Gaming Memory/Timings/Latency for z370 gaming MOBO

[Vayra86]

Yes, its a RAM test to show how it improves FPS in a game. So, wouldn't more valid testing be to at least use AA at 1080p (which is what most users do that have at least half a clue)? People don't run 1080Ti's at 1080p with no AA. To that end, it is exaggerating the results with an unrealistic testing environment. I don't care what card or CPU is used, 1080p with Ultra/High settings and AA his how people do (who aren't running a 1050, lol).

Why do you have to be CPU bottlenecked? You aren't when playing with normal settings? Why fabricate that environmental variable when testing?

You keep harping on low-end cards with high-end CPUs like it has something to do with this. I can't help how others test, nor is that any justification for testing in an equally lopsided manner (and to be clear, I am referring to HEDT CPUs excluding 7740K and its little brother - I believe testing with a 7700K or 8700K is absolutely fine. Plenty buy this way ). Can we focus on the discussion ?

There is no way everyone will be happy. My only concern is people see this data and think they will achieve the same results at their settings and resolution. They won't. IMO, the best way to test is more realistic running situations. If you are going to use a 1080Ti down to a 1060, at least run at 1080p Ultra/High with gobs of AA, or better yet, 2560x1440. I feel complete testing will test this fabricated environment and something more realistic so users have results closer to where they play. AS this testing stands, people look at it and think they will see these gains where they play, and they won't.

It isn't though. That is what I am saying. Because in many games, you are not cpu bottlenecked unless you rock Sandybridge or lower on Intel and every AMD cpu not named Ryzen. Some are, but then are they actually memory bandwidth limited which is what they are actually testing???? Lots of variables...


Oh well, a never-ending discussion again. Sorry I brought it up.

1. Market share of CPUs has NEVER been an influential factor for reviewers on gaming GPUs. A lot of them used HEDT class CPUs with mainstream GPUs, in fact.
2. You have a fundamentally different approach to 'reviews' it seems. Your reviews are aimed at 'what can I expect in real life' and not at 'what is the raw potential of this component'. Commendable but not realistic. How do you decide what is the 'average use case' and above all, what is 'average' or 'normal'? Its way too abstract especially in the PC world.
3. This different approach of yours, reduces the value of your style of reviewing to that of the simpleton: it is no more than a 'Can I Run this' benchmark focused on a specific component. I hate to break it to you, but in all fairness, we already have crappy websites for that basic info. Why write a whole review and analysis? I won't even read it, honestly. I care a lot more about that raw performance, because that is still relevant in other situations as an indicator of what the fastest solution will be.

We have butted heads on this on multiple occasions but the pattern is clear and again, I believe these are two radically different approaches at work here. Synthetic versus real world. While synthetic still is a valuable data-set 10 years from now, real world only matters today with the current driver, games, components and market reality. In my humble opinion, that is why we now see 'Performance Analysis' per game. Because really, for these real world performance numbers, the per-game basis is so much more valuable.

Focus on what's important for the subject of the article, and clearly divide the two, and there is no need for a never ending discussion on the subject, but instead we do gain more valuable information both for the NOW and future. When you review a component: focus on the component. When you review a game/bench a game: focus on the game.

 

[EarthDog]

1. Tell that to Cucker... I really don't care what CPU is used. It makes sense to me to use a 7700K/8700K for this kind of testing. I can also go along with a 7900X or whatever... but then we get into a testing environment for the 1%. While we arent talking 50% here for 7700K/8700K they are mainstream and affordable which I believe is his underlying point there.
2. I don't care about raw potential as I never see it. I see the potential at reasonable settings and resolutions.
3. I disagree completely with that assertion... particularly the simpleton part. It isn't what you say it is. My method gives more realisitic results, period. There just isn't getting around the fact that the results here are fabricated to show a difference which isn't there in a normal testing environment.

I care a lot more about that raw performance, because that is still relevant in other situations as an indicator of what the fastest solution will be.

Buuuuuuuuuuut, it isn't. And it seems that is what is a big hang up here. That fabricated results doesn't scale or tell you anything when it's in a realistic use scenario. Some seem to think isolating the RAM in this manner, the results are scalable to other settings or resolutions. It isn't. Not remotely so. So looking back at a bad result still yields a bad result. For example, if we see a result of 20 FPS (say 15%) difference in this test, but with 'proper' settings, shows 5 FPS or no gains? What does that result really tell you? Its only good for its unrealistic test enviro.

And it will be tested again when DDR5 comes out, or new CPUs or new, whatevers. Real world DOES matter with current driver, games, and components in the market... see the first sentence. This is why this fabricated test environment is so toxic as it skews perception and actual results replacing them with results unable to scale.

Yes, I agree, focus on the component, but not to the point where the testing environment isn't remotely realistic. This is why I am sticking with 1080p and AA.. because it can be a CPU bound resolution. Going lower or using Low/no AA, particularly with a high-end GPU, skews reality.

Anyhoo, continuing to pile on... again, the PM box is always open.

 

[Vayra86]

1. Tell that to Cucker... I really don't care what CPU is used. It makes sense to me to use a 7700K/8700K for this kind of testing. I can also go along with a 7900X or whatever... but then we get into a testing environment for the 1%. While we arent talking 50% here for 7700K/8700K they are mainstream and affordable which I believe is his underlying point there.
2. I don't care about raw potential as I never see it. I see the potential at reasonable settings and resolutions.
3. I disagree completely with that assertion... particularly the simpleton part. It isn't what you say it is. My method gives more realisitic results, period. There just isn't getting around the fact that the results here are fabricated to show a difference which isn't there in a normal testing environment.

Buuuuuuuuuuut, it isn't. And it seems that is what is a big hang up here. That fabricated results doesn't scale or tell you anything when its in a realistic use scenario. Some seem to think isolating the RAM in this manner, the results are scalable to other settings or resolutions. It isn't. Not remotely so. So looking back at a bad result, still yields a bad result.

And it will be tested again when DDR5 comes out, or new CPUs or new, whatevers. Real world DOES matter with current driver, games, and components in the market... see first sentence. This is why this fabricated test environment is so toxic as it skews perception and actual results replacing them with results unable to scale.

Yes, I agree, focus on the component, but not to the point where the testing environment isn't remotely realistic. This is why I am sticking with 1080p and AA.. because it can be a CPU bound resolution. Going lower or using Low/no AA, particularly with a high-end GPU, skews reality.

Anyhoo, continuing to pile on... again, the PM box is always open.

Let's pick up the example given in this thread, to turn it back on topic:

We see 1080p no AA testing. You say 'not realistic'... I say 'this is my use case'.

I use a strong GPU with the fastest CPU at a relatively low resolution, hell I even drop IQ settings if that is what's needed to achieve 120 fps. The linked review and situation is 100% relevant to me and I can also tell you that it does indeed make a huge difference increasing RAM speeds in a CPU bound scenario that still is gaming. I can also transplant these numbers to different games that are far more CPU/RAM bound and conclude that faster RAM will benefit me in those games as well. And, again, this is supported by my own experiences switching to faster RAMs in for example Guild Wars 2 World vs World with loads of players on screen.

So again; who gets to decide what is realistic or relevant? Should the writer do that, or the reader... perhaps I HATE AA because it can make games blurry.

EDIT: as to your below points on
1.; I am already running into games at 1080p where a (non-TI) 1080 won't do the trick as in below 60 FPS performance. No less than a year ago, people yelled 'why would you ever use a 1080 for 1080p, its a 1440p card'.
2. AA really isn't that much of an influence here on recent GPUs, and there are tons of AA methods too, TAA being more and more prevalent, even combined with internal upscale or downscale. The AA-line is becoming real blurry (lol!). Go look at Resident Evil 7 and you can see what I mean... Dynamic Resolution is a thing these days and it can be used in both directions.

Ah well. I'm dropping it, as before, let's agree to disagree ^^

 

[EarthDog]

If you (royal you note) are the person who runs a 1080Ti at 1080p without AA, than this is for you. If you have a clue on what you are doing, you......

1. Wouldn't be rocking a 1080Ti on 1080p without AA in the vast majority of situations.
2. Would be using higher settings and AA (otherwise, go console)

I have a 1080 that pounds through nearly everything at 60 FPS+ 2560x1440, well more than that in many titles (high/ultra 2xAA for this res). I have a 144 Hz monitor... I prefer the eye candy over getting another 20 FPS as its buttery smooth anyway.. In most titles, I am around 100 FPS.

As I said, this is realistic for the fabricated environment its in. Outside of that, it really isn't. It doesn't scale, it doesn't tell you anything about the sweet spot at more realistic settings. Id simply rather a review cover more realistic settings (again can't please everyone) to show a more accurate picture of what to expect. If the review is outside of your use case, it really doesn't tell users much.


EDIT:

Ah well. I'm dropping it, as before, let's agree to disagree ^^

We'll never see eye to eye on this point.

Dynamic resolution is also placing the load on the GPU, making ram speed matter less... as does AA....as does higher settings which most strive to run.

EDIT2: I mean, I said my PM box was open and you replied here, LOLOLOLOL!

 

[cucker tarlson]

Yes, its a RAM test to show how it improves FPS in a game. So, wouldn't more valid testing be to at least use AA at 1080p (which is what most users do that have at least half a clue)? People don't run 1080Ti's at 1080p with no AA. To that end, it is exaggerating the results with an unrealistic testing environment. I don't care what card or CPU is used, 1080p with Ultra/High settings and AA his how people do (who aren't running a 1050, lol).

Why do you have to be CPU bottlenecked? You aren't when playing with normal settings? Why fabricate that environmental variable when testing?

You keep harping on low-end cards with high-end CPUs like it has something to do with this. I can't help how others test, nor is that any justification for testing in an equally lopsided manner (and to be clear, I am referring to HEDT CPUs excluding 7740K and its little brother - I believe testing with a 7700K or 8700K is absolutely fine. Plenty buy this way ). Can we focus on the discussion ?

There is no way everyone will be happy. My only concern is people see this data and think they will achieve the same results at their settings and resolution. They won't. IMO, the best way to test is more realistic running situations. If you are going to use a 1080Ti down to a 1060, at least run at 1080p Ultra/High with gobs of AA, or better yet, 2560x1440. I feel complete testing will test this fabricated environment and something more realistic so users have results closer to where they play. AS this testing stands, people look at it and think they will see these gains where they play, and they won't.

It isn't though. That is what I am saying. Because in many games, you are not cpu bottlenecked unless you rock Sandybridge or lower on Intel and every AMD cpu not named Ryzen. Some are, but then are they actually memory bandwidth limited which is what they are actually testing???? Lots of variables...


Oh well, a never-ending discussion again. Sorry I brought it up.

Well it's still a good tets for people who know how to approach the outcome of such test. And people who you think might be confused or misinterpret these results don't read sources like that and delve enough into the subject anyway. I'm all for testing in fabricated environments but even I use a lot of common sense in then purchasing any components, just cause DDR4 3600 does scale over 3200 nicely, don't mean many people need it. If you're on a 240Hz monitor and got money to blow then why not though, it's good to know it will give you a slight boost.

 

[mouacyk]

Testing to differentiate two products that are alternatives to each other needs strict isolation. Suppliers love this, because they can charge for the "unrealistic" potential. Enthusiasts love this testing, because when other variables are controlled, they're free to chase and "realize" this potential or deny it.

It's a huge waste of time and money for your average reader.

 

[EarthDog]

Well it's still a good tets for people who know how to approach the outcome of such test. And people who you think might be confused or misinterpret these results don't read sources like that and delve enough into the subject anyway. I'm all for testing in fabricated environments but even I use a lot of common sense in then purchasing any components, just cause DDR4 3600 does scale over 3200 nicely, don't mean many people need it. If you're on a 240Hz monitor and got money to blow then why not though, it's good to know it will give you a slight boost.

That is the problem, you give people too much credit

Look around these (any) forums and see some of the questions people ask... and forums are where enthusiasts are supposed to be... can you imagine a normal user with a Dell/prebuilt gaming PC thinking of upgrading reading that? "zOMG......I can hz 20 moar FPS by t3h memor13s? Where do I sign up?" More than half the people here wouldn't know what that testing is showing and how it extrapolates (or doesn't) to a higher res/settings (AKA- more realistic and not fabricated to exaggerate differences).

just cause DDR4 3600 does scale over 3200 nicely

But does it at the resolution and settings you play at or only when you run low and no AA?

It's a huge waste of time and money for your average reader.

Exactly... they cannot discern like many here (but not a majority) can. This is another good reason why more testing at more appropriate settings is helpful to the average joe reading it.



Damnit.. stop replying people...trying to leave this thread alone and nobody seems to care to want to use the PM box!!!! hahaha! lololol!