Ryzen RAM speeds a myth?

[Mark Kaine]

6 minutes ago, AMD A10-9600P said:

You're using a 3200G in those screenshots, the limit there isn't the RAM but the CPU itself, as it is comparatively very weak. I would guess as well that you're using a lower end motherboard

Heh, so we have doesnt count because my CPU isnt maxed out and doesnt count because my CPU is already  maxed out... haha.

Interesting.

 

Also note Ive made the almost exact same comparison before (with more games and benchmarks) but with a 3600 (so not CPU bound in most scenarios) and also between 8GB 2933  and 16GB 3200 of the *same* RAM type down to the same chip manufacturer.

 

I agree, its not totally representative (and that wasnt my goal) but its still significant, Id have expected at least "some" improvements in a few games, there were none though, not even 3dmark / firestrike which is specifically designed to test these things.

 

I do agree, game engines can probably make a difference, but that wasnt relevant in my testing apparently.

 

Also I can exclude "variables" a far as possible, room temp is very steady 20C, Windows updates are *off* there arent any background tasks except hwinfo.

 

15 minutes ago, FakeKGB said:

Try running Minecraft Java. It's HUGELY CPU bound and when I had my Ryzen 5 3600 underclocked to 3.4 GHz (to avoid overheating) vs. when it was overclocked to 4.3 GHz there was a huge difference. At 4.3, I could easily get 600+.

At 3.4, I could barely crack 250.

For fun, I decided to set my RAM to 2133 and test both CPU clockspeeds.

2133 + 3.4 = ~140 FPS

2133 + 4.3 = ~300 FPS

3200 + 3.4 = ~450 FPS

3200 + 4.3 = ~800 FPS

Interesting, so yeah, it maybe hugely depends on the games, its still a bit odd none of mine show this.  

[mariushm]

If your cpu is close to being maxxed out, or maxxed out ... that means the game competes with the video card driver in resources ... so basically the video card drivers may not have enough cpu time to process the data it receives from the game and push it to the video card to produce more frames.

Basically, you have  [ game engine - does calculations on cpu ] ---- [ video card driver , takes data from game engine and translates / adapts to video card language ]  ---- [ video card does stuff ]

 

Some games are more single threaded, some use more threads .... some parts of the video card drivers are single threaded, which means they benefit from cores with higher frequencies, because higher frequency means they'll finish stuff faster on that single core.

If the game uses all cpu cores a lot, the video card driver doesn't have enough time to prepare stuff for the video card and you end up with a low number of frames, and underutilized video hardware, because the video card simply waits for the game engine and for the video card drivers to feed it with new stuff.

 

So too much cpu usage can be bad for games.

Too low gpu usage can mean either you're playing at low quality settings or that the game uses too much cpu leaving too little cpu for the video card driver to feed the video card

 

The particular game you chose is bad because it has networking, and it's optimized for shooting, high fps, doesn't have super complex maps, with lots of visual effects and so on ... but the cpu can be busy with the network data, predicting where the enemies are on the map, sending your position hundreds of times a second to the server, doing calculations and crap like that.

 

---

 

edit : as far as I know ... Minecraft can be bad at multithreading being Java based, so I can only assume the performance differences shown above are due to Minecraft using mostly one cpu core for everything... and therefore higher cpu core frequency gives better  results. It shouldn't be gpu heavy, but the video card driver still needs some cpu to work. 

[Mark Kaine]

14 minutes ago, mariushm said:

If your cpu is close to being maxxed out, or maxxed out ... that means the game competes with the video card driver in resources ... so basically the video card drivers may not have enough cpu time to process the data it receives from the game and push it to the video card to produce more frames.

Basically, you have  [ game engine - does calculations on cpu ] ---- [ video card driver , takes data from game engine and translates / adapts to video card language ]  ---- [ video card does stuff ]

--snip--

I mean I did ask for reasons why Im not seeing benefits (on 2 vastly different CPUs no less) and I think you gave some good possible explanations, but doesnt that all boil down to what Ive been suspecting, that the advantages of fast RAM are hugely overrated, and apparently highly subjective (depending on games/applications) too?

 

So while there maybe are benefits (which i cannot confirm as of now) giving a blanket statement of "Ryzen needs fast RAM" (which Im guilty myself) might not be universally true, and at least a bit exaggerated as its not guaranteed at all - only in certain cases.  

 

23 minutes ago, mariushm said:

The particular game you chose is bad because it has networking, and it's optimized for shooting, high fps, doesn't have super complex maps, with lots of visual effects and so on ... but the cpu can be busy with the network data, predicting where the enemies are on the map, sending your position hundreds of times a second to the server, doing calculations and crap like that.

Huh, Resident Evil 2 has none of that, its just a pretty demanding sp game on a highly popular enginge, which is why I think its a pretty good representation.

 

29 minutes ago, mariushm said:

Minecraft can be bad at multithreading being Java based, so I can only assume the performance differences shown above are due to Minecraft using mostly one cpu core for everything... and therefore higher cpu core frequency gives better  results. It shouldn't be gpu heavy,

 

Thats perhaps a good explanation for what is described with Minecraft, but Im sorry for repeating this, its also a good example that the blanket statement most likely is false - even AMD says Ryzen need 3000 minimum, maybe thats how the "myth" started, yet they allow lower speeds* and some boards dont veven support "3000mhz" (*maybe for technical reasons, im aware) and as said, I cant see this being true as a blanket statement afterall, if its just for certain scenarios and a lot of ifs and buts.

 

 

 

[AMD A10-9600P]

1 hour ago, Mark Kaine said:

Heh, so we have doesnt count because my CPU isnt maxed out and doesnt count because my CPU is already  maxed out... haha.

Interesting.

 

Also note Ive made the almost exact same comparison before (with more games and benchmarks) but with a 3600 (so not CPU bound in most scenarios) and also between 8GB 2933  and 16GB 3200 of the *same* RAM type down to the same chip manufacturer.

 

I agree, its not totally representative (and that wasnt my goal) but its still significant, Id have expected at least "some" improvements in a few games, there were none though, not even 3dmark / firestrike which is specifically designed to test these things.

 

I do agree, game engines can probably make a difference, but that wasnt relevant in my testing apparently.

 

Also I can exclude "variables" a far as possible, room temp is very steady 20C, Windows updates are *off* there arent any background tasks except hwinfo.

 

Interesting, so yeah, it maybe hugely depends on the games, its still a bit odd none of mine show this.  

The CPU doesn't need to be maxed out to be a limiting factor, thanks to IPC, clock speed, and a whole manner of other factors.

 

Comparing with a 3600 is a fairer test, I agree with that, as it is much more capable. I'm sorry, but I don't believe that you've not seen any difference between 8GB 2933 and 16GB 3200. And again, there are other factors there, such as using different amounts of memory, rather than solely comparing the speeds it is running at.

 

I'm glad you agree that it's not representative, because it really is not. It is significant to you, in the games that you have tried, with the two CPUs you have tried. That's a tiny, tiny sample size. And again, the games you are trying may well not benefit from higher RAM speeds, but many other games and engines benefit hugely from higher RAM frequencies. 3DMark and FireStrike are not designed to test your RAM alone, plus they like GPUs too, so again, another factor comes into play that can limit the performance of both RAM frequencies, especially if you're using an outdated GPU.

 

Yes, different engines can make an enormous difference, and if I were you I'd leave it to others to actually test with bigger sample sizes of games, which will prove your results entirely wrong.

 

No, you can't exclude variables, there are far more variables at play here than just room temperature, Windows Update and background tasks. There's a reason that people out there, specifically tech YouTubers, run in depth tests and specify what conditions were controlled during their tests. They also ensure to remove any massive bottlenecks, which it appears you have in the form of your older GPU. It'd also be interesting to know what motherboard this was on, and whether you were using DOCP or XMP, and whether you messed with any of the settings there.

 

Lastly, I would say it isn't odd. Different software has different requirements from the hardware, so they will tax the components differently, and this is to be expected.

[Mark Kaine]

44 minutes ago, AMD A10-9600P said:

It is significant to you, in the games that you have tried, with the two CPUs you have tried.

But you have to admit its remarkable none of the games and benchmarks have shown any notable differences whatsoever, right?  I honestly couldnt tell u which *exact* games anymore outside a few i remember for sure (Tomb Raider, Rise Of The Tomb Raider, Resident Evil 2, Tekken 7, and probably Monster Hunter World) but I *do* remember that there were no improvements because  that was what I was looking for obviously.

 

44 minutes ago, AMD A10-9600P said:

I'm sorry, but I don't believe that you've not seen any difference between 8GB 2933 and 16GB 3200

But, that happened, if you believe it or not ( it is definitely not what I expected either though)

 

44 minutes ago, AMD A10-9600P said:

The CPU doesn't need to be maxed out to be a limiting factor, thanks to IPC, clock speed, and a whole manner of other factors.

Possible, but not in my experience, i only see frame drops when theres 100% on at least one core / thread - which is kinda weird, you'd think the burden would be offloaded to another core then, but apparently thats not how it works (as of yet)

 

44 minutes ago, AMD A10-9600P said:

3DMark and FireStrike are not designed to test your RAM alone, plus they like GPUs too,

They are specifically designed to test gaming performance, and that surely includes RAM. (otherwise its just a glorified userbenchmark lol)

 

44 minutes ago, AMD A10-9600P said:

Yes, different engines can make an enormous difference, and if I were you I'd leave it to others to actually test with bigger sample sizes of games, which will prove your results entirely wrong.

I cant deny that possibility, but then I always have the feeling games are often specifically chosen to show  what the "tester" wants to show, my games are not, theyre simply chosen because  I play them, so the relevancy is a lot bigger than games I'll never play.

 

Also that would not make my results  wrong, they wouldnt change, only the perspective would (towards games I have no interest in)

 

44 minutes ago, AMD A10-9600P said:

There's a reason that people out there, specifically tech YouTubers, run in depth tests and specify what conditions were controlled during their tests.

What conditions?  I get they have more options to test things, but seriously what conditions? 

I kind of doubt theres a difference to what Im doing - except as said i have less options, which isnt a bad thing, its just more "realistic" and my job isnt to sell stuff so Im a lot less biased.

 

PS: the only thing i totally agree, I should test more games from different genres to be more representative because what I have is pretty limited, but then again its not my goal. My goal has been "lets see the difference of RAM speeds" and the results are baffling, and quite revealing in my opinion. To me "faster RAM = more performance" has not seen any relevancy, and I doubt Im the only one where thats the case.  Which can have many reasons, but saying faster RAM = better, seems pretty *subjective* after all, if it was an *objective* fact Id expect to see this reflected in my personal testing, at least a little bit!

 

 

 

[Chris Pratt]

2 hours ago, Mark Kaine said:

Oh, so a youtube show with live chat and merchandise is more "scientific" ? I also didnt claim what im doing is scientific, more like "wait a minute... why am I seeing zero difference when everybody and their subscribers says there *is* a difference?"

At what point is it relevant then? Ive already tested 10+ games (I do remember the first time well, it was two back to back runs of the same games/benchmarks over the course of two days, and I didnt want to test any more games and benchmarks because  the results were pretty clear cut at this point) And its relevant to me, either the whole faster RAM = more FPS thing is simply not true, or something is wrong with my motherboard - I dont think its my testing per say as setting  XMP and reading some numbers isnt exactly rocket surgery, its quite difficult to mess that up...

 

 

This has been tested by the likes of Steve from Gamers Nexus. You want to see what scientific looks like go watch his vids (hint: he showed definitively that there *is* a difference with RAM speed). And, you may have not used the word "scientific", but when you just call something a scam, you better have some science to back you up. You don't.

 

As far as statistical significance goes, is not about how many games you test with the same setup. It's about how many different configurations. In order to make a blanket claim like "RAM speed doesn't actually matter on Ryzen", you'd need to test multiple different memory configurations, across multiple boards, with multiple different Ryzen CPUs, etc. Two kits and two CPUs on one board, means nothing.

 

This is especially true given the lack of scientific rigor. For example, soft sciences like sociology make assertions based on statistics, because you can't (ethically at least) be truly scientific when it comes to people and behavior, but they then need thousands or millions of data points to tease out a conclusion.

 

What you have is sloppy testing with anecdotal evidence and a relatively tiny sample size, so your results are not indicative of anything but your results. They cannot be used to speak to anything broader, so your entire post amounts to a rant, with no more credence than you'd give the homeless guy shouting at no one in particular on the street.

[AMD A10-9600P]

33 minutes ago, Mark Kaine said:

But you have to admit its remarkable none of the games and benchmarks have shown any notable differences whatsoever, right?  I honestly couldnt tell u which *exact* games anymore outside a few i remember for sure (Tomb Raider, Rise Of The Tomb Raider, Resident Evil 2, Tekken 7, and probably Monster Hunter World) but I *do* remember that there were no improvements because  that was what I was looking for obviously.

 

But, that happened, if you believe it or not ( it is definitely not what I expected either though)

 

Possible, but not in my experience, i only see frame drops when theres 100% on at least one core / thread - which is kinda weird, you'd think the burden would be offloaded to another core then, but apparently thats not how it works (as of yet)

 

They are specifically designed to test gaming performance, and that surely includes RAM. (otherwise its just a glorified userbenchmark lol)

 

I cant deny that possibility, but then I always have the feeling games are often specifically chosen to show  what the "tester" wants to show, my games are not, theyre simply chosen because  I play them, so the relevancy is a lot bigger than games I'll never play.

 

Also that would not make my results  wrong, they wouldnt change, only the perspective would (towards games I have no interest in)

 

What conditions?  I get they have more options to test things, but seriously what conditions? 

I kind of doubt theres a difference to what Im doing - except as said i have less options, which isnt a bad thing, its just more "realistic" and my job isnt to sell stuff so Im a lot less biased.

 

PS: the only thing i totally agree, I should test more games from different genres to be more representative because what I have is pretty limited, but then again its not my goal. My goal has been "lets see the difference of RAM speeds" and the results are baffling, and quite revealing in my opinion. To me "faster RAM = more performance" has not seen any relevancy, and I doubt Im the only one where thats the case.  Which can have many reasons, but saying faster RAM = better, seems pretty *subjective* after all, if it was an *objective* fact Id expect to see this reflected in my personal testing, at least a little bit!

 

 

 

Again, no I don't think it's remarkable. I'm sorry but it's expected that certain games will take better advantage of the faster RAM than others. Some of those games will be more limited by other factors than by the RAM frequency anyway, as each of those games will demand different things from your hardware. Some will hit the CPU harder, some will hit the GPU harder, some will hit the RAM harder. Clearly you have just found games that don't hit the RAM as hard. I'm not disagreeing entirely that there were no improvements, but going from 8GB to 16GB, the frequency is no longer the only change you need to be measuring for, and your results will no doubt be skewed.

 

These benchmarks test everything as a whole set, so if you're adding faster RAM to a mediocre system with a 3200G, then yes, overall you'll get improved performance, but not noticeably, as other areas are holding you back. This is just in your use case that you're seeing this. I implore you to actually, properly search this up on YouTube and other tech forums, and you'll find that higher frequency RAM clearly offers benefits. There is a reason that people here commonly recommend 3200MHz and 3600MHz memory, and they are not all wrong, as they know the most important components, and where you will gain the most performance for your money. Yes, these benefits will not be seen 100% of the time, they depend on the rest of your hardware too, but these benefits are not to be denied.

 

Your results aren't wrong, your conclusion is wrong. Higher frequency RAM is not a scam. Higher frequency RAM has been proven time and time again to increase performance.

 

Motherboard, RAM configuration, power delivery, voltages, XMP, DOCP, CPU architecture, GPU. I could go on. There is a massive difference with what you're doing compared to people like LTT. The people who professionally test these things use balls to the wall hardware so they don't run into limits elsewhere. If I was to test this, I'd want a 5950X on the most expensive motherboard with an RTX 3090 and the highest wattage PSU I could get my hands on, just so I could really pinpoint where the differences lie. Look at LTT, I've watched many videos where they compare CPUs and other parts, and in these videos they use the same top of the range hardware for the rest of the build, such that they have no other limits to run into that may jeopardise their results. Having fewer options limits how watertight you can make your testing.

 

My job isn't to sell hardware either, so I have no bias here. I have, however, learned a lot about testing during my studies, and your testing has not encompassed a big enough sample size to be reliable. Furthermore, you are lacking controlled variables. Your performance may have been changed by another factor, but you haven't taken that into account.

 

Yes, your testing was very limited, which is why the conclusions you are drawing at this point are wrong. Again, faster RAM is not a scam. People buy it for a proven reason. And again, your results are baffling to you because you have failed to take the rest of the factors you need to into account; I'd like to specifically pinpoint your GPU and motherboard as issues here. All they reveal is that your hardware and software seemingly cannot make use of faster memory, and that is no fault of the faster memory or those selling it, but the rest of your hardware. I can tell you for a fact that higher frequency RAM has made a noticeable difference to my PC, we're talking like 20% in some areas, which is definitely not to be scoffed at.

 

Your overall conclusions are wrong, but I would like to mention that, no, for everybody, higher frequency memory is not the solution. Other components may be the limiting factor before the RAM, as is the case with your machine.

[Mark Kaine]

1 hour ago, AMD A10-9600P said:

Your overall conclusions are wrong, but I would like to mention that, no, for everybody, higher frequency memory is not the solution. Other components may be the limiting factor before the RAM, as is the case with your machine.

Oh, so I agree with your post, its not really conclusive, and the testing *was* limited... the point Im making is just that, faster ram alone will not necessarily give you better performance, and to me, all this extensive testing is Interesting but nothing more, its not really relevant to me, and id wager to most, because we cannot afford to buy 500 dollars godlike motherboards, 600 dollars bling RAM and 700 dollars CPUs  (usually itll be the cheapest boards, "ok" RAM and some 200 dollar CPU) 

 

I did actually realize that after watching several GN videos where he always talks about RAM tuning and timings and whatnot - it is certainly Interesting, but it is also falsifying the results heavily because  all the tuning makes them somewhat irrelevant, and probably misleading to the average consumer /  viewer.  I get there are people this is relevant for, but they will be the vast minority. This has surely its place, but its not a good way to say this is relevant to the average person - which is why I didnt want to take GN specifically into account. Hes more an overclocking guru than unbiased reviewer imo.

 

 

1 hour ago, Chris Pratt said:

hint: he showed definitively that there *is* a difference with RAM speed

at default timings, and on under $100 motherboards? you know, the condition most people will run their RAM at ?  See above, Ive seen these videos, theyre a huge reason I dont trust these results and i think they're mostly snakeoil, most people will be happy buying the fastest RAM, they will however *not* fact check in comparison to slower RAM, not to mention "tighten timings" and stuff ~

 

1 hour ago, AMD A10-9600P said:

Your results aren't wrong, your conclusion is wrong. Higher frequency RAM is not a scam. Higher frequency RAM has been proven time and time again to increase performance.

Right, i didnt say "scam" though i said myth, meaning it might not be the holy grail it is being sold as.

 

1 hour ago, Chris Pratt said:

As far as statistical significance goes, is not about how many games you test with the same setup. It's about how many different configurations. In order to make a blanket claim like "RAM speed doesn't actually matter on Ryzen", you'd need to test multiple different memory configurations, across multiple boards, with multiple different Ryzen CPUs, etc. Two kits and two CPUs on one board, means nothing.

 

I agree, for a really conclusive statement a bigger sample size is needed and I get the idea of using the best equipment to exclude bottlenecks and the like, Im just saying by doing that a huge percentage of relevance for the average person is being lost as they will not be able to afford that equipment. Thats also why there are so many unbalanced builts, cheap motherboard, 32GB 3600mhz RAM, the most expensive CPU, no name bronze PSU...

 

So maybe I should have made the title "fast RAM, how significant is it really to the average gamer?"  

Im not good with titles, but would that have been better?  I mean I dont think what youre saying is wrong, but its also not exactly the point of my enquiry.

 

The take away so far is, faster RAM might be beneficial if the rest of the system is up to snuff - and that id still like to see it myself, although maybe my motherboard is really the limiting factor, its max rated for 3200 RAM (b350 board btw) so Im not sure if faster RAM would be worth it? Probably not. 

 

 

PS: after all, I found a game that benefits from faster RAM.

Rise Of The Tomb Raider:

2133

 

2933

 

 

On average a whooping 1FPS! And it was like that over the course of 5 runs each, not great, but at least something！

 

CPU was maxed out on all cores btw.

 

 

[Chris Pratt]

14 minutes ago, Mark Kaine said:

Oh, so I agree with your post, its not really conclusive, and the testing *was* limited... the point Im making is just that, faster ram alone will not necessarily give you better performance, and to me, all this extensive testing is Interesting but nothing more, its not really relevant to me, and id wager to most, because we cannot afford to buy 500 dollars godlike motherboards, 600 dollars bling RAM and 700 dollars CPUs  (usually itll be the cheapest boards, "ok" RAM and some 200 dollar CPU) 

 

I did actually realize that after watching several GN videos where he always talks about RAM tuning and timings and whatnot - it is certainly Interesting, but it is also falsifying the results heavily because  all the tuning makes them somewhat irrelevant, and probably misleading to the average consumer /  viewer.  I get there are people this is relevant for, but they will be the vast minority. This has surely its place, but its not a good way to say this is relevant to the average person - which is why I didnt want to take GN specifically into account. Hes more an overclocking guru than unbiased reviewer imo.

 

 

at default timings, and on under $100 motherboards? you know, the condition most people will run their RAM at ?  See above, Ive seen these videos, theyre a huge reason I dont trust these results and i think they're mostly snakeoil, most people will be happy buying the fastest RAM, they will however *not* fact check in comparison to slower RAM, not to mention "tighten timings" and stuff ~

 

Right, i didnt say "scam" though i said myth, meaning it might not be the holy grail it is being sold as.

 

 

I agree, for a really conclusive statement a bigger sample size is needed and I get the idea of using the best equipment to exclude bottlenecks and the like, Im just saying by doing that a huge percentage of relevance for the average person is being lost as they will not be able to afford that equipment. Thats also why there are so many unbalanced builts, cheap motherboard, 32GB 3600mhz RAM, the most expensive CPU, no name bronze PSU...

 

So maybe I should have made the title "fast RAM, how significant is it really to the average gamer?"  

Im not good with titles, but would that have been better?  I mean I dont think what youre saying is wrong, but its also not exactly the point of my enquiry.

 

The take away so far is, faster RAM might be beneficial if the rest of the system is up to snuff - and that id still like to see it myself, although maybe my motherboard is really the limiting factor, its max rated for 3200 RAM (b350 board btw) so Im not sure if faster RAM would be worth it? Probably not. 

 

 

 

 

There's still an issue with the "average gamer" presumption. You and your results here are not representative of that, and that's been my point the entire time. Even if you go off the assumption that the "average gamer" is going to be using lower end hardware, it's still not indicative that they would have the same results as you, and faster RAM may in fact make a difference for their build, as it should it general.

 

Again, all you can say is it didn't make a difference for *you*. It cannot be applied any more broadly than that. Even then, I don't think your testing was rigorous enough to make that a definitive statement for you. Based on your limited testing, no difference was bore out, but even something as simple as a different RAM kit to test with may show a different result.

[Mark Kaine]

5 minutes ago, Chris Pratt said:

Even if you go off the assumption that the "average gamer" is going to be using lower end hardware

I agree with the other points youve made, but that is not an assumption, Steam surveys are the closest to statistical evidence we get and most will have rather low end hardware, and even laptops.  Currently GTX 1060 is the most popular card at 20% (iirc) for example.

 

9 minutes ago, Chris Pratt said:

it's still not indicative that they would have the same results as you, and faster RAM may in fact make a difference for their build,

Yeah, true, I cannot say that, which is also why Im asking. Its really difficult to choose the right motherboard for example, especially on the lower end, in my impression theyre kind of a mess, luck of the draw.

 

 

 

[Guest]

Testing = Not flawed, just incompleted.

 

Memory = 2933/3200mhz = Not really high speed.

Don't even consider 3600mhz high speed myself.

 

Needs testing done on processors that are not limited by thread/core count. (my opinion here)

- testing on all cpus at the same clock frequency. Only memory variances.

- testing on multiple software, including games and benchmarks (choices matter)

 

And on top of all that, listing the speeds and the memory timings/voltage would really gather up some information.

 

I'm sorry Mark sir. But this post lacks any real useful information to back any statements made here today.

Lastly, my Ryzen Athlon 220ge was benching memory at 3333mhz. 

 

PiFast, pretty sensitive to memory tweaking and frequency.

YOUR GOAL = beat this time. (probably want the cpu frequency similar for best results // 4350mhz)

GL dude!

 

Spoiler

 

 

[Mister Woof]

In every test I've seen with fast memory on both Ryzen and Intel it really depends on what's being tested and how.

 

Most games that are GPU limited won't matter at all. GPU only benchmarks don't matter much since they mostly ignore CPU, and CPU benchmarks don't ever seem to be memory sensitive.

 

The best example of fast memory speeds giving improvements are in heavily CPU limited scenarios like in MMO games.

 

 

[Guest]

1 minute ago, Mister Woof said:

In every test I've seen with fast memory on both Ryzen and Intel it really depends on what's being tested and how.

 

Most games that are GPU limited won't matter at all. GPU only benchmarks don't matter much since they mostly ignore CPU, and CPU benchmarks don't ever seem to be memory sensitive.

 

The best example of fast memory speeds giving improvements are in heavily CPU limited scenarios like in MMO games.

 

 

 

It's an OS thing really.

 

W7 - 3DMark06 = 3D and CPU dependant benching.

Memory = underclocked stock timings no tweaks

Cache = No OC- on auto - Cpu all core 5.2ghz no HT

Gpu 2060, little OC 1415/1782mhz

 

 

Spoiler

 

[Mark Kaine]

16 minutes ago, ShrimpBrime said:

Testing = Not flawed, just incompleted.

I never said anything else though.

My approach was literally simple, out of the box, default settings vs XMP, because I think thats what most people do and what I do.  Of course it would be better if I had higher speed RAM to test too, but then probably my motherboard would hold me back?

 

16 minutes ago, ShrimpBrime said:

PiFast, pretty sensitive to memory tweaking and frequency.

YOUR GOAL = beat this time. (probably want the cpu frequency similar for best results // 4350mhz)

Haha, I would probably try that but I think the mobo is really not meant for RAM overclocking? (b350m mortar)  i cant even get it to run 1hz above rated speed, immediate crash upon windows login lol (bios is "fine")

 

Literally the only reason I did these tests is because of my recent BIOS woes - I updated my BIOS and my R5 3600 wouldn't boot anymore ?? Can you believe it, MSI have a *broken* BIOS as the latest non beta BIOS… 

 

It basically removes compatibility with 3000 series lol, so luckily I had my 2200g  laying around and it boots with that! 

 

So i just thought new CPU Im going to test this, because I already knew the RAM makes zero noticable difference with my 3600 CPU, so in a way, if it did with the 2200g something would be probably wrong with the 3600 or the mobo - but it doesn't!

 

So whatever it is, its maybe just the motherboard being slow, or the RAM being subpar, its most likely not the CPU!?

 

13 minutes ago, Mister Woof said:

Most games that are GPU limited won't matter at all. GPU only benchmarks don't matter much since they mostly ignore CPU, and CPU benchmarks don't ever seem to be memory sensitive.

But that *is* basically what im saying, ive seen "tests" with overclocked RAM and throughout better FPS, like 10, 20 + FPS only with memory tuning! So A I dont believe it and B even if true, who does that, memory tweaking for days? Almost no one, people dont really have time for that!

So yes, its a good show, but not really applicable for every day / average gamer use - in fact most people I talked with are scared af about overclocking,  they think it blows up their PC!  (well, it might!)

 

 

18 minutes ago, Mister Woof said:

The best example of fast memory speeds giving improvements are in heavily CPU limited scenarios like in MMO games.

 

More believable imo, but thats not whats being shown by "influencers" , its the whole litany of popular games (or ones they think are popular)

 

 

[Guest]

Just now, Mark Kaine said:

I never said anything else though.

My approach was literally simple, out of the box, default settings vs XMP, because I think thats what most people do and what I do.  Of course it would be better if I had higher speed RAM to test too, but then probably my motherboard would hold me back?

 

Haha, I would probably try that but I think the mobo is really not meant for RAM overclocking? (b350m mortar)  i cant even get it to run 1hz above rated speed, immediate crash upon windows login lol (bios is "fine")

 

Literally the only reason I did these tests is because of my recent BIOS woes - I updated my BIOS and my R5 3600 wouldn't boot anymore ?? Can you believe it, MSI have a *broken* BIOS as the latest non beta BIOS… 

 

 

 

L@@K at the screen shot. MB tab is open - Running B450M-A Asus Prime. Bottom of the bottom of the very bottom before you buy a A320 chipset haha.

 

Excuses excuses.  

[Mister Woof]

Just now, Mark Kaine said:

I never said anything else though.

My approach was literally simple, out of the box, default settings vs XMP, because I think thats what most people do and what I do.  Of course it would be better if I had higher speed RAM to test too, but then probably my motherboard would hold me back?

 

Haha, I would probably try that but I think the mobo is really not meant for RAM overclocking? (b350m mortar)  i cant even get it to run 1hz above rated speed, immediate crash upon windows login lol (bios is "fine")

 

Literally the only reason I did these tests is because of my recent BIOS woes - I updated my BIOS and my R5 3600 wouldn't boot anymore ?? Can you believe it, MSI have a *broken* BIOS as the latest non beta BIOS… 

 

It basically removes compatibility with 3000 series lol, so luckily I had my 2200g  laying around and it boots with that! 

 

So i just thought new CPU Im going to test this, because I already knew the RAM makes zero noticable difference with my 3600 CPU, so in a way, if it did with the 2200g something would be probably wrong with the 3600 or the mobo - but it doesn't!

 

So whatever it is, its maybe just the motherboard being slow, or the RAM being subpar, its most likely not the CPU!?

 

But that *is* basically what im saying, ive seen "tests" with overclocked RAM and throughout better FPS, like 10, 20 + FPS only with memory tuning! So A I dont believe it and B even if true, who does that, memory tweaking for days? Almost no one, people dont really have time for that!

So yes, its a good show, but not really applicable for every day / average gamer use - in fact most people I talked with are scared af about overclocking,  they think it blows up their PC!  (well, it might!)

 

 

More believable imo, but thats not whats being shown by "influencers" , its the whole litany of popular games (or ones they think are popular)

 

 

You'd be surprised how much time people have on their hands.

 

Or how much time they'd rather be dicking off on their PCs than listen to nagging family.

 

This is just speculation.

[Guest]

Just now, Mister Woof said:

You'd be surprised how much time people have on their hands.

 

Or how much time they'd rather be dicking off on their PCs than listen to nagging family.

 

This is just speculation.

I work days, wife works nights.  

[Mister Woof]

Just now, ShrimpBrime said:

I work days, wife works nights.  

My wife and I work right next to each other in the home office all day every day.

 

[Mark Kaine]

8 minutes ago, ShrimpBrime said:

L@@K at the screen shot. MB tab is open - Running B450M-A Asus Prime. Bottom of the bottom of the very bottom before you buy a A320 chipset haha.

 

 

Oh, im on phone i didn't notice.

 

9 minutes ago, ShrimpBrime said:

Excuses excuses.  

UwU

 

 

Spoiler

I might try XMP2 which says "3200" but Im pretty sure it'll crash lol

 

[Guest]

47 minutes ago, Mark Kaine said:

 

I might try XMP2 which says "3200" but Im pretty sure it'll crash lol

 

 

If that's the case, then OC from 2933 and see where it lands.

 

[Mark Kaine]

5 minutes ago, ShrimpBrime said:

If that's the case, then OC from 2933 and see where it lands.

 

So I did "Try it!" settings, but I chose the slower timings, and it didnt crash!

Spoiler

Should I go for 3333 ?

Or tighter timings?

 

 i got another half FPS in Tomb Raider , and ~+50 points in C23, but that doesn't really say much I guess.

¯\_(ツ)_/¯ 

 

 

 

 

[Guest]

Just now, Mark Kaine said:

So I did "Try it!" settings, but I chose the slower timings, and it didnt crash!

  Reveal hidden contents

Should I go for 3333 ?

Or tighter timings?

 

 i got another half FPS in Tomb Raider , and ~+50 points in C23, but that doesn't really say much I guess.

¯\_(ツ)_/¯ 

 

 

 

 

I would "increase" the Cpu "ratio" to "accommodate" the memory "overclock" and greatly "increase" the "bandwidth".

 

Also set the Dram Write recovery latency to 8 opposed to 16. Trfc, maybe try 420, can tighten this one as you go along. 

[Mark Kaine]

59 minutes ago, ShrimpBrime said:

would "increase" the Cpu "ratio"

By how much?

59 minutes ago, ShrimpBrime said:

Also set the Dram Write recovery latency to 8 opposed to 16.

Which one is that? tcwl? maybe 

 

 

Edit: ok i set multiplier to 39 , tcwl to 10, 9 would be lowest and trfc to 420. 

 

It didnt crash!  (I think this BIOS is really better than the previous, unfortunately its only temporary because  no Ryzen 3600 support )

 

+210 or so points in c23

 

Not unsurprisingly so since I just OC the CPU by 200mhz effectively

 

@ShrimpBrime I still have the loose timings, shouldn't I tighten them now to see if it has any effect, or is that more involved?

 

^I mean after I did the Tomb Raider benchmark  

 

[MadAnt250]

2 hours ago, Mark Kaine said:

basically what im saying, ive seen "tests" with overclocked RAM and throughout better FPS, like 10, 20 + FPS only with memory tuning! So A I dont believe it and B even if true, who does that, memory tweaking for days? Almost no one, people dont really have time for that!

1. Going from 3000Mhz to 3200 with better timings got me an extra 5 FPS in a game were I was GPU limited, so your testing is limited.

2. It took me less than 5 minutes to adjust.

3. Not a myth.

[Guest]

46 minutes ago, Mark Kaine said:

By how much?

Which one is that? tcwl? maybe 

 

 

Edit: ok i set multiplier to 39 , tcwl to 10, 9 would be lowest and trfc to 420. 

 

It didnt crash!  (I think this BIOS is really better than the previous, unfortunately its only temporary because  no Ryzen 3600 support )

 

+210 or so points in c23

 

 

Not unsurprisingly so since I just OC the CPU by 200mhz effectively

 

@ShrimpBrime I still have the loose timings, shouldn't I tighten them now to see if it has any effect, or is that more involved?

 

^I mean after I did the Tomb Raider benchmark  

 

Of course tighten timings. One at a time though. 

 

200mhz = 210 CB score increase. 

Now run the memory stock 2133 + 200mhz OC and what is the difference?