Blur Busters Forums

Posted: **16 Feb 2025, 07:45**

9800x3d dont have a single stutter. Actually in cs2 deadlock marvels everything i tested, it is higher 1%lows by a bunch.
And i had z790apex encore well tuned.

Is Just that everyone is a clown when it comes tô configurating both bios hw os.

Nothing in this world that is good is free. U wont find free good info, and Who really know stuff dont sell themselves

So go learn

Posted: **16 Feb 2025, 10:14**

1. Rule #1: Be Nice. This is published forum rule #1. Even To Newbies & People You Disagree With!

I warn you (and all other ill-intentioned person that start to come here), I understand that we all don't agree but stop calling/implying that people are "incompetent" when you don't know them.
This is the only and last answer i will gave here, and i will not reading any other replies, you're arrogant and you're not worth debating.

kyube wrote: ↑
15 Feb 2025, 19:59
I don't see a problem in my statement.
I've been here for quite a decent time and I stand by my words.
Most people are clueless.

I think you don't understand, you are belittling and insulting the majority of users on this forum in a condescending way implying they don't know any subject but who are you to judge anyone here?
Go piss off someone else and leave me alone, I already told you, I don't want to get into your ragebait which makes no sense, I'm not coming here for explanations or to play the AMD vs Intel card, i'm just sharing a user experience, nothing more.

kyube wrote: ↑
15 Feb 2025, 19:59
Yes. I'm telling you that your experience is a user issue because you haven't done research on what components you should pick to not encounter stuttering. You'd encounter the same stuttering on Intel, if not picking the right settings.
Intel is far easier to make worse, hence why a myriad of people have been coming here looking for some unicorn to save their "floaty CS2 mouse"

Excuse me? you have absolutely no idea on what research i've done, do we know each other? no.
You don't know the ton of research i've done, research that that go beyond what you think, I don't allow you to judge me as a casual user here. I wouldn't do any research, I wouldn't already be on this forum. And like some users here, no worries with Intel, I received my 12900k kit today, light undervolt and XMP on > I have no stuttering and latency feels much better. No need to being denial, both Intel and AMD have issues (not to mention degradation etc) but that's just not the subject actually. If you have no worries that's great, have a good day and enjoy your system, I only believe my own tests.
Latency does not come from a "bad choice" of motherboard only as you imply but from a lots of things including hardware and software, unefortunately you won't teach me anything with your optimizations and even if that would be the case, I don't like your behavior of criticizing people in a condescending way.

kyube wrote: ↑
15 Feb 2025, 19:59
I'm not ragebaiting. The entire thread title "RYZEN VS INTEL" is flawed to begin with and I didn't want to let anyone derail this topic any further into the vicious circle of misinformation.

This is clearly what you do, you are only stating either the obvious or false statements when you know no one here, neither myself, nor my choices and what I have done in the past, you have no information but you continue to assert in a condescending manner, this thread is very informative, maybe there are some users errors, I didn't say the opposite but there is feedback, human feedback and not paper BS test or latencymon test only.
If you want to make your opinion true and hearable for anyone, make an informative youtube channel, make a topic here instead of coming to bother those who have nothing to ask for, there is no misinformation here, just feedback, if you have to do 3 months of benchmarking, software tweaking, spending your nights doing the developers' work, then it's a bad product, nothing more, nothing less.

kyube wrote: ↑
15 Feb 2025, 19:59
Yes, your experience is tainted by bad choices in the hardware & software stack.
I don't spend "my life benchmarking & testing", I just want to stop misinformation based on clear user error.

And again you continue your false statements, what are you waiting for doing white knight on all the forums who complains every day about the problems of AMDip?

kyube wrote: ↑
15 Feb 2025, 19:59
If you want to go by your definition of "buy a product = it must work", have you looked at Intel's side?
They are shipping a product with a defective socket that causes instabilities (causing macrostuttering & microstuttering) and even rapid CPU degradation due to incorrect voltages out of the box.
AMD's X3D lineup is miles ahead in comparison when it comes to this regard.
I don't care about either of the two companies, I care about good products & the experience.
I wouldn't be replying to you if I didn't want to steer you into the proper direction.

I had to work with 5 Intel CPUs during these 5 years and i don't have any issues with "defective socket" in my case.
Some of issues that you said also happening at AMD, not to mention the 9800x3D which just burned recently from a guy watching his TV but so, you're really just fueling the "Intel vs AMD" aren't you?
"AMD's X3D lineup is miles ahead in comparison when it comes to this regard." This is just your personnal opinion on the subject.
I reassure you, you have neither need nor obligation to put “someone in the right direction”, once again, you have no idea what you're talking about me and others experience.

kyube wrote: ↑
15 Feb 2025, 19:59
Not a single youtube reviewer knows how to benchmark properly and actively deceives it's audience under layers of makeup.
Take a look at GamersNexus—3 videos on PresentMon w/ it's new parameters such as AnimationTimeError, GPUTime, GPUBusy, GPUWait and not a single benchmark since then with these specific metrics.
Not to mention the fact he only tests D3D12 titles, without testing OpenGL, Vulkan, D3D9-D11.
Not to mention the fact he hides 0.01% lows & finer granularity statistics, which he himself is pretty aware (even HWunboxed)
They all actively choose to hide it to end-users, even though it's the #1 problem-solving tool we have to troubleshoot these things.
The other content creators are even worse (yes, techyescity is also clueless)

More seriously, I just think you're the biggest Troll on BlurBustersForum and you've lost your mind: you said "90% of people here don't know anything" but now you're attacking people on Youtube and "Gamer Nexus".
You seem really frustrated about this subject, I won't answer the rest, that's not the subject and you are completely off base.

kyube wrote: ↑
15 Feb 2025, 19:59
I've mentioned the fixes to these issues many times on this sub-forum, you're free to read through my post history.
It's not even rocket science that I'm spouting, just basic good HW choices & good SW choices which are known in semi-open communities.

Continue, go ahead continue

kyube wrote: ↑
15 Feb 2025, 19:59
I'll even re-list them for you in a short format:

1.) Buy X3D with a decent 8-layer motherboard with a ALC1220 on it, with non-switched GPU & M.2 slots.
Ideally with a free PCIe x4 slot for potential expansions later down
Examples: X670e Steel Legend (2x PCIe x4 slots via CPU), X670e Aorus Master (2x m.2 slots via CPU)
2.) Buy non-RGB RAM which has been verified by the motherboard vendor. They have a list for this in their respective support page.
3.) Don't run PBO, Curves or touch anything in the BIOS that isn't verified by AMD's server performance guide documentation
4.) Buy a good PSU, Corsair only. RM-x, HX, AX series only
5.) Buy a good SSD. Optane (P4800x, P5800x, 905p), 990 Pro, KC3000, P44 Pro, SN850X, Crucial T500 (always opt for a SSD with DRAM, unordered list)
6.) Install W11 23H2 (important if Intel) or older releases, depending on games you play.
Good Windows releases: W7, W10 1803, W10 1809, W10 1909, W11 23H2
Run valleyofdoom's .reg file after install (follow his post-install for easiest setup)

1. 6 layers motherboard don't cause stuttering or latency issues.
2. What's wrong with RGB ? / QVL is NOT something guaranteed.
3. So people's like Optimum who are very close to the public would be wrong? even ITX enthusiast field that prefer to use Curve Optimizer? incredible.
4. Won't make anysense.
5. As long as you don't take something catastrophic/Dramless, no worries.
6. You don't have to take 23H2 for Intel, my 13600k worked very well without it but it's better than Windows 10, "valleyofdoom's .reg" is not necessary either.

At this point i just think you are just using chatgpt and throwing random information that either everyone knows or false subjective statement because nothing above is a universal "fix" for latency and stuttering.

kyube wrote: ↑
15 Feb 2025, 19:59
See? Very simple steps. From here, you change things accordingly.
XMP/EXPO can be unstable on some configurations, be aware of this.
^ This is the #1 cause of all "EMI, FLOATY MOUSE, STUTTER, NOT SMOOTH" threads.

Obviously.
That was not the case here but i guess you don't know what Karhu or Memtest is.

kyube wrote: ↑
15 Feb 2025, 19:59
OC enthusiasts are the exact opposite of what you should be looking at.
Their goals don't align with people chasing latency & "good feel". Not to mention the fact that 99% of them have questionable "skills".
Everyone, whether they perceive it or not, desire good latency.
Good, consistent & low latency makes up the "smoothness feel" that you want.

This is how they claim to be, I could very well have said latency enthusiast or any other bullshit name, at least, they are enthusiasts, not random people who just want to throw around their pseudo science and impose it on others.
A person who says "everyone is wrong, even YouTubers, even enthusiasts, even overclockers" will never be taken seriously, no one really takes you seriously Kyube, we only agree to reassure you.
If you want to be heard, be humble, post evidence, videos, create your youtube channel, show your face and make your voice heard, climb the ladder, there, sorry but you are just a troll who criticizes the whole world saying "they are wrong, I am right", just your arrogant response throughout has no respect, you could have shortened to your message in 6 lines it would amount to the same thing, knowing that you make an unbearable pile of sophistry to read, keep your "skills" in private and don't come to piss off others.

kyube wrote: ↑
15 Feb 2025, 19:59
I just asked a simple question on what your hardware & software choices were... nobody is asking of you to justify yourself?...

No, you first come by implying that I don't know the subject "skill issues" while you have no information on what I did (that's the point that make you frustrated), I hate people that make this kind of place a hypocritical writing blocks filled with hatred and sophistry by hiding behind your so-called "questions" and "skills knowledge", I answered you above that I wasn't coming for that and to leave me alone and you continue to insist with your lame “optimizations”, that's called lack of respect.
You are so mad because i already said that i tried everything with other people's on other website (that you maybe knows and that you maybe consider as "clueless" and you're frustrated that I don't refer to you)
Remember, you are NO ONE to judge anyone here or anywhere esle, I shared my experience like the others did, I told you again that I don't need help and you started to say that not everyone knows about it, ok, we understand, you think you are above everyone by suffering from Dunning Kruger's symptom, you seem to have a problem with me, too bad, we could have avoided this if you would have come in a more intelligent way at first, I will avoid any contact with you in the future.

1. Rule #1: Be Nice. This is published forum rule #1. Even To Newbies & People You Disagree With!

Posted: **16 Feb 2025, 10:35**

Sirius wrote: ↑
16 Feb 2025, 10:14

kyube wrote: ↑
15 Feb 2025, 19:59

You can disagree with each other all you want, but do not start with personal attacks. Stick to tests, benchmarks and opinions about hardware.

Posted: **16 Feb 2025, 11:59**

So are just all streamers and pros playing on AMD extremely lucky and don't see any issues that would block them from playing?

Posted: **16 Feb 2025, 16:46**

Good evening, colleagues. I am glad to see you all.
I would like to share my opinion about the electrochemical hardware .
Recently, more and more often, test results and conclusions are distorted, electrochemical hardware located between the chair and the keyboard needs to be adjusted more precisely. It is necessary to develop a method for fine-tuning this electrochemical hardware. Taking into account the wishes of the administration. (but do not start with personal attacks.)
The horrific quality of the electrochemical hardware offends me to the depths of my soul. The agony, pain and suffering.

Sirius wrote: ↑
15 Feb 2025, 16:51
I know I’m not crazy because others have described the same "mouse latency feeling."

I know I’m not crazy because others have described the same "Aliens/Bigfoot."

Please share your opinion.

Posted: **17 Feb 2025, 16:10**

witega wrote: ↑
16 Feb 2025, 02:25
I agree with everything you said but I feel this part of your advice deserves more elaboration. As you said previously, most people are clueless, and if you suggest them buy RAM on the QVL list from the motherboard manufacturer, that doesn't guarantee it's going to be stable (say if you bought something like an approved 8800mt/s EXPO kit) and you're back right to the problem where people have bad 1%/0.1% lows, stuttering, floaty mouse, etc.

And then you have RAM modules that have XMP/EXPO profiles with too high voltages (like 1.35V or greater for SOC).

I appreciate your post for being no-nonsense, straight to the point and easy to follow, but the RAM I feel is super important because it affects performance a lot, especially in the lows.

Since 9800x3d is the hot new chip for gaming, what would you suggest as a guidance for RAM kits speeds to look for and primary timings? Obviously you want stability and lower latency and smoothness but without the need for extreme overclocking.

I fully agree with what you said, you're right.
RAM is the most important aspect which can make or break your experience.
It's why I also mentioned this:

...Very simple steps. From here on now, you change things accordingly.

I've only mentioned the bare minimum for newcomers to modern HW components, such as Sirius, have to do in order to have a good experience.
We're all served borderline unstable CPUs out of the box, as seen with the Intel Raptor-Lake CPU debacle.
(btw, I haven't seen any AM5 boards list 8800 modules for their QVL, at least for 2DPC boards, just wanted to mention this)

In terms of kits themselves, it's always best to follow a “lazy” approach, especially if someone doesn't have the willpower/ time to do in-depth benchmarking in various ways.
Changing frequency to 6000MHz (MCLK=3000 MHz), FCLK=2000 MHz and copying the XMP's primaries while not touching secondary or tertiary timings.
If this does not work (some XMP's can be “XOC”), I usually apply this:

NOTE: This is not my template, I've come across it only. I do not claim to be the author of this quoted text.
Do not apply it blindly, it's only anecdotal information from a X3D user

FCLK: 2k (its really rare but some cpus can do only 1800)
Ram freq: 6k (its really rare but some cpu cant do 6000mhz, x3d seems to fail more)
UCLK=MCLK
GDM enabled, DFE Read Training enabled, DDR5 Robust Training Mode enabled, DDR Training Runtime Reduction disabled (these are usually defaults)
Nitro 2-3-1 8x 8x (should be default on msi, dont touch it at all on msi otherwise u will have issues)
Memory context restore/Power down both enabled or both disabled at the same time, otherwise unstable (latest bios asus = MCR might be broken/some gigabyte bios = broken PDM too, the best strategy on all boards is to set both to disabled. was fixed on 9000)
DF C-states: disabled (should be "safe" to disable)
CPU Speculative Store Modes: More Speculative (test this one)
Soc uncore OC mode: enabled

tCL: convert from ur xmp (test with RTP beforehand since unstable xmp is a thing; should scale with V a bit)
tRCD-tRP: same as above (temp dependant, scale with V a bit)
tRAS: tRCD+tRTP (u will start to have issues with tras/trc if u go below this one) / tRCD+tRTP+8 ? / tCL+tRCD+8 (made up shit xd) / 32ns (jedec)
tRC: tRAS+tRCD
tRRD_s: 8
tRRD_l: 12 (should be safe for 24gb m-die: 14-16?)
tFAW: 32
tREFI: 12k (some RAM cant even run 16K+ (verified))
tRFC: 512 (scales with dram voltage, but it doesnt really matter with 170ns. 130ns works fine for me though for a long time, keep in mind the limit is way higher on mdie)
tWTR_s: 10/12 (even auto/xmp values can be unstable so we have to raise this one (verified), stuff like tWTR_s = 4 on 6000mhz = insta xoc, test with RamTestPro)
tWTR_l: 26
tRDRDSCL: 8 (lower = no profit/will do nothing)
tWRWRSCL: 8 (16 = 100% safe?)
tWR: 72 (maybe 48 here, test)
tRTP: 18 (maybe 12 here, test)
tCWL: tCL-2
tRDWR: 20
tWRRD: 4 (auto probably better)
everything else auto

vSOC: 1.25V (depends on ur bin a lot, 1.25v should work on any cpu? higher vsoc adds significant W consumption)
VDDIO>0.1V+vSOC (asrock bios formula, practice shows that it is desirable but not necessary. VDDIO>=VSOC)
DRAM VDD= (depends on ur bin/xmp basically)
DRAM VDDQ=VDDIO: (depends on ur bin/xmp basically)
use fixed VDDP (0.95v/1v) /VDDG's (0.9-1v) / VDD_misc (1.1v)

This text above is heavily inspired by JEDEC's 6000MHz profile and personal testing of a user who's more oriented in eliminating jitter caused by unstable HW.
JEDEC DDR5 documentation is a great point to start from.

One step towards the direction of safety is to choose AUTO:1:1 (FCLK:UCLK:MCLK), as per ex-AMD Robert Hallock's early Zen4 recommendation
For someone stability oriented, Robert suggested going for a 5200MHz "low CL" kit, as some CPUs can exhibit instabilities at these high frequencies.
One can always get a higher binned kit and lower the clock rate manually.

The final step you can take in the direction of maximum data integrity while foregoing RAM performance (bandwidth & R/W latency) is to opt for ECC UDIMM sticks on capable motherboards (ASUS, Asrock, Gigabyte)
I'm not sure whether they're SECDED or Reed-Solomon based ECC on AM5/DDR5...

“XOC” stands for “Extreme Overclocking” and is a term used to indicate instability in a user's system.
The effects of XOC can vary in severity, and they can be categorized as follows:
BSOD (Blue Screen of Death)— This is the most common and easily recognizable issue, indicating a critical system failure.

Macrostuttering—This is a noticeable performance issue that can be observed during gameplay or video playback. It can be detected using tools like PresentMon and may appear during some stress tests, but it’s not always guaranteed to show up.

Microstuttering—This is a more subtle and complex issue that often goes unnoticed in standard stress tests. It manifests as a lack of smoothness in motion, which can include symptoms like erratic mouse movement or inconsistent frame delivery. This one is harder to diagnose and requires a more rigid methodology to fix.

Disclaimer: Please note that this explanation is not official and the terminology used may vary. Different communities or individuals may refer to these effects using alternative names or classifications.

Always start from JEDEC's profiles as baseline, but don't treat it as unfallible
Don't treat stress tests as a 100% assurance that you're stable, all of them still miss errors

A good stress-test for RAM that I rarely see mentioned online is RamTestPro (rule above applies)
Other good stress-tests include:
GSAT (AMD's internal testing app sw of choice, as per Robert Hallock)
HCI MemTest (identical to RunMemtestPro according to RTP dev, used by V-COLOR internally for binning)
ycruncher,
TM5 (different configs, I avoid usmus & anta)
OCCT (can be CPU & FCLK-oriented)
Prime95 (can be CPU & FCLK-oriented) such as small FFT, large FFT (ok for mem. controller stability testing), Blend (also for IMC)
IntelBurnTest (ok for CPU testing, as per Robert Hallock. YMMV, I personally don't use it)

I avoid Karhu's test, as I find it unreliable. It seems to be written by an amateur, not preferred for serious stress-testing of RAM.
MemTest86 is not good for DDR4 and larger capacity era.
Linpack supposedly does not work on AMD, but I might be wrong here.
Never use AIDA64 to gauge latency benefits, use CLAM benchmark or RamTestPro's benchmark
One can also run a few app sw from an bootable Linux ISO, such as StresKit, which contains a set of these tests inside of it.
This is the most optimal approach to do, to avoid any corruption of the OS

A few other resources I've come across:
I always recommend people to refer to AMD's documentation or guides, such as:
https://www.youtube.com/watch?v=y8ezs_Si5nk — "GDC 2024 - AMD Ryzen™ Processor Software Optimization"

https://www.amd.com/content/dam/amd/en/ ... server.pdf - windows-specific tidbits

https://www.amd.com/content/dam/amd/en/ ... g-java.pdf

https://www.amd.com/en/developer/uprof.html — profiling tool by amd

EDA workloads are mostly ST, lightly MT

SMEE, TSME, DF C-state, SMT Control (depends on game), uCLK DIV 1 MODE, PCIe/FCH Spread Spectrum, IOMMU are settings I usually play around with.
PBO use should be avoided. “Static” clock rate & static Vcore is preferred for least theoretical jittering. Sticking to "Auto" can also be desirable or playing with voltage offsets can also be considered. YMMV.
(refer to AMD server docs for more information on each setting)

For DDR4 (AM4) rigs, this read may be beneficial:
https://web.archive.org/web/20201108001 ... -talk-dram

I hope this helps for anyone wanting to dive into optimizing the AM4/AM5 platform.

EDIT: I shall add these two quotes, which I find necessary to include for future references:

shrimps wrote: No amount of memory stress testing is enough, stress testing is just an approximation of stability.
Learn patterns of your memory/imc and apply an overclock that not only passes tests, but also has headroom for when conditions change.

RTP_Dev wrote: one note here, testing RAM is relatively straightforward. But testing the memory controller requires deep knowledge of the cpu’s architecture and a custom approach to apply the right kind of load by writing your own code. In practice, doing this properly is almost impossible. Most existing tests aren’t optimized for specific architectures. So you’re left with experimentation.
For example, Ryzen memory controllers are very sensitive to cache eviction (surprising isn’t it?). But to take proper advantage of that, you need asm skills and a lot of time… — and even then, there’s no guarantee it’ll be worth it

so even if you run every stress test, there’s still no guarantee that for example your cpu’s mc won’t fail, simply because there are no tools that can fully verify its stability

Posted: **23 Oct 2025, 03:19**

kyube wrote: ↑
15 Feb 2025, 19:59
I've mentioned the fixes to these issues many times on this sub-forum, you're free to read through my post history.
It's not even rocket science that I'm spouting, just basic good HW choices & good SW choices which are known in semi-open communities.

Multiple thoughts on this misinformation...

So recently I picked up 9800x3d, twice, same mobo however unfortunately, asus b850-g, although 1. I doubt that makes much difference between x890 & 2. if b850 experience can be this bad an argument of a so called poor 'research'/'choice' is an obscene one to make and likely bs in the first place and the same issues probably apply to a bunch (if not all) x890 chipsets as well.

I'm completely in agreement with Sirius that this must be some kind of chiplet + sh** kernel/sys devs on the part of amd. The most comical part is that I now can remember when I had 5800x that there were moments where I didn't understand where an occasional weird latency feel was coming from and mind that I had 5800x paired with used 3090 and at the time I just figured it might have either been 3090 or just bad interop between amd & nvidia. And ofc, before 5800x, I had intel 4790k/5960x... Also PC cafes I go to with my friends - intel systems. However, with 5800x system I just shrugged it off as it was somewhat bearable.

Now fast forward to today, 9800x3d, I infact think it's gotten worse, now like you just plug this sh** in and it's f*** obvious that something's wildly out of sync. And hey, no wonder, you look at all these modular amd system design diagrams and you see how there is like 4 different sections/chips that all run at different frequencies and somehow need to all f*** get in line eventually. I mean compare this to an intel chip for response and e2e consistency, good luck... Ok to clarify what I'm really referring to is on amd from what I understand the signal path is roughly: usb controller -> i/o die serialize/deserialize -> infinity fabric -> core. On intel it's usb controller -> ringbus -> core. And the thing is that some of the sections on i/o die have different clocks. There is i/o clock which is 1000mhz or so, FCLK that like tops out around 2000mhz realisticly. On latest monolithic intel there is just a single ringbus and it clocks above 4ghz.

Anyways, first I started off with win 10, I was hella confused as soon as I plugged it in to get this weird mouse delay feel out of nowhere. I'd install necessary drivers n all but it wasn't getting better. Then after tinkering a bunch with mem oc that would only get the system more unstable, and even moments where the mouse would actually start to lag - so frankly much of that didn't really get it to a better level in terms of latency.

Now, kind of an ironic part is that trying win11 maybe did improve 1-2ms input in the compositor but at a cost of higher base dpc oob, but the issue still persisted. Like literally, I'm using my keyboard to type this message right now and I swear to god the input on screen happens like 2-6ms later than on intel systems. I'm literally getting the sensation in my fingers and my body is thrown off at every moment when it doesn't get the proper timely e2e response. Like something's odd.

Ofc how can I tell the difference today even given the old systems that I talked about? Well atm I have a 13980hx laptop and also I briefly just a couple days ago tried i7 14700kf with z790 mobo. The only issue why I returned the intel pc platform is because the mobo I got was horrible with mem stability @ 7200mhz and I indeed did miss higher frames in some games specific to me + wtf is 150-200w for base performance.

However I can't overstate how wild of a difference there is in latency between these two platforms. And honestly like, for example I could imagine how in windows 10, CS2 specifically was borderline acceptable due to all the improvements that the valve devs made & nvidia specifically made over the years to patch all the sh**ty microsoft bloat. And hey, it works I can tell, even at the time of writing on win11 the irony of the fact that gsync enabled yields slightly lower latency in compositor than off. Like wtf.

Now I've completely given up on trying to find a magical solution in software so I decided to get a pci usb extension with VIA805 chip plugged in directly to CPU pcie. Because I'm losing trust now in the entire amd stack end-to-end.

You have to understand that the biggest failure here besides the tech grifer-nomics of playing both sides is that the companies themselves are okay with putting out disguising trash and call it applicable for use-case X that's as clear as day. It's like nvidia claiming 4090 performance with so called 'fps' counter being equiv on 5070. Yeah great, fps is the same, but you get scammed on: image quality, latency.

Also a quick note about techtubers. I can say one thing, regardless of how intentional or not is their lack of coverage is on this issue. I'm certain and positive that just doing a e2e click test is a lost cause. For example: The Finals on both win10 and win11. On win10, exclusive fullscreen mode doesn't work anymore because UE did some changes that for some reason now only work on win11, ie even if it's in fullscreen exclusive it still defaults to borderless. In win11 however fullscreen exclusive does work. So why am I saying all that. What am I trying to get to? Dude... I swear to god, if you are playing The Finals on borderless fullscreen on an amd platform you are done, cooked, finito. I mainly play on scout/sniper and let me tell you. If you are trying to do quick movements in borderless and you can tell how rapid mouse movements somehow feel like they cancel out or just wildy f***ed and it's not a framerate or even % lows issue. I can have doggo fps on an intel platform but the mouse will always land and transition from A to B perfectly. Here is the kicker, on win11 if I switch to fullscreen exclusive, most of that goes away. But latency is still there. Now, I can't thank the universe enough that gsync exists and due to nvidia's actual progress, to some extent gsync with fullscreen masks some of that disguising performance. But I can't overstate enough that this type of problem domain simply is not a thing on intel. When something is done right you don't notice it.

And I just have to remind some of the folk here. You may be getting billion fps in a game, but if you move a mouse and that buttery smooth movement occurs a second later trust me - your brain will have a crash out.

Anyways, there is much more I could say but overall what I wanted to conclude and add to convo mainly is this:

1) I completely reject anyone saying that the issues talked in here is a user error somewhere along the lines of picking a bad setting or making a bad choice during purchase of one of the parts. The only bad purchasing decision in 2025 is getting amd frankly. And I'm not saying intel is amazing. 200W+ for base performance is poor.
2) I specifically avoided entirely talking about 1% lows & 0.1% lows because genuinely in my view, there are ways to improve the situation with lows, but there are much less obvious ways to improve latency in a real way.
3) techtubers are churn boomers that don't have time to actually use the platforms they review. And also it's either likely part of a larger grift economy where if you don't play both sides you lose or unintentional fear mongering effort where not shilling amd (the small bean) will get intel (the greedy, bully one) somehow too powerful and the world will suddenly turn into an Orwellian hellscape. Realistically, nobody should care which company it is, the whole notion of team red blue green team is a 1st grader book coloring exercise.
4) I'm not interested in discussion with someone who is going to try to tell me how you are gonna get getting baseline stability at JEDEC w/ latency / input consistency.
5) I'm not interested in discussing how toggling C-States or enabling XMP profile will change your life or alike. Simple answer, if intel with similar oob settings can just work so should amd. Why the f*** does one need a phd in electronics to figure this out, but see, it's a failed statement already. No amount of tinkering will fix the lacking architectural decisions that amd's made.
6) The fact that amd still hasn't acknowledged this reality is deeply concerning

Posted: **23 Oct 2025, 23:10**

bbbe wrote: ↑
23 Oct 2025, 03:19

kyube wrote: ↑
15 Feb 2025, 19:59
I've mentioned the fixes to these issues many times on this sub-forum, you're free to read through my post history.
It's not even rocket science that I'm spouting, just basic good HW choices & good SW choices which are known in semi-open communities.
Multiple thoughts on this misinformation...

So recently I picked up 9800x3d, twice, same mobo however unfortunately, asus b850-g, although 1. I doubt that makes much difference between x890 & 2. if b850 experience can be this bad an argument of a so called poor 'research'/'choice' is an obscene one to make and likely bs in the first place and the same issues probably apply to a bunch (if not all) x890 chipsets as well.

I'm completely in agreement with Sirius that this must be some kind of chiplet + sh** kernel/sys devs on the part of amd. The most comical part is that I now can remember when I had 5800x that there were moments where I didn't understand where an occasional weird latency feel was coming from and mind that I had 5800x paired with used 3090 and at the time I just figured it might have either been 3090 or just bad interop between amd & nvidia. And ofc, before 5800x, I had intel 4790k/5960x... Also PC cafes I go to with my friends - intel systems. However, with 5800x system I just shrugged it off as it was somewhat bearable.

Now fast forward to today, 9800x3d, I infact think it's gotten worse, now like you just plug this sh** in and it's f*** obvious that something's wildly out of sync. And hey, no wonder, you look at all these modular amd system design diagrams and you see how there is like 4 different sections/chips that all run at different frequencies and somehow need to all f*** get in line eventually. I mean compare this to an intel chip for response and e2e consistency, good luck... Ok to clarify what I'm really referring to is on amd from what I understand the signal path is roughly: usb controller -> i/o die serialize/deserialize -> infinity fabric -> core. On intel it's usb controller -> ringbus -> core. And the thing is that some of the sections on i/o die have different clocks. There is i/o clock which is 1000mhz or so, FCLK that like tops out around 2000mhz realisticly. On latest monolithic intel there is just a single ringbus and it clocks above 4ghz.

Anyways, first I started off with win 10, I was hella confused as soon as I plugged it in to get this weird mouse delay feel out of nowhere. I'd install necessary drivers n all but it wasn't getting better. Then after tinkering a bunch with mem oc that would only get the system more unstable, and even moments where the mouse would actually start to lag - so frankly much of that didn't really get it to a better level in terms of latency.

Now, kind of an ironic part is that trying win11 maybe did improve 1-2ms input in the compositor but at a cost of higher base dpc oob, but the issue still persisted. Like literally, I'm using my keyboard to type this message right now and I swear to god the input on screen happens like 2-6ms later than on intel systems. I'm literally getting the sensation in my fingers and my body is thrown off at every moment when it doesn't get the proper timely e2e response. Like something's odd.

Ofc how can I tell the difference today even given the old systems that I talked about? Well atm I have a 13980hx laptop and also I briefly just a couple days ago tried i7 14700kf with z790 mobo. The only issue why I returned the intel pc platform is because the mobo I got was horrible with mem stability @ 7200mhz and I indeed did miss higher frames in some games specific to me + wtf is 150-200w for base performance.

However I can't overstate how wild of a difference there is in latency between these two platforms. And honestly like, for example I could imagine how in windows 10, CS2 specifically was borderline acceptable due to all the improvements that the valve devs made & nvidia specifically made over the years to patch all the sh**ty microsoft bloat. And hey, it works I can tell, even at the time of writing on win11 the irony of the fact that gsync enabled yields slightly lower latency in compositor than off. Like wtf.

Now I've completely given up on trying to find a magical solution in software so I decided to get a pci usb extension with VIA805 chip plugged in directly to CPU pcie. Because I'm losing trust now in the entire amd stack end-to-end.

You have to understand that the biggest failure here besides the tech grifer-nomics of playing both sides is that the companies themselves are okay with putting out disguising trash and call it applicable for use-case X that's as clear as day. It's like nvidia claiming 4090 performance with so called 'fps' counter being equiv on 5070. Yeah great, fps is the same, but you get scammed on: image quality, latency.

Also a quick note about techtubers. I can say one thing, regardless of how intentional or not is their lack of coverage is on this issue. I'm certain and positive that just doing a e2e click test is a lost cause. For example: The Finals on both win10 and win11. On win10, exclusive fullscreen mode doesn't work anymore because UE did some changes that for some reason now only work on win11, ie even if it's in fullscreen exclusive it still defaults to borderless. In win11 however fullscreen exclusive does work. So why am I saying all that. What am I trying to get to? Dude... I swear to god, if you are playing The Finals on borderless fullscreen on an amd platform you are done, cooked, finito. I mainly play on scout/sniper and let me tell you. If you are trying to do quick movements in borderless and you can tell how rapid mouse movements somehow feel like they cancel out or just wildy f***ed and it's not a framerate or even % lows issue. I can have doggo fps on an intel platform but the mouse will always land and transition from A to B perfectly. Here is the kicker, on win11 if I switch to fullscreen exclusive, most of that goes away. But latency is still there. Now, I can't thank the universe enough that gsync exists and due to nvidia's actual progress, to some extent gsync with fullscreen masks some of that disguising performance. But I can't overstate enough that this type of problem domain simply is not a thing on intel. When something is done right you don't notice it.

And I just have to remind some of the folk here. You may be getting billion fps in a game, but if you move a mouse and that buttery smooth movement occurs a second later trust me - your brain will have a crash out.

Anyways, there is much more I could say but overall what I wanted to conclude and add to convo mainly is this:

1) I completely reject anyone saying that the issues talked in here is a user error somewhere along the lines of picking a bad setting or making a bad choice during purchase of one of the parts. The only bad purchasing decision in 2025 is getting amd frankly. And I'm not saying intel is amazing. 200W+ for base performance is poor.
2) I specifically avoided entirely talking about 1% lows & 0.1% lows because genuinely in my view, there are ways to improve the situation with lows, but there are much less obvious ways to improve latency in a real way.
3) techtubers are churn boomers that don't have time to actually use the platforms they review. And also it's either likely part of a larger grift economy where if you don't play both sides you lose or unintentional fear mongering effort where not shilling amd (the small bean) will get intel (the greedy, bully one) somehow too powerful and the world will suddenly turn into an Orwellian hellscape. Realistically, nobody should care which company it is, the whole notion of team red blue green team is a 1st grader book coloring exercise.
4) I'm not interested in discussion with someone who is going to try to tell me how you are gonna get getting baseline stability at JEDEC w/ latency / input consistency.
5) I'm not interested in discussing how toggling C-States or enabling XMP profile will change your life or alike. Simple answer, if intel with similar oob settings can just work so should amd. Why the f*** does one need a phd in electronics to figure this out, but see, it's a failed statement already. No amount of tinkering will fix the lacking architectural decisions that amd's made.
6) The fact that amd still hasn't acknowledged this reality is deeply concerning

I've made an account on here just to say i have the exact same issues as you.
9800x3d
4090
X870E Aorus Pro Ice Rev 1.0

Its been driving me insane for about 5 months ever since i replaced by 14900k which was previously replaced twice, you mention about the youtubers i agree, it took them a year to clock on that there was an issue with the 14900k and the other processors effected. Everyone was just shrugging those issues of as well blasting stupid ideas out like disable C-States same way they're doing now for the 9800X3D its only a matter of time before this becomes a widespread issue all of a sudden. The amount of people having the same problems is crazy and its not the boards either since it happens across various manufactures B series and X series

I'm debating buying a PCIE card myself although i think its a greater issue than that. Were just little mice to AMD and boards so theres no way we can really raise this issue but the state of PCs for the last few years has been a joke, its completely moved away from the Plug n Play aspect. I shouldn't have to mess with BIOS to make my PC stable, even after messing with BIOS it leads to instability in other areas so you cant win lol

Blur Busters Forums

Ryzen vs Intel Input Lag/Latency

Re: Ryzen vs Intel Input Lag/Latency

Re: Ryzen vs Intel Input Lag/Latency

Re: Ryzen vs Intel Input Lag/Latency

Re: Ryzen vs Intel Input Lag/Latency

Re: Ryzen vs Intel Input Lag/Latency

Re: Ryzen vs Intel Input Lag/Latency

Re: Ryzen vs Intel Input Lag/Latency

Re: Ryzen vs Intel Input Lag/Latency