Is it true Ryzen has higher input lag than Intel? Is there conclusive data to prove this?

[purist]

Memory latency is arguably the biggest bottleneck in interactivity in modern systems. It's just hidden to most behind frame buffers, poorly designed operating systems, system hardware & their drivers, peripheral latency - and then overlooked.

The benefits of lower memory latency extend beyond benchmarks in 2D and 3D games and can be seen on a single system without comparing between platforms. Simply tightening certain memory timings will lead to a more responsive experience - even secondary and advanced timings. Aiming becomes more natural as cursor/cross-hair movements are closer to where we intended them to go - and in less time.

It's up to the user to determine whether or not the change had an effect on their quality of usage. The user's level of coordination/skill and experience in interactivity-reliant tasks or games will determine how easily they are able to notice a change. Most users do not have a way of properly quantify or provide evidence of how these changes effect a system. They cannot be revealed by mere tests with 1 mouse click and measuring the time it takes to see a response to that single event on the screen.

Lowering CAS Latency by a mere 1 step results in a very noticeable increase in system responsiveness. Other, seemingly less important, timings too also impact system responsiveness (Some lesser known: tRRD, tWR, even small reductions in tRFC provide a noticeable change. Having properly set RTL/IOL timings is important, and some advanced timings that are hidden in the majority of Taiwanese vendor boards are critical to change. Unfortunately Intel hides many more IMC timings and leaves no easy way to change or even view them).

These are all changes that only slightly alter the results in latency that show up in AIDA64 and other micro-benchmarks. Intel's memory latency checker provides a slightly more in-depth analysis, but is still not all-revealing:

https://software.intel.com/content/www/ ... ecker.html

No tool currently exists that allows for easily calibrating memory timings optimally for specific gaming workloads.

Linus Torvalds has a tool on Github but I've yet to use it: https://github.com/torvalds/test-tlb

Hunting for memory overclocking scores, you'll occasionally find a forum user that states their PC felt more responsive after changing a certain timing. It's rare to see in a scene where the participants are more than likely not so coordinated, but suggests memory does impact interactivity significantly.

By taking interest in memory overclocking and experiencing how changing timings affects quality of system usage, you learn quickly how Ryzen having 2x the memory latency of a tuned Intel system is very hindering. Ryzen is the byproduct of processors intended for the server industry.

Nanosecond results in benchmarks accumulate to much more.

[1000WATT]

Calypto
For me, decreasing DPC latency values. The goal was one - to raise 1% low framerate and 0.1% low framerate. And the difference sometimes reached 60%. But I did not feel any changes in the input delay. (old age ).
Can you record a video example of a working d3d and LatencyMon? So that we can make sure that Windows does not work in safe mode and that the gpu driver is enabled.

[Anonymous768119]

@purist

So you want to tell us, that changing from 3200@CL16 to 3800@CL14 would make a significant difference in responsiveness and mouse movement? I was playing little bit with frequency and timings and honestly I haven't noticed any changes beside FPS. You write quite bold theses not supported by any surveys BUT if there are some, don't hesitate to share with us.

[schizobeyondpills]

@purist

So you want to tell us, that changing from 3200@CL16 to 3800@CL14 would make a significant difference in responsiveness and mouse movement? I was playing little bit with frequency and timings and honestly I haven't noticed any changes beside FPS. You write quite bold theses not supported by any surveys BUT if there are some, don't hesitate to share with us.

to notice such changes your entire system both in hardware(BIOS/CPU/GPU/RAM/display) as well as software (OS/interrupts/debloated/performance plan) need to be tuned/debloated since latency counts from start to finish of the pipeline and adjusting/improving something wont be noticeable if its bottlenecked by other items in the latency pipeline.

@purist didnt state his opinion about this, its clear objective logic backed with facts, every single CPU from AMD and Intel have 3 ( 1 2 3 ) levels of cache(which reduces RAM latency by storing data closer to CPU) between CPU and RAM, however due to their small size being on die they help but reducing RAM latency for read/write is beneficial for all cases which dont hit the cache.( 3900X has 64MB of L3 cache, while game usually uses from 2GB to 16GB of RAM, see how big the ratio of cache:RAM is??)

so to recap

arguments for RAM latency nanoseconds being useless
- people in this thread dont feel it on their unoptimized setups

arguments for RAM latency nanoseconds having huge effect
- billions of dollars poured into R&D to reduce it, result is 3 small sized CPU caches on die
- every single CPU in last 5-10 years has not 1, not 2, but 3 levels of cache to shave off nanoseconds
- Calypto posted proof how worse CPU with GPU eDRAM, "the 5775C is capable of beating the 6700K in average FPS as well as % lows, thanks to its 128MB eDRAM/L4." beats better CPU due to cache.
- paper about CPUs from 20 years ago attached below which shows that even back then CPUs were starved for more data from RAM, 20 years later we have 8+ cores of 4+ GHz CPUs, RAM latency matters now more than ever.

[Anonymous768119]

Quite interesting and amazing how our skill is dependant on our hardware. Doesn't matter how much we spend for our rig, there will be something that is bottlenecking us. It gives food for thought how many players wasted their potential because they were not aware how much bad "synergy" between RAM and CPU can interfere with their hitreg, reaction etc. I don't want it to sound like showing off but I have pretty good aiming skills and I finish games with highest HS ratio but I feel like I am somehow limited. Just out of curiosity I was watching twitch cs:go streamers and I was astonished how easy they get kills on highest level even if they don't seem to be so focused.

In the past I was in inet cafe for gamers and I could not belive, that I can peek with deagle with such precision. After that I bought my PC and I tried to do the same (expecting same results) and I was like 'what the hell?'. Everything was so sluggish and my movement rather resembled ice-skating than sharp and responsive moves.

Another puzzling case is that almost every pro player is using Intel for CS:GO. Is this a coincidence?

[schizobeyondpills]

Quite interesting and amazing how our skill is dependant on our hardware. Doesn't matter how much we spend for our rig, there will be something that is bottlenecking us. It gives food for thought how many players wasted their potential because they were not aware how much bad "synergy" between RAM and CPU can interfere with their hitreg, reaction etc. I don't want it to sound like showing off but I have pretty good aiming skills and I finish games with highest HS ratio but I feel like I am somehow limited. Just out of curiosity I was watching twitch cs:go streamers and I was astonished how easy they get kills on highest level even if they don't seem to be so focused.

In the past I was in inet cafe for gamers and I could not belive, that I can peek with deagle with such precision. After that I bought my PC and I tried to do the same (expecting same results) and I was like 'what the hell?'. Everything was so sluggish and my movement rather resembled ice-skating than sharp and responsive moves.

Another puzzling case is that almost every pro player is using Intel for CS:GO. Is this a coincidence?

Yeah, the faster you go on the road, the more smaller things such as shrapnels,debris or fly hitting you affect you, same can be said for CPU running at 5 billion cycles a second(per core), and then add in GPU/RAM. Doesnt matter how much you spend because its optimized for top 50% use case, meaning, to be stable for all possible use cases including overheating, all possible boards and their variations in quality, all possible RAM variations combined with motherboards, combined with variations and different wattages/quality of PSU, so that 50% is optimized to cover all that.

All of your local hardware and software tweaks fall apart if your internet has a lot of latency, jitter, packet delay/coalescing/packet steering on ISP level or at server/datacenter/routing level. CS:Go is a good game to use as example of that due to it being 64/128 tickrate and from personal experience its very very sensitive to internet quality. (theres a reason why a lot of pro players are from scandinavia ^_^, and its not viking blood). Internet is optimized for TCP and bandwidth to serve content rather than low latency and UDP which is complete opposite of TCP/bandwidth. So you can be playing a match and on your local ISP exit, the ISP can decide to delay your packet in round robbin fashion because your neighbour or anyone on the same node has higher priority than you, if you think about it, why would you be able to send/receive 128 packets a second and your neighbour wouldn't get priority over you to send/receive 1 packet?
Then there's routing both from you to ISP, from ISP to datacenter from datacenter to server through 10 firewalls/load balancers/local networks in rack, until it finally hits the server (if its even hosted on a dedi rather than VPS). And I didnt even mention electricity/error correction or any other thing which there's plenty of, but i cba. Also a lot of ISPs pull tricks to combine packets to show higher download speeds since they are measured in seconds rather than in ms for high tickrate servers(I was even a victim of this, never ever ask your ISP to tune your internet for increased speeds).

Also I should add that the miliseconds in your game and/or what you see in traffic analysis tools locally @ your home doesnt really mean much since they dont show even 1% of what actually happens with the packet when it leaves your house.

Also your internet can affect input lag and FPS, A LOT. Because some packets may arrive, some may not, some may arrive out of order, all of those things can cause game engine to discard what it processed and start over, causing stutters, FPS drops, microstutters, input lag. Depending how its programmed/architected.

[ptuga]

I've noticed this aswell, having tight ram makes the mouse feel more snappy, and I can easily benchmark it using kovaaks. It's not placebo.

But having memory tweaked to it's limits is not the best either, mouse can start to feel inconsistent even when stress test don't show instability.
I've found tcl, rrd&tfaw and trfc to have the biggest impact.
Lower twr on my system gives me less consistent mouse movement, also doesn't improve fps at all.
Higher trefi gives me higher fps but can me mouse sometimes slower/faster while being stable on stress tests.

I'm running a 8700k with 2*8gb tridentz 3600mhz cl16.

[Brainlet]

I've noticed this aswell, having tight ram makes the mouse feel more snappy, and I can easily benchmark it using kovaaks. It's not placebo.

But having memory tweaked to it's limits is not the best either, mouse can start to feel inconsistent even when stress test don't show instability.
I've found tcl, rrd&tfaw and trfc to have the biggest impact.
Lower twr on my system gives me less consistent mouse movement, also doesn't improve fps at all.
Higher trefi gives me higher fps but can me mouse sometimes slower/faster while being stable on stress tests.

I'm running a 8700k with 2*8gb tridentz 3600mhz cl16.

Yes, RAM affects mouse input a lot.
Stress tests generally don't seem to detect error correction which can lead to really bad mouse input. For starters I suggest ditching AIDA as the main "on paper" benchmark tool and using https://software.intel.com/content/www/ ... ecker.html instead.

[ptuga]

Yes, RAM affects mouse input a lot.
Stress tests generally don't seem to detect error correction which can lead to really bad mouse input. For starters I suggest ditching AIDA as the main "on paper" benchmark tool and using https://software.intel.com/content/www/ ... ecker.html instead.

In Kovaak's close long strafes invincible i get probably +10% score with tweaked ram. I can get to 70% acc and 16k+ points farily easy, but with xmp i can't break 60ish%. It's pretty staggering and i don't understand how most people don't notice this.

I still feel the mouse feel it's holdming back a little, my previous system had better mouse feel to it and my aim felt more crisp. It's probably my motherboard, it's and msi z370m gaming pro...

[Anonymous768119]

FYI I changed mobo from B450 Tomahawk to Asus TUF Gaming X570 Plus and I didn't notice difference in mouse movement which is sluggish and inaccurate.

I don't have comparison so can you summarize my results?

I have Ryzen 7 3700X, HyperX Predator 3000MHz 15-17-17-36 (XMP), Samsung SSD 970 EVO Plus if it helps you with something.