What amount of latency is considered good pc latency for a regular gamer, and does it always require expensive parts?

[davidl41]

Would I need a Ryzen CPU that ends in x3d, or an equivalent Intel? Also, is Intel CPUs better with latency in general because they aren't taking short cuts in how they engineer as it seems like when AMD Ryzen has issues they just make it more powerful which surely creates flaws? Does the motherboard matter for pc latency?

Basically, is the general idea to just make the PC 80th percentile or better, and it's pc latency should be fine?

[crack]

Would I need a Ryzen CPU that ends in x3d, or an equivalent Intel? Also, is Intel CPUs better with latency in general because they aren't taking short cuts in how they engineer as it seems like when AMD Ryzen has issues they just make it more powerful which surely creates flaws? Does the motherboard matter for pc latency?

Basically, is the general idea to just make the PC 80th percentile or better, and it's pc latency should be fine?

It depends what you want to do with your pc, ryzen cpus are generally great for gaming only, low latency, lower with tweaking, don't ask for too much in cooling. I wouldn't choose an x3d cpu, it's a gamble if it runs flawless or not, any ryzen ending in X is fine without tweaking, choose mono CCD ryzens if you want the lowest latency out of ryzen cpus.

Intel cpus are a jack of all trades: gaming and productivity, lower latency without tweaking than ryzen most of the time, lowest possible latency with tweaking, require investment in cooling.

Motherboard choice matters, don't cheap out especially with Intel.

[davidl41]

I wouldn't choose an x3d cpu, it's a gamble if it runs flawless or not, any ryzen ending in X is fine without tweaking, choose mono CCD ryzens if you want the lowest latency out of ryzen cpus.

It is for gaming use. I am trying to use my AMD AM4 motherboard. The Ryzen 5900x with a 3070 GPU will get me 120 to 140 FPS on 1080p low setting in the game Battlefield 6 and the Finals. I think. My Ryzen 3600 and 3070 gets me around 85 to 105 FPS on those same settings.

The two upgrade options is a x3d or a 14600k or better. They both get above the 200fps range. A 5700 x3d CPU or 14600k with a 3070 would get me 250 FPS or more on those games. I hear there are things about Intel that are better. Is there less end to end latency with the 14600k with a 3070 compared to a 5700x3d with a 3070? What advantages are there that most are unaware of? Also, what are the issues with the x3d CPU?

[crack]

I wouldn't choose an x3d cpu, it's a gamble if it runs flawless or not, any ryzen ending in X is fine without tweaking, choose mono CCD ryzens if you want the lowest latency out of ryzen cpus.

It is for gaming use. I am trying to use my AMD AM4 motherboard. The Ryzen 5900x with a 3070 GPU will get me 120 to 140 FPS on 1080p low setting in the game Battlefield 6 and the Finals. I think. My Ryzen 3600 and 3070 gets me around 85 to 105 FPS on those same settings.

The two upgrade options is a x3d or a 14600k or better. They both get above the 200fps range. A 5700 x3d CPU or 14600k with a 3070 would get me 250 FPS or more on those games. I hear there are things about Intel that are better. Is there less end to end latency with the 14600k with a 3070 compared to a 5700x3d with a 3070? What advantages are there that most are unaware of? Also, what are the issues with the x3d CPU?

The issue with x3d cpus is stuttering high performance. The advantage of x3d cpus is the large cache they provide which boosts their performance significantly compared to non x3d. Also the latency will never match an Intel equivalent.

On Intel side, while x3d does have more raw gaming performance, even with less FPS Intel cpus feel better to play with, way smoother image on screen, if you pair your 3070, which I also happen to have, 3070 aorus master to be exact, with an Intel cpu you will have the lowest end to end latency and smoothest image. Between 13th and 14th gen Intel there is a small difference in performance, but the big difference in 14th gen is heat output, they work extra hard to squeeze what are really diminishing returns.

[dervu]

It was proven that end to end latency is the same for intel or ryzen.

[Sz3ypy]

The issue with x3d cpus is stuttering high performance.

Can you explain ??

[kyube]

Would I need a Ryzen CPU that ends in x3d, or an equivalent Intel? Also, is Intel CPUs better with latency in general because they aren't taking short cuts in how they engineer as it seems like when AMD Ryzen has issues they just make it more powerful which surely creates flaws?
Basically, is the general idea to just make the PC 80th percentile or better, and it's pc latency should be fine?

You also shouldn't forget your network latency to the server.
The question is very nuanced, but is generally answered with — get the best possible HW if you want to compete at a high level or be the best.

Think of it from this perspective:
You want the absolute lowest latency possible, with minimum variation, to minimise hardware bottlenecking your potential (skill ceiling)
You want to be the player with the lower latency than your opponents in a particular game to have an advantage over them.
This, due to severe levels of regression in user application software optimizations, requires getting the highest performing hardware such as Zen 5 based 8 core microprocessors with a large L3 cache (e.g.: 9800X3D) & a newest GPU.

Is there a 'optimal value' or perhaps a value that would be at the top of the bell curve of a normal distribution (which is how I'd classify the term "regular gamer")? I haven't come across such data to give you an answer.

For finding that answer, it might be best to look at the PCLatency metric. You could also look at RTSS' reflex.ovl overlay in D3D11/D3D12 titles or use SK (Special-K, in games without a kernel-level AC) to evaluate the pipeline a bit deeper.

Does the motherboard matter for pc latency?

Yes, it absolutely does. This is one of the bigger problems that users around here forget to account for.

The issue with x3d cpus is stuttering high performance.

Please stop equating anecdotes (likely based on ignorance, either knowingly or unknowingly) with the objective reality without having any concrete evidence to back it up, thanks
I'll give you a easy 'proof' of how ridiculous all of your claims are — name me the sample size of users which allowed you to form such a conclusion?
Better yet, how many of them had proper side-band ECC UDIMMs to account for 'overclocking' mishaps? I can confidently claim, from empirical evidence, that is the root cause for the vast majority of users who have "an issue" around here. Especially the "EMI" users.
This is all without going into the validity/trustworthyness of these samples (being prone to lying/deception/trolling, having medical conditions etc.)

I will state this, in case anyone else sees this:
99,5% of easy-to-acquire resources / data online in regards to comparing (the 0,5% is private information or information I haven't come across, as claiming absolutes is foolish) CPU architectures (e.g. 9800X3D vs 14900K/285K) cannot claim absolutes without equating or minimising the variation in the following variables when discussing them:
• Human bias (one of the most important non-HW variables; e.g.: HWUnboxed pushing AMD-first content)
• Microprocessor unit variance
• Power supply unit
• Motherboard PCB layer count,
• I/O controllers on the motherboard (NIC, USB, audio),
• GPU
• Cooling solution,
• Side-band ECC UDIMMs (good luck doing that with Intel, as that requires either choosing X79, X99 or W680-based motherboards. Though, one could circumvent this by avoiding EXPO/XMP use as a compromise)
• Software (very vague term, can be a needle in the haystack situation due to potential CPU&GPU&WinVer interplay, but let's simplify this by claiming use same WinVer & same GPU driver)

This also completely disregards the topic of power quality (impossible to avoid) & networking quality variance (assuming one would devise a online-based test)
Not to mention the fact that users here are seemingly unable to look past past PresentToPresent time, which is what's measured by overlays when users online claim ""FPS boosts"" (It's not representative of what the end-user feels)
Not to mention that users here aren't able to utilise statistical techniques to evaluate this topic objectively.
Not to mention that psychological & physiological bias exists (feeling "off")
Not to mention that networking & electricity can impact the experience (I'll throw a random number here without any basis — <0,5% of users are experiencing this)

All of these variables require extensive reading into & extreme caution before coming to any conclusion.

At this point in time, X3D's (7800X3D & 9800X3D) absolutely clear current Intel microprocessor offerings in terms of Present2Present time ("FPS") in the vast majority of eSports games. The large L3 cache is a savior to the unoptimized mess of games we have on the market.
With the current Nova Lake rumors of 8p16e (no HT) 144MiB microprocessor coming from Intel, this will likely shift in favor to Intel.
Disclaimer: I completely disregard averages when discussing "FPS" (present-to-present time), unless that testing data is the only one available from a particular reviewer.
I only take into account games which scale properly with good CPUs (e.g.: SoTR, Doom Eternal), ideally ones with a repeatable environment.
I try to find seldom data which has the least likelyhood of being "GPU limited" (having a render queue depth > 1 on each frame).
This requires a lot of assumptions/speculation from my side and I don't have a hard stance on any of my "X > Y" claims due to the lack of good testing data.


A visualized example of how different "FPS" values can be, depending on what you 'choose' to constitute a "frame":

[crack]

The issue with x3d cpus is stuttering high performance.

Please stop equating anecdotes (likely based on ignorance, either knowingly or unknowingly) with the objective reality without having any concrete evidence to back it up, thanks
I'll give you a easy 'proof' of how ridiculous all of your claims are — name me the sample size of users which allowed you to form such a conclusion?
Better yet, how many of them had proper side-band ECC UDIMMs to account for 'overclocking' mishaps? I can confidently claim, from empirical evidence, that is the root cause for the vast majority of users who have "an issue" around here. Especially the "EMI" users.
This is all without going into the validity/trustworthyness of these samples (being prone to lying/deception/trolling, having medical conditions etc.)

I will state this, in case anyone else sees this:
99,5% of easy-to-acquire resources / data online in regards to comparing (the 0,5% is private information or information I haven't come across, as claiming absolutes is foolish) CPU architectures (e.g. 9800X3D vs 14900K/285K) cannot claim absolutes without equating or minimising the variation in the following variables when discussing them:
• Human bias (one of the most important non-HW variables; e.g.: HWUnboxed pushing AMD-first content)
• Microprocessor unit variance
• Power supply unit
• Motherboard PCB layer count,
• I/O controllers on the motherboard (NIC, USB, audio),
• GPU
• Cooling solution,
• Side-band ECC UDIMMs (good luck doing that with Intel, as that requires either choosing X79, X99 or W680-based motherboards. Though, one could circumvent this by avoiding EXPO/XMP use as a compromise)
• Software (very vague term, can be a needle in the haystack situation due to potential CPU&GPU&WinVer interplay, but let's simplify this by claiming use same WinVer & same GPU driver)

This also completely disregards the topic of power quality (impossible to avoid) & networking quality variance (assuming one would devise a online-based test)
Not to mention the fact that users here are seemingly unable to look past past PresentToPresent time, which is what's measured by overlays when users online claim ""FPS boosts"" (It's not representative of what the end-user feels)
Not to mention that users here aren't able to utilise statistical techniques to evaluate this topic objectively.
Not to mention that psychological & physiological bias exists (feeling "off")
Not to mention that networking & electricity can impact the experience (I'll throw a random number here without any basis — <0,5% of users are experiencing this)

All of these variables require extensive reading into & extreme caution before coming to any conclusion.

At this point in time, X3D's (7800X3D & 9800X3D) absolutely clear current Intel microprocessor offerings in terms of Present2Present time ("FPS") in the vast majority of eSports games. The large L3 cache is a savior to the unoptimized mess of games we have on the market.
With the current Nova Lake rumors of 8p16e (no HT) 144MiB microprocessor coming from Intel, this will likely shift in favor to Intel.
Disclaimer: I completely disregard averages when discussing "FPS" (present-to-present time), unless that testing data is the only one available from a particular reviewer.
I only take into account games which scale properly with good CPUs (e.g.: SoTR, Doom Eternal), ideally ones with a repeatable environment.
I try to find seldom data which has the least likelyhood of being "GPU limited" (having a render queue depth > 1 on each frame).
This requires a lot of assumptions/speculation from my side and I don't have a hard stance on any of my "X > Y" claims due to the lack of good testing data.


A visualized example of how different "FPS" values can be, depending on what you 'choose' to constitute a "frame":

Get off your high horse buddy, at least word your replies in a non-condescending way, from the several other posts you have made, your solution is throw way more money at the problem than it's needed.

ECC memory being the solution for a lot of people in this forum, get outta here.

I wish I didn't bother reading your reply.