games performance and try and play the same level on each PC when you do, do a few tests on each rig and work out the average that way it will smooth out differences in recorded figures if you are getting scores from actual play rather than repeatable scripted play.

For startup times you can look into your system logs to get an accurate figure on how long it takes for your PC to boot and can be directly compared to other PCs using the same logs.

Here is a list of games with benchmark tools (taken from this link):

Alan Wake
Aliens vs. Predator
Arma 2
Batman: Arkham Asylum
Batman: Arkham City
BioShock Infinite
Company of Heroes
Crysis
DiRT 3
DiRT Showdown
Doom 3
Far Cry 2
Grand Theft Auto IV
GRID 2
Hard Reset
Hitman: Absolution
Just Cause 2
Lost Planet 2
Mafia II
Metro 2033
Metro: Last Light
Mirror's Edge
Natural Selection 2
Quake III Arena
Quake IV
Remember Me
Resident Evil 5
Resident Evil 6
Return to Castle Wolfenstein
Sleeping Dogs
Sniper Elite V2
Street Fighter IV
Thief
Tomb Raider (2013 version)
Total War: Shogun 2
Unreal Tournament 2004
Warhammer 40,000: Dawn of War 2
World in Conflict

Holy crap, that's awesome, thanks.

This is 100% about justifying my expenditure though the medium of a meaningless series of lines going up. A little bit.

I've made my peace with that.

I'm really not bothered about comparing to other peoples systems, it's all for internal use only!

I'm also switching from Radeon to Geforce and doing a hard drive transplant from one motherboard chipset to another and I'm just really craving all possible data to see what changes and how :-)

Thanks for the tips, I'll probably go though the logs if thats the only datum points available and then graph them myself.

Oh God, I'm a geek.

I didn't know.

[KeyboardDemon]

I'd like it to track startup time, time to open a browser, time in and out of hibernation, track and graph memory and processor usage throughout AND then test gaming performance.

It's in your logs.

Startup times
Wake Source log

I've never looked at individual application startup times, but I'm sure it's in the event viewer logs somewhere.

[kalniel]

Sisoft Sandra plus one of these:


And the job's a goodun.

[Spuzzell]

Sisoft Sandra plus one of these:


And the job's a goodun.

Yeah, but.

How do I print that out to then fold in my wallet to then pull out and show people in the pub?

[KeyboardDemon]

Yeah, but.

How do I print that out to then fold in my wallet to then pull out and show people in the pub?

My advice is not to show people in the pub, unless you want to stop them from talking to you that is.

[Spuzzell]

My advice is not to show people in the pub, unless you want to stop them from talking to you that is.

You think it's possible that other people might not care? Even if all the lines go up?

Maybe if they went up MORE.

[KeyboardDemon]

You think it's possible that other people might not care? Even if all the lines go up?

Maybe if they went up MORE.

I think it's possible that the people you drink with might not be as passionate about your results as you are, sharing them on here will usually generate more interest, unless you happen to drink with a bit of an IT crowd, but your mileage may vary.

I remember explaining how a graphics card upgrade made my 3dMark scores rocket to a non-geek friend who asked meif that made it possible to type faster into Word.

Usually you can tell when a person has gone from actively engaged in a conversation to feigning interest while they scout the room for an unattached person to rescue them from the conversation, at least that is until alcohol has been introduced into the conversation. I guess what I'm trying to say is just don't let it get to you if the people you drink with are more interested in football scores then they in benchmark scores.

Also with me, if I am out with friends and my mind is on my PC it usually tells me that I am in the wrong place, just as if I am playing a game and my mind is on my friends then I would most likely shut down the PC and go out.