Bottlenecks.. still not clear cut like people think they are.

[Potato*Salad]

 

Interesting watch, the G4560 was able to power up to a GTX 1080, it bottlenecks it in games where it needs more cores, but not nearly as much as many would think.

 

 

[Crunchy Dragon]

Yeah, I thought that was pretty interesting as well. It's shocking how a Pentium wouldn't bottleneck a 1080.

[PlayStation 2]

Bottlenecking is a blanket statement which generally means "I don't know what I'm talking about."

[Jurrunio]

Bottleneck is never an accurate measure anyway. It depends on what you are doing with it.

[RGProductions]

Bottlenecking is real. But it doesn't even matter 90% of the time.

[Mira Yurizaki]

A lot of people are too sensitive about "not getting the best out of their hardware"

 

But unless you are getting the best hardware, you're not going to get the best out your hardware anyway.

[Zyndo]

this is a terrible test done with terrible reasoning done by a guy with the resources and the experience to do better.... Using the Superposition benchmark to prove a 1080 isn't limited by a pentium is an incredibly stupid notion. the unigine benchmarks (Valley, Haven, Suporposition) are specifically and intentionally designed to be as GPU intensive as possible and as least CPU intensive as possible. they are GPU specific benchmarks. they are made specifically for testing the GPU and only the GPU as much as possible. You pick ANY game, any ACTUAL game, with a normal amount of CPU/GPU scaling and you start seeing bottlenecking in and around the GTX 1060 area. This video perpetuates the notion that a 1080 and a pentium are an okay pairing.... which is just not at all the case. I'm not saying pentium based gaming systems are bad, not at all. and if you're gaming at 4k a pentium/1080 might be a more sane pairing for high graphical gaming on a budget, but on the topic of how much a pentium actually bottlenecks specific GPU's at 1080p gaming, its WAYYY less than a GTX 1080's worth of power.

 

I'm with the Jay in the sense that bottlenecking is a craze which people should really not be so paranoid about especially since most people don't really understand the concept of bottlenecking.... but I really feel that this video does more harm than good to the conversation due to an ignorant testing methodology.

[HKZeroFive]

The words "bottleneck" and "future-proofing" will continue to trigger me for as long as I live.

[i_build_nanosuits]

1 hour ago, Potato*Salad said:

 

Interesting watch, the G4560 was able to power up to a GTX 1080, it bottlenecks it in games where it needs more cores, but not nearly as much as many would think.

 

 

well dahhh...superposition is a GPU benchmark...that require very little to be done from the CPU...no wonder the CPU a dual core do just fine...Jay is such a numbnut...

[Princess Luna]

2 minutes ago, i_build_nanosuits said:

well dahhh...superposition is a GPU benchmark...no wonder the CPU a dual core do just fine...Jay is such a numbnut...

His videos are even worse than LTT's worse videos... frankly, his qq about not getting free stuff from nvidia annoys me.

 

A G4560 is a good pair up to the GTX 1060, beyond that it is a clear bottleneck to where you're wasting potential. Balancing the build is a must we need videos explaining further the best pairings and not go and say "hey if you only do synthetic gpu benchmarks you could pair a TITAN Xp with a Celeron!!! much wow.... facepalms >_>

[i_build_nanosuits]

Just now, Princess Cadence said:

His videos are even worse than LTT's worse videos... frankly.

 

A G4560 is a good pair up to the GTX 1060, beyond that it is a clear bottleneck to where you're wasting potential. Balancing the build is a must we need videos explaining further the best pairings and not go and say "hey if you only do synthetic gpu benchmarks you could pair a TITAN Xp with a Celeron!!! much wow.... facepalms >_>

G4560 is a place holder...a CPU that you purchase because you are broke...a CPU design for an office PC...a CPU that, has a gamer, you should be ashamed of and want to replace ASAP...it's ...ok...i've said enough

 

[PCGuy_5960]

1. He used Superposition to look for CPU bottlenecks. (Superposition is GPU bound)
2. He used a 7600K as a best case scenario

GG Jay 

[Zyndo]

9 minutes ago, i_build_nanosuits said:

G4560 is a place holder...a CPU that you purchase because you are broke...a CPU design for an office PC...a CPU that, has a gamer, you should be ashamed of and want to replace ASAP...it's ...ok...i've said enough

 

to be fair..... a pentium g4560 isn't this bad. its actually not horrible for what it is and can honestly do a decent job of gaming.... but there are very few instances where it could actually truly keep up with a gtx 1080 (maybe if you were after a 'budget' 4k ultra gaming build lol)

[i_build_nanosuits]

Just now, Zyndo said:

to be fair..... a pentium g4560 isn't this bad. its actually not horrible for what it is and can honestly do a decent job of gaming.... but there are very few instances where it could actually truly keep up with a gtx 1080 (maybe if you were after a 'budget' 4k ultra gaming build lol)

GPU load is one thing...FRAME TIME is another...playing any modern demanding titles that can load 4 to 6 or more CPU threads with this...your GPU load MIGHT not be that low or that bad...but you WILL defenetly not get a smooth experience...it will be ponctuated by micro-stuttering and wild fluctuations in frametimes and those are not reflected in average FPS you see everywhere, but an experienced gamer will defenetly notice them and it drives you nuts...and those occurs even if you're using a GTX 1050.

[Princess Luna]

4 minutes ago, Zyndo said:

to be fair..... a pentium g4560 isn't this bad. its actually not horrible for what it is and can honestly do a decent job of gaming.... but there are very few instances where it could actually truly keep up with a gtx 1080 (maybe if you were after a 'budget' 4k ultra gaming build lol)

 

2 minutes ago, i_build_nanosuits said:

GPU load is one thing...FRAME TIME is another...playing any modern demanding titles that can load 4 to 6 or more CPU threads with this...your GPU load MIGHT not be that low or that bad...but you WILL defenetly not get a smooth experience...it will be ponctuated by micro-stuttering and wild fluctuations in frametimes and those are not reflected in average FPS you see everywhere, but an experienced gamer will defenetly notice them and it drives you nuts...and those occurs even if you're using a GTX 1050.

Imagine the stuttering hell...

[Zyndo]

3 minutes ago, i_build_nanosuits said:

Snip

yeeeah.... but not really. I mean, I have a 6700k + 980ti (bought them when they were king of the hill) so I certainly understand wanting to have the best of the best. But honestly, a G4560 gaming experience is not as horrible as people think. I'm not saying it can handle something like GTA V on max settings without a hitch.... but if you go into a game with a reasonable expectation of what the pentium can do it won't be a disappointment.

[PhaseShift]

30 minutes ago, i_build_nanosuits said:

GPU load is one thing...FRAME TIME is another...playing any modern demanding titles that can load 4 to 6 or more CPU threads with this...your GPU load MIGHT not be that low or that bad...but you WILL defenetly not get a smooth experience...it will be ponctuated by micro-stuttering and wild fluctuations in frametimes and those are not reflected in average FPS you see everywhere, but an experienced gamer will defenetly notice them and it drives you nuts...and those occurs even if you're using a GTX 1050.

This!!! There is much more to cpu bottlenecking than just synthetic benchmarks and average framerate. A large variance in fps caused by the cpu constantly hanging around 90-100% usage results in a terrible stuttery gaming experience (even though the avg fps is not that badly affected).

[Masada02]

As soon as I heard him say he was going to use Superposition to test CPUs any of his findings became completely irrelevant. Superposition is completely based around the GPU and it's kind of impressive how little it cares about whatever CPU you run. That makes it completely useless for a test like this. Even more so, trying to test "bottlenecking" without a huge swath of games is pointless. Take for instance BF1 that loves more threads more than IPC and would play awesome on something like a 1700, but then another game like Arma 3 that thrives on single core performance will suck ass on that same CPU when compared to something like a 7700k. 

 

My point is, it's impossible to define whether a certain CPU is enough to drive a certain GPU in a broad sense. The best way to pick these two components is to put together a "use case" (the games or tasks you want to do and HOW you want to do them) and then pick based on what gives you the best average.

 

A 7700k at *pick a frequency* will ALWAYS hold a *pick your favorite GPU* back at SOMETHING. You just have to decide whether that something means something to you.

 

 

Jesus I hate the term "bottlenecking."

[Mira Yurizaki]

Maybe the whole point of using a GPU intensive task is to show that the CPU is not inherently limiting the GPU in any way and that it entirely depends on what kind of game you throw at the CPU.

 

i.e., this is to debunk any conception of by sticking a Pentium with a GTX 1080, you automatically cut your performance by say 40% and there's nothing you can do about that than if you went with an i7-7700K

[Masada02]

2 minutes ago, M.Yurizaki said:

Maybe the whole point of using a GPU intensive task is to show that the CPU is not inherently limiting the GPU in any way and that it entirely depends on what kind of game you throw at the CPU.

Maybe, but it just doesn't seem very responsible to show a G4560 driving a 1080 to high levels because realistically it's not going to perform that well. Luckily he did at least reign back the first set of results with some actual games, but I still think it is going to confuse uneducated gamers into thinking a lowly Pentium is all they will need.

 

In the end though, it's all subjective really. If the owner of a system like that is happy, who am I to be irritated? And really, the best answer is always an engineers answer: "it depends..."

[Mira Yurizaki]

4 minutes ago, Masada02 said:

Maybe, but it just doesn't seem very responsible to show a G4560 driving a 1080 to high levels because realistically it's not going to perform that well. Luckily he did at least reign back the first set of results with some actual games, but I still think it is going to confuse uneducated gamers into thinking a lowly Pentium is all they will need.

 

In the end though, it's all subjective really. If the owner of a system like that is happy, who am I to be irritated? And really, the best answer is always an engineers answer: "it depends..."

I think as long as it's made clear that the task was purely a GPU one and that the actual results vary depending on what you throw at it, it doesn't hurt to run said task to show that the hardware configuration by itself is not the problem, but what kind of software you run on it. I mean yes, eventually it'll come back to "your CPU can't run this software fast enough", but I feel like people think bottlenecking means that they have a performance cap, period. Like a CPU is only good for 60 FPS at 1080p on medium or something silly like that.

 

But yeah, I feel most PC gamers don't really know what's going on under the hood and try as we might to educate anyone on the subject, at the end of the day they just want the cheatsheet to "best performance ever (without giving up a kidney to do so)", which can't really exist.

[Intelfreak]

I can some what comfirm this. Not with a pentium or 1080, but with and old i7 980x @ 4.25 ghz and a 1080 ti.

 

The bottleneck with this setup is far less than i exspected. 1080P yes bottleneck in some games gpu load can be from 80 to 99 % but 2560 x 1600 and up it is still the gpu that hits the limit before my old cpu does. I exspected the bottleneck to be far worse than this. Because of this i have droped the plans for a newer cpu i else planed for then intel released there 14, 16 and 18 core cpu's.

[DragonTamer1]

GamesNexus already did something like this and found that unless you are playing Sniper Elite 4 the 1060 is the most powerful card you should pair with a pentium. Beyond that is diminishing returns and there is almost no performance gains from 1070 to 1080. Hence, the CPU is a bottleneck in those games.