Benchmarking - When is The Most Suitable Time?

[jiyeon]

Lately I've been getting into the overclocking side of things and getting a nice score of 1141 with my 4.7GHz 8600K, provided with Asus multicore enhancement enabled.

 

I was wondering, when is the best time to be doing the benchmarks? I personally prefer to conduct runs after a few minutes from a cold PC so there isn't any background processes in the way, but I feel almost criminal going from a 25C CPU straight to a 70C one, particularly on my 4.7GHz run. Also because of my car mind that you shouldn't run mechanics cold straight to hot.

 

What are your guys' preparations for benchmarking? Do you just do them straight from a running system, or let the PC cool down and benchmark right after a cold boot?

[EPPHO]

If CPU isn't thermal throttling (it has good cooling) it doesn't matter when you benchmark it. It will instantly jump to almost max temps when you press the benchmark button.

[AskTJ]

Start the PC, and boot into Windows. Leave it on for around 10 minutes and hit start. It would've had enough time to get its oil flowing.

[jiyeon]

2 minutes ago, AskTJ said:

Start the PC, and boot into Windows. Leave it on for around 10 minutes and hit start. It would've had enough time to get its oil flowing.

Right, this is what I do. Although I leave it for merely five minutes compared to ten.

[AskTJ]

On 17/06/2018 at 10:28 PM, MiNy said:

Right, this is what I do. Although I leave it for merely five minutes compared to ten.

When you switch on your PC, all the code, C++ and C# has to properly load in, as well as your startup programs which can be Fraps, Spotify and others. If I am benchmarking, I do a restart and let all the apps settle and make sure I am using the least amount of ram and 0% CPU usage.

 

 

[Froody129]

At 4am in winter with the windows open, your case panel off and a large fan blowing at it.

[jiyeon]

1 minute ago, Froody129 said:

At 4am in winter with the windows open, your case panel off and a large fan blowing at it.

I mean, it would definitely eliminate all possibilites of thermal throttling, to be fair!

[Christophe Corazza]

2 minutes ago, Froody129 said:

At 4am in winter with the windows open, your case panel off and a large fan blowing at it.

 

Yeah... but that doesn't work righ now, is it? 

[Majestic]

Catfish warning!

 

Anyway, don't compare cars to computers. Except for the fans where are no moving parts, but those are not stressed.

 

The only thing to keep in mind is that starting background processes can take up some CPU cycles. But leaving it too long at idle can cause the system to start scanning.

I've long since stopped caring about benchmark scores, because there is always someone using more exotic cooling methods. 

 

Also that stark increase in temperature probably means your autovoltage at multi-core enhancement is way too high, plus it's not delidded. A 45-50C instant increase in temperature tells me the thermal transfer is poor on your setup.

[done12many2]

1 hour ago, MiNy said:

What are your guys' preparations for benchmarking? Do you just do them straight from a running system, or let the PC cool down and benchmark right after a cold boot?

 

You'll need to determine what it is that you are trying to accomplish with your benchmarks.  Are you doing it to determine how your PC runs as is or are you trying to have some friendly competition with fellow enthusiasts?  

 

If you are just trying to get a baseline of your PC's performance, simply run it as it is.  Remain consistent in your testing so that in the future, if you need to re-verify something, you can do so consistently.   

 

If you are trying to be somewhat competitive with others, there are tons of hardware and software based tricks to increase your score.  

[jiyeon]

27 minutes ago, Majestic said:

The only thing to keep in mind is that starting background processes can take up some CPU cycles. But leaving it too long at idle can cause the system to start scanning.

Yeah, that's my concern is the amount of processes from a fresh boot.

 

28 minutes ago, Majestic said:

I've long since stopped caring about benchmark scores, because there is always someone using more exotic cooling methods.

It's a nice self-system evaluation if you ever need to overclock and stuff though, right?

 

28 minutes ago, Majestic said:

Also that stark increase in temperature probably means your autovoltage at multi-core enhancement is way too high, plus it's not delidded. A 45-50C instant increase in temperature tells me the thermal transfer is poor on your setup.

My voltage is set to 1.25V, and I now have MCE off, since I don't feel it benefits me enough. As for my temperature spike, I blame my Hyper 212, it's a budget solution after all.

 

29 minutes ago, Majestic said:

Catfish warning!

That's just rude. Not every girl on the internet is a false one. That's including me.

[jiyeon]

1 minute ago, done12many2 said:

You'll need to determine what it is that you are trying to accomplish with your benchmarks.  Are you doing it to determine how your PC runs as is or are you trying to have some friendly competition with fellow enthusiasts?  

 

If you are just trying to get a baseline of your PC's performance, simply run it as it is.  Remain consistent in your testing so that in the future, if you need to re-verify something, you can do so consistently.

I'm the type of benchmarker which is purely to see how my system runs, so I suppose running as is makes much more sense than what I do currently.

[done12many2]

1 minute ago, MiNy said:

I'm the type of benchmarker which is purely to see how my system runs, so I suppose running as is makes much more sense than what I do currently.

 

Then just run it as it.  If you want to see how it performs in its normal capacity, test it in its normal capacity. 

 

Have fun and good luck.

[jiyeon]

Just now, done12many2 said:

Then just run it as it.  If you want to see how it performs in its normal capacity, test it in its normal capacity. 

 

Have fun and good luck.

Thank you for the advice!

[Majestic]

5 minutes ago, MiNy said:

My voltage is set to 1.25V, and I now have MCE off, since I don't feel it benefits me enough. As for my temperature spike, I blame my Hyper 212, it's a budget solution after all.

Even at that budget there are better solutions. It's underpowered for a 8600K on OC.

 

7 minutes ago, MiNy said:

That's just rude. Not every girl on the internet is a false one. That's including me.

Rule 30

[Rubin Chen]

To add to what Majestic said about computers are not like cars, I like to point out why cars need a some to get everything warmed up. Cars has a lot of moving metal parts that are nearly air tight with one another. If you ever tried rubbing 2 pieces of metal against each other you will find that no matter how smooth they are they get really hot and the friction is enormous. Therefore it needs the oil to lubricate it. However back in the day oil consistencies are not what it is like today. Especially on colder days the oil is thick and not very fluid. Thus you needed to heat the oil to increase the fluidity before you go 60 on the freeway, since at those speeds the friction might be too much for the car to handle long term wise. Now computers are not the same. There's no oil, or metal rubbing against each other. Computers hit up because of electrical resistance. If you think about light bulbs you will understand this more. When you connect a light bulb to power you see it give light due to the temperature the filament is at. Same with a CPU and GPU. When you power on the PC, it is like turn on the light bulb. it actually gets hot immediately. In fact the die of the CPU or GPU is always hotter than the metal touching it. The only reason why it doesn't overheat is because there is a cooling system attached to it spreading the heat to a large surface area be cooled off by the fans. Also due to the fact that at higher temperature differences heat transfer is also increased, your CPU will loose heat at a faster rate the hotter it gets. 

 

In summary your CPU will never heat up too quick for your cooling solution to handle unless its thermal throttling. There can be too much power(energy/time) for the coolers to handle but not too quick to handle. 

 

Also side note if you want to see how you system runs. Depending on if you have liquid or air cooling if might be best to give it like 30 mins after startup to run benchmarks. This has nothing to do with damaging the system or anything it will just give you more "real" numbers. Liquid coolers has water inside and water has a high specific heat, meaning they can absorb a lot of heat before going up 1 degree Celsius. So you kind of want your system to warm up the liquid before running benchmarks to not fool yourself thinking you may have a better cooling solution than you actually do.

[jiyeon]

5 minutes ago, Majestic said:

Even at that budget there are better solutions. It's underpowered for a 8600K on OC.

I just did a Cinebench run at 4.4GHz 1.25V (with MCE off) and got max temps of 60C, it's not so bad, but I do agree somewhat.

 

7 minutes ago, Majestic said:

Rule 30

TIL

[Majestic]

2 minutes ago, MiNy said:

I just did a Cinebench run at 4.4GHz 1.25V (with MCE off) and got max temps of 60C, it's not so bad, but I do agree somewhat.

Yeah it'll suffice at gaming at reasonable voltages. Just don't push it on that cooler, nor run Prime95 past version 26.6.

[jiyeon]

6 minutes ago, Rubin Chen said:

To add to what Majestic said about computers are not like cars, I like to point out why cars need a some to get everything warmed up. Cars has a lot of moving metal parts that are nearly air tight with one another. If you ever tried rubbing 2 pieces of metal against each other you will find that no matter how smooth they are they get really hot and the friction is enormous. Therefore it needs the oil to lubricate it. However back in the day oil consistencies are not what it is like today. Especially on colder days the oil is thick and not very fluid. Thus you needed to heat the oil to increase the fluidity before you go 60 on the freeway, since at those speeds the friction might be too much for the car to handle long term wise. Now computers are not the same. There's no oil, or metal rubbing against each other. Computers hit up because of electrical resistance. If you think about light bulbs you will understand this more. When you connect a light bulb to power you see it give light due to the temperature the filament is at. Same with a CPU and GPU. When you power on the PC, it is like turn on the light bulb. it actually gets hot immediately. In fact the die of the CPU or GPU is always hotter than the metal touching it. The only reason why it doesn't overheat is because there is a cooling system attached to it spreading the heat to a large surface area be cooled off by the fans. Also due to the fact that at higher temperature differences heat transfer is also increased, your CPU will loose heat at a faster rate the hotter it gets.

That's very interesting, I had known about car oils but never really understood how it worked, and the added comparison to CPU/GPU systems was one heck of an education.

 

8 minutes ago, Rubin Chen said:

Also side note if you want to see how you system runs. Depending on if you have liquid or air cooling if might be best to give it like 30 mins after startup to run benchmarks. This has nothing to do with damaging the system or anything it will just give you more "real" numbers

I have air cooling since I don't trust AIO with leakage rates and whatnot, I usually run my benchmarks straight from a cold boot so I'll try when I've been on my PC for say an hour of work sometime and then I'll see what happens. I am still new to this whole benchmarking thing after all!

[jiyeon]

2 minutes ago, Majestic said:

Yeah it'll suffice at gaming at reasonable voltages. Just don't push it on that cooler, nor run Prime95 past version 26.6.

Yeah, I don't stress my CPU past 70% usage in my typical games anyway, but clearly my cooler is next to be upgraded.

[Rubin Chen]

1 minute ago, MiNy said:

I have air cooling since I don't trust AIO with leakage rates and whatnot, I usually run my benchmarks straight from a cold boot so I'll try when I've been on my PC for say an hour of work sometime and then I'll see what happens. I am still new to this whole benchmarking thing after all!

For air cooling the effect is far less. I would say 5 minutes is more than enough. As for "benchmarking", if you are really serious about it, get another hard-drive/ssd and have a clean copy of windows and all the drivers you need on it and nothing else. Then use this ssd/hard-drive as a control over your usual setup.