CPU Benchmark in realistic condition (multiple programs running)

[mikeymikec]

I don't know about you guys, but I always launch both Prime95 and Furmark, before I start gaming.

Why?

 

[Dresdenboy]

Why?

To make fun of the topic by exaggeration.

 

[mikeymikec]

To make fun of the topic by exaggeration.

Maybe I've been on these forums for too long, my "not sure if serious face" is more or less a default position when I encounter strange forum posts

 

[frozentundra123456]

Well the thread title does assume that gaming with a lot of background tasks is "realistic", which implies that this is what is the normal situation. Clearly that is not true for myself and others who have posted in this thread that they close out background programs when gaming. So the thread title is a presumptive over-generalization at best.

 

[Erenhardt]

Even consoles do not close background tasks when playing games... talk about peasantry... sheesh!

 

[Dresdenboy]

Even consoles do not close background tasks when playing games... talk about peasantry... sheesh!

Don't they have dedicated cores (e.g. 2 of 8 in PS4)?

On desktop I see problems like equally prioritized background tasks while multitasking (e.g. a browser utilizing a full core due to stupid JS/Flash code - in the ads...) stealing a full core and mem B/W.
Or a HT machine, where even a low priority thread reduces another threads performance by 20-40%.

 

[TheELF]

On desktop I see problems like equally prioritized background tasks while multitasking (e.g. a browser utilizing a full core due to stupid JS/Flash code - in the ads...) stealing a full core and mem B/W.

Why would an add utilize a full core? it would be more like 2-3% of a core,meaning 2-3% of slower gameplay (if there is 0% GPU bottleneck)
unless it's an athlon 5150 (consoles) or something.

Or a HT machine, where even a low priority thread reduces another threads performance by 20-40%.

But HT is specifically made to run the main thread (of the core) at full speed and utilize whatever is left to run whatever other thread needs running at the same time.
A low priority thread would mean 0% performance loss if the other thread has normal priority.

 

[Dresdenboy]

Why would an add utilize a full core? it would be more like 2-3% of a core,meaning 2-3% of slower gameplay (if there is 0% GPU bottleneck)
unless it's an athlon 5150 (consoles) or something.

I did't profile the tons of scripts, just saw the CPU usage. No video or audio playback btw.

But HT is specifically made to run the main thread (of the core) at full speed and utilize whatever is left to run whatever other thread needs running at the same time.
A low priority thread would mean 0% performance loss if the other thread has normal priority.

Without thread priority information in the processor HT cannot know how to prioritize. Does it have priority bits? My i7-5600 seems to have none.

 

[TheELF]

Without thread priority information in the processor HT cannot know how to prioritize. Does it have priority bits? My i7-5600 seems to have none.

You said low priority thread which made me think that you where talking about a thread that has it's priority set to low.

But anyway, HT does not work by slowing a thread down,one thread runs at 100% and if it doesn't use up every command available in a cycle then the surplus gets used to run commands from a second thread.

http://software.intel.com/en-us/articles/performance-insights-to-intel-hyper-threading-technology

 

[Excessi0n]

Why *would* you?

Because closing and restarting everything is a hassle at absolute best. That goes double if you switch back and forth between the game and something else, or you have something which must be run continuously or for a long duration.

I treat games like I treat any other program: I leave everything else running unless one of the background programs directly conflicts with whatever I'm focusing on. This typically means pausing music/video while gaming, and pausing uploads/downloads during online games.

 

[Dresdenboy]

You said low priority thread which made me think that you where talking about a thread that has it's priority set to low.

There is no need to explain SMT to me. I just provocatively asked for a bit.

It was a low priority thread, because I set priority to "idle". Only changing affinity helped.

My point is, that HT has no main thread.
1T -> 1C = 100%
2T -> 1C ~= 2x65%.

 

[Atreidin]

Using a modern general-purpose PC to do multiple tasks at once is not an amazingly novel idea. There is no intrinsic reason that everyone needs to turn their PCs into single-purpose consoles every time they want to start a game. Obviously people would only do this when their CPU is at least "good enough," since people generally have brains and would close programs that were interfering too much with game performance. The biggest issue would be testing it reliably (edit- and with sufficient repeatability). Also, review sites might not like this idea because to satisfy all the separate use cases for all of their multi-tasking readers, they might see their number of test runs explode unless they pick specific program combinations very carefully; and even then they will be constantly accused of bias for pairing certain programs together and not others.

 

[Phynaz]

You said low priority thread which made me think that you where talking about a thread that has it's priority set to low.

But anyway, HT does not work by slowing a thread down,one thread runs at 100% and if it doesn't use up every command available in a cycle then the surplus gets used to run commands from a second thread.

http://software.intel.com/en-us/articles/performance-insights-to-intel-hyper-threading-technology

That's not how SMT works. The first word is "symmetrical" for a reason.

 

[VirtualLarry]

That's not how SMT works. The first word is "symmetrical" for a reason.

Except, it isn't.

https://en.wikipedia.org/wiki/Simultaneous_multithreading

Edit: Just in case you think Wikipedia is incorrect, this is also in his Intel link:

Also known as Simultaneous Multi-Threading,

 

[Magic Carpet]

turn_pike,

You have a point. Personally, I like to watch a movie or some crazy tv documentary about nazis on the second screen. Multi-monitor performance would be nice to test in such circumstances. I know when I tried to play GTA V at the same time, after a while there were occasional moments of slow downs for whatever reasons (GPU/CPU was never pegged at 100%). I ended up using iGPU for that purpose, no problems since.

But it is certainly very niche to encode videos and game at the same time. That's why you see no reviews with that.

 

[Dresdenboy]

turn_pike,

You have a point. Personally, I like to watch a movie or some crazy tv documentary about nazis on the second screen. Multi-monitor performance would be nice to test in such circumstances. I know when I tried to play GTA V at the same time, after a while there were occasional moments of slow downs for whatever reasons (GPU/CPU was never pegged at 100%). I ended up using iGPU for that purpose, no problems since.

But it is certainly very niche to encode videos and game at the same time. That's why you see no reviews with that.

That multi monitor scenario should be added to the "dirty benchmark" set.

Encoding while gaming happens by streaming or recording the game (twitch, etc.). PS4 has a dedicated chip for that.

 

[TheELF]

Encoding while gaming happens by streaming or recording the game (twitch, etc.). PS4 has a dedicated chip for that.

So do all three major graphics companies quicksync/nvenc/vce

And personally I never saw any decrease in performance from anything running with idle priority.
And that's with the slowest haswell dual core.
But I guess there are enough badly written things out there for it to maybe happen.
https://www.youtube.com/watch?v=b8jcX7rXVdk

 

[Dresdenboy]

So do all three major graphics companies quicksync/nvenc/vce

That's right, but quicksync might steal some power budget then:

https://software.intel.com/en-us/ar...k-sync-video-h264-codec-with-sorenson-squeeze

And personally I never saw any decrease in performance from anything running with idle priority.
And that's with the slowest haswell dual core.
But I guess there are enough badly written things out there for it to maybe happen.

Agreed. Not all softwares and website scripts are perfect or even close to that.

And there is a difference between a HLT instruction and instructions doing real work at "idle priority".

 

[TheELF]

That's right, but quicksync might steal some power budget then:

So you use a bit more power but are done in 1/3 the time... so you actually use less power.

Not to mention (but I'm gonna anyway) it doesn't even reach 50 Watts,whichever CPU this might be,there are CPUs that idle at that.

 

[Dresdenboy]

So you use a bit more power but are done in 1/3 the time... so you actually use less power.

Not to mention (but I'm gonna anyway) it doesn't even reach 50 Watts,whichever CPU this might be,there are CPUs that idle at that.

Of course the integral counts and it won't be that much during gaming (adapting to target stream resolution).

During gaming with well threaded game code, Quick Sync shouldn't eat too much away.

Edit: I did some measurements on my i7-5600U work notebook. This diagram shows the Prime95 FFT times when using 4 threads.

I have multiple taskmanager screenshots of the different situations, but don't want to spam the thread for now.

The background task is my genetic programming trading system software, which runs with 4 threads and usually a priority one level below "normal". It does lots of floating point calculations, pointer handling, memory accesses and branching. For the tests I set it to the lowest level.

Prime95 sets priority to "normal" for each subtest and does heavy AVX work.

What we're seeing here was my vague impression before based on uarch and is a solid fact now. There is no priority in HT. Skylake might have improved that, but I didn't hear about this.

 

[ElFenix]

You guys close down your browsers when you game?

 

[dionasaur]

You guys close down your browsers when you game?

My thoughts exactly... I usually game in windowed mode as well.

 

[frozentundra123456]

Yes, I do close down browsers, (and all other programs except windows background tasks) when gaming except for perhaps a walkthrough on a second screen. I guess I am old school (well I know I am, my first computer was 200 mhz or something like that) but I just dont see why some are so shocked that one would want to devote as many resources as possible to a demanding task like gaming, both computer resources and your own attention. I mean is it really *that* hard to open and close a browser or other program?

 

[Erenhardt]

Yes, I do close down browsers, (and all other programs except windows background tasks) when gaming except for perhaps a walkthrough on a second screen. I guess I am old school (well I know I am, my first computer was 200 mhz or something like that) but I just dont see why some are so shocked that one would want to devote as many resources as possible to a demanding task like gaming, both computer resources and your own attention. I mean is it really *that* hard to open and close a browser or other program?

I did the same, but that was in 1995 or so...

Honestly? I have to admit to it... I am guilty of running multiple game clients (which was prohibited by the server rules in certain games) to boost farming items/skills.

I know it is unethical and probably ruined gaming experience for some. I want you to know that I am very sorry for what I did, and I will not do this again.

But still, I find myself bored of WoW, PoE, Diablo etc. leave it alt+tabed and give a spin in NFS, LFS, AC. After a race I can go back to farming again. Thankfully I don't have that much time recently, so I keep more focused when I'm playing games.

 

[turn_pike]

OP here.

I feel that many here on the forum as well as Anandtech's reviewers are perhaps of the older demographics that might be out of touch with what younger (15-30 yo) gamers do with their computers.

Go to twitch.tv and see what those streamers do. As you get better and better at online competitive games, the queue time often gets longer. What some people do is play another game while waiting. So they often have stuff like Fallout 4, Overwatch, Heathstone being paused in the background while they're playing Dota or LoL. That, on top of streaming, browser open to interact with viewers, skype-ing with teammates in the games, recording the gameplay for review and upload to youtube, the list goes on.

I've seen quite a few also play games like Osu or I Wanna Be the Guy during queue to train / retain their mouse mechanic. These are what modern gamers do. The very notion of closing the browser and other programs is the antithesis of the whole exercise.

PS
As a side note, I'm happy that AT finally included games that are actually being played, like Rocket League, on their recent benchmark of the AMD 7670. Hopefully they would add something like Dota2 or LoL being played under the harshest condition (like 12 players fighting on the screen)

PPS
Oh and gaming in windowed mode. Yes, that's also something that'd be great to benchmark. Some do that so they can still multitask while gaming (social media stuff usually). Personally I only do it while on a single player games like EU4 or Football Manager. But I've seen progamers play competitive matches on windowed mode because he keeps a notepad window open to jot down important information during the game. TLDR; it's becoming more prevalent and a benchmark should be considered