dgianstefaniOK, set your memory to 4800 MT JEDEC and report back afterwards if you still believe this.

You're on a 144 Hz monitor, so your averages probably wouldn't be affected too much, but I'd wager almost anything you'd notice your minimum FPS dipping below 60 FPS quite frequently.

Regardless, RAM speed/latency objectively matters to gaming performance, even if you have an X3D chip, this is a statement of fact. Depending on your monitor refresh rate, your GPU power and the type of gamer you are, you may notice this more or less, but the underlying technical data is irrefutable.

kapone32I can't see I see that and I run my RAM at 5200. Could be your 12GB Vram buffer effecting that. Maybe it is more CPU cores I have and more VRAM to not experience that.

dgianstefaniRAM speed/latency objectively matters to gaming performance, even if you have an X3D chip, this is a statement of fact. Depending on your monitor refresh rate, your GPU power and the type of gamer you are, you may notice this more or less, but the underlying technical data is irrefutable.

dgianstefaniCompletely missing the point as usual, and somehow conflating GPU bottlenecking with CPU bottlenecking, well done.

dgianstefaniOK, set your memory to 4800 MT JEDEC and report back afterwards if you still believe this.

You're on a 144 Hz monitor, so your averages probably wouldn't be affected too much, but I'd wager almost anything you'd notice your minimum FPS dipping below 60 FPS quite frequently.

Regardless, RAM speed/latency objectively matters to gaming performance, even if you have an X3D chip, this is a statement of fact. Depending on your monitor refresh rate, your GPU power and the type of gamer you are, you may notice this more or less, but the underlying technical data is irrefutable.

AusWolfI am running my RAM at 4800 MHz JEDEC (because my CPU needs a lot less SoC voltage this way), and my gaming experience is just fine. :)

Here is the underlying technical data:


My own take on it is that if I have 218/166 avg/min FPS or 204/149 is an insignificant difference, as the amount of info reaching my eyes and brain is exactly the same.

Also, this test was done with a 4090 at 1080p. With my 7800 XT at 1440 UW, I am infinitely more GPU-limited, therefore, my RAM speed matters even less.

dgianstefaniAnd here's some other testing.

In your own testing for a 7700X, the difference between 4800 and 6000 is 30 FPS.

Some games such as Tarkov, Factorio or Minecraft are intensely CPU bottlenecked almost at all times, and will have more significant differences regardless of GPU used.

Other games will be GPU bottlenecked at all times, mostly single player games.

Regardless of how significant you consider 10-30 FPS, or if you consider it to be the "same amount of info reaching your eyes and brain", it's a difference. Your earlier statement of "RAM speed and latency don't matter at all" is therefore subjective, not objective.




If you have low expectations of your hardware, that's fine. If you don't notice the difference between 120 and 150 FPS, that's fine, but lets not pretend that then translates into "RAM speed doesn't matter".

Because people read technical threads like these, and misinformation is then propagated.

dgianstefani"RAM speed and latency don't matter at all" is therefore subjective, not objective.

kapone32I did not know that a 13900K comes with V cache otherwise this chart means nothing. In case you did not know AMD Overlay also shows 1% data and the X3D is famous for holding 1% numbers. The fact that you have a 7800X3D and did not use your own data speaks volumes.

Yep you can definitely tell the difference between 130 and 100 FPS on a Freesync panel....not.

kapone32The fact that you have a 7800X3D and did not use your own data speaks volumes.

dgianstefaniOK, set your memory to 4800 MT JEDEC and report back afterwards if you still believe this.

You're on a 144 Hz monitor, so your averages probably wouldn't be affected too much, but I'd wager almost anything you'd notice your minimum FPS dipping below 60 FPS quite frequently.

Regardless, RAM speed/latency objectively matters to gaming performance, even if you have an X3D chip, this is a statement of fact. Depending on your monitor refresh rate, your GPU power and the type of gamer you are, you may notice this more or less, but the underlying technical data is irrefutable.

TigerfoxIn theory, yes. In real life, for most people it doesn't matter.

I'm not talking, JEDEC 4800. Im talking 5600 or 6000 Cl32/CL30 with EXPO. And I'm talking 1440p@144Hz or even UHD. Even with a 4080, you will run into GPU limit most of the time, and when you don't, there wont be any tangible difference between 5600, 6000 ord 6400. I definiately won't consider anything abvoe that for AM5, because I don't have the time I would need to invest into optimzing for so little gain.

dgianstefaniWhat you consider "tangible" or "real life" are subjective analyses, the fact is that faster/lower latency RAM affects performance, even on an X3D platform, unless you think the entire game data fits in ~100 MB of cache. Considering this is a hardware enthusiast forum, I would expect less resistance to this statement.

dgianstefaniThe benchmarks use a 4090 at 1080p because that's an easy way to force a CPU bottleneck, instead of playing for an hour to get 10 minutes of usable data demonstrating the differences you want to focus on. Unless you think CPU/RAM benchmarking should use GPU limited scenarios? This is the basis of scientific testing, you exclude the control variables and test the variable you're interested in.

1440p is considered a CPU limited resolution these days BTW. Especially with the (now cheap) popular 240 Hz monitors, or the emerging 360 Hz/480 Hz 1440p monitors. 4K is pretty much the only resolution where you're GPU bound all the time, but even then you can see improved minimum FPS with better RAM. 144 Hz monitors have been around for more than a decade now, it's pretty entry level.

dgianstefaniNoted. I now know that Kapone cannot tell the difference when looking at a 30% performance improvement. Thanks for the quote, will be useful for future reference when you start discussing performance.

I'm sure you do indeed have "the best AMD computer", but how could you tell? What if another PC was 29% faster?


It speaks for the fact I don't need to waste time logging my own testing to disprove anything you say, benchmarks off TPU are more than sufficient.

Or the general web, as AusWolf showed with his data.

TigerfoxThat's what I meant with "in theory". I don't doubt it does. What I doubt is that RAM speed affects performance @1440p or, UHD to such an extent that it warrants investing the time and/or money needed especially over somethin like 6000 CL30 or 6400 CL32 with EXPO that isn't much more expensive than 5600 and is usable out of the box.

If you can't afford at least a 4080, worrying about highend RAM is nonsense. Even then, you should much rather invest in a 4090. If you can't even afford a 7800X3D and 4080, even more so.

What I do think helps performance a lot, but won't do myself for time reasons, is buying cheap RAM and OCing it to 6000/6400 or lowering bad timings to good.

So, for people useing a 7800X3D/7950X3D or 13900K/14900K(S), 4090 and a monitor supporting 240Hz+ wanting to get the framerate as high as possible, especially the lows, RAM OC might be important, but those people are few.

dgianstefani[...] there are issues with the "20 FPS doesn't matter" attitude. Especially as this performance % difference sticks across all levels of hardware (moving your minimums from 100 to 130 is noticable even if you're only using a 4070/7700 class GPU, which is easily capable of 120 FPS. In a CPU limited scenario, the FPS is limited by the CPU, so whether or not you are using a 4090 or a 4070, those minimum lows will still be very similar, as long as the 4070 is capable of reaching a higher average than the CPU dictates, which it is).

For instance, ~20-25 FPS is the difference between a 4070 Ti and a 4080 (which used to be $400 difference), yet I don't see anyone saying that a faster GPU won't make a difference? Yet when that 20-25 FPS is coming from a RAM tune (free, or maybe $50 more expensive if you want to buy a faster stock kit), suddenly it's imperceivable or "not tangible". A 7900XTX vs 7900XT is even less than 20 FPS, does that render AMD's flagship pointless? No.

dgianstefaniNoted. I now know that Kapone cannot tell the difference when looking at a 30% performance improvement. Thanks for the quote, will be useful for future reference when you start discussing performance.

I'm sure you do indeed have "the best AMD computer", but how could you tell? What if another PC was 29% faster?

TigerfoxFirst, I have yet to see bencmarks in 1440p were RAM-tuning above DDR5-6000 CL30 from the shelf on a 7800XD does even make a difference of a two digit percentage, in low fps if you like. The benchmark you posted is, again, unrealistic 1080p with cards designed for 1440p and UHD. Then, even if there were games were that was the case, it would have to be in a fps-range were I would see the difference. I don't play fast shooters and I didn't get to play on my 34" 144Hz-monitor, but I really doubt I would see a difference between ~120fps and 130fps.

AusWolfIt's all relative. 30% over 50 FPS is noticeable. 30% over 200 FPS is not (unless the game you're playing is called FRAPS or Afterburner).

Edit: The only difference is that you don't get 30% extra with high graphical settings / resolutions, and with mid-range or lower GPUs.

Your own sensitivity matters a lot as well. I know someone who demands a constant 360 FPS on his 360 Hz monitor. As for me, anything above ~40 is smooth enough, especially with Freesync.

dgianstefaniNoone tests at 1440p, not because you wouldn't see a difference, but because it's a mixed benchmark. At 1440p sometimes you'll be CPU limited, sometimes you'll be GPU limited, this isn't helpful when you're trying to evaluate the effect of RAM performance. At 4K we know CPU and therefore RAM speed becomes less meaningful, because only high end cards can start to push more than 120 FPS, but 1440p is significantly easier to drive for the GPU than 4K.

dgianstefaniA 10-20% difference in FPS matters more at 100 FPS+ than it does at 50 FPS, as AusWolf has just said.

dgianstefaniThe point I've been trying to make for the past 30 minutes, which seems to have significant resistance (for some reason), is that you don't ever want to be CPU bottlenecked, because that is what people notice as stuttering, or dips. It's irrelevant if you're playing at 100 FPS or 200 FPS, that number suddenly halving or going down by an appreciable amount because your CPU is struggling to keep up is noticable and immersion breaking, and coincides with massively increased input lag. If you want to talk esports, then "muscle memory" is tied to frame rates, you want consistency, not high averages.

dgianstefaniLike I said, subjective analysis is fine, but lets call it for what it is.

Not much point buying a high refresh monitor if you barely ever hit or sustain that high FPS.

AusWolfThat's why I'm saying that RAM speed doesn't matter to me, as I'm playing at 1440 UW with as high graphical details as my GPU allows. This way, I'm always GPU limited.

Artificially inducing a CPU-limited scenario purely for the sake of science is fine, but why should I give the results more credit than they're worth?


I actually said the opposite. At low FPS, any small extra can help, but at high FPS, I couldn't care less if there's any difference.


I agree with this sentiment, but which situations your experience is limited by your CPU/RAM is highly dependent on your hardware and your sensitivity. I doubt that I could ever notice when a game running at 200 FPS dips into 100 for a microsecond, and I also doubt that any mid-range GPU is CPU limited at 1440p unless you pair it with a Celeron.


Exactly - it's all subjective. That's my point all along. :)

The point is not necessarily the high refresh rate, but VRR, which eliminates any screen tearing at the appropriate performance levels.

dgianstefaniNo. You're not.

I'm busy doing other things now, but good chat I guess.

dgianstefaniScreen tearing is an entirely different problem than stuttering or framerate dips. They may have similar causes, but they're different issues entirely.

dgianstefaniI mean, my current X670E mobo already has two USB4 ports, so hopefully there's also other improvements or the refresh is pretty boring.

Mandating it is good I suppose, but surely there's more interesting stuff to improve.

Thunderbolt 4? WiFi 7? Better memory traces?

dgianstefaniWhile some of what you're saying is true (but exaggerated, you don't need a 4090 and a 240 Hz+ panel to notice differences in RAM and therefore CPU performance), there are issues with the "20 FPS doesn't matter" attitude. Especially as this performance % difference sticks across all levels of hardware (moving your minimums from 100 to 130 is noticable even if you're only using a 4070/7700 class GPU, which is easily capable of 120 FPS. In a CPU limited scenario, the FPS is limited by the CPU, so whether or not you are using a 4090 or a 4070, those minimum lows will still be very similar, as long as the 4070 is capable of reaching a higher average than the CPU dictates, which it is).

For instance, ~20-25 FPS is the difference between a 4070 Ti and a 4080 (which used to be $400 difference), yet I don't see anyone saying that a faster GPU won't make a difference? Yet when that 20-25 FPS is coming from a RAM tune (free, or maybe $50 more expensive if you want to buy a faster stock kit), suddenly it's imperceivable or "not tangible". A 7900XTX vs 7900XT is even less than 20 FPS, does that render AMD's flagship pointless? No.

Tigerfox@dgianstefani : Just out of interest I looked up recent benchmarks about the benefits of fast RAM on Ryzen 7000 and found this. While DDR5-7400 is possible since AGESA 1.0.0.7b, it performs even slightly worse than DDR5-6200, because of 1:2 mode. Same for 7950X in 1440p and 7800X3D in 1080p. So what's your argument again? There is no performance to be gained by fast RAM. You can buy RAM fast enough to max out Ryzen 7000 for just slightly more than cheap RAM. Only thing you can do ist buy cheap RAM and tune it to 6400 CL30/32 yourself, if you have time but no money.